article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
the particle spectrum is a basic quantity directly measured in the experiments and it can reveal the information of particle production mechanism in heavy - ion collisions .
many physicists have devoted themselves to studying the particle spectra produced in the heavy - ion collisions using thermodynamical approaches , phenomenological methods , transport models _
recently , the tsallis distribution , which was first proposed about twenty - seven years ago as a generalization of the boltzmann - gibbs distribution @xcite , has attracted many theorists and experimentalists attention in high energy heavy - ion collisions @xcite .
the excellent ability to fit the spectra of identified hadrons and charged particles in a large range of @xmath2 up to 200 gev , which covers 15 orders of magnitude , is quite impressive .
this spectacular result was first shown by wong _
_ @xcite . in refs .
@xcite , we have shown that tsallis distribution can fit almost all the particle spectra produced in p+p , p+a and a+a at rhic and lhc . from the phenomenological view , there may be real physics behind the prominently phenomenological work , e.g. , regge trajectory for particle classification @xcite .
we also note that there are different versions of tsallis distribution in the literature and we classify them as type - a , b and c to clarify the comparison in ref .
type - a tsallis distribution is obtained without resorting to thermodynamical description , but it has been adopted to analyze the particle spectra by star @xcite , phenix @xcite collaborations at rhic and alice @xcite , cms @xcite collaborations at lhc . in ref .
@xcite , we applied it to do the systematic analysis of identified particle spectra in p+p collisions at rhic and lhc and proposed a cascade particle production mechanism . on the other hand ,
type - b tsallis distribution is derived by taking into account the thermodynamical consistency and is widely used by j. cleymans and his collaborators to study the particle spectra in high - energy p+p collisions @xcite .
it is also used by the other authors @xcite .
type - a and b are the most popular tsallis distributions in the literature but they give quite different temperatures while fitting the same particle spectra , e.g. , for pion , type - a gives @xmath3 gev while type - b gives @xmath4 gev . in this paper , we would like to systematically address the question what results in the discrepancies of the temperatures for the two types of tsallis distribution using particle spectra in p+p collisions . the data produced in p+p collisions with different @xmath2 ranges and different rapidity cuts are collected from the experimental collaborations at rhic and lhc @xcite . the paper is organized as follows . in sec .
ii , we introduce the two types of tsallis distribution without and with thermodynamical description in our comparison . in order to make our discussion clear
, we also introduce another three transient distributions which are very similar to type - a and type - b tsallis distributions . in sec .
iii , the results of particle spectra from the different distributions in p+p collisions and the comparisons are shown .
a brief conclusion is given in the sec .
type - a tsallis distribution has been widely adopted by star @xcite , phenix @xcite collaborations at rhic and alice @xcite , cms @xcite collaborations at lhc @xmath5}(1+\frac{m_t - m}{nc})^{-n } , \label{exptsallis}\ ] ] where @xmath6 is the transverse mass .
@xmath7 was used as a fitting parameter in ref .
@xcite , but it represents the rest mass of the particle studied in refs .
@xmath8 , @xmath9 and @xmath10 are fitting parameters .
when @xmath11 , we can ignore the @xmath7 in the last term in eq .
( [ exptsallis ] ) and obtain @xmath12 .
this result is well known because high energy particles come from hard scattering and they follow a power law distribution with @xmath2 .
when @xmath13 which is the non - relativistic limit , we obtain @xmath14 and @xmath15 , i.e. , a thermal distribution .
the parameter @xmath10 in eq .
( [ exptsallis ] ) plays the same role as temperature @xmath16 . in refs .
@xcite , we have obtained the simpler form of eq .
( [ exptsallis ] ) @xmath17 where @xmath18 .
@xmath19 , @xmath9 and @xmath16 are free fitting parameters in eq .
( [ tsallisus ] ) .
we note that it has been used by cms collaboration @xcite and by wong _
_ in their recent paper @xcite .
the star collaboration also applied a formula which is very close to eq .
( [ tsallisus ] ) @xcite .
we adopt the eq .
( [ tsallisus ] ) in the following study . in the framework of tsallis statistics ,
the distribution function is @xmath20^{-\frac{1}{q-1}}.\ ] ] taking into account the self - consistent thermodynamical description , one has to use the effective distribution @xmath21^q$ ] . therefore , the tsallis distribution is obtained @xcite @xmath22^{-\frac{q}{q-1 } } , \label{tsallisb}\end{aligned}\ ] ] where @xmath23 is the degeneracy of the particle , @xmath24 is the volume , @xmath25 is the rapidity , @xmath26 is the chemical potential , @xmath16 is the temperature and @xmath27 is the entropic factor , which measures the non - additivity of the entropy .
we dubbed it as the type - b tsallis distribution @xcite . in eq .
( [ tsallisb ] ) , there are four parameters @xmath28 .
@xmath26 was assumed to be 0 in refs .
@xcite which is a reasonable assumption because the energy is high enough and the chemical potential is much smaller than the temperature . in the mid - rapidity @xmath29 region ,
( [ tsallisb ] ) is reduced to @xmath30^{-q/(q-1)}. \label{tsallisbr}\ ] ] it becomes very similar to eq .
( [ tsallisus ] ) , but there are some differences , e.g. , @xmath0 replaces @xmath1 in the bracket and there is an extra term @xmath0 in front of the bracket .
it should be pointed out that there is no direct match between @xmath9 and @xmath27 in eqs .
( [ tsallisus ] , [ tsallisbr ] ) .
but we could find a connection between @xmath9 and @xmath27 in the limit at large @xmath2 .
when @xmath11 , from eq .
( [ tsallisbr ] ) , we can obtain @xmath31 recalling that @xmath32 when @xmath11 from eq .
( [ exptsallis ] ) , therefore the relation between @xmath9 and @xmath27 is @xmath33 another treatment to find the relation between @xmath9 and @xmath27 can be found in ref .
@xcite . for the other tsallis distributions in the literature ,
we refer them to refs .
we noted that type - a and type - b tsallis distributions can reproduce the particle spectra in p+p collisions very well but type - b gives lower temperatures than the ones given by type - a . in this paper , we would like to address this discrepancies between the two types of tsallis distribution . to make our discussion clear , another three transient distributions are used to bridge the type - a and type - b distributions . in ref .
@xcite , a tsallis - like distribution is obtained in the framework of non - extensive statistics for the particle invariant yield at midrapidity @xmath34^{-\frac{1}{q-1 } } , \label{transeq1}\ ] ] where @xmath19 , @xmath27 and @xmath16 are fitting parameters . comparing eq .
( [ transeq1 ] ) and eq .
( [ tsallisbr ] ) , the only difference is the power of the distribution function , i.e. , @xmath27 for eq .
( [ tsallisbr ] ) and @xmath35 for eq .
( [ transeq1 ] ) .
we also introduce another two forms of distribution .
one is @xmath36^{-\frac{q}{q-1 } } , \label{transeq2}\ ] ] where we neglect the term @xmath0 outside of the bracket in eq .
( [ tsallisbr ] ) and the constants are absorbed into the parameter @xmath19 .
the other one is @xmath36^{-\frac{1}{q-1}}. \label{transeq3}\ ] ] similar to eq .
( [ transeq2 ] ) , we neglect the term @xmath0 in front of the bracket in eq .
( [ transeq1 ] ) but keep the rest . using the relation eq .
( [ nq ] ) , we find that eq .
( [ transeq2 ] ) becomes eq .
( [ tsallisus ] ) in the limit of massless particle .
let us put the four distributions in order of eqs .
( [ tsallisbr ] , [ transeq1 ] , [ transeq3 ] , [ tsallisus ] ) and eqs .
( [ tsallisbr ] , [ transeq2 ] , [ transeq3 ] , [ tsallisus ] ) , one can see that any adjacent distributions have only one term different .
we successfully bridge the type - a and type - b tsallis distributions and are able to conduct our investigation .
we need to point out that even though we use the same symbols for the parameters in the five distributions , they may have different values when fitting the experimental data . in next section
, we will systematically apply the five distributions to the particle spectra in p+p collisions , similar to our previous work @xcite .
but we update some experimental data which have larger @xmath2 ranges and will focus on the temperature differences among the five distributions . [
cols="^,^,^ " , ]
we fit the particle spectra with different @xmath2 ranges and different rapidity cuts from p+p collisions at @xmath37 and @xmath38 gev with the five distributions discussed in the last section .
the fitting process is the same as that in refs .
@xcite . compared with ref .
@xcite , the identified particle spectra data at @xmath39 and @xmath38 gev have been updated . in this work ,
we are interested in the differences of the parameter @xmath16 from the five distributions . as we argued in refs .
@xcite , @xmath16 is one free fitting parameter and can be different for different particles even they are produced in the same colliding system . in order to distinguish the parameter @xmath16 in the five distributions and make our discussion clear , we assign @xmath40 , @xmath41 , @xmath42 , @xmath43 and @xmath44 to eqs .
( [ tsallisbr ] ) , ( [ transeq1 ] ) , ( [ transeq2 ] ) , ( [ transeq3 ] ) and ( [ tsallisus ] ) , respectively .
all the five distributions can fit the particle transverse momentum spectra very well .
the values of @xmath45 ( @xmath46 ) and corresponding @xmath47 for @xmath48 , @xmath49 , @xmath50 and charged particles are shown in table [ fitpara ] . here
, we only show the fitting results with the five distributions for four cases : ( 1 ) @xmath51 at @xmath52 gev , ( 2 ) @xmath51 at @xmath53 gev , ( 3 ) @xmath54 at @xmath39 gev , ( 4 ) @xmath55 at @xmath56 gev . in figs .
[ fig1 ] and [ fig2 ] , we show the fitting results for @xmath51 at @xmath52 gev from star collaboration and @xmath53 gev from alice collaboration using the five distributions eq .
( [ tsallisbr ] ) ( solid line ) , eq .
( [ transeq1 ] ) ( dashed line ) , eq .
( [ transeq2 ] ) ( dotted line ) , eq .
( [ transeq3 ] ) ( dash - dotted line ) and eq .
( [ tsallisus ] ) ( long dash - dotted line ) , respectively . since the lines are so close to each other , they are almost indistinguishable in the figures .
to visualize the fitting quality better , we also plot the ratios of experimental data and fitting results at the bottom of the figures .
we can see that the five distributions can describe the experimentally measured @xmath51 transverse momentum spectra very well .
the errors are within 20% . to show the fitting results other than pions , we select the @xmath54 at @xmath39 gev and @xmath55 at @xmath56 gev at lhc in figs .
[ fig3 ] and [ fig4 ] , respectively .
for the five distributions , we can not distinguish them at all in the two cases .
similar to pions , the fitting qualities are also very good .
[ fitpara ] from the above results , we can see that the five distributions have the same fitting power to the particle spectra .
but as we can see from table .
[ fitpara ] , the parameter @xmath45 are different .
it is the main purpose of this work to target what causes the temperature discrepancies among different tsallis distributions , with and without thermodynamical description . in order to avoid confusion ,
we emphasis the meaning of each @xmath45 . using the tsallis distribution classification in ref .
@xcite , @xmath40 refers to type - b tsallis distribution and @xmath44 refers to type - a tsallis distribution .
@xmath41 , @xmath42 and @xmath43 are from the transient distributions to bridge the type - a and type - b tsallis distributions . to have a clear picture for the parameter @xmath45 from the five distributions , we plot @xmath45 versus the colliding energy @xmath57 for @xmath48 , @xmath49 and @xmath50 , as shown in fig .
we can clearly see that the type - b tsallis distribution gives lower @xmath16 than the type - a tsallis distribution , as we mentioned in ref .
for all the particles , @xmath40 and @xmath41 which are from the distributions , eqs .
( [ tsallisbr ] , [ transeq1 ] ) , with extra @xmath0 term are lower than @xmath42 , @xmath43 and @xmath44 which are from the distributions , eqs .
( [ transeq2 ] , [ transeq3 ] , [ tsallisus ] ) , without it . as we see that @xmath40 is larger than @xmath41 , since eq .
( [ tsallisbr ] ) and eq .
( [ transeq1 ] ) are similar except the power in the distributions , we conclude that the power @xmath27 in eq .
( [ tsallisbr ] ) causes larger @xmath16 .
this can be verified by comparing @xmath42 with @xmath43 . with the same argument , since @xmath43 is smaller than @xmath44 , the @xmath0 in eq .
( [ transeq3 ] ) causes smaller @xmath16 . to see the effects of @xmath27 and @xmath0 in the distribution
, we can compare @xmath42 with @xmath44 .
for the light particles , i.e. , pions ( left panel ) , @xmath42 and @xmath44 are similar .
the effects of the power @xmath27 and @xmath0 cancel each other .
but for the heavier particles , i.e. , kaons and protons ( middle and right panels ) , the effect of @xmath0 in eq .
( [ transeq2 ] ) wins and @xmath42 is smaller than @xmath44 .
we note that the effect of the extra @xmath0 term is crucial for the temperature difference between type - a and type - b tsallis distributions , especially for the light particles .
while for the heavier particles , the effect of the choice of @xmath0 or @xmath1 in the tsallis distribution becomes more important as we can see that @xmath44 is larger than the other @xmath45 for kaons and protons in fig .
in this paper , we have presented a detailed investigation of two types of tsallis distribution , with and without the thermodynamical description , by the @xmath2 spectra measured from star , phenix collaborations at rhic and alice , cms collaborations at lhc .
the power @xmath27 in the tsallis distribution with thermodynamical description is responsible for the thermodynamical consistency . to show a clear and complete comparison , another three transient distributions to bridge the two types of tsallis distribution
are also given .
the five distributions have the similar fitting power to the particle spectra .
the good agreements are obtained , but they give different temperatures . agreed with our previous work @xcite , the tsallis distribution with thermodynamical description gives lower @xmath16 than the ones given by the distribution without it .
the extra term @xmath0 in the tsallis distribution with thermodynamical description is responsible for the discrepancies of the temperatures .
but for the heavier particles , the effect of the choice of @xmath0 or @xmath1 in the tsallis distribution beats the effect of the extra term @xmath0 . the data for p+p collisions at 8 and 13 tev are coming .
we wish these data would also support our conclusion presented here .
the authors declare that there is no conflict of interests regarding the publication of this article .
we thank the anonymous referee who brought us to this topic . this work is supported by the nsfc of china under grant no . 11205106 .
j. cleymans and d. worku , `` the tsallis distribution in proton - proton collisions at @xmath57 = 0.9 tev at the lhc '' , m. d. azmi and j. cleymans , `` transverse momentum distributions in proton - proton collisions at lhc energies and tsallis thermodynamics '' , j. cleymans , g.i .
lykasov , a.s .
parvan , a.s .
sorin , o.v .
teryaev and d. worku , `` systematic properties of the tsallis distribution : energy dependence of parameters in high energy p - p collisions '' , m. rybczyski and z. wodarczyk , `` tsallis statistics approach to the transverse momentum distributions in p - p collisions '' , p.k .
khandai , p. sett , p. shukla and v. singh , `` hadron spectra in p + p collisions at rhic and lhc energies '' , c.y .
wong , g. wilk , l.j.l .
cirto and c. tsallis , `` from qcd - based hard - scattering to nonextensive statistical mechanical descriptions of transverse momentum spectra in high - energy @xmath61 and @xmath62 collisions '' , g. wilk and z. wodarczyk , `` quasi - power laws in multiparticle production processes '' , chaos , solitons and fractals , vol . 81 , pp .
487 - 496 , 2015 . c. beck ,
`` non - extensive statistical mechanics and particle spectra in elementary interactions '' , a. adare _
( phenix collaboration ) , `` identified charged hadron production in p + p collisions at @xmath57 = 200 and 62.4 gev '' , b. abelev _
( alice collaboration ) , `` neutral pion and @xmath63 meson production in proton - proton collisions at @xmath64 tev and @xmath65 tev '' , k. aamodt _
( alice collaboration ) , `` production of pions , kaons and protons in pp collisions at @xmath57 = 900 gev with alice at the lhc '' , s. chatrchyan _
( cms collaboration ) , `` study of the inclusive production of charged pions , kaons , and protons in pp collisions at @xmath57 = 0.9 , 2.76 , and 7 tev '' , s. chatrchyan _ et al . _
( cms collaboration ) , `` charged particle transverse momentum spectra in pp collisions at @xmath57 = 0.9 and 7 tev '' , v. khachatryan _ et al . _
( cms collaboraction ) , `` transverse - momentum and pseudorapidity distributions of charged hadrons in pp collisions at @xmath57 = 0.9 and 2.36 tev '' , v. khachatryan _ et al . _
( cms collaboraction ) , `` transverse - momentum and pseudorapidity distributions of charged hadrons in pp collisions at @xmath65 tev '' , a. adare
_ et al . _
( phenix collaboration ) , `` inclusive cross section and double helicity asymmetry for @xmath67 production in p+p collisions at @xmath57=200 gev : implications for the polarized gluon distribution in the proton '' , | we compare two types of tsallis distribution , i.e. , with and without thermodynamical description , using the experimental data from the star , phenix , alice and cms collaborations on the rapidity and energy dependence of the transverse momentum spectra in p+p collisions . both of them can give us the similar fitting power to the particle spectra .
we show that the tsallis distribution with thermodynamical description gives lower temperatures than the ones without it . the extra factor @xmath0 ( transverse mass ) in the tsallis distribution with thermodynamical description plays an important role in the discrepancies between the two types of tsallis distribution .
but for the heavy particles , the choice to use the @xmath0 or @xmath1 ( transverse energy ) in the tsallis distribution becomes more crucial .
# 1#2#3#4#1 , # 2 , # 3 , # 4 . |
convection bounded by a stable region is important in all aspects of stellar evolution .
the depth to which convective motions can overshoot into an adjacent stable layer has important consequences for many stages of stellar evolution , from dredge up in late stages of stellar evolution @xcite to nucleosynthesis in classical novae @xcite . in the sun
, the overshoot region at least partially overlaps the tachocline ( the shear layer at the base of the convection zone where the rotation changes from being differential in the convection zone to solid body in the radiation zone and the likely site of toroidal field generation in the dynamo process ) .
therefore , understanding the extent of the overshoot is necessary for understanding the nature of the tachocline ; is it a violent region , constantly bombarded by overshooting plumes , or is it a laminar region just below the overshoot ?
the answer to this question can constrain models for the solid body rotation of the solar interior .
in addition to the uncertainty of penetrative convection , which has plagued stellar evolution theory , there are other noteworthy aspects of convective - radiative boundaries .
convection adjacent to a radiative region generates internal gravity waves which can have dynamical affects . in the earth
s atmosphere , the nonlinear interaction between convection , shear and gravity waves is the cause of the quasi - biennial oscillation ( qbo ) @xcite ; this shear - gravity wave interaction has been demonstrated in a remarkable experiment by @xcite . the same process responsible for the qbo has been invoked to explain the solar magnetic cycle @xcite .
internal gravity waves can share angular momentum with the mean flow when they are attenuated .
the transport and deposition of angular momentum by gravity waves is one explanation for the solid body rotation observed in the radiative interior of the sun @xcite .
however , a vast literature in atmospheric science tells us that gravity waves have an anti - diffusive nature ; they enhance local shear rather than smooth it . in addition , internal gravity waves have also been implicated in the mixing of species @xcite , and in order to explain the lithium depletion @xcite .
most studies of convection bounded by a stable layer have been focused on the penetration depth .
the depth that convection overshoots into an adjacent radiative region has been studied extensively by numerical simulations in two @xcite and three dimensions @xcite . in general , these simulations find that the penetration depth is a sensitive function of the stiffness of the subadiabatic layer and find that it is also sensitive to the peclet number ( @xmath0 ) @xcite .
early 2d studies with `` soft '' interfaces ( mild subadiabaticity ) @xcite found that overshooting convection could extend the adiabatic region .
three - dimensional simulations @xcite and simulations with very stiff ( large subadiabaticity ) radiative regions @xcite find no extended adiabatic region , in agreement with helioseismic observations .
however , in all of the previous studies a constant ( with height ) subadiabaticity was used and none were as stiff as the sun .
analytic models generally predict extended adiabatic regions , but rely on highly uncertain parametrizations .
the penetration depth in the sun and its dependencies will be discussed in greater detail in a forthcoming paper .
while internal gravity waves have been studied analytically for some time , e.g. @xcite , they have not received much attention in numerical simulations of stellar interiors . in the few instances where they have been simulated
they have not been analyzed in detail and have been unlike those in the sun because the subadiabaticity was much less than solar .
however , @xcite did extensive numerical studies in 2d , similar to those by @xcite and analyzed the wave spectrum produced and compared it to previous analytic predictions for g - mode spectra by @xcite .
they found that the spectrum of waves produced in numerical simulations was more broadband in frequency and that numerical simulations always predicted larger fluxes in waves than analytic studies .
the overestimate of energy in waves in that study was probably due to the mild subadiabaticity used in those calculations .
the subadiabaticity of the stable region is extremely important in these types of simulations @xcite ; the mild values used in all previous simulations allow plumes to penetrate much more deeply , causing larger nonlinearities and velocities than would otherwise occur .
however , since internal gravity waves may be maintaining the solid body rotation of the solar interior @xcite , may be driving a qbo like oscillation ( coined slo by @xcite ) at the base of the convection zone @xcite and may someday be observed with helioseismology @xcite , we have begun a detailed study via numerical simulations of their excitation and evolution .
the way in which gravity waves redistribute angular momentum and the likelihood of their observation depends on the spectrum of gravity waves produced by the overlying convection , their nonlinear interaction with convection and rotation , and on their radiative damping . here
we discuss the spectrum of gravity waves produced by overlying convection and penetration , radiative damping and the energy flux of waves beneath the convection zone .
the impact gravity waves have on the angular velocity in the radiative interior will be discussed in a forthcoming paper .
we discuss the numerical method in section 2 , convection in section 3 and gravity waves in section 4 .
we solve the navier stokes equations with rotation in the anelastic approximation @xcite for an ideal gas . the anelastic approximation is appropriate when the fluid flow is sufficiently subsonic and the perturbations relative to the mean ( reference ) thermodynamic state are small .
the radially ( r ) dependent reference state is taken from a 1d solar model ( christensen - dalsgaard private communication ) . unlike previous formulations of the anelastic equations @xcite , here our reference state temperature gradient is not adiabatic , and we choose to use the temperature perturbation as our working thermodynamic variable , instead of the specific entropy .
this formulation has advantages and disadvantages compared to the standard formulation @xcite .
the equations are solved in 2d cylindrical coordinates , with the radiation zone extending from 0.001r@xmath1 to 0.718r@xmath1 and the convection zone occupying the region from 0.718r@xmath1 to 0.9r@xmath1 .
this geometry represents an equatorial slice of the sun that includes most of the radiative and convective regions . for numerical reasons we put another stable region from 0.9r@xmath1 to 0.93r@xmath1 .
the exclusion of @xmath2 of @xmath3 around the centre has little effect since gravity waves internally reflect where their frequency equals the brunt - vaisala frequency , which vanishes at the centre .
we solve the following anelastic equations for perturbations to the reference state : @xmath4 @xmath5 @xmath6+}\nonumber\\ & & \gamma\overline{\kappa}[\nabla^{2}\overline{t}+(h_{\rho}+h_{\kappa}){{\frac{\partial{\overline{t}}}{\partial{r } } } } ] + \frac{\overline{q}}{c_{v}}\end{aligned}\ ] ] equation ( 1 ) represents the continuity equation in the anelastic approximation , where @xmath7 is the reference state density and @xmath8 is the fluid velocity .
equation ( 2 ) is the momentum equation , assuming a constant dynamic viscosity , @xmath9 . the gravitational acceleration , -@xmath10 is taken from the 1d solar model .
the rotation rate , @xmath11 is set to the mean rotation rate of the sun @xmath12 rad / s .
the reduced pressure , p @xcite is defined as @xmath13 where @xmath14 is the pressure perturbation and u is the gravitational potential perturbation .
the co - density perturbation is defined as : @xmath15 equation ( 3 ) is derived from @xmath16 where here the thermodynamic variables are the full variables ( sum of the reference state , denoted by an overbar , and the perturbation ) and an ideal gas is assumed .
the thermal expansion coefficient @xmath17 , the adiabatic exponent @xmath18 and @xmath19 , the thermal diffusivity are all functions of radius . the inverse density scale height and thermal diffusivity scale height are defined as : @xmath20 the first term on the right hand side of equation ( 3 ) can also be written as @xmath21 which illustrates how this term depends on the local difference between the nonadiabatic reference state and an adiabatic state .
the last two terms in equation ( 3 ) , which are only a function of the reference state , are assumed zero at the beginning of our simulations .
we neglect viscous heating for these low viscosity simulations .
the top and bottom boundary conditions are isothermal , stress - free and impermeable .
the technique used here is similar to that of @xcite , but for a rotating disk instead of a box . for numerical simplicity
we take the curl of the momentum equation , leaving us with a vorticity equation : @xmath22 where the vorticity , @xmath23 is in the @xmath24 direction .
we have neglected viscous terms involving @xmath25 .
we calculate the pressure term using the longitudinal component of the momentum equation ( 2 ) : @xmath26 .\ ] ] where @xmath27 and @xmath28 represent the longitudinal and radial velocity components , respectively .
the time derivative of @xmath27 is from the previous timestep and we neglect the perturbation to the gravitational potential in ( 10 ) .
these equations are solved using a fourier spectral transform in the longitudinal ( @xmath29 ) direction and a finite difference scheme on a non - uniform grid in the radial ( r ) direction .
time advancing is done using the explicit adams - bashforth method for the nonlinear terms and an implicit crank - nicolson scheme for the linear terms .
we do a 21st order polynomial fit to the reference state density ( @xmath7 ) , temperature ( @xmath30 ) , mass ( which gives us gravity , @xmath31 ) , opacity ( @xmath32 ) and @xmath33 given in the 1d solar model .
figure 1 shows density and temperature as a function of radius in our model ( solid line ) compared to the standard model ( dotted line ) .
analytic derivatives of these ( fitted ) values give us inverse scale heights ( @xmath34 and @xmath35 ) and the temperature gradient .
the subadiabaticity in the radiative region is then defined as : @xmath36 where the second term in equation ( 11 ) is the adiabatic temperature gradient .
figure 2 shows the effective brunt - vaisala frequency ( subadiabaticity ) , defined in ( 16 ) as a function of radius .
the superadiabaticity in the convection zone is specified to be a constant , @xmath37 .
the thermal diffusivity is that given by the solar model , multiplied by a constant factor for numerical stability : @xmath38 where @xmath39 is the stefan - boltzman constant and the multiplying factor , @xmath40 , is @xmath41 ( and therefore , the convective flux is @xmath41 larger than solar ) .
this gives us the proper radial profile of the radiative diffusivity , albeit increased by a large factor for numerical stability ; this is a `` turbulent '' diffusivity .
the viscous diffusivity is specified to be : @xmath42 keeping the dynamic viscosity constant . since both @xmath43 and @xmath19 vary with height , the pr ( prandtl number = @xmath44 ) varies with height from @xmath45 ( at base of the convection zone ) to @xmath46 ( at top of convection zone ) and is @xmath47 near the centre .
therefore , the ra ( rayleigh number = @xmath48 , where d is the depth of the convection zone and @xmath49 is the superadiabatic temperature gradient ) also varies from @xmath50 at the base of the convection zone to @xmath51 at the top of the convection zone .
the ekman number ( ek=@xmath52 ) varies from @xmath53 to @xmath54 .
the resolution used in these models is 2048 longitudinal zones x 1500 radial zones , with 500 radial zones dedicated to the radiative region .
here we discuss the convection in this model . in order to illustrate the differences between convection bounded by a radiative region and convection bounded by a hard boundary we also present a model that has an impenetrable bottom boundary .
the model with an impenetrable boundary has the same reference state as the model with a stable region and therefore , has the same ra , pr and ek . we note here that the simulations with impenetrable boundaries usually require greater spatial resolution near the boundaries and smaller timesteps because the hard boundaries deflect fast descending ( ascending ) plumes into shallow , high - speed horizontal flows .
this problem is more severe at the top of the computational domain because the density is lower there , resulting in larger velocities .
this is one reason we have added a very shallow stably stratified region at the top boundary .
another is that it is more realistic to have rising plumes stopped by a stable layer rather than an artificial hard boundary .
snapshots of the vorticity in the two models are shown in figure 3 .
there are many differences immediately obvious in this figure .
most noticeable is the nature of the convective cells . in the model with an impenetrable ( hard ) bottom boundary ,
there are fewer convective cells and those cells are more orderly and appear more laminar . because descending and ascending plumes are diverted horizontally by the hard , stress - free boundary nearly all of their vertical kinetic energy is converted into horizontal kinetic energy .
these ( relatively ) large horizontal velocities allow the flow to travel further before cooling ( heating ) enough to fall ( rise ) . in the model bounded below by a stable layer ,
a small fraction of the vertical energy in descending plumes is transferred to the stable region generating gravity waves . because of
this transfer of energy and less rigid horizontal trajectories , horizontally diverted fluid does not travel as far before cooling ( heating ) and sinking ( rising ) .
the time dependent convective penetration in the model bounded by a stable region also results in more chaotic interaction between convective cells , albeit with lower overall kinetic energy .
figure 4 shows the kinetic energy density ( @xmath55 ) , horizontal kinetic energy flux @xmath56 and vertical kinetic energy flux @xmath57 , all scaled to the maximum values in the impenetrable case .
it is obvious in this figure that in the convective region there is significantly more energy ( by 2 - 3 orders of magnitude ) in the model with hard boundaries .
this is due almost entirely to a difference in magnitude of horizontal velocity ( 4b , note magnitudes ) , as explained above . in models bounded by a stable region a fraction of the kinetic energy
is shared with the radiative region .
descending plumes overshoot the convective - radiative boundary and energy is transferred to gravity waves .
the simplest estimate of the penetration depth is given where the kinetic energy density drops to 1% of its maximum value in the convection zone .
in this model this results in a penetration depth of approximately 0.1 pressure scale heights , but no extended adiabatic region .
several things can affect this penetration depth .
first , the `` stiffness '' of the subadiabatic layer has been shown to be of utmost importance in several previous numerical studies @xcite . as stated above , this model has the ( realistic ) subadiabatic structure of the sun and hence , a very high `` stiffness '' , which makes the penetration depth small .
the dimensionality of the model also has an effect ; 3d models have smaller penetration depths than 2d models @xcite .
therefore our estimate may be an upper limit .
in addition , rotation has been shown to hinder penetration @xcite , while the effect of the rayleigh number is still unclear @xcite and may depend on the dimension .
kinetic energy spectra in the convection zones of the two models also differ .
figure 5 shows the kinetic energy as a function of horizontal wavenumber at a radius half way through the convection zone in both models . as illustrated in figure 4 , and now again in 5 , the overall kinetic energy in the convection zone is larger for the model with an impenetrable lower boundary ( shown as the dotted line in figure 5 ) , especially at low wavemode ( large scales ) .
both models are fit rather well with a @xmath58 power law . while @xmath59 spectrum is expected in 2d turbulence ,
the geometry of the flow has not been previously considered , nor the effect of adjacent stable regions ( both models have stable regions on top ) . fully convective simulations of convective motion in a 2d disk give a variety of scaling laws ( between @xmath60 and @xmath58 ) depending on the ekman number , the density stratification and the depth at which the spectrum is calculated .
the model with hard boundaries shows less of a dissipation range and some build - up at small scales .
the model bounded by a stable region shows a dissipation range and no such build - up at large wavemodes .
this difference is probably due to the issues mentioned above : vorticity generation and large horizontal velocities along the ( impenetrable ) boundary . in summary , both the energetics and the character of convection
are changed when a bounding stable layer is used instead of the traditional ( artificial ) impenetrable boundary .
gravity waves can be generated by reynolds stresses in a convective region overlying a radiative region as well as by descending plumes that overshoot the convective - radiative boundary .
the gravity waves generated by such processes have been studied previously in analytic studies . in particular , @xcite studied the spectrum of gravity waves generated by reynolds stresses in the overlying convective region and @xcite studied the spectrum of gravity waves generated by overshooting plumes .
both models are analytic and assume some combination of mixing length theory ( mlt ) and a kolmogorov spectrum for turbulence .
gravity waves have the ability to mix species and transport angular momentum and thus , may have profound effects on the dynamics of the radiative zone in the sun . in the following ,
we discuss results from both linear and nonlinear numerical simulations .
first , however we briefly review basic properties of gravity waves in the asymptotic regime without rotation ( for a more thorough review see , e.g. @xcite ) .
the asymptotic , linear dispersion relation for pure gravity waves is : @xmath61 where @xmath62 represents the horizontal wavenumber and @xmath63 where @xmath64 is the radial wavenumber . the brunt - vaisala frequency ,
n , represents the upper limit on the frequency of gravity waves .
it depends on the thermal stratification of the fluid and is defined as : @xmath65 from the basic linear dispersion relation ( 14 ) it is seen that the frequencies of gravity waves depend only on the gas stratification and the direction of phase propagation . from the above dispersion relation it can also be shown that the group velocity and phase velocity are perpendicular , a peculiar trait of internal gravity waves .
in addition to the standing g - modes ( in the frequency range of @xmath66hz to @xmath67hz ) , propagating waves from the overlying convection and penetration are produced .
these propagating gravity waves , generally with much lower frequencies , have been suggested as the source of the oscillating shear layer at the base of the convection zone , the net angular momentum transport in the radiative interior @xcite and the mixing of species @xcite .
the spectrum of these waves has been predicted in those analytic studies .
however , here we present the first self - consistent nonlinear simulation and analysis of these waves in the sun s radiative interior . while our main simulation here is nonlinear in both the convective and radiative regions of the computational domain , we have computed linear cases for comparison with previous asymptotic results and to illustrate the differences between linear and nonlinear solutions .
we first discuss the linear calculations .
figure 6 shows a snapshot of the vorticity in the radiation zone produced with a linear version of our model , excluding the convection zone and initially perturbed at several horizontal wavenumbers .
while there are several different horizontal wavenumbers depicted in this image , their modes do not interact .
this is similar to that predicted from ray - tracing @xcite . in our partially linear simulation ,
the full set of nonlinear equations is solved in the convection zone , but the nonlinear terms in equations ( 3 ) , ( 9 ) and ( 10 ) are decreased linearly between 0.718@xmath3 ( the radius of the transition between convective and radiative layers ) and 0.69@xmath3 . therefore , below 0.69@xmath3 only the linear set of equations is solved .
a snapshot of vorticity looks similar to that in 8a ( the fully nonlinear case ) .
however , there is a remarkable difference between this gravity wave pattern ( seen in 8b ) and that in figure 6 .
when constantly being driven by penetrative convection at many different frequencies and wavenumbers g - modes form a much simpler spiral pattern in the radiative interior .
a simple reason for this behaviour can be explained as followed .
if convection is considered as nearly equally spaced convective cells , descending plumes drive similar wavenumbers , albeit slightly out of phase .
the nonlinear wave - wave interactions generate new waves with wavenumbers equal to both the sum and difference of the original waves .
the high wavenumber waves are damped in a very short distance ; wherease the small wavenumber ( long wavelength ) waves survive through most of the radiative interior .
this wave - wave interaction only occurs in nonlinear simulations .
figure 6 has no nonlinear interactions and thus all initially excited wavenumbers remain .
frequencies of gravity waves in the solar interior in the linear , non - dissipative , non - rotating case have been calculated previously ( @xcite and generally range from 100@xmath68hz at low horizontal mode ( large wavelength ) to 500@xmath68hz for higher horizontal mode ( small wavelength ) depending on the radial order . figure 7 shows the power spectra of gravity waves in the radiative interior in our partially linear simulation shortly after the simulation was started .
significant power occurs at the base of the convection zone , up to the highest frequencies for the low modes and up to about 200@xmath68hz for the high modes ( figure 7a ) . at this radius
the fluid motion is fully nonlinear and the expected broadband nature of the convection is seen .
note that at the convective - radiative boundary there is power in frequencies much higher than the typical convective turnover frequency of 2 - 10@xmath68hz because descending plumes pummel the stable region at many locations slightly out of phase .
power is not concentrated at the convective turnover frequency as is often assumed .
moving into the stable layer , figure 7b shows the dispersion relation for gravity waves at 0.575 @xmath3 . the peak power in figures 7b and
c is three orders of magnitude smaller than the peak power shown in figure 7a .
there is ( significant ) power at these smaller radii only out to horizontal mode 25 , indicating the rapid decrease in energy due to radiative damping . in figure
7b there is power in both low frequency propagating g - modes generated by the overlying convection and penetration ( below 100@xmath68hz ) and high frequency ( above 100@xmath68hz ) standing waves .
ridges similar to those calculated by asymptotic analysis ( christensen - dalsgaard 1986 ) for high radial order are apparent . at any given height
there are several radial orders contributing to the power at that horizontal mode and frequency , giving rise to the width of the ridges seen in this dispersion relation . at radius
0.360 @xmath3 ( figure 7c ) , it is seen that power at higher modes is diminished relative to figure 7b and that the lowest frequency / high wavenumbers have been completely damped ( see section 4.2.2 ) .
again , there is significant energy in the propagating waves at frequencies much higher than the convective turnover time .
figure 8a shows a snapshot of the vorticity produced with our fully nonlinear model for the full computational domain ( as in figure 3b ) .
figure 8b shows the temperature perturbation just in the radiation zone to illustrate the salient features of the gravity waves produced ; waves spiral toward the centre due to rotation and geometry .
while both small and large horizontal modes are produced at the interface only small mode numbers survive at the centre with substantial amplitude .
figure 9 shows the power spectra for our fully nonlinear case after the convection has reached a statistically steady state . the peak amplitude in figure 9
is scaled the same as in figure 7 .
there is little energy in the high frequency standing waves .
that is , there is little power at frequencies above 100@xmath68hz for modes
greater than 10 , in figures 9b and c. the lack of standing modes after the convection has reached steady state is seen in both linear and nonlinear calculations , indicating that the energy in these frequencies is radiatively damped and is not continually generated by the overlying convection . the appearance of the high frequency standing waves is a numerical artifact and we show them here purely as a proof that the model can reproduce the asymptotic limit ( given a large enough perturbation ) .
the simulations are started with an initial perturbation to the convection zone .
this initial perturbation has much higher amplitude than steady state convection .
the large amplitude perturbation hits the stable region hard enough to excite all resonant frequencies .
once the convection has reached a statistically `` steady '' state , high frequencies still exist in the convection , but at a lower amplitude .
these lower amplitudes are unable to drive the high frequency waves .
the ( dis)appearance of the high frequency standing g - modes was investigated in several different models .
we ran a series of models including : partially linear , fully nonlinear , increased rotation rate , decreased rotation rate , larger and smaller diffusion coefficients . in general any `` large '' perturbation to the momentum equation , like increasing or decreasing the rotation rate by a large value , or turning off nonlinear terms gives rise to the high frequency standing modes ; however , these modes disappear on a timescale of @xmath69s .
it appears that statistically `` steady '' convection does not provide a large enough perturbation to continually generate these modes . in the sun
the thermal diffusivity is @xmath41 times smaller than in our simulation and therefore , one would expect these high frequency waves to damp on the order of @xmath70s rather than the @xmath71s seen in our simulation .
however , this is still a relatively short time ; and therefore these high frequency standing waves likely do not exist in the sun .
in addition , it has been found in previous studies that two dimensional simulations produce larger perturbations to the underlying stable region than 3d simulations . in 3d these modes
would be excited with lower amplitude and hence , would be even less likely to exist . in order for these modes to exist , they will have to be continually generated by something other than steady convection .
as waves propagate in a diffusive medium they are radiatively damped .
this damping leads to the transfer of their angular momentum , positive or negative , to the differential rotation .
the radiative damping of waves depends on the frequency and wavenumber of the waves .
low frequency and high wavenumber ( small wavelength ) waves are damped more quickly ( over a shorter distance ) than high frequency / low wavenumber waves . because thermal diffusion has more time and larger temperature gradients for the former . in this geometry group velocity
is directed inward and phase velocity is outward .
the inward propagation of wave packets , indicative of group velocity , can be seen in figure 10 .
this figure shows the temperature perturbation as a function of radius for a given horizontal mode ( in this case m=20 and 100 , where the horizontal mode number m is defined as @xmath62r ) for two sequential times , with the dotted line representing the earlier time and the solid line representing a later time ( @xmath72s later ) .
it can be seen that wave packets move inward in time .
as wave packets move inward they are radiatively damped and therefore , their amplitudes decrease . in figure 10 , not only does the lower m case have higher amplitude overall , it is damped less in time relative to the higher m case . in animations of these data
a group velocity can be ( crudely ) measured and we find that at these radii ( far away from the transition ) this group velocity agrees well with that predicted by the linear dispersion relation . while we are able to pick out wave packets and a group velocity for higher values of the horizontal mode m , we see only standing waves in lower values of m. it is likely that there is a propagating wave component , however , it has significantly less amplitude than the standing waves and is therefore , hard to discern .
this has implications for energy in propagating waves ( section 4.2.2 ) .
the radiative damping of waves can also be seen in figure 11 . in that figure
we show the kinetic energy density in waves as a function of radius for three different frequencies and horizontal mode numbers .
the energy at the convective - radiative interface varies by no more than an order of magnitude across the three different frequencies and modes shown .
however , moving radially inward it is seen , as expected , that the lowest frequency waves are damped more rapidly with depth , with the 2@xmath68hz , m=1 wave amplitude falling by nearly 6 orders of magnitude and the 10 and 20@xmath68hz , m=1 waves dropping by only around 3 orders of magnitude . also as expected ,
higher values of horizontal mode numbers are damped more than m=1 and their energy in the lower frequencies is damped over a shorter distance . for comparison , the dot - dashed line shows the analytic prediction for the damping of the m=1 wave @xcite . using amplitudes for the kinetic energy at the convective - radiative interface from these simulations ,
we damp them with depth using the analytic prescription in @xcite ; there is a substantial difference , particularly for the lowest frequencies . in figure 11
it is seen that ( 1 ) only the very low m waves make it to the centre with any appreciable energy , ( 2 ) all three frequencies are excited with remarkably similar amplitudes and ( 3 ) m=1 looks like a standing wave .
gravity wave energy and flux is of great interest for mixing and angular momentum transport in the solar radiative region .
the time averaged kinetic energy density as a function of radius in the stable region is shown in figure 12 .
the large kinetic energy just below the convection zone is due to convective overshoot .
the local peak in energy density at the centre may be due to the focusing affect discussed by press ( 1981 ) .
we can also calculate the wave kinetic energy as a function of horizontal wavenumber ( @xmath62 ) and frequency ( @xmath73 ) at a particular radius : @xmath74 this can be compared with the energy calculated using ( 14 ) and ( 15 ) : @xmath75 where @xmath76 .
the ( dis)agreement of these two equations gives us an estimate of the applicability of the linear dispersion relation at different radii , wavenumbers and frequencies .
figure 13 shows these two relations at two radii and two horizontal modes as a function of frequency .
it is seen in that figure that the two formulations do not agree near the convective - radiative interface , with the linear dispersion relation underestimating the energy by 2 + orders of magnitude for both m=1 and m=10 .
well away from the convective - radiative boundary , the linear dispersion relation is a good approximation , particularly at m=10 where the two formulations match exactly ( the dotted line is under the solid line ) .
it is expected that the linear dispersion relation would break down near the convective - radiative interface , given that overshooting plumes cause significant nonlinearity . since the linear dispersion relation does not hold near the convective - radiative boundary it is difficult to justify the use of the group velocity calculated from that dispersion relation at those radii .
therefore , in order to estimate the energy flux in waves at the base of the convection zone ( defined as energy density in waves times the group velocity times the surface area ) , we use the group velocity at a deeper radius , where the linear dispersion relation is a good approximation . the wave flux calculated this way is shown in figure 14 .
the peak flux here ( m=1,low frequency ) is three orders of magnitude smaller than the peak wave flux predicted in @xcite for l=1 .
however , in our simulations this flux is distributed evenly in frequency for all ( relatively low ) wavemodes unlike the calculations by @xcite where the flux drops off rapidly with frequency .
this broadband distribution of energy produced is seen in figure 11 and was found in the simulations by @xcite .
these estimates should be considered rough at best , as the group velocity at radii close to the convective - radiative boundary is very uncertain .
the kinetic energy of the fluid , calculated using ( 17 ) is likely not just wave energy .
however , this energy , while not completely wave energy , still contributes to the angular momentum transport and mixing below the convection zone .
in addition , these estimates should be considered as upper limits on the wave flux for a couple of other reasons ; ( 1 ) in two dimensions energy is transferred to the stable region more readily than in 3d , because of the alignment of convective cells and the `` flywheel '' motion that tends to occur in 2d convection because of the inverse cascade , ( 2 ) our thermal diffusivity is signficantly larger than the solar value ( by a factor of @xmath41 ) , causing the convective velocity to be larger than those expected in the sun and therefore , the wave flux could also be larger and ( 3 ) most of the energy in low frequency/ low mode waves are in the standing wave component rather than the propagating wave ( figure 11 ) .
we have presented self consistent calculations of gravity wave excitation by overlying convection and penetration .
these simulations produce power spectra of the gravity waves as a function of radius .
we find that both high frequency standing g - modes and low frequency propagating modes are excited depending on the amplitude of the perturbation . in general
the standing waves are not continually driven by the convection and only larger perturbations to the momentum equation produce these modes .
nonlinear effects need to be considered in the radiation zone for several reasons .
these effects broaden the frequency ridges in the dispersion relation , they allow for transfer of energy from high frequency to low frequency modes .
in addition , just beneath the convection zone , where plumes overshoot , nonlinear interactions increase the energy there by more than two orders of magnitude over what the linear dispersion relation would predict for energy in waves .
while this energy may not be completely wave energy it still contributes to the dynamics of the region just below the convection zone .
the low frequency modes are driven by the overlying convection and penetration .
typical frequencies of these waves can be ten times larger than the convective turnover frequency , which varies from 2 - 10@xmath68hz .
we also find that the energy in these waves is distributed rather evenly in frequency . the steep drop off in energy at larger frequencies
is not observed . while these are self - consistent calculations of gravity waves in a realistic solar simulation , there are many shortcomings that still need to be addressed .
first is the three dimensional effect on the waves and probably more importantly on the nature of the convection which drives the waves .
in addition , any numerical simulation will be more laminar than the sun .
presumably , a more turbulent convection zone should provide a broader spectrum ( in frequency and wavenumber ) of waves .
what effect this will have on the spectrum and dissipation of the waves is unclear .
we thank j. christensen - dalsgaard and d. gough for helpful discussions and guidance as well as an anonymous referee for insightful comments . t.r .
would like to thank the npsc for a graduate student fellowship and the institute of astronomy , cambridge university for a travel grant .
support has also been provided by doe scidac program de - fc02 - 01er41176 , and the ucsc institute of geophysics and planetary physics .
computing resources were provided by nas at nasa ames and by an nsf mri grant ast-0079757 . ) for a convection model with impermeable boundaries ( a ) and for convection model with a stable layer below ( b ) , blue represents negative vorticity while white / yellow represent positive vorticity.,width=480 ] ) ( a ) & ( d ) , horizontal kinetic energy flux ( @xmath77 ) ( b ) & ( e ) and vertical kinetic energy flux ( @xmath77 ) ( c ) & ( f ) .
all plots are scaled to the peak value in the impermeable case .
the qualitative nature of the energy and energy fluxes are not vastly different , however , their amplitudes are , with the total energy density in the convective region of the impermeable case 2 - 3 orders of magnitude larger than in the model bounded by a stable layer .
a similar difference in magnitude is seen in the horizontal flux , while the vertical flux is closer in the two models . ] ) of gravity waves at the base of the convection zone ( a ) and deeper into the radiative interior ( b)/(c ) for our partially linear model .
white represents large energy , blue represents low energy .
peak energy in ( a ) is three orders of magnitude larger than in ( b)/(c ) .
energy is broadband in frequency and horizontal wavenumber near the base of the convection zone .
moving deeper into the radiative interior ridges from standing g - modes are seen at high frequencies , while low frequencies are occupied by propagating waves from the overlying convection .
these spectra were taken shortly after the simulation was begun.,width=480 ] ) in the full computational domain . in ( b ) we show the temperature perturbation ( white represents positive - hotter than reference state temperature , while black represents negative - colder than reference state temperature ) in the radiative region only to emphasize the gravity waves .
it is seen there that the amplitude of the waves falls off quickly and that while many scales of g - modes are produced at the interface , only long wavelength waves survive in the center .
this gravity wave pattern is significantly different than that shown in figure 6.,width=480 ] ) as a function of radius for three frequencies : 2@xmath68hz ( a ) , 10@xmath68hz ( b ) and 20@xmath68hz ( c ) and three different horizontal modes : m=1 ( solid line ) , m=10 ( dotted line ) and m=20 ( dashed line ) and for an analytic prediction of m=1 ( dot - dashed line ) . for m=1 ,
2@xmath68hz waves are damped significantly more than their 10 and 20 @xmath68hz counterparts .
higher values of m are damped more rapidly with depth . only high frequency , low wavenumber waves make it to the center with any appreciable energy .
note that all three frequencies are excited with similar amplitudes .
note m=1 looks like a standing wave for at least 20 @xmath68hz , although it could be argued for all m=1 waves shown . ]
solid line and as @xmath78 ( estimate from the linear dispersion relation ) dotted line . near the convection zone
the linear dispersion relation significantly underestimates the energy density for both modes shown . however ,
deeper in the interior it is a better approximation . ] | we present numerical simulations of penetrative convection and gravity wave excitation in the sun .
gravity waves are self - consistently generated by a convective zone overlying a radiative interior .
we produce power spectra for gravity waves in the radiative region as well as estimates for the energy flux of gravity waves below the convection zone .
we calculate a peak energy flux in waves below the convection zone to be three orders of magnitude smaller than previous estimates for m=1 .
the simulations show that the linear dispersion relation is a good approximation only deep below the convective - radiative boundary .
both low frequency propagating gravity waves as well as higher frequency standing modes are generated ; although we find that convection does not continually drive the standing g - mode frequencies .
[ firstpage ] sun : mixing , internal gravity waves , convection |
in financial mathematics , the fair value @xmath0 of a european style option on two underlying assets is modelled by the two - dimensional _ black scholes partial differential equation _ ( pde ) , see e.g. @xcite , @xmath1 for @xmath2 . here
, @xmath3 denotes the time to maturity @xmath4 and we assume real parameters @xmath5 .
the pde is provided with an initial condition that is defined through the payoff of the option .
the mixed spatial derivative term in represents the correlation between both asset prices in the two - dimensional black scholes model .
mixed spatial derivative terms are very important , notably , in the field of financial option valuation theory . here
they arise due to the correlation between underlying stochastic processes . a well - known approach for determining the fair values @xmath6 consists of numerically solving pde ( [ eq:2dblackscholes ] ) by the _ method - of - lines _
, whereby one first discretizes in space and subsequently in time . in this paper
we consider a uniform cartesian grid and second - order central finite difference schemes in space .
this semidiscretization is second - order convergent with respect to the spatial mesh width if the initial and boundary data is smooth , see e.g. @xcite . when the pde is multidimensional , then the application of classical implicit time discretization methods to the obtained semidiscrete systems can be computationally very intensive . in view of this , for the effective time discretization , operator splitting schemes of the _ alternating direction implicit _ ( adi ) type are widely considered . in this paper
we consider the _ modified craig
sneyd ( mcs ) scheme _
@xcite , which is a prominent scheme of the adi type . in the past years various positive stability results for the mcs scheme
have been derived relevant to multidimensional convection - diffusion equations with mixed derivative terms , see e.g. @xcite .
recently , in t hout and wyns @xcite proved that , under some natural stability and smoothness assumptions , the mcs scheme is second - order convergent with respect to the time step whenever it is applied to semidiscrete two - dimensional convection - diffusion equations with mixed derivative term .
the temporal convergence result from @xcite has the crucial property that it holds uniformly in the spatial mesh width .
hence , the fully discrete numerical solution is second - order convergent in space and time for smooth initial and boundary data .
a relevant convergence analysis for the mcs scheme and nonsmooth data is still open in the literature . in financial applications , however , the initial function is in general nonsmooth .
it is well - known that convergence can then be seriously impaired . as an illustration ,
consider a two - asset cash - or - nothing option with strikes @xmath7 and @xmath8 , so that @xmath9 where @xmath10 denotes the indicator function .
in the upper left plot in figure [ fig : digitalcall2 ] , the numerical solution for @xmath6 is shown for ( natural ) financial parameter values @xmath11
@xmath12 @xmath13 @xmath14 @xmath15 @xmath16 @xmath17 .
irregularities can be observed around the strikes , leading to a loss of accuracy in the maximum norm . for hedging purposes it is important to consider also the greeks , for example the cross gamma @xmath18 .
the corresponding pde is given by @xmath19 for @xmath2 .
this is supplemented with initial function @xmath20 where @xmath21 is the _ dirac delta function_. the lower left plot in figure [ fig : digitalcall2 ] shows the numerical solution for the cross gamma at maturity @xmath4 for the same financial parameter values as above . around the point
@xmath22 strong , spurious erratic behaviour shows up and , hence , this approximation is useless in practice .
if the cross gamma is approximated by applying finite difference schemes directly to the numerical solution for the option value , which is a common alternative technique in practice , the same observations are found . for one - dimensional applications in finance ,
the impact of nonsmooth initial data on convergence has already been studied extensively and various techniques have been proposed in order to recover standard convergence results , see e.g. @xcite .
a common technique consists of first applying several implicit euler ( sub)steps and then continue with the time stepping scheme under consideration , @xcite .
this is called _ rannacher time stepping _ or _ implicit euler damping_. consider again pdes ( [ eq:2dblackscholes ] ) and for the two - asset cash - or - nothing option . replacing the mcs scheme in the first two timesteps by four half - timesteps of the implicit euler scheme , the two right plots in figure [ fig : digitalcall2 ] are obtained . clearly , there are no longer irregularities or oscillations present . in many other multidimensional applications ,
see e.g. @xcite , the same observations were made . to the best of our knowledge , however , there are no theoretical results available in the literature concerning the favourable effect of rannacher time stepping on the convergence of the mcs scheme if the initial data is nonsmooth . in the present paper
we will prove a useful convergence bound for the mcs scheme when it is applied to a model two - dimensional convection - diffusion equation with mixed derivative term , provided with dirac delta initial data . here
, semidiscretization is performed with second - order central finite difference schemes . the precise influence of rannacher time stepping on the order of convergence will be investigated .
our analysis in this paper is inspired by that of giles and carter @xcite , who deal with the crank - nicolson scheme applied to a model one - dimensional convection - diffusion equation .
we make use of a two - dimensional mixed discrete / continuous fourier transformation and analyse the asymptotic behaviour of the fourier transform . by applying then the inverse transformation we arrive at an error bound for the total error .
the sharpness of the error bound is confirmed by ample numerical experiments .
.,title="fig : " ] .,title="fig : " ] .,title="fig : " ] .,title="fig : " ]
semidiscretization by finite difference methods of initial - boundary value problems for time - dependent convection - diffusion equations leads to large systems of stiff ordinary differential equations ( odes ) , @xmath23 with given operator @xmath24 and given initial value @xmath25 .
assume the pde is two - dimensional and the semidiscrete operator @xmath24 is decomposed into a sum @xmath26 where @xmath27 represents the mixed spatial derivative term and @xmath28 , represent all spatial derivative terms in the first , respectively , the second spatial direction .
let @xmath29 be a given parameter , @xmath30 the number of timesteps and set @xmath31 with @xmath32 .
then the _ modified craig sneyd ( mcs ) scheme _ generates , in a one - step fashion , approximations @xmath33 to @xmath34 successively for @xmath35 through @xmath36 the mcs scheme was introduced by in t hout & welfert @xcite for general multidimensional convection - diffusion problems with mixed derivative terms .
it can be viewed as an extension of the craig
sneyd ( cs ) scheme , proposed in @xcite . for @xmath37 , the mcs scheme reduces to the cs scheme . besides @xmath37 , common choices for @xmath38 in the literature are @xmath39 and @xmath40 .
scheme starts with an explicit euler stage applied to the full system , which is followed by two implicit corrections corresponding to each of the two spatial directions .
subsequently an explicit update is performed , followed again by two implicit unidirectional corrector stages .
note that both @xmath24 and @xmath27 , which contain the mixed derivative term , are always treated in an explicit manner .
each implicit stage handles spatial derivatives in only one spatial direction .
this can lead to a major computational advantage in comparison to classical non - splitted implicit time stepping methods .
consider the coordinate transformation @xmath41 and @xmath42 .
the pde ( [ eq:2dblackscholes ] ) is then transformed into @xmath43 for @xmath44 .
this provides a motivation for considering a constant coefficient model convection - diffusion equation with mixed derivative term @xmath45 for @xmath46 and with @xmath47 .
we supplement equation ( [ eq : modelpde ] ) with the initial condition @xmath48 which arises for example in the case of the cross gamma of a two - asset cash - or - nothing option .
the dirac delta initial function , however , has other important applications as well .
for instance , it arises naturally in the adjoint equation for the joint density . by using the _
fourier transform _
pair @xmath49 an exact closed - form analytical solution will be derived .
here @xmath50 denotes the imaginary unit . taking the fourier transformation of equation ( [ eq : modelpde ] ) yields the ode @xmath51 subject to initial condition @xmath52 .
the solution of this transformed equation is given by @xmath53 next , if @xmath54 is a multivariate normal distributed random variable with mean @xmath55 and covariance matrix @xmath56 , its characteristic function is defined by @xmath57 = \exp(\imi\kappa \mu_{1 } + \imi \eta \mu_{2 } - \tfrac{1}{2 } ( \kappa \ \eta ) \sigma ( \kappa \ \eta)^{\top}).\ ] ] by exploring the connection between the characteristic function of a random variable and the fourier transform of its density function , it follows that @xmath58 can be seen as the density function of a two - dimensional normal distributed random variable with mean @xmath59 and covariance matrix @xmath56 given by @xmath60 since @xmath47 , this yields the closed - form analytical solution @xmath61 \right).\ ] ]
as mentioned in section [ intro ] , spatial discretisation of ( [ eq : modelpde ] ) will be performed on a uniform cartesian grid with second - order central finite difference schemes . for the time integration
the mcs scheme will be considered .
let @xmath62 denote the spatial mesh width in the @xmath63-direction , @xmath64 the spatial mesh width in the @xmath65-direction and define spatial gridpoints @xmath66 for all @xmath67 .
semidiscretization of ( [ eq : modelpde ] ) with second - order central finite difference schemes then gives rise to approximations @xmath68 of the exact solution value @xmath69 which are defined by the system @xmath70 where @xmath71 and @xmath72 with @xmath73 the usual second - order central finite difference operators .
for example , @xmath74 semidiscrete system ( [ eq : semidiscretesystem ] ) is provided with initial data @xmath75 in order to approximate the dirac delta initial function . for convenience
we define @xmath76 and we denote by @xmath77 the identity operator .
then , starting from @xmath78 , application of the mcs scheme to semidiscrete system ( [ eq : semidiscretesystem ] ) yields approximations @xmath79 of @xmath80 successively for @xmath81 through @xmath82 concerning the rannacher time stepping , let @xmath83 denote the number of initial mcs time steps replaced by @xmath84 half - time steps of implicit euler integration . whenever @xmath85 scheme ( [ eq : mcsonmodel ] ) is replaced by @xmath86 for @xmath87 .
this provides a numerical approximation @xmath88 of the exact solution .
the goal of our convergence analysis consists of quantifying the _ total error _
@xmath89 to do so , we will analyse the asymptotic behaviour of a mixed discrete / continuous fourier transform for @xmath90 simultaneously tending to zero . applying
the inverse fourier transformation on the resulting error in fourier space will yield a useful bound for the total error .
special attention will be paid to the influence of @xmath83 , i.e. the influence of rannacher time stepping , on the total error .
we consider a mixed discrete / continuous fourier transform pair , cf .
e.g. @xcite , @xmath91 for ease of presentation , the dependency of the fourier transform on @xmath92 and @xmath93 will be omitted in the notation .
fourier transformation of @xmath94 yields @xmath95 .
concerning operator @xmath96 it follows that @xmath97\widehat{v } \\ & = & - \tfrac{2 \rho \delta t}{h_{1}h_{2 } } ( \sin\vartheta_{1}\sin\vartheta_{2 } ) \widehat{v}.\end{aligned}\ ] ] analogously one finds @xmath98 define functions @xmath99 and @xmath100 .
then , fourier transformation of the implicit euler scheme ( [ eq : rannacherstartup ] ) gives @xmath101 after some calculations , fourier transformation of the mcs scheme ( [ eq : mcsonmodel ] ) yields @xmath102 with @xmath103 where @xmath104 assume that @xmath105 .
since @xmath106 it follows that @xmath107 by applying the inverse fourier transformation , the numerical approximation at @xmath108 can be written as @xmath109 where we made use of the substitutions @xmath110 from section [ modelproblem ] it can be seen that the exact solution is given by @xmath111 in our analysis , we will examine the _ fourier error _
@xmath112 for @xmath113 tending to zero , the _ total error _ is approximated by @xmath114 note that expression can be viewed as the inverse mixed discrete / continuous fourier transform of the fourier error . in figure
[ fig : normtransformexact ] , @xmath115 is shown in the @xmath116-domain for parameter values @xmath117 .
this has to be compared with figure [ fig : normtransform ] where @xmath118 is shown for the same parameter values .
discretization is performed with @xmath119 and well - known mcs parameters @xmath120 . for the rannacher time stepping we considered values @xmath121 from figure [ fig : normtransformexact ] and figure [ fig : normtransform ] it is clear that the difference @xmath122 has different properties in different regions of the fourier domain .
these regions are illustrated in figure [ fig : regions ] .
first there is a _ low - wavenumber region _
, where both @xmath123 and @xmath124 are small , in which there is a good agreement between @xmath125 and @xmath126 .
next , if either @xmath123 or @xmath124 is medium and the other one is small or medium ( region ) , then both the fourier transforms of the numerical solution and analytical solution are negligible . in the _ high - wavenumber region _ , i.e. where both @xmath127 are large , we observe that the modulus of the fourier transform @xmath126 is close to zero .
the modulus @xmath128 , however , is strongly dependent on @xmath83 and the mcs parameter @xmath38 . for larger values of @xmath38
we see that @xmath125 has a larger magnitude in the high - wavenumber region .
hence , a larger high - wavenumber error can be expected for larger values of @xmath38 .
further we observe that the modulus of @xmath125 in the high - wavenumber region is always damped whenever rannacher time stepping is applied .
this matches our observations from figure [ fig : digitalcall2 ] where unwanted erratic behaviour was avoided by using rannacher time stepping .
finally , we have the case where either @xmath123 or @xmath124 is large but the other one is not . in our analysis , the region where @xmath123 is large and the region where @xmath124 is large will be treated separately . in both regions
the fourier transform @xmath126 is negligible but @xmath125 has to be further analysed .
in particular , we will show that @xmath125 is not negligible if the mcs scheme reduces to the cs scheme .
following giles & carter @xcite we will perform an asymptotic analysis of the fourier error @xmath122 in each of these ( five ) disjoint regions which form a partition of the fourier domain .
we consider the limit @xmath129 and since the same discretization is performed in both spatial directions , @xmath130 _ is held fixed_. for ease of presentation we denote @xmath131 .
further , since both the semidiscretization and the time integration are convergent of order two for smooth initial data , it seems natural to keep @xmath132 _
constant_. substitutions @xmath133 yield & & z_0 = - h c h = - ( ( ( - c)h ) - ( ( + c)h ) ) , + & & z_1 = - ^2 + a_1 h = - ( 1-h ) + a_1 h , + & & z_2 = - ^2 + a_2 ch = - ( 1-ch ) + a_2 ch .
[ eq : z0z1z2 ] the expressions in ( [ eq : z0z1z2 ] ) will be used to analyse the asymptotic behaviour of ( [ eq : fouriertransformnumericalsolution ] ) as @xmath134 .
_ throughout the analysis , by the notation @xmath135 we shall always mean that the modulus @xmath136 of the term under consideration is bounded by a positive constant times @xmath137 where the constant is independent of @xmath138 and the mesh width @xmath139 .
_ in order to deal with the powers in expression a @xmath140-transformation of @xmath125 will be considered . since @xmath141 , thus @xmath142
, it follows that _ n & = & ( n - n_0 ) ( r ) + 2n_0 ( ) + & = & + & & + n_0 . for parameter values @xmath117.,title="fig : " ] + with @xmath143 ( left ) and @xmath144 ( right ) for mcs parameter @xmath39 ( top ) , @xmath37 ( middle ) and @xmath145 ( bottom ) .
the other parameter values are : @xmath146 .
, title="fig : " ] with @xmath143 ( left ) and @xmath144 ( right ) for mcs parameter @xmath39 ( top ) , @xmath37 ( middle ) and @xmath145 ( bottom ) .
the other parameter values are : @xmath146 .
, title="fig : " ] with @xmath143 ( left ) and @xmath144 ( right ) for mcs parameter @xmath39 ( top ) , @xmath37 ( middle ) and @xmath145 ( bottom ) .
the other parameter values are : @xmath146 .
, title="fig : " ] with @xmath143 ( left ) and @xmath144 ( right ) for mcs parameter @xmath39 ( top ) , @xmath37 ( middle ) and @xmath145 ( bottom ) .
the other parameter values are : @xmath146 .
, title="fig : " ] with @xmath143 ( left ) and @xmath144 ( right ) for mcs parameter @xmath39 ( top ) , @xmath37 ( middle ) and @xmath145 ( bottom ) .
the other parameter values are : @xmath146 .
, title="fig : " ] with @xmath143 ( left ) and @xmath144 ( right ) for mcs parameter @xmath39 ( top ) , @xmath37 ( middle ) and @xmath145 ( bottom ) .
the other parameter values are : @xmath146 .
, title="fig : " ] multiple regions will encounter values @xmath147 with certain @xmath148 . by
_ taylor expansion _ of ( [ eq : z0 ] ) it directly follows that @xmath149 } h + z_{0}^{[3 ] } h^{3 } + z_{0}^{[5 ] } h^{5},\end{aligned}\ ] ] where @xmath150 } & = & -2\rho \lambda \kappa \eta , \\
z_{0}^{[3 ] } & = & \tfrac{1}{3}\rho \lambda ( \kappa^{2 } + c^{2}\eta^{2})\kappa\eta , \\
\vert z_{0}^{[5 ] } \vert & \leq & \tfrac{4}{6!}\tfrac{\rho\lambda}{c } ( \vert \kappa \vert + c \vert \eta \vert ) ^{6}. \end{aligned}\ ] ] analogously as above , taylor expansion of ( [ eq : z1 ] ) and ( [ eq : z2 ] ) yields @xmath151 } h + z_{1}^{[3 ] } h^{3 } + z_{1}^{[5 ] } h^{5 } , \\
z_{2}(h ) & = & z_{2}^{[1 ] } h + z_{2}^{[3 ] } h^{3 } + z_{2}^{[5 ] } h^{5 } , \end{aligned}\ ] ] where @xmath152 } & = & -\lambda \kappa^{2 } + \imi a_{1 } \lambda \kappa ,
\\ z_{1}^{[3 ] } & = & \tfrac{1}{12}\lambda\kappa^{4 } - \tfrac{1}{6 } \imi a_{1 } \lambda \kappa^{3 } , \\
\vert z_{1}^{[5 ] } \vert & \leq & \tfrac{2}{6!}\lambda\kappa^{6 } + \tfrac{1}{5 ! }
\vert a_{1 } \vert \lambda \vert \kappa \vert^{5 } , \\\\ z_{2}^{[1 ] } & = & - \lambda \eta^{2 } + \imi a_{2 } \lambda \eta , \\ z_{2}^{[3 ] } & = & \tfrac{1}{12}\lambda c^{2 } \eta^{4 } - \tfrac{1}{6 } \imi a_{2 } \lambda c^{2 } \eta^{3 } , \\
\vert z_{2}^{[5 ] } \vert & \leq & \tfrac{2}{6 ! }
\lambda c^{4}\eta^{6 } + \tfrac{1}{5!}\vert a_{2 } \vert \lambda c^{4 } \vert \eta \vert^{5}. \end{aligned}\ ] ] since @xmath148 , it is ensured that all terms in the above expansions stay bounded as @xmath139 tends to zero . using these expansions and the definition of @xmath153 it follows that @xmath154}h + p^{[2]}h^{2 } + p^{[3]}h^{3 } + p^{[4]}h^{4 } + p^{[5]}h^{5},\ ] ] where @xmath155 } & = & -\theta ( z_{1}^{[1 ] } + z_{2}^{[1 ] } ) , \\
p^{[2 ] } & = & \theta^{2}z_{1}^{[1]}z_{2}^{[1 ] } , \\
p^{[3 ] } & = & - \theta ( z_{1}^{[3 ] } + z_{2}^{[3 ] } ) , \\ p^{[4 ] } & = & \theta^{2 } ( z_{1}^{[1]}z_{2}^{[3 ] } + z_{1}^{[3]}z_{2}^{[1 ] } ) , \\ p^{[5 ] } & = & \mathcal{o}\left(1+(\kappa^{2 } + c^{2}\eta^{2})^{3}\right).\end{aligned}\ ] ] under the condition @xmath147 with certain @xmath148 , the variables @xmath156 and @xmath157 can become very large as @xmath139 tends to zero . in this case the highest powers of @xmath138 will dominate the order term in @xmath158}$ ] .
under the same condition , however , @xmath156 and @xmath157 can both be very small and then the lowest powers of @xmath159 will dominate . by considering the sum of @xmath160 and the highest powers of @xmath159 in the remaining order term , we ensure that both cases are covered .
as mentioned above we will make use of @xmath140-transformation ( [ eq : logtransformfouriertransformnumercialsolution ] ) to analyse the asymptotic behaviour .
let @xmath161 be a strictly positive and sufficiently smooth function and set @xmath162 taylor expansion yields @xmath163}h + g^{[2]}h^{2 } + g^{[3]}h^{3 } + g^{[4]}h^{4},\ ] ] where @xmath164 } & = & \tfrac{f'(0)}{f(0 ) } , \\
g^{[2 ] } & = & \tfrac{1}{2}\left ( \tfrac{f''(0)}{f(0 ) } - \tfrac{f'(0)^{2}}{f(0)^{2 } } \right ) , \\
g^{[3 ] } & = & \tfrac{1}{6}\left ( \tfrac{f'''(0)}{f(0 ) } - 3\tfrac{f'(0)f''(0)}{f(0)^{2 } } + 2 \tfrac{f'0)^{3}}{f(0)^{3 } } \right ) , \\ g^{[4 ] } & = & \tfrac{1}{4 ! } \left ( \tfrac{f^{(4)}(\xi)}{f(\xi ) } - \tfrac{4f'(\xi)f'''(\xi ) + 3 f''(\xi)^{2}}{f(\xi)^{2 } } + 12 \tfrac{f'(\xi)^{2}f''(\xi)}{f(\xi)^{3 } } - 6 \tfrac{f'(\xi)^{4}}{f(\xi)^{4 } } \right ) \quad \mbox{with certain } \ 0 <
\xi < h.\end{aligned}\ ] ] in order to encounter the first part of ( [ eq : logtransformfouriertransformnumercialsolution1 ] ) consider @xmath165 so that @xmath166 and thus @xmath167 concerning the rannacher time stepping , define @xmath168 such that @xmath169 all of these expressions will be used in the forthcoming subsections , where we analyse the asymptotic behaviour of @xmath125 in five different regions of the fourier domain , i.e. the @xmath170-domain with @xmath171 .
in order to analyse @xmath174 in this region , the parts stemming from the mcs scheme and rannacher time stepping will be considered separately .
write ( [ eq : logtransformfouriertransformnumercialsolution1 ] ) as @xmath175 = \tfrac{1}{\lambda h } \left [ \log(f_{m}(h ) ) - 2 \log(p(h))\right].\ ] ] using the analysis above it follows that @xmath176 = s^{[0 ] } + s^{[1]}h + s^{[2]}h^{2 } + s^{[3]}h^{3},\ ] ] where @xmath177 } & = & 2p'(0 ) + z'(0 ) - 2p'(0 ) , \\ \lambda s^{[1 ] } & = & \tfrac{1}{2}\left [ 2p'(0)^{2 } + 2p''(0 ) + 2p'(0)z'(0 ) + 2\theta z_{0}'(0)z'(0 ) + 2(\tfrac{1}{2 } - \theta ) z'(0)^{2 } - ( 2p'(0 ) + z'(0))^{2 } \right ] \\ & & - \ \left [ p''(0 ) - p'(0)^{2}\right ] \\ \lambda s^{[2 ] } & = & \tfrac{1}{6 } \left [ 6p'(0)p''(0 ) + 2p'''(0 ) + 3p''(0)z'(0 ) + z'''(0 ) \right . \\ & & - \
3(2p'(0 ) + z'(0))(2p'(0)^{2 } + 2p''(0 ) + 2p'(0)z'(0 ) + 2\theta z_{0}'(0)z'(0 ) + 2(\tfrac{1}{2 } - \theta)z'(0)^{2 } ) \\ & & \left .
+ \ 2 ( 2p'(0 ) + z'(0))^{3 } \right ] - \tfrac{1}{3 } \left [ p'''(0 ) -3p'(0)p''(0 ) + 2p'(0)^{3 } \right ] \\ s^{[3 ] } & = & \mathcal{o}\left ( 1+(\kappa^{2 } + c^{2}\eta^{2})^{4 } \right).\end{aligned}\ ] ] by using the expansions in subsection [ subsec : taylor ] and after simplifying the resulting expressions , one gets @xmath178 } & = & - \kappa^{2 } - 2\rho\kappa\eta - \eta^{2 } + \imi a_{1}\kappa + \imi a_{2 } \eta , \\ s^{[1 ] } & = & 0 , \\ s^{[2 ] } & = & \tfrac{1}{12 } \kappa^{4 } + \tfrac{1}{3 } \rho ( \kappa^{2 } + c^{2}\eta^{2 } ) \kappa \eta + \tfrac{1}{12 } c^{2}\eta^{4 } - \tfrac{1}{6 } \imi a_{1 } \kappa^{3 } - \tfrac{1}{6 } \imi a_{2 } c^{2 } \eta^{3 } \\ & & - \ \lambda^{2}\theta^{2 } ( - \kappa^{2 } + \imi a_{1}\kappa ) ( - \eta^{2 } + \imi a_{2 } \eta ) ( -\kappa^{2 } -2\rho \kappa \eta - \eta^{2 } + \imi a_{1}\kappa + \imi a_{2 } \eta ) \\ & & + \
\tfrac{\lambda^{2}}{12 } ( -\kappa^{2 } -2\rho \kappa \eta - \eta^{2 } + \imi a_{1}\kappa + \imi a_{2 } \eta)^{3 } \\ & & - \ \lambda^{2 } ( -\kappa^{2 } -2\rho \kappa \eta - \eta^{2 } + \imi a_{1}\kappa + \imi a_{2 } \eta ) ( - \rho \kappa \eta + ( \tfrac{1}{2 } - \theta ) ( -\kappa^{2 } - \eta^{2 } + \imi a_{1}\kappa + \imi a_{2 } \eta ) ) ^{2}.\end{aligned}\ ] ] as for the part stemming from the rannacher time stepping , write ( [ eq : logtransformfouriertransformnumercialsolution2 ] ) as @xmath179.\ ] ] using the same analysis as above one gets @xmath180 = n_{0}^{[1]}h + n_{0}^{[2]}h^{2 } + n_{0}^{[3]}h^{3},\ ] ] where @xmath181 } & = & n_{0 } \left [ - \lambda s^{[0 ] } - 2(-\tfrac{1}{2 } z'(0 ) ) \right ] = 0 , \\ n_{0}^{[2 ] } & = & n_{0 } \left [ - \lambda s^{[1 ] } - ( -\tfrac{1}{2 } z''(0 ) - \tfrac{1}{4 } z'(0)^{2 } ) \right ] = \tfrac{1}{4 } n_{0 } z'(0)^{2 } \\ & = & \tfrac{1}{4 } n_{0 } \lambda^{2 } ( -\kappa^{2 } -2\rho \kappa \eta - \eta^{2 } + \imi a_{1}\kappa + \imi a_{2 } \eta)^{2 } ,
\\ n_{0}^{[3 ] } & = & \mathcal{o}\left ( 1+(\kappa^{2 } + c^{2}\eta^{2 } ) ^{3 } \right).\end{aligned}\ ] ] by combining ( [ eq : logtransformfouriertransformnumercialsolution ] ) , ( [ eq : mcstaylor ] ) and ( [ eq : rannachertaylor ] ) it directly follows that @xmath182 } + n_{0}^{[2]})h^{2 } + ( s^{[3 ] } + n_{0}^{[3]})h^{3},\ ] ] and hence @xmath183 } + n_{0}^{[2]})h^{2 } + ( s^{[3 ] } + n_{0}^{[3]})h^{3}).\ ] ] next , we will expand the second exponential in order to compare this expression with the fourier transform @xmath126 from ( [ eq : fouriertransformexact ] ) at @xmath184 .
let @xmath185 } h^{2 } + c^{[3 ] } h^{3}),\ ] ] where @xmath186 } = \mathcal{o}\left ( 1+(\kappa^{2 } + c^{2}\eta^{2 } ) ^{3 } \right ) , \
c^{[3 ] } = \mathcal{o}\left ( 1+(\kappa^{2 } + c^{2}\eta^{2 } ) ^{4 } \right),\ ] ] then @xmath187 } h + 3 c^{[3 ] } h^{2 } ) e(h ) , \\
e''(h ) & = & ( 2c^{[2 ] } + 6c^{[3]}h)e(h ) + ( 2c^{[2 ] } h + 3 c^{[3 ] } h^{2 } ) ^{2 } e(h ) , \\
e'''(h ) & = & 6c^{[3 ] } e(h ) + 3(2c^{[2 ] } + 6c^{[3 ] } h)(2c^{[2 ] } h + 3 c^{[3 ] } h^{2 } ) e(h ) + ( 2c^{[2 ] } h + 3 c^{[3 ] } h^{2 } ) ^{3 } e(h).\end{aligned}\ ] ] since @xmath188 with @xmath189 , we have that @xmath190 and @xmath191 whenever @xmath139 is sufficiently small .
hence it follows that @xmath192 } h^{2 } + e^{[3 ] } h^{3},\ ] ] with @xmath193 } & = & c^{[2 ] } , \\
e^{[3 ] } & = & \mathcal{o}\left ( 1+(\kappa^{2 } + c^{2}\eta^{2})^{4 } \right ) + \mathcal{o}\left ( 1+(\kappa^{2 } + c^{2}\eta^{2})^{6 } \right)h + \mathcal{o}\left ( 1+(\kappa^{2}+c^{2}\eta^{2})^{9 } \right)h^{3 } \\ & = & \mathcal{o}\left ( 1+(\kappa^{2 } + c^{2}\eta^{2})^{4 } \right ) + \mathcal{o}\left ( 1+(\kappa^{2 } + c^{2}\eta^{2})^{6 } \right)h,\end{aligned}\ ] ] where the latter equality follows from the assumption @xmath188 with @xmath189 .
finally , for this region , one arrives at the following expression for the fourier error : @xmath194 } + n_{0}^{[2 ] } ) + \mathcal{o}\left ( 1+(\kappa^{2 } + c^{2}\eta^{2})^{4 } \right)h + \mathcal{o}\left ( 1+(\kappa^{2 } + c^{2}\eta^{2})^{6 } \right)h^{2}\right ) .
\label{eq : lowwavenumbertransform}\ ] ] note that @xmath195}$ ] and @xmath196}$ ] actually depend on @xmath156 and @xmath157 .
for ease of presentation , this is omitted in the notation .
first consider the case where both @xmath201 and @xmath202 .
based on the analysis in subsection [ region1 ] , expression ( [ eq : logtransformfouriertransformnumercialsolution1 ] ) can be rewritten as @xmath203 = s^{[0 ] } + s^{[2']}h^{2},\ ] ] where @xmath178 } & = & -\kappa^{2 } - 2\rho\kappa\eta - \eta^{2 } + \imi a_{1}\kappa + \imi a_{2 } \eta , \\ s^{[2 ' ] } & = & \mathcal{o}\left ( ( \kappa^{2 } + c^{2 } \eta^{2})^{3 } \right).\end{aligned}\ ] ] since either @xmath156 or @xmath157 becomes large in this region as @xmath139 tends to zero , only the highest powers of @xmath159 are taken into account in the order term in @xmath204}$ ] . from @xmath205 one
gets @xmath206 such that @xmath207 } ) \leq -(1- \vert \rho \vert ) ( \kappa^{2 } + \eta^{2 } ) < 0.\ ] ] using that both both @xmath201 and @xmath202 , it directly follows that @xmath208 and thus @xmath209 hence , for @xmath210 @xmath211 and since @xmath212 or @xmath213 we may conclude @xmath214 next , consider the case where at least one of the equalities , @xmath215 or @xmath216 , holds . for analysing the asymptotic behaviour of @xmath217 we then make use of the following proposition .
its proof is a direct modification of the proof of one of the statements in ( * ? ? ? * theorem 1 ) and is therefore omitted .
[ prop : stabiliteit ] let @xmath218 denote real numbers with @xmath219 and @xmath47 . set @xmath220 and @xmath221 . if @xmath222 or @xmath223 , then @xmath224 whenever @xmath225 and @xmath226 recall that in the current region of the fourier domain the assumption @xmath227 with @xmath228 holds .
this yields @xmath229 since @xmath230 or @xmath231 it follows that @xmath232 or @xmath233 .
hence , all the assumptions on @xmath234 in proposition [ prop : stabiliteit ] are fulfilled such that @xmath235 and thus @xmath236 for @xmath237 further , it always holds that @xmath238 such that @xmath239 by combining this with and , and by using that @xmath83 is independent from @xmath139 , one may conclude that in this region it holds that @xmath240 under restriction on @xmath38 .
this means that @xmath118 quickly becomes negligible as @xmath139 tends to zero .
it decays faster to zero than any polynomial in @xmath139 .
here we reconsider the substitutions @xmath243 in order to get @xmath244 further , in this region @xmath245 are different from zero and since we consider values @xmath246 , we may write @xmath247 and thus @xmath248 making use of an expansion similar to ( [ eq : taylorexpansionlogfunction ] ) it follows that @xmath249 \\ & = & \log \theta^{4 } \\ & & + \
\left [ 2 \theta^{3 } \left ( \frac{1}{4\lambda \sin^{2 } \tfrac{\vartheta_{1}}{2 } } - \theta \imi a_{1 } \frac { \cot \tfrac{\vartheta_{1}}{2}}{2 } + \frac{c^{2}}{4\lambda \sin^{2 } \tfrac{\vartheta_{2}}{2 } } - \theta \imi a_{2 } \frac { c \cot \tfrac{\vartheta_{2}}{2}}{2 } \right ) \right . \\ & & - \ \left .
\rho \frac{c \cot \tfrac{\vartheta_{1}}{2}\cot \tfrac{\vartheta_{2}}{2}}{2\lambda } - \theta^{2 } \frac{1}{4\lambda \sin^{2}\tfrac{\vartheta_{1}}{2 } } - \theta^{2 } \frac{c^{2}}{4\lambda \sin^{2}\tfrac{\vartheta_{2}}{2 } } \right ] \frac{h}{\theta^{4 } } \\ & & + \
\mathcal{o}\left ( \left ( \frac{1}{\vert \sin \tfrac{\vartheta_{1}}{2 } \vert } + \frac{c}{\vert \sin \tfrac{\vartheta_{2}}{2 } \vert } \right)^{4}h^{2 } \right),\end{aligned}\ ] ] and @xmath250 & = & \log \theta^{2 } \\ & & + \ \left ( \frac{1}{4\lambda \sin^{2 } \tfrac{\vartheta_{1}}{2 } } - \theta \imi a_{1 } \frac { \cot \tfrac{\vartheta_{1}}{2}}{2 } + \frac{c^{2}}{4\lambda \sin^{2 } \tfrac{\vartheta_{2}}{2 } } - \theta \imi a_{2 } \frac { c \cot \tfrac{\vartheta_{2}}{2}}{2 } \right ) \frac{h}{\theta } \\ & & + \ \mathcal{o}\left ( \left ( \frac{1}{\vert \sin \tfrac{\vartheta_{1}}{2 } \vert } + \frac{c}{\vert \sin \tfrac{\vartheta_{2}}{2 } \vert } \right)^{4}h^{2 } \right).\end{aligned}\ ] ] combining both expressions yields @xmath251 further , recall that in this region @xmath245 are both different from zero such that @xmath252 and hence @xmath253 as for the implicit euler time stepping scheme we note @xmath254 using once again an expansion analogous to ( [ eq : taylorexpansionlogfunction ] ) it follows that @xmath255 which yields @xmath256 making use of relationship ( [ eq : fouriertransformnumericalsolution ] ) one becomes an expression for the logarithm of the fourier transform @xmath125 : @xmath257 such that in this region @xmath258^{2n_{0 } } } \exp\left ( - \frac{1}{4\lambda^{2 } \theta^{2 } } \frac{\iota(\vartheta_{1},\vartheta_{2 } ) } { \sin^{2}\tfrac{\vartheta_{1}}{2 } \sin^{2 } \tfrac{\vartheta_{2}}{2 } } \right ) \left ( 1 + \mathcal{o}\left ( \frac{h}{\left ( \vert \vartheta_{1 } \vert + \vert \vartheta_{2 } \vert \right)^{2 } } \right ) \right ) .
\label{eq : highwavenumbertransform}\ ] ] in figure [ fig : normtransform ] we noticed that in the high - wavenumber region , i.e. where both @xmath259 are large , the norm @xmath118 is highly dependent on the mcs parameter @xmath38 .
this is confirmed by ( [ eq : highwavenumbertransform ] ) since inequality ( [ eq : positivitydenominator ] ) holds in the high - wavenumber region .
hence , _ for larger values of the mcs parameter @xmath38 one can expect a larger high - wavenumber error_. reconsider the substitution @xmath262 and recall that @xmath263 . then , as @xmath92 is non - zero in this region , one may write @xmath264 such that @xmath265 where @xmath266 denotes a positive real number . hence , concerning @xmath217 it follows that @xmath267 for the latter expression we obtain the following positive result .
[ prop : regio5 ] if @xmath268 and @xmath269 , then @xmath270 for all real numbers @xmath271 . if @xmath272 or @xmath37 , then the inequality holds for numbers @xmath273 .
let @xmath266 be a positive real number .
first , it is clear that the inequality holds whenever both & & -(2+)+ < 0 , + & & 2 ( 1 + ) ^2 ^2 - ( 2 + ) + > 0 .
it is readily seen that ( [ eq : proposition2ineq1 ] ) is satisfied for @xmath268 . for strictly positive @xmath266
the inequality is satisfied whenever @xmath274 .
regarding inequality ( [ eq : proposition2ineq2 ] ) we consider the left - hand side as a second - order polynomial in @xmath38 with discriminant @xmath275 if @xmath276 , then @xmath277 and the polynomial is strictly positive for all real numbers @xmath38 .
if @xmath278 , the polynomial reduces to @xmath279 which reaches its minimum ( zero ) in @xmath37 .
let @xmath268 and @xmath269 .
then , applying proposition [ prop : regio5 ] in this region yields @xmath280 and thus @xmath281 since @xmath83 is independent from @xmath139 and @xmath282 one may conclude that @xmath283 next , consider the case @xmath37 . recall that the mcs scheme then reduces to the original cs scheme .
if @xmath231 , it follows that @xmath284 with @xmath285 such that proposition [ prop : regio5 ] can be applied and @xmath286 for all @xmath287 .
now , assume @xmath288 with @xmath202 .
an expansion similar to ( [ eq : taylorexpansionlogfunction ] ) yields @xmath289 \\ & = & \log ( -\theta^2 + 2\theta - \tfrac{1}{2 } ) \\ & & - \ \left [ 2\theta \left ( \frac{1}{4\lambda \sin^{2}\tfrac{\vartheta_{1}}{2 } } - \tfrac{1}{2 } \theta \imi a_{1 } \cot \tfrac{\vartheta_{1}}{2 } - \theta^{2 } z_{2}^{[1 ] } \right ) \right
. + \theta \left ( -\rho \cot \tfrac{\vartheta_{1}}{2 } \eta + \tfrac{1}{2 } \imi a_{1 } \cot \tfrac{\vartheta_{1}}{2 } \right ) \\ & & - \ \left ( \frac{1}{4\lambda \sin^{2 } \tfrac{\vartheta_{1}}{2 } } - \tfrac{1}{2 } \theta \imi a_{1 } \cot \tfrac{\vartheta_{1}}{2 } - \theta^{2 } z_{2}^{[1 ] } \right ) + \theta \rho \cot \tfrac{\vartheta_{1}}{2 } \eta \\ & & - \ 2(\tfrac{1}{2 } - \theta ) \left ( - \rho \cot \tfrac{\vartheta_{1}}{2 } \eta + \tfrac{1}{2 } \imi a_{1 } \cot \tfrac{\vartheta_{1}}{2 } \right ) \bigg ] \frac{h}{-\theta^{2 } + 2\theta - \tfrac{1}{2 } } \\ & & + \ \mathcal{o}\left ( \left ( \frac{1}{\sin^{2 } \tfrac{\vartheta_{1}}{2 } } + \frac{\vert \eta \vert}{\vert \sin \tfrac{\vartheta_{1}}{2 } \vert } + \eta^{2 } \right)^{2 } h^{2 } \right),\end{aligned}\ ] ] and @xmath290 & = & \log \theta^{2 } \\ & & + \
2\theta \left ( \frac{1}{4\lambda \sin^{2}\tfrac{\vartheta_{1}}{2 } } - \tfrac{1}{2 } \theta \imi a_{1 } \cot \tfrac{\vartheta_{1}}{2 } - \theta^{2 } z_{2}^{[1 ] } \right ) \frac{h}{\theta^{2 } } \\ & & + \
\mathcal{o}\left ( \left ( \frac{1}{\sin^{2 } \tfrac{\vartheta_{1}}{2 } } + \eta^{2 } \right)^{2 } h^{2 } \right).\end{aligned}\ ] ] making use of @xmath37 it follows that @xmath291 } \right ) h + \mathcal{o}\left ( \left ( \frac{1}{\sin^{2 } \tfrac{\vartheta_{1}}{2 } } + \frac{\eta}{\vert \sin \tfrac{\vartheta_{1}}{2 } \vert } + \eta^{2 } \right)^{2 } h^{2 } \right ) \\ & = & \left ( \frac{-1}{\lambda \sin^{2}\tfrac{\vartheta_{1}}{2 } } - \lambda \eta^{2 } + \imi a_{2 } \lambda \eta \right ) h + \mathcal{o}\left ( \left ( \frac{1}{\sin^{2 } \tfrac{\vartheta_{1}}{2 } } + \frac{\vert \eta \vert}{\vert \sin \tfrac{\vartheta_{1}}{2 } \vert } + \eta^{2 } \right)^{2 } h^{2 } \right ) , \\\end{aligned}\ ] ] and hence @xmath292 in order to analyse the rannacher time stepping we note @xmath293 which yields @xmath294 by exploring relationship ( [ eq : fouriertransformnumericalsolution ] ) one may conclude that @xmath295 whenever @xmath231 , the right - hand side of ( [ eq : cswavenumbererror ] ) is @xmath296 for all @xmath297 such that we can use expression ( [ eq : cswavenumbererror ] ) for the whole region in the case of @xmath37 .
the analysis for this region is completely analogous to the analysis in subsection [ region4 ] .
hence , for @xmath268 and @xmath300 it follows that @xmath283 whenever the cs scheme is considered , i.e. @xmath37 , one gets the expression @xmath301 in the above analysis natural bounds on the mcs parameter @xmath38 arise under which the asymptotic results are valid .
these bounds can be interpreted as stability bounds . in particular , the conditions @xmath302 are needed to ensure that the fourier transform @xmath125 is negligible in the second region .
this restriction is only slightly stronger than the lower bound on @xmath38 derived in @xcite , guaranteeing unconditional stability of the mcs scheme in the von neumann sense pertinent to two - dimensional diffusion equations with mixed derivative term .
this is , indeed , not very surprising . in @xcite
it is stated that the stability analysis of the mcs scheme in this case reduces to bounding by one of the modulus of the scalar expression @xmath303 where @xmath304 and @xmath218 denote real numbers satisfying the condition ( [ eq : scalarrestrictions ] ) .
this explains why proposition [ prop : stabiliteit ] is just a slight modification of one of the statements in ( * ? ? ?
* theorem 1 ) .
in this section we will use the asymptotic results in fourier space from section [ results ] to perform an error analysis in physical space . first note that the fourier transform @xmath126 is only sizeable in region 1 of the fourier domain . in the other regions it holds that @xmath305 or @xmath306 and hence @xmath307 based on equalities , ( [ eq : lowwavenumbertransform ] ) and ( [ eq : highwavenumbertransform ] ) we define @xmath308}(\kappa,\eta ) + n_{0}^{[2]}(\kappa,\eta))\ ] ] and @xmath309^{2n_{0 } } } \exp\left ( - \frac{1}{4\lambda^{2 } \theta^{2 } } \frac{\iota(\vartheta_{1},\vartheta_{2 } ) } { \sin^{2}\tfrac{\vartheta_{1}}{2 } \sin^{2 } \tfrac{\vartheta_{2}}{2 } } \right).\ ] ] recall that @xmath310 @xmath311 and @xmath312 . as a consequence , @xmath313 is only sizeable in region 1 and @xmath314 is only sizeable in region 3 . in the other regions of the fourier domain @xmath125
is negligible whenever @xmath315 and @xmath300 .
hence , for these values of @xmath38 , the results can be combined to @xmath316 when @xmath37 , i.e. when the mcs scheme reduces to the cs scheme , @xmath125 is also sizeable in region 4 and region 5 of the fourier domain .
this case will be treated separately .
consider the case where the mcs scheme is different from the cs scheme , i.e. @xmath269 , and suppose that the restriction @xmath315 is satisfied .
approximation ( [ eq : fouriererror ] ) is then valid and based on we have for the total error : @xmath317 where @xmath318}(\kappa,\eta ) + n_{0}^{[2]}(\kappa,\eta ) ) \exp(\imi \kappa x_{j } ) \exp ( \imi \eta y_{k } ) d\kappa d\eta \label{eq : lowwavenumbererror}\ ] ] and @xmath319^{2n_{0 } } } \exp\left ( - \frac{1}{4\lambda^{2 } \theta^{2 } } \frac{\iota(\vartheta_{1},\vartheta_{2 } ) } { \sin^{2}\tfrac{\vartheta_{1}}{2 } \sin^{2 } \tfrac{\vartheta_{2}}{2 } } \right ) d\kappa d\eta \\ &
= & \frac{h^{2n_{0}-2}c^{4n_{0}-1}}{4\pi^{2 } } \int_{-\pi}^{\pi}\int_{-\pi}^{\pi } \frac{\exp(\imi j \vartheta_{1})\exp(\imi k \vartheta_{2})}{\left [ 2 \lambda \iota(\vartheta_{1},\vartheta_{2 } ) \right]^{2n_{0 } } } \exp\left ( - \frac{1}{4\lambda^{2 } \theta^{2 } } \frac{\iota(\vartheta_{1},\vartheta_{2 } ) } { \sin^{2}\tfrac{\vartheta_{1}}{2 } \sin^{2 } \tfrac{\vartheta_{2}}{2 } } \right ) d\vartheta_{1 } d\vartheta_{2}.\end{aligned}\ ] ] first , consider the low - wavenumber error .
the inverse mixed discrete / continuous fourier transform of @xmath313 is given by @xmath320}(\kappa,\eta ) + n_{0}^{[2]}(\kappa,\eta ) ) \exp(\imi \kappa x_{j } ) \exp ( \imi \eta y_{k } ) d\kappa d\eta,\ ] ] which can be approximated by as @xmath113 tend to zero .
let @xmath321 the density function of a two - dimensional standard - normally distributed random variable with correlation @xmath322 .
its fourier transform is @xmath323 hence , for all positive integers @xmath324 the fourier transform of @xmath325 is @xmath326 such that the inverse fourier transform of @xmath327 is given by @xmath328 recalling the formulas for @xmath195}$ ] and @xmath329}$ ] from subsection [ region1 ] , this leads to the following expression for the low - wavenumber error : @xmath330 with @xmath331 \phi_{\rho}\left ( \frac{x_{j}+a_{1}}{\sqrt{2}},\frac{y_{k}+a_{2}}{\sqrt{2 } } \right).\end{aligned}\ ] ] next , consider the high - wavenumber error and note that @xmath332 symmetry yields @xmath333 where @xmath334^{2n_{0 } } } \exp\left ( - \frac{1}{4\lambda^{2 } \theta^{2 } } \frac{\iota(\vartheta_{1},\vartheta_{2 } ) } { \sin^{2}\tfrac{\vartheta_{1}}{2 } \sin^{2 } \tfrac{\vartheta_{2}}{2 } } \right ) d\vartheta_{1 } d\vartheta_{2 } \\ & & + \ \frac{c^{4n_{0}-1}}{2\pi^{2 } } \int_{-\pi}^{0 } \int_{0}^{\pi } \frac{\cos ( j\vartheta_{1 } + k \vartheta_{2})}{\left [ 2 \lambda \iota(\vartheta_{1},\vartheta_{2 } ) \right]^{2n_{0 } } } \exp\left ( - \frac{1}{4\lambda^{2 } \theta^{2 } } \frac{\iota(\vartheta_{1},\vartheta_{2 } ) } { \sin^{2}\tfrac{\vartheta_{1}}{2 } \sin^{2 } \tfrac{\vartheta_{2}}{2 } } \right ) d\vartheta_{1 } d\vartheta_{2}. \end{aligned}\ ] ] combining both expressions and gives an approximation for the _ total error _ : @xmath335 the values @xmath336 are only dependent on the position @xmath337 , the parameter values of the problem and the ratios @xmath338 and @xmath339 .
the constants @xmath340 only depend on the index @xmath341 , the correlation parameter @xmath322 and the ratios @xmath342 .
for the numerical experiments , cf .
infra , the values @xmath336 are calculated by determining all the partial derivatives . the integrals in @xmath340 are approximated by numerical integration .
it is readily seen that @xmath343 so @xmath344 has a maximum magnitude where @xmath345 .
_ this is exactly at the position of the discontinuity of the initial function .
_ at the end of subsection [ subsec : highwavetransf ] it was conjectured that for larger values of the mcs parameter @xmath38 one can expect a larger high - wavenumber error .
this conjecture is confirmed by the above analysis given that @xmath346 is always positive . in order to avoid spurious erratic behaviour in the numerical solution ,
_ it is therefore recommended to use smaller values of the parameter @xmath38 .
_ however , one has to take into account the lower bound on @xmath38 described in subsection [ subsec : stability ] .
we showed in that the total error is @xmath347 so that @xmath144 is a lower bound on @xmath83 for the rannacher time stepping in order to ensure convergence of the numerical solution to the exact solution .
this is confirmed by the plots in figure [ fig : convergence13 ] which display total errors ( in the maximum norm ) in actual numerical experiments for model problem ( [ eq : modelpde ] ) as a function of @xmath348 , with parameter values @xmath117 , mcs parameter @xmath39 and with @xmath349 , @xmath350 .
since it is not possible to handle infinite domains in numerical experiments , the computational domain is restricted to spatial gridpoints @xmath351 \times [ -10 , 10]$ ] . at the boundaries ,
homogeneous dirichlet boundary conditions are applied . in the left plots
the case @xmath143 is considered , whereas the right plots show the corresponding results for @xmath144 . in the upper plots
the maximum error between our numerical solution and the exact solution is shown as a function of @xmath348 for different values of @xmath339 . in the lower plots we show the same maximum error for one value of @xmath339 , together with our theoretical estimates for the corresponding low - wavenumber error and high - wavenumber error . in these lower plots
it is clearly seen that our theoretical estimates for the total error are sharp . for the case where no rannacher time stepping is applied , the left plots in figure [ fig : convergence13 ] reveal second - order convergence behaviour until @xmath139 reaches a critical value where the high - wavenumber error starts exceeding the low - wavenumber error .
it can be observed that this value of @xmath139 , and thus the high - wavenumber error , is highly dependent on the ratio @xmath352 . for smaller values of @xmath339 , @xmath344 is only sizeable whenever @xmath139 is very small , whereas for larger values of @xmath339 , @xmath344 already dominates the total error for larger values of @xmath139 .
moreover , the error constant for the low - wavenumber error is also dependent on @xmath339 .
however , this is much less pronounced than for the high - wavenumber error .
the right plots in figure [ fig : convergence13 ] show the corresponding results in the case where the first two mcs timesteps are replaced by four backward euler half - timesteps , thus @xmath144 .
one observes that the numerical approximations now exhibit second - order convergence for all values of @xmath339 . in the bottom right plot
the high - wavenumber error is not visible since it is strongly dominated by the low - wavenumber error .
the same observation is made for other values of @xmath339 .
hence , whenever rannacher time stepping is applied with @xmath144 , the total error can be approximated by @xmath353 , which is of second - order in @xmath139 .
we find that the error constant for the low - wavenumber error is mildly dependent on the ratio @xmath352 .
this can be explained through the fact that for a fixed value of @xmath139 but smaller value of @xmath339 the same semidiscrete system is solved with a smaller timestep @xmath354 .
finally , we notice that the latter error constant is slightly larger than for the case where @xmath143 .
thus , by applying rannacher time stepping with @xmath144 , second - order convergence can be recovered at the small cost of a marginally larger error constant for the low - wavenumber error .
as stated above , the high - wavenumber error is very sensitive to the mcs parameter @xmath38 . to illustrate this , figure [ fig : convergence1 ]
shows the same plots as in figure [ fig : convergence13 ] but with the mcs parameter replaced by @xmath40 .
it can be seen that all the conclusions from figure [ fig : convergence13 ] remain valid . in order to get decent plots , however , it is necessary to consider smaller values for @xmath339 .
this confirms that , for fixed @xmath339 , @xmath344 is strongly increasing as a function of @xmath38 .
moreover , by comparing the upper plots of figure [ fig : convergence13 ] and figure [ fig : convergence1 ] for @xmath355 it can be seen that the error constant of the low - wavenumber error is substantially larger for mcs parameter @xmath40 than for @xmath356 .
we conjecture that for fixed @xmath339 and fixed @xmath139 , the low - wavenumber error is also increasing as a function of @xmath38 .
therefore , regardless of the number of rannacher timesteps @xmath83 , it seems more favourable to consider smaller values of @xmath38 .
in particular , the lowest value of @xmath38 which satisfies the restrictions from subsection [ subsec : stability ] for all values @xmath357 is given by @xmath39 .
( left ) and @xmath144 ( right ) .
the parameter values are : @xmath358.,title="fig : " ] ( left ) and @xmath144 ( right ) .
the parameter values are : @xmath358.,title="fig : " ] ( left ) and @xmath144 ( right ) .
the parameter values are : @xmath358.,title="fig : " ] ( left ) and @xmath144 ( right ) .
the parameter values are : @xmath358.,title="fig : " ] ( left ) and @xmath144 ( right ) .
the parameter values are : @xmath359.,title="fig : " ] ( left ) and @xmath144 ( right ) .
the parameter values are : @xmath359.,title="fig : " ] ( left ) and @xmath144 ( right ) . the parameter values are : @xmath359.,title="fig : " ] ( left ) and @xmath144 ( right ) .
the parameter values are : @xmath359.,title="fig : " ] for @xmath37 , the mcs scheme reduces to the cs scheme and @xmath125 is not negligible in region 4 and region 5 .
based on equalities ( [ eq : cswavenumbererror ] ) and ( [ eq : cswavenumbererror2 ] ) we define @xmath360 and @xmath361 since @xmath310 @xmath311 and @xmath312 , @xmath362 only has to be considered in region 4 and @xmath363 is only not negligible in region 5 .
hence , the fourier error can be approximated by @xmath364 the inverse mixed discrete / continuous fourier transform of @xmath365 is given by @xmath366 as @xmath113 tend to zero this can be approximated by @xmath367 making use of a symmetry argument and a one - dimensional inverse fourier transformation , @xmath368 can be rewritten as @xmath369 with @xmath370 where @xmath371 denotes the density function of a standard normally distributed random variable .
it is readily seen that @xmath372 , respectively @xmath373 , reaches its highest magnitude near the points @xmath341 where @xmath374 , respectively @xmath375 .
for the numerical experiments , the integrals in @xmath372 and @xmath373 are approximated by numerical integration . combining the expressions , and leads to the following approximation of the total error : @xmath376 from approximation ( [ eq : totalerrorapproxcs ] )
it can be concluded that the total error is also @xmath377 when cs time stepping is considered .
this matches the observations from the plots in figure [ fig : convergencecs ] which show convergence results for the same problem as in subsection [ convergencemcs ] but with mcs parameter @xmath37 .
the lower plots indicate again that our theoretical estimates for the total error are sharp . without rannacher time stepping , i.e. @xmath143 , the results in figure [ fig : convergencecs ] show second - order convergence in @xmath139 until @xmath368 starts exceeding the low - wavenumber error .
then the total error increases in a first order way until the high - wavenumber error starts dominating . from there
the total error is @xmath378 . in case
the mcs scheme is replaced in the first two timesteps by four half - timesteps of the implicit euler scheme , i.e. @xmath144 , figure [ fig : convergencecs ] reveals unconditional second - order convergence in @xmath139 .
note that both @xmath368 and @xmath379 are not visible in the lower - right plot because they are strongly dominated by the low - wavenumber error .
the same observation as in subsection [ convergencemcs ] can be made concerning the dependency of the low- and high - wavenumber error on the parameter @xmath339 .
( left ) and @xmath144 ( right ) . the parameter values are : @xmath380.,title="fig : " ] ( left ) and @xmath144 ( right ) .
the parameter values are : @xmath380.,title="fig : " ] ( left ) and @xmath144 ( right ) .
the parameter values are : @xmath380.,title="fig : " ] ( left ) and @xmath144 ( right ) .
the parameter values are : @xmath380.,title="fig : " ]
if the initial data is nonsmooth , application of the mcs scheme for multidimensional time - dependent convection - diffusion equations with mixed - derivative terms can cause spurious erratic behaviour in the numerical solution .
a motivating example , with the two - dimensional black scholes equation for a two - asset cash - or - nothing option , shows that this undesirable feature can be resolved by replacing the very first @xmath83 mcs timesteps by @xmath84 half - timesteps of the implicit euler scheme , with @xmath144 .
we proved , by fourier analysis , that for a model two - dimensional convection - diffusion equation with mixed - derivative term and with dirac delta initial data , the total error can be approximated by the sum of a low - wavenumber error of @xmath381 and a high - wavenumber error of @xmath382 . in case
the mcs scheme reduces to the cs scheme , i.e. when @xmath37 , this has to be augmented with an extra error term of @xmath383 .
hence , @xmath144 is the minimum on @xmath83 in order to guarantee ( second - order ) convergence of the numerical solution to the exact solution , in the maximum norm . in general this choice for @xmath83
is optimal since larger values will increase the low - wavenumber error .
our convergence analysis and numerical experiments further indicate that it is favourable to consider small values of the mcs parameter @xmath38
. however , it is necessary to take into account the lower bounds on @xmath38 in order for our asymptotic analysis to be valid . the smallest value which satisfies all the restrictions , independent of the parameters of the model , is given by @xmath39 .
this is , indeed , also the most common value of @xmath38 for the mcs scheme considered in the literature .
the author is grateful to karel in t hout for his constructive and valuable remarks .
this work has been supported financially by a phd fellowship of the research foundation flanders . in t hout , k. j. , mishra , c. ( 2010 ) a stability result for the modified craig
sneyd scheme applied to 2d and 3d pure diffusion equations . _
anal . and appl .
, aip conf .
_ , * 1281 * , 20292032 .
in t hout , k. j. , welfert , b. d. ( 2009 ) unconditional stability of second - order adi schemes applied to multi - dimensional diffusion equations with mixed derivative terms .
_ , * 59 * , 677692 . | in this paper we consider the modified craig sneyd ( mcs ) scheme which forms a prominent time stepping method of the alternating direction implicit type for multidimensional time - dependent convection - diffusion equations with mixed spatial derivative terms .
when the initial function is nonsmooth , which is often the case for example in financial mathematics , application of the mcs scheme can lead to spurious erratic behaviour of the numerical approximations .
we prove that this undesirable feature can be resolved by replacing the very first mcs timesteps by several ( sub)steps of the implicit euler scheme .
this technique is often called rannacher time stepping .
we derive a useful convergence bound for the mcs scheme combined with rannacher time stepping when it is applied to a model two - dimensional convection - diffusion equation with mixed - derivative term and with dirac - delta initial data .
ample numerical experiments are provided that show the sharpness of our obtained error bound .
* key words : * convection - diffusion equations , adi splitting schemes , convergence analysis , rannacher time stepping . |
the number of wide field optical imagers with a large fov ( @xmath0 square degree ) available at astronomical telescopes is growing .
these instruments perform surveys which can cover thousands of square degrees of sky aimed at a wide range of galactic and extragalactic science goals .
the surveys are made public often promptly allowing astronomers from around the globe to use the data for their own specific scientific needs .
the imagers contain many ccds and often elaborate components to correct for optical distorsions over the fov .
accurate and robust photometric calibration over the complete area of such surveys is required to successfullly use the data for the many different ( and sometimes unforeseen ) science projects . in this article
we discuss the photometric calibration of a new wide - field imager : omegacam .
first we discus the main properties of the instrument and its telescope .
omegacam has been constructed by a consortium consisting of leiden observatory , kapteyn astronomical institute , the university observatories of munich , gttingen and bonn , the padova and capodimonte astronomical observatories and the european southern observatory ( eso ) .
it has an array of 32 thinned , low - noise ( 5e@xmath1 ) 2x4k e2v ccds which cover an area of 1x1 deg@xmath2 at 0.21 arcsec / pixel ( kuijken et al .
the key filters used will be the sloan @xmath3 , @xmath4 , @xmath5 , @xmath6 and @xmath7 filters ( see figure [ f : filters ] for measured throughputs ) .
besides this , the list of filters includes a filter segmented into a @xmath3 , @xmath4 , @xmath5 , and @xmath6 quadrant , various segmented narrow - band filters and johnson @xmath8 and @xmath9 filters .
the omegacam wide - field imager will be the sole instrument on the eso vlt survey telescope ( vst , capaccioli et al .
2005 ) and will be mounted at the cassegrain focus .
the vst is a 2.6 m telescope being built by eso and the capodimonte astronomical observatory .
its modified ritchey- crtien set - up is specifically designed for wide - field imaging , and has been optimized for excellent image quality in natural seeing .
three large eso public surveys ( see http://www.eso.org/observing/webone.html ) and many guaranteed - time surveys will be carried out covering thousands of square degrees in total .
the surveys address galactic science ( e.g. , ultra - compact binary systems , transiting planets ) and extragalactic science ( e.g. , gravitational weak lensing , galaxy evolution in superclusters ) . telescope and instrument are expected to start operations at paranal , chile in 2007 .
omegacam is expected to have a data rate which is equivalent to the rate for observing the southern hemisphere in about three years through one filter .
this will result in over hundred terabyte of raw plus processed data .
a prime concept of the omegacam calibration plan is to derive the photometric calibration for each of the 32 chips completely independently .
this will improve the photometric accuracy compared to a method which averages over ( a subset of ) the chips .
another important concept of the omegacam photometric calibration is to continuously maintain the photometric scale in the sloan bands , even when the science programme does not require the usage of these passbands during a particular night or period .
this continuity will ease maintaining an accurate photometric system .
we plan to use a scalable standard extinction curve which facilitates ( i ) recognizing deviating data , ( ii ) processing of incomplete data and ( iii ) performing quality checks .
omegacam will monitor the atmosphere , in partciular its extinction , via observations of a fixed near - polar field about three times per night .
the near - polar field is observable throughout the year .
nightly observations of standard star fields then allow accurate photometric calibration .
further details on the photometric calibration plan for omegacam can be found in the calibration plan manual available at http://www.astro.rug.nl/ omegacam / documents .
the result of this pipeline is that , during automated processing , the photometric calibration of the broad - band filters of omegacam achieves the required accuracy of 5% or better on the photometric scale in ` instrumental magnitudes ' as assigned to the units of a resultant output image of the image pipeline . in dedicated processing
the accuracy is expected to be 2% or better .
the accuracy of the colour transformation terms of instrumental to standard systems will be better than 10% .
another relevant number is the accuracy of flats .
laboratory tests have shown that the internal calibration unit for dome flats has a stability which is better than 1% over at least a period of 4 days .
the unit lamp is a type no .
14612 philips brilliant lamp produced in poland .
tests have shown that , with the help of this high stability , domeflats can be used even to determine accurately ( @xmath10 ) chip - to - chip variations in the gain .
a photometric calibration for each chip independently requires a spatially dense coverage of standard stars which covers the complete 1 square degree fov ( not the least to be efficient in terms of observing time ) .
catalogs of photometric standard stars that meet this requirement do not exist currently .
we will obtain eight of such 1 square degree fields of secondary standards , each centered on a landolt equatorial field , during the first year of omegacam operations ( see table [ t : standard_fields ] ) .
the result will be the omegacam secondary standards catalog .
the chosen approach is to observe each of the eight landolt fields centered on each individual omegacam ccd .
these landolt equatorial fields already contain standards measured by landolt ( 1992 ) and by stetson ( 2000 ) in the @xmath11 johnson - cousins system and measurements from data release 5 ( dr5 , adelman - mccarthy 2007 in the sloan @xmath12 system .
the observations will also be used to detect and assess effects that influence the photometric accuracy as a function of position on the chip , e.g. , illumination variations .
this poses a bootstrap problem : determining accurate zeropoints of the secondary standards requires that an illumination correction is already known . to address this bootstrap problem we need to obtain a quick and efficient initial determination of the illumination correction for omegacam based on a preliminary catalog of standard stars
we have constructed this preliminary catalog from observations of 1 square degree on the equatorial landolt fields using the wfc at the int on la palma .
the widefield - camera ( wfc ) on the 2.5 m isaac newton telescope at la palma is a mosaic camera consisting of four ccds covering an area of about @xmath13 with a pixel scale of @xmath14 .
the runs were performed in june 2002 and february 2003 .
all landolt fields listed in table [ t : standard_fields ] except sa95 were observed .
every landolt field was observed in 5 partially overlapping pointings covering @xmath15 on the sky . to cover an adequate range of magnitudes both a short ( @xmath16 ) and a long exposure ( @xmath17 )
were made for each pointing .
all the observations were done in the @xmath3 , @xmath4 , @xmath5 , @xmath6 and @xmath7 bands .
lccrrrr field & alpha & delta & # pc & # landolt & # stetson & # sdss dr5 + & ( deg ) & ( deg ) & & & & + + sa92 & 13.946 & @xmath180.949 & 6475 & 41 & 210 & 882 + sa95 & 58.500 & @xmath180.000 & 0 & 45 & 417 & 1154 + sa98 & 103.021 & @xmath190.328 & 23840 & 46 & 1116 & 0 + sa101 & 149.112 & @xmath190.386 & 5591 & 35 & 117 & 1842 + sa104 & 190.592 & @xmath190.553 & 5701 & 34 & 76 & 1964 + sa107 & 234.896 & @xmath190.252 & 12006 & 28 & 728 & 4143 + sa110 & 280.679 & @xmath180.348 & 38562 & 39 & 589 & 0 + sa113 & 325.533 & @xmath180.493 & 13947 & 42 & 477 & 4044 + the wfc data has been reduced using the astro - wise environment ( valentijn & kuijken 2001 , valentijn & verdoes kleijn 2006 ) .
the scientific exploitation of many of the omegacam surveys will be carried out using the astro - wise environment .
the astro - wise environment provides a novel way to deal intelligently with the large amount of data coming from optical wide field imagers in general and omegacam in particular .
it is a fully scalable science information system which unifies the archiving , the processing and advanced analysis tools .
the hardware and software that make up the environment are internationally federated facilitating collaborators in many places to share , validate and combine processed data and to pool hardware resources . all information about raw and reduced frames and the derived results from analysis on them
are stored in catalogues in the federated database which constitutes the heart of the system .
the astro - wise environment has been developed in the last five years by a partnership between omegacen - nova/ kapteyn institute ( coordinator ) , osservatorio astronomico di capodimonte , terapix at iap ( france ) , eso , universitts - sternwarte & max - planck institut fr extraterrestrische physik ( munich , germany ) .
the wfc data reduction consisted of the usual steps of de - biasing ( no overscan was used ) , flat - fielding and astrometric calibration .
the masterflat was constructed from domeflats and/or twilightflats .
defringing has been applied on the @xmath7 band data .
no illumination variation has been detected in accordance with other users of wfc ( m. irwin private communication ) .
one of the wfc chips ( a5506 - 4 ) turned out to have a measurable non - linear countrate in agreement with the non - linearity report by the casu int wide field survey ( mcmahon et al .
the non - linearity turned out to be too weak to affect the overall photometric calibration and hence no correction was made .
finally , the astrometry was determined per individual frame using the usno catalog .
the aperture flux measurements and its errors were obtained in an automated fashion using sextractor with a weight frame .
the stellar fluxes were measured through a circular aperture with a diameter of 30 pixels ( @xmath20 ) .
stars not suited for accurate photometry were removed from the list ( e.g. , saturated stars , stars with overlapping apertures , stars on the edge of chips ) .
after this selection many thousands of stars remained .
the number of stars in the preliminary catalog and the various literature sources are given in table [ t : standard_fields ] .
the goal is to put the wfc data for the preliminary catalog on the sloan photometric system .
ideally one would like to tie the standard in the preliminary catalog to the fundamental standards that define the sloan photometric system .
unfortunately , the sparseness of the sloan standards in the landolt equatorial fields and in the southern hemisphere did not allow us to do this within reasonable observing time .
moreover the brightness of these standards poses technical problems as well .
therefore we followed the strategy of observing landolt fields which contain both landolt standards in the johnson - cousins photometric system and stars in the sloan digital sky survey .
we photometrically calibrate the wfc data twice : once with the landolts stars as calibrators and once with the dr5 stars as calibrators . the difference in the results puts constraints on the systematic and random errors in the photometric zeropoints as will be discussed below .
it will turn our that we can not prefer clearly one calibration over the other at this point . in the end we decided , for mostly practical reasons , to use the calibration based on dr5 for the preliminary catalog . for the landolt calibrator
set we use the transformations determined by jester et al .
( 2005 ) to convert the @xmath21 magnitudes to the sloan @xmath12 .
this transformation is based on a comparison of magnitudes for standards that are used in the definition of both photometric systems . in this way , the photometric calibration of the wfc data with landolt stars is indirectly tied to the sloan standards . using the stars from sdss dr5 as calibrators offers many more stars ( see table [ t : standard_fields ] ) but far less accurate individual calibrator magnitudes .
however , the sloan photometry of dr5 is much more directly tied to the standards that define the sloan photometric system ( tucker et al .
2006 ) . to convert the wfc natural system to the sloan @xmath12 photometric system we solve the following equation for each filter : @xmath22 where @xmath23 is the sloan magnitude of the calibrator ( either from sdss dr5 or the landolt catalog ) in filter @xmath6
, @xmath24 is our instrumental magnitude , @xmath25 is the zeropoint , @xmath26 the atmospheric extinction coefficient , @xmath27 the airmass , @xmath28 the color coefficient and @xmath29 the color defined by filter @xmath6 and @xmath30 .
the following fixed values were used for the atmospheric extinction coefficients : @xmath31 $ ] .
these values are based on the values derived by the casu int wide field survey ( mcmahon et al .
they agree within the errors for the determination on our observing nights as done by the carlsberg meridian telescope on la palma ( only @xmath5 band measurements ) .
color terms for all bands have been determined for our data .
the analysis was based mainly on catalogs derived from the data of all the five pointings in the fields sa101 and sa107 ( 2003 run ) and sa 113 ( 2002 run ) .
the color terms were all found to be small or even non - existent .
this is not surprising , because the sloan filters at the int / wfc closely match the original ones .
the color terms for @xmath4 and @xmath6 are @xmath32 and @xmath33 , respectively , using ( @xmath34 ) and ( @xmath35 ) as the colors .
no evidence for any color term was found for the @xmath3 , @xmath5 and @xmath7 bands .
the zeropoint @xmath25 was determined by taking the average of the typically 10 zeropoints from individual observations of the fields ( i.e. , 5 pointings @xmath36 ( 1 short+1 long exposure ) ) .
we removed outliers among them .
finally , a star was admitted to the standard star catalog provided it had a @xmath5 magnitude and magnitudes for at least two more sloan bands .
the @xmath5 magnitude distribution of the resulting catalog is shown in figure [ f : aw2srmags ] .
random errors in the photometric calibration arise partly from random errors in the calibrator sets .
the scatter in the zeropoints of different sloan fields , which are 0.03 square degrees , for the photometric calibration of the dr5 itself is estimated to have a rms width of @xmath37mag , @xmath38mag in @xmath39 and @xmath40mag ( ivezi et al .
for the photometric calibration based on the landolt catalog a main source of random errors is expected to be the usage of transformation equations . the equations from jester et al .
( 2005 ) were derived from stars which have @xmath41 .
they report a rms scatter in the conversion of individual stars of 0.06mag in @xmath42 , 0.04mag in @xmath34 , 0.03mag in @xmath35 , 0.03mag in @xmath43 , 0.02mag in @xmath4 and 0.03mag in @xmath5 .
these transformation equations made also use of the very small transformation between the sloan @xmath44 photometric system ( on which the sloan fundamental standards were observed ) and the @xmath12 sloan photometric system ( on which the dr5 stars were observed ; see tucker et al .
the rms scatter in this transformation seems to be @xmath45mag typically .
it is difficult to determine how these errors for individual stars fields / small patches of sky propagate into zeropoint errors for the wfc data .
in fact , a more direct estimate of the overall random error in the zeropoint of the instrumental magnitudes per ccd ( @xmath46 square degrees ) can be obtained from the scatter in the individual zeropoints determined per chip plus filter combination over the different fields throughout the photometric part of the night .
the measured @xmath47 scatter in these is @xmath48mag , @xmath49mag , @xmath50mag , @xmath51mag and @xmath52mag .
we estimated the systematic error of our photometric calibration by comparing the sloan magnitudes for the @xmath53 stars in the eight landolt equatorial fields which appear both in sdss dr5 and in the landolt catalog .
table [ t : sloanvslandolt ] shows that the residuals of the magnitudes indicate systematic differences @xmath54mag except for the @xmath3 band . [ cols="^,>,^ " , ] another estimate of the systematic error in the photometric calibration is to compare directly the zeropoints as determined using the landolt catalog and dr5 catalog respectively .
this is effectively an independent estimate from the previous one as the @xmath53 sources which appear in both catalogs form a very small subset of both catalogs .
the resulting residuals in zeropoint , defined as the zeropoint using dr5 stars minus the one based on the landolt catalog have the following mean values : @xmath55mag in @xmath3 , @xmath56mag in @xmath4 , @xmath57mag in @xmath5 , @xmath58mag in @xmath6 and @xmath56mag in @xmath7 . comparing these results to those in ( table [ t : sloanvslandolt ]
we see that both methods yield very similar results .
given that the estimates of the systematic error with and without using the wfc observations are very similar it seems likely that the systematic error is due to inconsistencies between the dr5 photometric calibration and the calibration via the landolt catalog plus the transformation equations from jester et al .
only a handful of stars in the landolt catalog are present typically on a ccd .
however it seems very unlikely that the systematic error , present in eight landolt fields , has its origin in small number statistics given the small size of the random errors ( see section [ s : randomerrors ] ) relative to the systematic error estimates .
it is more likely that for example the usage of a single set of transformation equations might contribute to the systematic error . in the end , it turns out that we can not prefer clearly one calibration over the other .
we decide , for mostly practical reasons , to use the calibration based on dr5 for the preliminary catalog for the early verification of omegacam+vst . which calibrator set to use for the omegacam secondary standards catalog remains to be determined .
we inspected the residuals in the magnitudes obtained from the wfc data and the sdss dr5 catalog for overlapping stars in the five landolt fields which contain measurements of both sources .
small constant offsets ( @xmath59 ) between the magnitudes were noted in some filters in some fields which were not used for the determination of the zeropoints . from inspection of the stellar locus on color - color plots
we conclude that the sdss dr5 most likely has consistent photometric calibration over all landolt fields .
thus we ascribe the small constant offsets due to initially unnoticed atmospheric variations during our wfc observations and correct the magnitudes from the wfc data to bring them into accordance with the dr5 data .
no systematic trends as function of e.g. magnitude or position have been detected once these constants have been applied .
figure [ f : aw2svssdss ] shows the residuals as a function of magnitude . for @xmath3 the residuals between the wfc and dr5 measurements
are as expected from the measurement errors estimated for both magnitudes . for the other bands
the scatter in the residuals is typically 2 times that expected on the basis of our measurement errors .
given the simplicity of our aperture photometry and error computation for the wfc data we suspect that this is due to underestimation of the error in our data .
figure [ f : colorvscolor ] shows the color distributions for the dr5 stars and the preliminary catalog stars .
these figures illustrate the larger random error in the preliminary catalog stars . however , overall no systematic offsets are present .
the main goal of the preliminary catalog is to verify the performance of omegacam+vst and in this way bootstrap the construction of the catalog of secondary standard stars from omegacam observations in the first year of operations .
figures [ f : aw2svssdss ] and [ f : colorvscolor ] indicate that we will be able to use the wfc data set to statistically detect and quantify photometric effects as e.g. , a function of location on the detectors for omegacam+vst .
the analysis discussed here shows that the random error in the zeropoint of instrumental magnitudes using the preliminary catalog is @xmath54mag .
the systematic errors involved in putting the wfc data on the sloan photometric system are @xmath60mag .
the scatter in individual magnitudes from the wfc observations is too large to have the preliminary catalog stars qualify as secondary standards in the usual sense .
the observations of the landolt fields with omegacam should be able to achieve the level of accuracy required to establish the catalog of omegacam secondary standard stars .
adelman - mccarthy , j. , et al .
, 2007 , submitted to apj supplements capaccioli , m. , et al . , 2005 ,
eso messenger 120 , also see http://www.na.astro.it/oacweb/vst/ kuijken , k. , et al .
2004 , proc spie , 5492 , 484 , also see http://www.astro.rug.nl/megacam/ landolt , a. u. 1992 , aj , 104(1 ) , 340 mcmahon , r. g. , walton , n. a. , irwin , m. j. , lewis , j. r. , bunclark , p. s. , & jones , d. h. 2001 , new astronomy review , 45 , 97 , see also http://www.ast.cam.ac.uk/ wfcsur/ ivezi , .
, et al .
2004 , astronomische nachrichten , 325 , 583 stetson , p. b. 2000 , , 112 , 925 tucker , d. l. , et al .
2006 , astronomische nachrichten , 327 , 821 valentijn , e. a. , & kuijken , k. 2004 , toward an international virtual observatory , 19 , see also http://www.astro-wise.org
valentijn , e. a. , verdoes kleijn , g. a. 2006 , ercim news 65 , 20 , see http://www.ercim.org/publication/ercim_news/enw65/valentijn.html | the omegacam wide - field imager will start operations at the eso vlt survey telescope at paranal , chile , in 2007 .
the photometric calibration of omegacam data depends on standard star measurements that cover the complete 1 square degree fov of omegeacam .
a catalog fullfilling this requirement for eight landolt equatorial fields , denoted the omegacam secondary standards catalog , will be constructed from omegacam observations during the first year of operations . here
we present the preliminary catalog which will be used to bootstrap the construction of the omegacam secondary standards catalog .
thus the preliminary catalog will be used to assess the performance of omegacam+vst early - on .
the catalog is based on wfc data from the int at la palma . |
properties of different types of networks have begun to attract the interest of statistical physicists .
the spectrum of network related problems include , among many others , ordinary traffic in a city @xcite , distribution of nutrients in the vascular system @xcite , distribution of goods and wealth in economies , queuing problems on information networks @xcite , biochemical- and gene expression pathways @xcite .
it has been recognized that most natural and man - made networks evolve in time and that their structure emerged as result of microscopic evolution rules @xcite . in many networks a scale - free linking structure
was found @xcite which emerges from complex processes of self - organization with specific constraints .
maybe the most prominent example of such a scale - free network is the world - wide web @xcite and the internet @xcite .
transport processes on networks , such as traffic of information packets on the internet or diffusion of signaling molecules in a biological cell , are physical processes which are closely related to the network geometry .
it is likely that biological networks have been optimized in structure by selections in evolution .
the vast research on information traffic on technological networks is motivated to optimize network structure for practical reasons such as , e.g. , optimum connection strategies and information flow at minimal costs and risk .
technological networks like the internet or the world - wide web are relatively easily accessible for empirical research in contrast to e.g. biological networks such as metabolic pathways .
a further important motivation for studying networks is to gain theoretical insight in the role of an underlying geometry on the function of a network .
recent empirical studies on internet traffic suggest a complex interplay of queuing behavior @xcite , temporal correlation in packet streams @xcite , and broad distributions of travel times of packets @xcite .
moreover , evidence is gathered that both traffic parameters ( traffic intensity , buffer sizes , link capacities ) and network connectivity play a role in the overall network s performance .
often two general regimes of traffic are recognized in networks : free flow and jammed traffic .
a jamming transition is expected when traffic density exceeds a critical value . in numerical work
recently a model of packet transport was introduced @xcite which studies simultaneous random walks to specified destinations on scale - free networks .
the results suggest that fat - tailed transit - time distributions and temporal correlations may occur for a wide range of values of packet densities as a consequence of network structure @xcite .
other theoretical models were able to capture essential properties of the jamming transition on simple network geometries , like hierarchical trees @xcite and square lattice models @xcite .
in this work we implement a general numerical model to study packet traffic on several classes of networks with diverse structural characteristics .
these include a highly organized cyclic scale - free graph and two types of tree - like graphs . by working in a stationary traffic regime below the jamming transition ,
our aim is to relate a given network topology to characteristic diffusion parameters such as probability distributions of travel times and packet velocities .
further we want to quantitatively characterize the diffusion on given network topologies within a general framework of non - linear diffusion phenomena . quite generally ,
transport on networks can be seen as a diffusion on a geometry which is not plane- or space filling .
diffusion on such topologies has been a matter of interest over the past decades .
many problems like the diffusion of water particles in a sponge , or crude oil in porous media are phenomenologically well described by the anomalous diffusion equation .
anomalous diffusion is naturally linked to power laws in the corresponding spatio - temporal probability distributions .
it has been shown that the probability density which maximizes entropy of a version of a non - extensive entropy definition @xcite is an exact solution of the anomalous diffusion equation @xcite .
it is part of this work to notice the connections between transport on networks and anomalous diffusion and to use the latter to quantify topology - dependence of network performance .
queuing processes on networks are currently subject of intensive study .
mathematically , being composed of many random processes , network queues are a natural laboratory to study limit stochastic processes .
the existence of stable laws in these processes is expected @xcite .
however , the connection between the character of the limit process and network structure remains an open theoretical problem . having numerically determined large series of transit - times of packets we also attempt to recognize their convergence to a stable law and explore the consequences of it on specific network topologies . in section [ secii ]
we specify the model of network diffusion and describe details of network structures and the implementation of diffusion rules .
we determine conditions of the stationary ( un - jammed ) regime for all network topologies . in section [ seciii ]
the results for transit - times distributions are presented in the stationary regime .
we show results for the limit of infinitely low packet density on networks and discuss the statistical regularity of transit - timeseries .
section [ seciv ] motivates the use of anomalous diffusion methods in network transport and contains the results of the velocity distributions on different networks . in section [ secv ]
we study temporal correlations of workload timeseries and in section [ secvi ] we summarize and discuss the results .
a fundamental problem in non - linear diffusion is to understand how the topological structure or the sparseness of space affects the dynamics that it supports .
transport on a complex network with local navigation rules is a prototype for diffusion on a sparse structure .
further reasons for non - linear terms in diffusion equations can be related to the diffusion process itself , i.e. , to the specific way of local navigation between individual nodes and queuing properties . in this work
we introduce a numerical model in which we can control both , graph topology and microscopic diffusion details ( transport rules ) .
we consider graphs with link structure that emerges from controlled microscopic linking rules , from which the graphs are grown .
we grow three different types of networks ( described in detail below ) consisting of an equal number of nodes ( @xmath0 ) and links ( @xmath1 ) but having different linking patterns . after specifying the diffusion rules we monitor the _ simultaneous _ transport of packets on these graphs , and analyze local and statistical properties of the resulting traffic .
a packet ( particle ) carries information from the node where it is created to an in - advance specified node ( address ) on the network .
these can be , for instance , a piece of information transported on the internet or a protein in a biological cell . to determine how the structure of links in the network influences diffusion , we study three topologies with emergent link structures .
their main structural characteristics are the following : ( i ) scale - free directed graph with closed cycles ( cyclic web graph , wg ) ; ( ii ) scale - free graph with no closed cycles ( scale - free tree , sf ) ; and ( iii ) absence of a scale - free structure and absence of closed cycles ( randomly grown tree , rg ) . for all cases we stop growth of the graph when @xmath2 nodes and links have been added .
all the graphs that we consider have the same number of nodes and links , they differ in the adjacency matrix . after growing the network , the corresponding adjacency matrix is not changed during the diffusion processes .
we first consider the scale - free cyclic graph ( wg ) shown in fig .
[ figgraph]a .
such a graph is grown by adjusting a single free parameter governing microscopic linking rules proposed in @xcite .
these rules have been suggested as an evolution model of the world - wide web .
the minimal set of rules consists of : _ growth _ by adding of a new node at each time step and _ attachment _ ( with probability @xmath3 ) of the new node to a pre - existing node , or _ rewiring _
( with probability @xmath4 ) of a pre - existing node .
each link is oriented originating from a node , @xmath5 , as an out - going link , and pointing to another node , @xmath6 , where it is seen as in - coming link . in the model
@xcite both the target node @xmath6 and and the origin @xmath5 of a link are selected by shifted preferential probabilities given by @xmath7 which depend on the number of respective in - coming links @xmath8 , and out - going links @xmath9 at current time @xmath10 .
the shift parameter @xmath11 is the initial probability , where @xmath12 .
@xmath13 is the average number of links _ added _ per time step . as shown in @xcite
the biased attachment and rewiring in the growth rules lead to a scale - free organization of both in- and in out - links with the local connectivity at node @xmath14 at time @xmath10 building up in time with a power - law @xmath15 \ ; q_{out } ( s , t ) = a_2\left[\left(t / s\right)^{\gamma _ { out } } -b_2 \right ] \label{local_connectivity}\ ] ] where the scaling exponents are @xmath16 , and @xmath17 . the power - law growth of the local connectivity in eq .
( [ local_connectivity ] ) ensures an emergent scale - free structure of the graph after long evolution time . the global graph structure at @xmath18
is thus characterized by the power - law distributions of links @xcite , @xmath19 and @xmath20 with @xmath21 and @xmath22 .
here we use the original one - parameter model @xcite with @xmath23 and @xmath24 . apart from the scale - free organization of in- and out - links ,
the emergent graph structure is characterized by the occurrence of closed cycles of directed links and a single giant connected component .
two types of hubs ( nodes with a large in - link and nodes with a large out - link connectivity ) appear together with a substantial degree of correlations between local in- and out - connectivity @xcite .
recent analysis of empirical data on the structure of the internet @xcite reveal similar topological properties on the level of routers .
this type of graph topology has also importance in biological scale - free networks ( metabolic networks ) @xcite , where the presence of closed cycles enables feedback regulation . as a further example we mention gene co - regulation networks @xcite .
after a gene is expressed its product molecules ( mrna or proteins ) are in part addressed to specific genes in the network where they bind and regulate the expression of the other genes .
another biological network which shares some characteristics of the wg is the cell crawling network , where cells move along dynamic paths traced by previous cells @xcite . [ cols="^,^ " , ] in fig .
[ fig_spectra ] correlators @xmath25 and power - spectra @xmath26 are shown for stationary flow at @xmath27 for the three graph topologies and diffusion rules .
the corresponding scaling exponents have been obtained by least square fits to the data .
results are found in table [ tab_fits ] .
for rd both @xmath11 and @xmath28 suggest practically uncorrelated workload processes , regardless of the network type .
the situation changes for cs where signs for anti - persistence are found in the sf ( @xmath29 , @xmath30 ) and especially in the wg ( @xmath31 , @xmath32 )
. observed anti - persistence ( @xmath33 ) may be related to some kind of regulatory effects due to powerful hubs in the structured networks .
as discussed above , the advanced search often directs packets via hub nodes , and even for low global workload , there can emerge long queues at large hub nodes .
while waiting in a queue , packets increase the network s workload , but they do not contribute to randomness of the process .
moreover , when a waiting packet is on turn to be processed , it gets quickly to the target ( which is often found in the neighborhood of that hub ) and disappears from the traffic ; in other words , the system corrects itself by quickly decreasing the workload .
connected with this scenario is the fact that networks with more powerful hubs , such as the wg , are more efficient than networks with only one hub ( e.g. sf ) or no hubs at all ( e.g. rg ) .
in this work by use of large - scale numerical simulations and theoretical concepts we demonstrated how packet transport between nodes on different structured networks can be quantified as an anomalous diffusion process . in our model
we control both graph topology and microscopic diffusion rules .
our main aim was to explore how topology of graphs affects properties of the diffusion process .
for this purpose we ensured a stationary un - jammed flow on each graph topology by an externally fixed posting rate @xmath34 .
we based our work primarily on the analysis of transit - times of packets .
we find power - laws in the transit - time distributions which imply non - negligible probabilities for very long delays .
we find a marked difference in power - law exponents depending on graph topology and search algorithm .
these power - laws are rather independent of the posting rate , suggesting a statistical regularity of the diffusion process that can be related to graph topology when below the jamming transition . for the range of posting rates studied ,
we find that the particular queuing discipline is of minor importance .
the type of queuing is known to play a more important role for larger posting rates where waiting - times of packets increase when approaching the jamming transition .
our main conclusions which characterize the diffusion process on structured networks can be summarized in the following points : we have demonstrated that the traffic of packets on structured networks can be made stationary without imposing any global constraint .
we find the stationary jammed flow for a range of values for the posting rate which balances the output rate ( number of packets delivered per time step ) .
the output rate strictly depends on the network s topology and the search algorithm to navigate packets .
generally , for local search algorithms more structured networks , e.g. , networks with more powerful hubs and closed cycles have larger output rates .
theoretically , the stationarity of the traffic guarantees that we are dealing with self - similar stochastic processes , which have major advantages for theoretical analysis ( scaling ) .
we found that the traffic of information packets is super - diffusive on any non - trivial graph topology ( emergent topologies in evolving networks ) . in formal terms , this means that distances between nodes in a large graph can be traversed in times that scale sub - quadratically with the graph size ( the dynamic exponent @xmath35 is @xmath36 ) .
this conclusion remains true even if packets are left to diffuse randomly .
this is in marked contrast to random diffusion on space - filling structures , where random diffusion is represented by brownian motion .
another finding is that in the free - flow regime , differences between graph topologies are marginal when packets diffuse randomly .
this underlines the necessity for improving and adapting local diffusion rules in order to optimally use the topology of an underlying graph . to this end
we studied transport with the improved local search algorithm which looks for the destination node in next - near - neighbor surrounding of the node .
when using the improved local search , transport remains super - diffusive but marked quantitative differences due to graph topologies emerge .
the improved local search in combination with some network topologies performs strikingly more efficient than for others .
this is especially true for highly structured networks .
the efficiency of the network with improved local search is directly related to regulatory effects played by large hub nodes , in particular by the number of hubs and the number of closed cycles . although the distributions of transit - times of packets in all structured networks are markedly different ( generally shorter times in highly structured networks ) , they still exhibit power - tails
. the optimum strategy for diffusion on a network would be the knowledge of the shortest paths between the initial point and its destination .
this knowledge would however require costly global navigation , which is often not available in real networks .
traffic along geodesics may also be inconvenient in the case of queuing , given that in the scale - free graphs a majority of shortest paths pass through hub nodes , while other nodes carry much less traffic .
to understand this better we mention that the probability distribution of the number of paths through a node is a fast decaying function @xmath37 in sf and @xmath38 in wg @xcite . an advanced local search such as
cs is certainly better in this respect , since it disperses the activity over more nodes in the network .
this is achieved through the random part of the search , i.e. , when the destination node appears not to be in the locally searched nnn area .
this feature of the local search implies , however , that some packets may get stuck in remote areas of the network , leading to the long tail in the transit - time distribution . to improve search algorithms on structured graphs without extending their local character one can think of a statistical improvement which might be realized in many natural systems . to this end one
sends multiple copies of the same packet to the same destination .
since all packets move independently and randomly there would be differences between path lengths selected by the different copies .
we monitor the shortest transit - time and ignore later arrivals .
the first - arrival statistics of multiple copies of the packet is summarized in fig .
[ fig_minpath ] .
it shows that a reasonable improvement is achieved already by sending ten copies of the packet .
statistically , the fastest arriving packets use the minimum path up to four - link separation and longest transit - times are reduced by an order of magnitude compared to the one - copy case .
the asymptotic tail has the slope close to 1.9 .
similar effects reduced transit - times by first arrivals are found in the other two graph topologies , with the slope of the curve practically unchanged , compared to the one - packet distribution .
the efficiency of information transport on biological networks
e.g. , signaling molecules sent by a node to a particular destination where they can bind chemically probably utilizes the above statistical methods for improvement .
many identical molecules are being sent from an area of production ( e.g. the nucleus ) to a target position via a network ( cytosceleton).the first arriving molecule binds and chemically acts on the target node . in technological networks
the statistical improvement method may be used in low density traffic .
for instance , in the stationary flow regime on the web graph with cs at rate @xmath39 one can copy each packet by a factor of ten without yet causing jamming on the network . other topologies are more sensitive to jamming and the number of copies needs to be adjusted to the basic driving rate @xmath34 .
maybe our most interesting result is that in several measures ( transit - times , @xmath40 and correlation exponents ) we found that neither graph topology nor the search algorithm alone are sufficient to explain for a drastic increase of efficiency of network traffic .
we rather find that the combination of both in form of a good matching
are the source of this improvement .
it is most intriguing that for ordinary diffusion very little dependence of network structure was found , meaning for a unguided diffusion no advantage from more structure in a network is gained .
in other words either the structure of the network should be adjusted to the processes that the network supports , or the network has to be formed to use these processes most effectively . with this finding it becomes possible to speculate that evolution has adopted certain network structures in nature to cheap and robust search mechanisms , and/or has adapted search strategies to existing networks . | we study diffusion of information packets on several classes of structured networks .
packets diffuse from a randomly chosen node to a specified destination in the network . as local transport rules we consider random diffusion and an improved local search method .
numerical simulations are performed in the regime of stationary workloads away from the jamming transition .
we find that graph topology determines the properties of diffusion in a universal way , which is reflected by power - laws in the transit - time and velocity distributions of packets . with the use of multifractal scaling analysis and arguments of non - extensive statistics
we find that these power - laws are compatible with super - diffusive traffic for random diffusion and for improved local search .
we are able to quantify the role of network topology on overall transport efficiency .
further , we demonstrate the implications of improved transport rules and discuss the importance of matching ( global ) topology with ( local ) transport rules for the optimal function of networks . the presented model should be applicable to a wide range of phenomena ranging from internet traffic to protein transport along the cytoskeleton in biological cells .
pacs : 89.75.fb , 89.20.hh , 05.65.+b , 87.23.ge |
floquet topological insulators , a new kind of topological phase , have been attracting much attention in recent years @xcite .
these phases are produced by periodically time - dependent hamiltonians . by tuning the amplitude , phase , and frequency of the periodic drive myriad topological phases
may be realized . among floquet topological insulators ,
effective time - reversal breaking floquet chern insulators ( fcis ) have even been experimentally realized in cold - atoms @xcite and optical wave - guides @xcite .
these states contain unique @xmath0 edge states that have no equivalent in a static system @xcite .
however , characterizing a driven system as topological is non - trivial , as the usual adiabatic arguments @xcite for topological properties can not be applied .
naively , the topological nature of these systems could be deduced by analyzing the ground state of the floquet hamiltonian @xcite , which we call the `` floquet ground state '' ( fgs ) .
however the fgs has no general connection to the non - equilibrium state realized in experiment , and therefore it can not be used to determine the existence of topological properties . in this work ,
we study the topological properties of fcis using the entanglement spectrum ( es ) @xcite .
entanglement statistics have already been demonstrated to detect ground states @xcite , critical states @xcite , topological states @xcite , and universal exponents @xcite .
the es in particular is known to have edge states for topological states that mimic the edge states appearing in the physical boundaries of the same system @xcite . as
the es is a function of the state of the system at a given time , it may be used to detect topological properties for a time - evolving state .
we calculate the es for four different phases of the fci generated by a sudden turn on of the driving laser both numerically and analytically , and we complement these results with that for a slow turn on of the laser .
we label states obtained this way , i.e. , via unitary time evolution , as physical or true states , and we compare the es of these states with the es of the fgs .
we present three main findings .
first , the es of the fgs correctly detects both the usual and @xmath0-edge states of the system .
second , in the case of a resonant laser , the fgs is qualitatively different from the physical state obtained from unitary time - evolution .
third , the es in the physical resonant state does not show all the expected edge states .
therefore , the system driven by a resonant laser does not have the naively expected topological properties .
these results do not depend on how rapidly the laser was switched because , unlike for an off - resonant laser , for a resonant laser there is no adiabatic limit @xcite .
we also explicitly show this lack of adiabaticity of the resonant laser via the properties of the es . in ref .
it was shown that a chern number constructed out of unitarily evolved states is conserved .
thus if the initial state before the periodic drive was turned on had a zero chern number , it would stay zero always .
however , this chern number is a property of the full density matrix and is not related to physical observables directly .
rather , local physical observables probe a local spatial region and therefore a reduced density matrix for which the arguments of ref . do not hold .
evidence that the chern number of the full density matrix is not relevant also comes from the fact that that unitarily evolved states show a non - zero dc and ac hall conductivity even for an initial state with zero hall conductivity @xcite , although this conductivity is smaller in magnitude than @xmath1 , as would be expected for the fgs .
thus our result that the es of a unitarily evolved state reflects some of the topological properties of the fgs is consistent .
the paper is organized as follows . in section [ sec1 ]
the model is presented , and the topological properties of the fcis of interest to us are summarized . in section [ sec2 ]
the methods for obtaining the es are described , while the results are presented and discussed in section [ sec3 ] .
many details are relegated to the appendices .
other than figure [ fig : slowfastturnon ] , all other plots are for the sudden quench .
appendix [ app1 ] contains results for the es for the slow turn on of the laser .
appendix [ app2 ] gives the bulk occupation probabilities for the quench switch on protocol of the laser , as well as the projections of these occupation probabilities on the translationally invariant axis in order to highlight their relation to the bulk states of the es .
plots showing edge states , their chiralities , and decay lengths in the es for fgs and the unitarily evolved states are given in appendix [ app3 ] .
a key equation whose solution yields the es is derived in appendix [ app4 ] , while analytic solutions for the edge states in the es of the dirac model are given in appendix [ app5 ] , and they help to provide physical intuition for the more complex es structure of the unitarily evolved state in the presence of the laser .
we study the quench from the ground state of graphene at half - filling to a time - periodic hamiltonian corresponding to driving by a circularly polarized laser .
the transport and optical properties following such a quench has been extensively studied @xcite .
the hamiltonian before the quench is that of a half - filled infinite sheet of graphene , @xmath2 where @xmath3 is the n.n .
spacing , and @xmath4 . at @xmath5 , the hamiltonian
is changed by substituting @xmath6 , so that the @xmath7 is now periodically time - dependent .
a state @xmath8 of momentum @xmath9 evolves under this hamiltonian for @xmath10 according to @xmath11 where @xmath12 is the initial state and @xmath13 , @xmath14 are periodic in time and given by the floquet - bloch equation @xmath15|a_k(t)\rangle = 0 ; \label{eq : floquetbloch}\ ] ] and likewise for @xmath16 .
the @xmath17 are the two quasi - energies , which are only defined up to integer multiples of @xmath18 .
we take them to lie between @xmath19 calling this range of @xmath17 and @xmath9 the floquet brillouin zone ( fbz ) .
we take @xmath20 .
the information of the initial state is encoded in the overlaps with @xmath3 and @xmath21 and may be quantified by the excitation density @xmath22 where @xmath23 is the occupation probability of the lower floquet band and likewise for @xmath24 .
_ topology _ the wavefunctions at fixed @xmath25 and quasi - energies @xmath17 may be interpreted as the band - structure of some underlying hamiltonian , @xmath26 and we may define a ground state " of this hamiltonian by completely filling the lower band @xmath27 it is found that for some choices of parameters , the corresponding hamiltonian has a non - trivial chern number and is therefore topological . as in the static case , the chern number is related to the existence of edge bands @xcite . in the fci , there are two kinds of edge states : `` conventional '' edges that disperse through @xmath28 and `` anomalous '' @xmath0 edges that disperse through @xmath29 . the chern number is given by @xmath30 where @xmath31 is the difference between the number of left - moving and right - moving conventional ( anomalous ) edge modes .
however , the natural physical question is whether the many - body state generically produced by experiment actually displays any topological properties .
note that despite appearances , there is no reason for the fgs to be produced in a generic experiment . as @xmath32 is periodic , there is no notion of a lowest - energy state and hence no notion of relaxation to a ground state .
in fact , it is believed that generic disorder and interactions cause a driven closed system to reach infinite temperature @xcite .
therefore we must consider the physical state as determined by time evolution under a reasonable experimental protocol , here a quench , both sudden and slow .
having produced the physical state , we now must decide if it is topological . here
several standard arguments fail .
as the system is time - dependent , there is no good notion of adiabatic flux threading @xcite .
the application of a physical edge will qualitatively change the evolution of the system and therefore does not provide good information about the system in the absence of such an edge . as the application of generic disorder drives the system to infinite temperature , there is no notion of static localization .
these problems are solved by using the entanglement spectrum ( es ) @xcite .
the es is calculated by imposing a fictitious boundary on the density matrix at a particular time .
the degrees of freedom outside the boundary are traced out , and the spectrum of the reduced density matrix is the es .
the fictitious entanglement boundary functions similarly to a physical boundary , and it may host chiral edge states that can be used as evidence of topological properties . for a free system @xcite the entanglement spectrum may be derived from the eigenvalues of the correlation matrix @xmath33,\ ] ] where @xmath34 is the operator that annihilates an electron at site @xmath35 , and the lattice sites @xmath35 , @xmath36 are restricted to lie in the sub - system of interest .
the eigenvalues @xmath32 of this matrix lie between @xmath37 and @xmath38 with a value of @xmath37 or @xmath38 representing an unentangled or pure state mode and @xmath39 representing a maximally entangled mode .
.[tab : phases]summary of the four phases analyzed .
columns from left to right : name of the phase , @xmath40 the amplitude of the driving laser in units of the inverse lattice spacing , @xmath41 laser frequency in units of the hopping strength , whether or not the laser is resonant with the static spectrum , @xmath42 number of usual edge modes of the fgs , @xmath43 number of anomalous @xmath0 edge modes , @xmath44 the total chern number . [ cols="^,^,^,^,^,^,^",options="header " , ]
we consider four representative phases ( @xmath45 ) of the fci summarized in table [ tab : phases ] .
the first phase @xmath46 is the simplest case with an off - resonant laser @xmath47 and a single conventional edge mode .
the other three ( @xmath48 ) are for resonant laser frequencies of @xmath49 , where @xmath0 edge states may appear .
@xmath50 has one conventional edge mode and two counter - propagating @xmath0 modes so that the total chern number is 3 .
@xmath51 has a single conventional mode like @xmath46 but is produced by a high - amplitude resonant laser .
finally , @xmath52 is an unusual topological phase where there are two conventional and two @xmath0 modes so that the total chern number is zero .
the occupation of these edge states following the laser quench in a system with boundaries was recently discussed in ref . .
the occupation probabilities for the infinite system are discussed and plotted in appendix [ app2 ] .
c we compute the long - time behavior of the es and ee for the four phases @xmath53 .
the time is fixed to be @xmath54 periods for the sudden quench , whereas the results are at non - stroboscopic times for the slow quench ( see appendix [ app1 ] ) .
the floquet - bloch wave - vectors are computed by expanding eq . in a finite number of fourier modes and solving the linear system .
the result is used to construct @xmath55 .
we take the entanglement boundary to be the zigzag edge of a strip of width @xmath56 sites so that the momentum @xmath57 along the strip remains a good quantum number .
thus , @xmath55 may be diagonalized at fixed @xmath57 , and the results are plotted in figure [ fig4 ] .
the results are compared with the es of the fgs a single statistic that may be calculated from the es is the entanglement entropy , given by @xmath58,\ ] ] the sum being taken over all eigenvalues . the results for the entropy are shown in figure [ fig3 ] for the fgs and the physical state
. the fgs state shows area - law scaling , @xmath59 .
this is because the floquet state is the ground state of @xmath60 , which is local and gapped , and such states are known to show area law@xcite .
however the physical state shows volume - law scaling ( @xmath61 ) as is expected of a generic state with ballistically propagating excitations .
this behavior , therefore , does not provide detailed behavior about the phases .
the full es for both states , namely the fgs and the physical state , for phases @xmath53 is shown in figure [ fig4 ] for the sudden quench and figure [ fig : slowfastturnon ] in appendix [ app1 ] for the slow quench .
the figure is symmetric around @xmath62 as a consequence of particle - hole symmetry .
let us first discuss the es for the fgs .
the es shows a bulk of bands clustered near @xmath37 and @xmath38 , with a gap between the two bands .
additionally , there are a small number of bands that disperse through @xmath62 that appear suggestively like chiral edge states . on comparing with table [ tab : phases ] , one finds that the number of edge states that cross the gap in the es of the fgs is the same as the chern number .
this means that when anomalous edge states appear , as in phases @xmath63 , the chiralities of the edge states in the es reverses in such a way that the chern number as calculated from the es agrees with the chern number of the phase .
thus for @xmath50 , for example , there are 3 chiral right - movers in the es even though the anomalous modes are left moving at the physical edge
. a similar reversal of chiralities is observed in @xmath52 .
therefore , the es `` understands '' that the @xmath0 modes contribute to the chern number with opposite sign .
while the es is a snapshot at a particular time , at other times , the location of the edge states changes , but their number and chirality are maintained so as to preserve the chern number .
thus the phase @xmath52 can show odd behavior due to the fact that it corresponds to @xmath64 .
this is highlighted in figure [ fig : p4ent ] , where within a laser period , the edge - modes can appear with canceling chiralities ( @xmath5 in figure ) , and totally disappear ( @xmath65 in figure ) , both of these cases being consistent with @xmath64 .
next we turn to the es of the physical state . in the off - resonant case @xmath46
, the floquet and quench es generally agree . in the phases
@xmath48 , the two appear drastically different .
while the central crossing at @xmath66 and the gap remain intact , a continuum of eigenvalues that pass through @xmath39 appears in the quench case .
these continua of states cover the region where the two edge states of @xmath50 located at @xmath67 appear in the fgs spectrum and completely cover the gap in @xmath52 .
this picture qualitatively holds for a slow quench .
as shown in figure [ fig : slowfastturnon ] , the edge modes that are absent for the sudden quench continue to be absent for a slower quench . in the next section ,
we explain these results .
this behavior of the es may be explained as follows . in the long - time limit
, the contribution to the correlation function @xmath55 of the overlap between the upper and lower bands oscillates as @xmath68 with @xmath69 . as @xmath70
, @xmath71 changes rapidly with @xmath9 when @xmath72 and the overlap terms may be neglected .
straightforward manipulation ( see appendix [ app4 ] ) then gives @xmath73 where , @xmath74 if the size @xmath56 of the sub - region @xmath75 is taken to infinity , @xmath76 may be diagonalized by fourier transformation , and its eigenvalues may be read off from eq .
as @xmath77 with eigenvectors @xmath78 and @xmath16 .
therefore we may think of @xmath79 as a hamiltonian with two bands , where the gap between the two bands is @xmath80 , and the wavefunctions have some topological properties .
note that the smaller the quench - induced excitation density is , the larger is the gap @xmath80 in the es .
the problem of finding the spectrum of @xmath79 at finite but large @xmath56 is then akin to the problem of finding the spectrum of a hamiltonian in a finite geometry .
we expect two kinds of states : bulk states and edge states .
the bulk states should have the `` energies '' , @xmath81 , but where @xmath9 is quantized in the finite direction .
therefore , the `` bulk '' states can be produced by projecting the occupation number onto the @xmath82 plane ( the translationally invariant direction ) ; see figure [ fig : proj ] in appendix [ app2 ] .
in addition to these bulk states , there may be edge states .
if the floquet bands are topological , that is , if the chern number is non - zero , _ and _ @xmath81 is non - zero , then the topological protection of edge modes of @xmath79 applies and there will necessarily be edge modes in the es .
this property is topological in the sense that a small deformation of the state , which is equivalent to a deformation of @xmath81 and @xmath83 , can not change the net number of edge modes . for the fgs , where @xmath84 by definition , there will exist edge modes that agree with the chern number of the bands .
this also appears to be the case in @xmath46 , where @xmath85 only at isolated points , and the edge modes of the fgs and physical states agree , although perhaps the issue is more delicate when the entanglement cut is very rough so that momentum @xmath57 is not conserved .
that these states are indeed topological can be seen in the numerics .
their number does not scale with @xmath56 and their crossings are not lifted by varying the phase of the driving laser .
moreover , a direct analysis of the wavefunctions associated with the eigenvalues shows that they are chiral and located on the edge .
figures [ fig : edgef12 ] and [ fig : edgef34 ] plot the wavefunctions for the edge states for the fgs , while figures [ fig : edgeq12 ] and [ fig : edgeq34 ] plot the same for the true state , and figure [ fig : edgedecay ] highlights the decay length of the edge states . if @xmath86 , then this is equivalent to a gap closing in the hamiltonian , and topological protection arguments no longer hold .
indeed , we find that the edge states do not coexist with the bulk states in the numerics . since a resonant process will co - occur with a population inversion ( see figure [ fig13 ] ) , we do not expect edge states to be protected in the resonant case .
further , since @xmath85 precisely where the resonant laser causes band crossings , the resulting edge modes are not robust at all .
indeed , this is what is seen in the phase @xmath50 where the two edge modes seen in the floquet bands at the boundary of the fbz do not appear in the es of the true state .
this is not an artifact of the quench protocol , and similar population inversions appear in the slow turn on of the resonant laser ( see appendix [ app1 ] ) .
although the edge modes lose topological protection when @xmath86 , there may still be non - protected edge states .
since we study a perfectly clean system with translational invariance in the @xmath87 direction , an edge mode will appear under the much weaker condition that @xmath88 for the same @xmath57 at which the edge mode appears .
therefore , we find that the central mode appears in @xmath50 and @xmath51 even though @xmath81 is zero over a large arc in the bz .
we expect that this edge mode would disappear for other geometries . for @xmath52 ,
the high - amplitude laser produces population inversions throughout the bz , and there are no edge states whatsoever in the physical state . for phase @xmath46 ,
on the other hand , one has a robust gap , with gap closings only at some special points ( rather than large arcs ) .
these excitations can be made even smaller for a slow laser switch on , as for the off - resonant case , an adiabatic theorem can hold .
this provides a complete description of the entanglement spectrum at long times after a floquet quench .
the key quantity that determines the applicability of the chern number calculation to the edge modes of the quench is the population difference @xmath81 .
it is interesting to ask how the results of the es compare with physical observables . in ref .
, the occupation of the edge states in a system with boundaries was studied , and it was found that the edge state for the off - resonant case @xmath46 was occupied at a low effective temperature .
in contrast , only one set of edge modes in the phase @xmath50 was occupied at a low effective temperature , while the two new sets of @xmath0-edge modes appearing due to resonant band crossings were occupied at a high effective temperature and coexisted with bulk excitations . the dc conductivity for these phases
was studied in ref . , where it was found that the conductance was very close to the maximum , @xmath89 , for @xmath46 , and it persisted at this value for @xmath50 even though new edge modes appear for @xmath50 .
this observation implies that the high effective temperature of the resonant edge modes prevent them from contributing much to transport , in contrast to the off - resonant edge mode .
thus our results show that the study of the es is a useful way to understand the topological properties of a periodically driven system .
an important direction of research would be to extend this analysis to interacting floquet topological insulators .
_ acknowledgements : _ this work was supported by the us department of energy , office of science , basic energy sciences , under award no .
de - sc0010821 .
to generate occupation probabilities for the slowly turned on laser , the schrdinger equation at fixed @xmath57 is solved by implementing the commutator free exponential time method following ref . .
the occupation probabilities and entanglement spectrum are calculated from these wavefunctions . due to the computational cost of this procedure
, the system width was taken to be @xmath90 sites .
the results for the es shown in figure [ fig : slowfastturnon ] , are in qualitative agreement with the quench protocol discussed in the body of the paper .
band inversions appear near momenta where the laser is resonant with the band gap .
these band inversions are accompanied by a high bulk excitation density , leading to a closing of the entanglement gap , and ruining the topological protection of edge states .
figure [ fig : slowfastturnon ] shows data for two different laser switch - on rates for the two resonant phases @xmath91 .
@xmath50 has three pairs of chiral edge states in the ground state ( fgs ) of the floquet hamiltonian .
however , we find that for all switch - on rates of the laser , two of the three pairs of edge states ( located at @xmath92 ) are absent from the es of the unitarily evolved states . only the central edge state located at @xmath66 survives . for phase @xmath51
, there is only one pair of chiral edge states in the floquet ground state , and this is visible in the es of all the unitarily evolved states as well .
however this edge state coexists with bulk excitations that survive both for a fast and slow laser quench .
paradoxically , a slower laser switch on creates more bulk excitations when the laser frequency is resonant .
we understand this as follows .
a large - amplitude laser can modify the band structure considerably , making the bands flatter .
thus a resonant laser can become effectively off - resonant when the amplitude of the laser becomes too large .
thus when a laser amplitude is switched on slowly , ( while its frequency is kept unchanged ) , since for a longer period the effective amplitude of the laser is smaller for a slower switch on than a faster switch on , more resonant excitations are created for the slower switch on as the resonance condition is obeyed for a longer duration of time . that is the reason why in figure [ fig : slowfastturnon ] , there are more bulk excitations for the slower switch on than for the faster switch on .
the only reason why the central edge state at @xmath66 is unaffected by these bulk excitations is because our entanglement cut is smooth and momentum @xmath57 is conserved .
this prevents hybridization between the central edge state and the bulk excitations , since the latter are located at other values of @xmath57 .
however a rough entanglement cut will wipe out even this central edge state .
the occupation probabilities of the floquet bands for phases @xmath94 are shown in figure [ fig : drhop12 ] , and those for the phases @xmath95 are shown in figure [ fig : drhop34 ] . except for @xmath46 ,
all the phases correspond to a resonant laser . for @xmath46 ,
the quasi - energy of the floquet hamiltonian is similar in appearance to static graphene with the important modification of becoming gapped at k and k. thus , the occupation probabilities follow intuitively from the static picture , and the sharp spikes in figure [ fig : drhop12 ] correspond to excitations around the k and k points .
the laser being off - resonant is equivalent to the statement that the periodicity of the quasi - energy does not play a role since corresponding boundaries in the fbz are far from the maximum and minimum of the bands .
both @xmath91 have lasers of the same frequency , but the amplitude of the laser for @xmath51 is larger than that for @xmath50 . in @xmath50 there is a population inversion at @xmath96 in addition to the excitations at k and k. this is simply the resonance condition , as now the frequency of the laser is such that it connects the peak and trough of the static graphene bands . in the quasi - energy picture
, this translates to the original static bands extending beyond the fbz and thus forming quasi - energy bands that avoid one another and bend away at the boundary of the fbz .
a large amplitude laser modifies the effective band structure to such a degree that a resonant laser becomes effectively off - resonant .
the bands are drawn toward the center of the fbz , away from the boundaries .
it is for this reason that @xmath51 has fewer bulk excitations than @xmath50 , despite the two phases sharing the same laser frequency .
the flattening of the bands explains the broadening of the excitations at the k and k points . for phase @xmath51
, even though the laser is resonant , the excitation density around @xmath66 , where the central edge mode in the es appears , is still fairly low .
the laser creates pockets of excitations in regions symmetrically located around the central edge mode .
momentum conservation prevents the central edge mode from mixing with these pockets of bulk excitations .
the amplitude of the laser for @xmath97 is much larger than all the other phases , and its frequency is much smaller as well . as a consequence ,
the @xmath52 phase is highly excited , with the two floquet bands being almost equally occupied .
figure [ fig : proj ] is the projection of the occupation probabilities onto the @xmath93 axis .
these projections are generated by selecting constant @xmath98 slices of the occupation probabilities that correspond to the modes that satisfy the boundary conditions of the strip .
these slices are then superimposed on top of one another to give the projection image . as explained in the main text
, these projected plots reproduce the bulk states of the es .
there are some small discrepancies between the projection plot and the true spectrum .
as seen in figure [ fig4 ] , the bulk excitations on either side of the central edge state in the quenched @xmath50 phase are smooth , but the analogous bands constructed from figure [ fig : proj ] appear ragged and cross @xmath99 , where they do not in the former .
we explain this difference below . as the width
is increased , we first expect the bulk excitation bands of the es to take on the rough appearance as predicted in the projection plot .
the larger width will also lead to more bands coalescing towards @xmath39 .
thus in the very large limit , we expect a continua of excitations on both sides of the central edge state of the quenched @xmath50 phase .
the reason for the differing appearances in the large , but not extreme width limit is due to the bulk bands having residual knowledge of the edges and thus undergoing a smoothening and repulsion procedure , as one expects for perturbations .
thus as the width is increased , this smoothening will diminish and bands from both figures [ fig4 ] and [ fig : proj ] will agree .
the reason the repulsion argument does not extend to the edge states is precisely due to their local nature and exponentially small overlap . in the @xmath51 phase ,
@xmath98 projections originate from occupation probabilities that have a relatively smooth structure throughout the fbz , as seen in figure [ fig : drhop34 ] .
thus we do not run into this problem for the @xmath51 phase because the projection already creates a continuum that agrees with the spectrum in figure [ fig4 ] .
however , for the @xmath50 phase , the probability occupations have sharp , nearly vertical transitions that lead to discrete bands that single out and amplify this otherwise small feature .
the continuum of excitations will only be created by the projection plot for @xmath50 if the strip width is so large that the spacing between @xmath98 states is narrow enough to allow for many states to be selected along the cliff faces ( see figure [ fig : drhop12 ] ) found in the occupation probability of the @xmath50 phase .
the entanglement spectrum for phase @xmath52 is shown in figure [ fig : p4ent ] .
the floquet state shows the behavior expected of a @xmath64 floquet chern insulator .
the `` bulk '' bands are clustered around @xmath28 and @xmath100 .
there are edge states seen around @xmath39 , however these do not have a net chirality .
this can be seen either from direct inspection of the wavefunctions in figure [ fig : edgef34 ] , or by noting that the edge states do not connect with the bulk bands .
therefore , these edge states are not topologically protected , and they should disappear under disorder .
in fact figure [ fig : p4ent ] shows that the edge states appear at certain times in the laser cycle , and they disappear for certain other times .
this does not happen for phases @xmath46 , @xmath50 , and @xmath51 ; when probed at times away from an integer number of laser periods , the precise location of the edge states is shifted in the es , but the results for the number and chirality of the edge states that are visible remain the same .
the true quench state for phase @xmath52 does not show any edge states , and the bulk bands approximately fill the space ( see figure in main text ) .
this is in agreement with the occupation number difference shown in figure [ fig : drhop34](b ) , which is found to vary widely between @xmath101 and @xmath38 and crosses @xmath37 on large arcs through the bz . from the perspective of the entanglement and occupation number , therefore , the true quench state for @xmath52 appears similar to a thermal state .
the amplitudes of the edge states are shown in figures [ fig : edgef12 ] , [ fig : edgef34 ] for the floquet ground state , and in figures [ fig : edgeq12 ] , [ fig : edgeq34 ] for the true state . to be clear , by state " we mean the eigenvector of the correlation matrix @xmath102 .
as discussed in the main text , we expect these edge states , at least for the floquet ground state , to behave as the usual edge states of a topological hamiltonian .
in fact , the states are highly localized to the edge with a decay length on the order of one lattice site .
moreover , we can directly verify the chirality of the edge states by analyzing how the wavefunctions vary with momentum @xmath57 .
in figures [ fig : edgef12 ] , [ fig : edgef34 ] , [ fig : edgeq12 ] , [ fig : edgeq34 ] the two states with @xmath32 closest to @xmath62 are shown .
the higher @xmath103 state - analogous to the higher energy state - is colored dark while the lower state ( @xmath104 ) is colored light . as can be seen at @xmath57 near the crossing point ,
the edge states are well defined and localized on opposite edges . moreover , at the crossing , the high @xmath32 state switches with its counterpart .
this is expected in a chiral edge because the two bands cross as @xmath57 is varied .
therefore , by analyzing the pattern of switching , the chirality of the bands stated in the main text can be verified . as an example of this analysis we consider the @xmath105 edge in figure [ fig : edgef12 ] for the fgs of phase @xmath50 .
we see a pattern of black to gray for all three crossings , which indicates that all the downward arching bands in figure [ fig4 ] correspond to chiral edges that reside on the @xmath106 edge of the strip .
the edge states corresponding to the downward trending bands at each crossing will be termed left movers " and the upward arching bands will be termed
right - movers " .
thus , by the same analysis , the @xmath107 edge is found to contain all right movers at each crossing .
the same can be done for the fgs of the @xmath52 phase to verify the assignment of @xmath108 .
both sides contain two crossings of both types , black to gray and gray to black .
thus each side contains an equal number of left movers and right movers .
note that these edge states vanish altogether at other times during the laser period , as pointed out in figure [ fig : p4ent ] .
in contrast , for all other phases with @xmath109 , the edge - states slightly shift their @xmath57 positioning in a periodic manner .
figures [ fig : edgeq12 ] , [ fig : edgeq34 ] contain states corresponding to @xmath32 near @xmath62 for the unitarily evolved system .
these figures show that unlike the fgs , not all the states at this eigenvalue are edge states .
while phase @xmath46 is similar to the fgs , for phases @xmath91 only one pair of edge modes exist , with a large fraction of states around @xmath110 being bulk states . the decay of the edge states is shown in figure [ fig : edgedecay ] .
the decay lengths of the edge states are determined by the inverse of the entanglement gap , which is controlled by the bulk excitation density .
thus @xmath46 which has few bulk excitations in the quenched state has a similar decay length to the fgs .
in contrast , since the quenched state of @xmath51 has more bulk excitations , it shows a longer decay length than the edge state of the fgs .
the surviving edge state of @xmath50 in the quenched state has a similar decay length as the fgs because , as seen in fig .
[ fig4 ] , in the vicinity of this edge state the entanglement gap of the quenched state is almost the same as that of the fgs .
c + + c + c + + c c + c + c + + c
generally , the correlation matrix is given by @xmath111,\end{aligned}\ ] ] where @xmath112 and likewise for @xmath113 , and @xmath114 is the initial state .
we make the assumption that the third and fourth off - diagonal terms proportional to the exponential vanish at long times because of the rapidly varying phase .
we emphasize that this is an acceptable assumption because we are considering @xmath35 to be restricted to a sub region of the lattice . for general @xmath35
the stationary phase approximation implies that the exponentials will contribute in a region where @xmath115 where @xmath116 is the group velocity of the floquet dispersion .
therefore , as long as we consider times that are much larger than @xmath117 , where @xmath118 is some characteristic velocity , we may neglect the oscillating terms .
we have also checked the validity of our argument from the numerical results .
we should mention that the order of limits here does not commute .
the limit @xmath119 must be taken before width @xmath120 .
physically this corresponds to only analyzing the interior of the light - cone .
if we neglect these off - diagonal terms , we obtain @xmath121.\end{aligned}\ ] ] now employing @xmath122 from the completeness of the basis , and the fact that @xmath123 , we obtain the stated equation where we identify @xmath124 .
before we discuss the es , let us review how edge states appear in the physical boundaries of the dirac hamiltonian .
the dirac hamiltonian in a semi - infinite geometry with a boundary at @xmath106 is , @xmath125 + m(x)\sigma_z.\end{aligned}\ ] ] the boundary is controlled by the behavior of @xmath126 .
we may write the eigenstate as @xmath127 , which should obey @xmath128 + m(x)\sigma_z\biggr)\psi_k(x , y ) \nonumber\\ & & = e_k\psi_k(x , y),\end{aligned}\ ] ] and implies the eigenvalue equation , @xmath129\chi + v_f k_y \sigma_y \chi = e_k \chi.\end{aligned}\ ] ] the solution is @xmath130 , @xmath131 . thus the eigenstate is , @xmath132 the above is an edge state with a decay length controlled by @xmath133 .
now let us consider how edge states emerge in the entanglement spectrum .
for this , we will consider a spatially uniform system ( @xmath134 ) with eigenstates , @xmath135 it is convenient to define @xmath136 where the correlation function in real space is @xmath137 for any generic state , be it one generated at long times after a quench when dephasing has set in , or a mixed state at finite temperature , we may write , @xmath138 we will assume the system to be at half - filling .
note that since , @xmath139 , a filled band will give , @xmath140 the above corresponds to an unentangled product state with eigenvalues of the correlation matrix being @xmath38 .
it is useful to note that @xmath141 is a matrix with eigenvalues @xmath142 .
we write @xmath143 , where @xmath144 are pure states .
since we are interested in the es of a semi - infinite geometry ( @xmath145 ) , we may study the @xmath57 fourier component of @xmath44 in the translationally invariant @xmath87 direction , @xmath146 note that the above correlation matrix is @xmath147 added to the fourier transform of @xmath148 times a matrix with eigenvalues @xmath142 .
therefore , in the infinite system limit it has eigenvalues @xmath149 corresponding to plane - waves eigenvectors . in the presence of a boundary
we expect some linear combinations of these plane waves to form extended scattering states .
the boundary is now determined by the entanglement cut at @xmath106 , rather than @xmath126 . if @xmath152 and smoothly varying around @xmath93 , we may approximate it by it s value at @xmath82 .
this may then be pulled out of the integral . then using the fact that @xmath153 and
since @xmath154 decays exponentially away from @xmath106 , this term can be replaced by a delta - function , @xmath155 .
thus , @xmath156,\end{aligned}\ ] ] up to terms which are higher order in derivatives . note that if @xmath157 the second term vanishes and @xmath44 is controlled by the higher order terms which will not generically have edge states .
this is solved by the same ansatz as for the usual physical edge mode of the dirac hamiltonian , @xmath160 where , @xmath161 and the step function to explicitly impose the boundary conditions .
the entanglement eigenvalues are , @xmath162 in particular this shows that the edge state with @xmath39 exists at @xmath163 , and it disperses linearly with @xmath57 about this point .
now we return to the assumption we made that @xmath164 . if @xmath165 were to hold at some point in @xmath98 , then there would be an extended state with @xmath166 . as this extended state , and edge mode ,
both have equal @xmath57 momentum , and the same eigenvalue under @xmath44 , we expect them to mix . therefore , in this case expanding around @xmath82 is not valid , the argument above does not hold , and we do not expect edge states .
while this analytic argument was shown for the dirac model , and the floquet chern insulator is far more complicated , especially for large chern numbers , yet we find that the same principles apply .
for example figure [ fig13 ] shows how @xmath167 varies along @xmath98 for the @xmath50 phase .
we have chosen @xmath168 which coincides with the edge mode in the es of the floquet eigenstate that is located to the left of the central edge mode ( see figure [ fig4 ] ( b ) in main text ) .
this particular edge mode is absent in the true wavefunction .
the reason now is clear .
figure [ fig13 ] shows that @xmath169 vanishes at two points in @xmath98 , implying extended states with the same energy as the edge state .
these extended and edge states can now mix , resulting in the absence of the edge state at @xmath170 .
it is interesting to note that the resonant laser results in a rather precise location of the vanishing of @xmath81 relative to the avoided band crossings of the upper and lower floquet bands .
33ifxundefined [ 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.79.081406 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.105.017401 [ * * , ( ) ] link:\doibase 10.1103/physrevb.84.235108 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevlett.113.236803 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.113.266801 [ * * , ( ) ] link:\doibase 10.1103/physrevb.91.155422 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrevx.3.031005 [ * * , ( ) ] link:\doibase 10.1103/physrevb.23.5632 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.101.010504 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/j.aop.2010.09.012 [ * * , ( ) ] , link:\doibase 10.1103/revmodphys.82.277 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.114.170505 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/j.nuclphysb.2006.12.012
[ * * , ( ) ] link:\doibase 10.1103/physrevb.80.115122 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.96.110405 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.96.110404 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.110.170605 [ * * , ( ) ] link:\doibase 10.1103/physrevb.94.024306 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.104.130502 [ * * , ( ) ] link:\doibase 10.1103/physrevb.93.241406 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevb.92.165111 [ * * , ( ) ] link:\doibase 10.1103/physrevb.90.195429 [ * * , ( ) ] link:\doibase 10.1103/physrevb.93.205437 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.115.030402 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.114.140401 [ * * , ( ) ] link:\doibase 10.1103/physrevx.4.041048 [ * * , ( ) ] http://stacks.iop.org/1751-8121/42/i=50/a=504003 [ * * , ( ) ] @noop ( ) http://stacks.iop.org/1367-2630/14/i=10/a=105008 [ * * , ( ) ] | results are presented for the entanglement entropy and spectrum of half - filled graphene following the switch on of a circularly polarized laser .
the laser parameters are chosen to correspond to several different floquet chern insulator phases .
the entanglement properties of the unitarily evolved wavefunctions are compared with the state where one of the floquet bands is completely occupied .
the true states show a volume law for the entanglement , whereas the floquet states show an area law .
qualitative differences are found in the entanglement properties of the off - resonant and on - resonant laser .
edge states are found in the entanglement spectrum corresponding to certain physical edge states expected in a chern insulator .
however , some edge states that would be expected from the floquet band structure are missing from the entanglement spectrum .
an analytic theory is developed for the long time structure of the entanglement spectrum .
it is argued that only edge states corresponding to off - resonant processes appear in the entanglement spectrum . |
brown dwarf spectra contain a wealth of information about the physical and chemical processes occurring in their atmospheres .
such processes include , but are not limited to : vertical atmospheric energy balance , disequilibrium chemistry , cloud formation / distribution and subsequent impact on molecular composition , and links to formation and evolution ( see marley & robinson 2015 review ) .
the main diagnostic quantities that provide insight into these processes , that can be obtained directly from a spectrum are : vertical thermal structure , molecular abundances , spectroscopic radius , and cloud properties .
deriving these diagnostic quantities is no simple task .
progress in characterization always involves the interplay between observational and modeling efforts .
it is from the comparison with models that we aim to understand the underlying physics and chemistry in substellar atmospheres .
the classic method for interpreting brown dwarf spectra is generating a set of self - consistent radiative- convective - thermochemical equilibrium models to which the data can be compared ( allard et al .
1996 , marley et al .
1996 , tsuji et al .
1996 , burrows et al .
grids of model spectra are typically generated by varying only a few parameters , usually gravity and effective temperature ( kirpatrick 2005 ) with all of the physics and chemistry in the atmospheres assumed ( e.g. , self - consistent radiative equilibrium plus equilibrium chemistry is used to predict the thermal structures and molecular abundances ) .
these models are then interpolated along these dimensions via a chi - squared search until a best fit , or range of best fit parameters are found ( e.g. , cushing et al .
2008 , rice et al .
2010 , stephens et al .
2009 , liu et al .
2011 ) . in some cases these two parameters
are not enough to adequately explain the spectra .
often times the fits are rather poor making any inference about these quantities and their corresponding uncertainties invalid ( czekala et al .
this suggests that the physical assumptions made within the grid models are not adequate enough to describe the data and that the grid models are underutilizing the full information content of the data . to remedy these issues some modelers chose to add additional physical parameters to their models to accommodate for new assumptions not previously ( as of then ) taken into account such as disequilibrium chemistry via an eddy diffusivity ( saumon et al .
2006 ; stephens et al . 2009 ) and more sophisticated cloud parameterizations ( ackerman & marley 2001 ; cushing et al . 2008 ) . furthermore , often - unknown differing assumptions within the various flavors of grid models from different groups , given the same sets of basic parameters , do nt agree resulting in inconsistent interpretations of the same datasets ( figure 8 , patience et al .
2012 ) . motivated by these shortcomings , line et al .
( 2014 ) & ( 2015 ) ( part 1 hereafter ) built upon the success of atmospheric retrieval methods applied to earth ( rodgers 1976 , towmey 1996 , rodgers 2000 , crisp et al .
2004 ) , solar system ( conrath et al .
1998 , irwin et al . 2008 , fletcher et al .
2007 , greathouse et al .
2011 ) , and exoplanet atmospheres ( lee et al .
2012 ; line et al .
2012;13 ; ; benneke & seager 2012 ; barstow et al .
2013;14 , waldman et al .
2015a , b ) to develop a framework for retrieving the molecular abundances , vertical temperature structure , gravity , and other quantities .
such an approach is agnostic as to the detailed physical and chemical mechanisms governing the atmosphere ; rather it allows the data to directly determine the fundamental quantities that impact the spectra , such as the opacities ( gases or clouds ) and radiative source functions .
part i leveraged the power of benchmark brown dwarf - stellar systems ( gl570 and hd3651 ) of which we know the host star ages , metallicities , and a new dimension , carbon - to - oxygen ratios ( c / o s ) .
the goal of part i was to demonstrate the feasibility of the approach by retrieving these quantities from the companion brown dwarfs .
assuming co - evolution , the brown dwarfs should have the same ages , metallicities , and c / o s as the host star . with low resolution
near infrared spex data of gl570d and hd3651b , line et al .
( 2015 ) was able to show that the retrieval method was indeed were able to reproduce these quantities to within the errors , thus validating the approach .
they also showed , for the first time , that one could robustly detect the presence of ammonia in near infrared low resolution data , without seeing " obvious spectroscopic features .
this detection demonstrated the power of using a bayesian approach to detect molecules that would have only otherwise been detected at higher resolution ( canty et al .
2015 ) or at longer wavelengths ( e.g. , saumon et al . 2006 ; hubeny & burrows 2007 ) .
furthermore , they were able to rule out the presence of optically significant clouds within the near infrared photospheres of these objects .
finally , via the retrieved temperature structures we were able to validate the radiative equilibrium assumption commonly made in self - consistent grid models . to illustrate the differences between retrieval " and grid - model " fits we compare a representative best fit retrieval model ( e.g. , line et al .
2015 ) to a representative grid model best fit ( morley et al .
2012 ) in figure [ fig : figure0 ] .
the grid models are based on morley et al .
( 2012 ) with the spectra being computed on a grid as a function of 4 fairly standard free parameters : effective temperature , gravity , clouds parameterized with an f - sed " ( ackerman & marley 2001 ) , and a radius - to - distance scale factor ( see morley et al .
2012 for details ) .
note that the retrieval fit ( red ) is significantly better than the grid model fit ( blue ) .
more quantitatively , the retrieval produces a chi - square per data - point of 5 , where - as the best - fit grid model a chi - square per data - point of 14.5 .
the retrieval model in part i included 27 free parameters , much more than the 4 used in the grid model .
are the retrievals justified in using this large number of parameters ?
we turn to the bayesian information criterion ( see figure [ fig : figure0 ] caption ) which penalizes models with more free parameters . in this case , the retrieval model is overwhelmingly favored , with a bic difference of @xmath51200 ( way off of the jefferey s scale ( e.g. , kass & raftery 1995 ) ) .
one potential drawback of the retrieval models is , however , the possibility of unphysical solutions . as shown in part i , the solutions obtained with our methodology are indeed chemically and physically plausible .
we elaborate on this further in a later section .
comparison between retrieval fits and self - consistent grid model fits .
the fit residuals are shown at the bottom of the plot .
the grid model fit ( in blue ) is based on a standard set of assumptions ( e.g. , saumon & marley 2008 ; morley et al .
2012 ) involving 4 free parameters : the effective temperature , gravity , cloud sedimentation ( ackerman & marley 2001 ) , and a flux scaling factor . the nominal best fit grid model ( effective temperature=900k , log - gravity=5 .
, no cloud ) results in a chi - square per data point of 14.5 , with 139 data points . with 4 free parameters
, this results a bayesian information criterion value of 2035.2 .
in contrast the retrieval model ( red ) with 27 free parameters produces a chi - square per data point of 5.0 and a bic of 832.4 . despite the factor of @xmath6 in the number of free parameters ,
the retrieval model is overwhelmingly favored with a @xmath7bic of @xmath51200 , which would be considered very strong " on the jeffery s scale .
, scaledwidth=50.0% ] in this manuscript , part ii , we apply the part i retrieval methodologies to a small sample ( 11 ) of late t - dwarfs ( t7-t8 ) within the spex prism library ( burgasser et al .
2006a ) in order to identify diagnostic physical / compositional trends .
we start with the late t - dwarfs as they are presumed to be free of optically thick clouds ( and demonstrated on two objects in part i ) ( e.g. , marley et al .
1996 , allard et al .
1996 ) , significantly reducing cloud degeneracies with abundances and thermal structures .
the long term goal is to extract molecular abundances , thermal structures , cloud properties , etc .
, for a large number ( @xmath5 hundreds to thousands ) of objects by working our way up ( and down ) the brown - dwarf spectral sequence , starting with these late t s then y - dwarfs , followed by more cloudy mid - to - early t s and finally l - dwarfs . by investigating
simpler " objects first , we can continue to identify the strengths and weaknesses of our model parameterization , and make the necessary adjustments along the way .
the manuscript is organized as follows : [ sec : methods ] briefly summarizes the methodology outlined in part i , and identifies small changes , [ sec : sample ] introduces the 11 t - dwarf sample , [ sec : results ] presents our retrieved quantities and the useful derived diagnostic quantities and trends , and finally in [
sec : conclusions ] we summarize our findings and discuss implications .
here we briefly summarize our methodologies extensively described in part i. we couple a simple thermal emission radiative transfer forward model ( line et al . 2013a ; 2014a , b ; 2015 ) with the powerful emcee sampler ( foreman - mackey et al .
the job of the forward model is to compute the the observable , in this case , the disk - integrated top - of - atmosphere thermal flux with wavelength .
we neglect scattering due to clouds and treat everything as pure absorption .
this assumption may break down in the presence of strongly forward scattering clouds .
the inputs into the forward model are the vertically uniform volume mixing ratios for h@xmath0o , ch@xmath1 , co ,
co@xmath0 , nh@xmath2 , h@xmath0s , na , and k , gravity , temperature - pressure ( tp ) profile , and the radius - to - distance scaling factor .
we use the same novel tp - profile retrieval scheme described extensively in part i. all other nuisance parameters from part i are included ( see table [ tab : table1 ] ) .
.[tab : table1 ] free parameters in the forward model ( 31 in total ) [ cols="^,^",options="header " , ]
here we present the retrieval results on the spectra of 11 objects ( figure [ fig : figure2 ] ) .
the relevant retrieved and subsequently derived quantities are summarized via marginalized posterior distributions ( histograms , figure [ fig : figure3 ] ) .
the retrieved temperature profiles are summarized in figure [ fig : figure4 ] .
we then present the molecular abundance results , their chemical plausibility , derived elemental abundances , and resulting diagnostic trends , implications for cloud formation , and finally comment on the retrieved evolutionary related parameters ( gravity , effective temperature , and radius ) .
understanding atmospheric energy balance requires knowledge of the temperature - pressure profiles ( tp - profile ) .
the tp - profile of a sub - stellar atmosphere is largely governed by the opacity structure and internal heat flux , which is in turn dictated by the mass and age ( e.g. , allard et al . 1996 ; burrows 2001 ; saumon & marley 2008 ) .
the two dominant energy transport mechanisms are radiation and convection .
radiative energy transport tries to eliminate any vertical radiative flux divergence .
thermal structures that meet this criterion are said to be in radiative - equilibrium . at deep enough layers , where the mean opacities are large , radiative energy transport is no longer effective at carrying the heat flux . at this point
, convective energy transport dominates .
the location of this radiative convective boundary is predicted to occur at pressures deeper than @xmath550 bars , or at atmospheric temperatures of @xmath51500k , for late t - dwarfs ( burrows et al .
additional heating due enhanced opacity from cloud particulates can sometimes result in multiple convection zones separated by a radiative zone ( burrows et al .
2006 ) .
other non - traditional sources of heating / cooling that could impact the tp - profiles and/or their time dependence ( e.g. , robinson & marley 2014 ) have been hypothesized .
latent heat release due to cloud condensation can result in a shallower adiabat than predicted from pure gas phase chemistry alone ( tan & showman 2016 ) , though the latent heat of typical l - t dwarf clouds are not particularly large ( say , compared with water or ammonia , which are important in the solar system gas - giant atmospheres ) .
double diffusive convection resulting from vertical composition gradients can also result in shallower tp - profiles in the mid to deep atmospheres ( tremblin et al .
chromospheric heating can cause a more isothermal behavior in the upper ( @xmath8 ) atmospheres ( sorahana et al .
additional dynamical effects like gravity wave breaking and horizontal dynamics can also cause perturbations to the tp - profile beyond the traditional assumption of 1d - radiative convective equilibrium . determining tp - profiles over a wide range of altitudes / pressures and comparing these derived profiles to radiative - convective equilibrium profiles allows us to diagnose additional atmospheric energetic processes .
figure [ fig : figure4 ] shows the tp - profiles for the 11 t - dwarfs in our sample . as in part i , we summarize the tp - profiles with a median , 68%- , and 95%-confidence intervals .
we also show a nominal ( median retrieved effective temperature and gravity ) pre - computed grid model ( saumon & marley 2008 ) .
thermal emission contribution functions are shown to indicate where in the atmosphere we are sensing the photosphere , typically between @xmath550 - 0.3 bar .
the y- , and j - band peaks probe the deepest levels , while the high opacity regions between these bands probe the lowest pressure levels ( highest altitudes ) .
the tp - profiles within the photosphere are considered the most robust .
typical 1-sigma photospheric temperature uncertainties are @xmath530 - 100 k. tp - profile information outside of those regions ( @xmath9 bar and @xmath10 bar ) are largely driven by the tp - profile parameterization ( via the smoothing prior described in detail in part i ) and should not be too heavily interpreted .
finally , we show the equilibrium condensation curves for kcl , na@xmath0s , and mgsio@xmath2 . in general ,
as we found for the two objects explored in part i , the retrieved tp - profiles and predicted radiative - convective equilibrium grid models agree quite well over the photosphere .
deviations begin to occur outside of the photosphere .
in part i we suggested that the shallower tp - profile gradient below the @xmath550 bar level maybe indicative of the double - diffusive convection suggested by tremblin et al .
however , there is very little contribution to the overall emission at these deeper levels due to the increasing opacity of the collision induced molecular hydrogen along with the alkali metal and water opacities at the j- and y - band peaks . above the photosphere ( @xmath10bar )
, we typically find that the tp - profiles become more isothermal than predicted from radiative - convective equilibrium , possibly hinting at some sort of additional heating , but again , temperature constraints are unreliable in this region . late t - dwarfs ( and y - dwarfs ) offer the best chances of learning bout the tp - profiles of brown dwarfs because of the large spectral dynamic range sculpted by strong water vapor absorption and predominately cloud free atmospheres .
this high spectral dynamic range provides leverage on the tp - profile over a wide range of photospheric pressures .
however , as seen here , it is difficult to push towards deeper pressures due to the increasing significance of collision induced opacities .
higher resolution spectra near the water band - heads ( between the y - j - h bands ) or longer wavelengths with higher opacity could allow us to probe _ lower _ pressures ( higher altitudes ) than probed here , permitting more precise upper atmosphere temperature constraints .
hotter objects , mid t - s and beyond , will present a challenge as their photospheres narrow due to the presence of clouds , preventing us from accessing the deeper atmospheric levels . any temperature profile information below where the cloud opacities become large will have to come from extrapolation via parameterizations or assumed from physically motivated models .
regardless , assumptions will have to be made and these assumptions will have to be tested .
we show , for the first time , the molecular abundances derived for an ensemble of t - dwarfs . molecular abundance determinations are diagnostic of the atmospheric chemical processes as well as the underlying elemental abundances .
we present the retrieved uniform - with - altitude molecular abundances for 7 gases : h@xmath0o , ch@xmath1 , co , co@xmath0 , nh@xmath2 , h@xmath0s , na+k , where na+k represents the combination of sodium and potassium , where their ratio is assumed to be solar .
the first 7 columns of figure [ fig : figure5 ] show the marginalized gas posterior distributions for the 11 objects , sorted in effective temperature ( low - high ) . as in part
i , we find that h@xmath0o , ch@xmath1 , nh@xmath2 , and
the alkalis are the only species with bounded constraints . we can derive only upper limits for the other species ( co , co@xmath0,h@xmath0s ) , consistent with a non - detection of these gases ( e.g. , see the bayes factor analysis in part 1 ) .
this is unsurprising given their low abundances predicted from thermochemical equilibrium ( lodders et al .
2002 ) . because we are retrieving each gas phase absorber individually , it is possible to obtain chemically implausible combinations ( e.g. , heng & lyons 2016 ; stevenson et al .
it is therefore prudent to check the plausibility of our results against a chemical model . for simplicity
, we compare our results to predictions from a thermochemical model ( chemical equilibrium with applications code 2 , gordon & mcbride 1996 , see part i for implementation details and relevant references ) in figure [ fig : figure5 ] .
briefly , the model computes the thermochemical equilibrium molecular ratios at a given p - t pair and set of elemental abundances ( lodders & fegley 2002 ) . the model accounts for both gas phase and condensed phase species in local thermochemical equilibrium , but does not take into account the rain - out paradigm whereby elemental species are depleted due to their sequestration into condensates that form large droplets and rain " out into the deeper unobservable atmosphere ( e.g. , fegley & lodders 1994 ; marley & robinson 2015 ) .
we do accommodate for the influence of main rain out process ( enstatite and forsterite condensation , fegley & lodders 1994 ) on the elemental abundances ( and subsequent equilibrium molecular abundances ) by manually removing 3.28 oxygen atoms per silicon atom ( sharp & burrows 1999 ) . as in part
i , we compute the equilibrium molecular abundances along the median of the retrieved tp - profiles for each object ( from figure [ fig : figure3 ] ) .
we compare the retrieved molecular abundances for only the well constrained species ( shaded regions in figure [ fig : figure5 ] ) to the equilibrium composition computed with solar elemental ratios ( solid curves in figure [ fig : figure5 ] ) . in general we find that the retrieved mixing ratios qualitatively agree with what is expected from thermochemical equilibrium at these temperatures : ch@xmath1 and h@xmath0o are the most abundant species followed by nh@xmath2 and then the alkalis .
we emphasize that both the retrieved and thermochemically calculated mixing ratios are really only relevant over the photosphere ( @xmath530 - 0.3 bar , figure [ fig : figure3 ] ) . upon more careful inspection
we find some inconsistencies between the thermochemical predictions at solar composition and the retrieved mixing ratios .
we then adjust the metallicity ( [ fe / h ] ) by rescaling all elements heavier than h and he and the carbon - to - oxygen ratio ( c / o ) to by - hand fit " the thermochemical model ( dashed curves in figure [ fig : figure5 ] corresponding to the inset values ) to the retrieved mixing ratios , exactly as was done in part i. this is by no means a rigorous fit ( see [ sec : element_abund ] for a more quantitative analysis ) , but is meant to show that we can obtain consistency between the thermochemical prediction and the retrieved mixing ratios by adjusting these two quantities , suggesting that we have retrieved _ chemically plausible _ molecular abundance combinations .
we do find some preliminary , though not robust , evidence of disequilibrium ammonia chemistry ( as discussed in saumon et al .
2006 ; cushing et al . 2008
; stephens et al .
2009 with respect to mid infrared data ) .
for most objects the thermochemically computed value of ammonia over the photospheric pressures probed by spex is lower than the retrieved values .
however , the thermochemical predictions at the artificial base of the atmosphere are consistent with the retrieved values , suggesting a deep quench level ( @xmath11100 bars ) could be possible in explaining the ammonia mixing ratios .
finally , we find some inconsistencies in between predicted and retrieved alkali abundances . this could be due to the difficulties in properly modeling the far pressure broadened wings of the alkali lines ( allard et al .
2009 ) and/or the details of condensate rainout chemistry ( of which our model does not account for ) , though it need not be that all elemental abundance ratios are solar . in the next section we will use this thermochemical modeling approach to derive the intrinsic bulk metallicity and carbon - to - oxygen ratios for each object .
an advantage of retrieval methods over grid models is that we can directly determine the molecular abundances rather than assume / compute them , and from trends in those abundances , infer physical / chemical processes . as an initial look
, we focus on molecular abundance trends with effective temperature ( figure [ fig : figure7 ] ) . chemical processes ( e.g. , equilibrium chemistry , condensate processes ) tend to depend strongly on temperature ( e.g , burrows & sharp 1999 , lodders & fegley 2002 ) ; we would therefore expect molecular abundances to change with temperature . specifically , we focus on trends associated with the well constrained species , e ..
g , those with bounded constraints , which include h@xmath0o , ch@xmath1 , nh@xmath2 , and the alkali metals .
the other species ( co , co@xmath0 , and h@xmath0s ) , while chemically interesting , simply do not have the precisions required to identify diagnostic trends .
we would predict water and methane to have little or no trend with temperature over our sample temperature range ( @xmath5 600 - 850 k ) as they are thermochemically constant ( e.g. , burrows & sharp 1999 figure 5 ) and largely unaffected by disequilibrium processes ( saumon et al .
2006 ) at these temperatures .
unsurprisingly , we indeed do not find any statistically robust trends of water and methane with temperature , as both are consistent with the null , no - trend , hypothesis ( figure [ fig : figure7 ] , to panels ) .
ammonia ( figure [ fig : figure7 ] , bottom left ) , on the other hand , is thermochemically predicted to decrease by nearly an order of magnitude over this temperature range over the photospheric pressure levels ( burrows & sharp 1999 ) . given the uncertainties , we should be able to identify such a trend if it existed ; the lack of trend is interesting .
perhaps the lack of trend is speaking to perturbations due to disequilibrium chemistry dredging up ammonia from these deeper layers . depending on the degree of dispersion in the vertical mixing strength amongst these objects (
can depend on internal heat flux , rotation rate , magnetic field strength , flasar and gierasch , 1977 ; visscher et al .
2010 ) , it is not surprising that the thermochemically predicted trend of decreasing abundance with temperature is not so clearly borne out .
a more striking trend is that of the alkali metal abundances with temperature ( figure [ fig : figure7 ] , bottom right ) , with the null hypothesis being rejected at @xmath325@xmath4 .
over the @xmath5 600 - 850 k range , the alkali metal abundance increases by nearly an order of magnitude .
we attribute this trend to the condensation of salt sulfide ( na@xmath0s ) and potassium chloride ( kcl ) clouds ( dominated by the kcl condensation as k dominates the alkali absorption neary y - band by a few to @xmath51 order magnitude over na at solar na / k ratios)20 times more abundant than k , the absorption cross section of k is @xmath5100 greater at shorter wavelength resulting in k dominating the opacity by nearly an order of magnitude ] .
the general hypothesis being , that as these objects cool , the tp - profile - condensate intersection point moves deeper into the atmosphere ( as seen in figure [ fig : figure4 ] where the dashed curves intersect the blue solid blue curve ) .
gas phase sodium and potassium only exist in appreciable quantities _ below _ the tp - profile - condensate intersection point . therefore , as the tp - profile - condensate intersection point moves deeper into the atmosphere , the overall integrated column abundance of gas phase elemental sodium and potassium decreases over the photospheric regions .
this is hypothesis is further confirmed in figure [ fig : figure8 ] . the left panel shows the pressure levels where the na@xmath0s and kcl condensate curves intersect the retrieved tp - profiles for each object ( as can be readily determined from the tp - profiles in figure [ fig : figure4 ] ) , as a function of effective temperature . as expected , as cooler temperatures occur at lower pressures for hotter objects , the condensates are in turn expected to form at these lower pressures ( higher altitudes ) .
we can then map back the retrieved alkali abundance trend with the pressure level of of the tp- profile - condensate intersection .
indeed we do find a trend , thus confirming our argument that hotter objects have a greater column of gas phase alkalis , of which our ensemble analysis is sensitive to .
there have long been two ` schools ' of how condensate chemistry should be treated in brown dwarf atmospheric models . in the pure equilibrium school condensates
are considered to continue to be in chemical equilibrium with their surrounding gas , even after condensation ( for example burrows & sharp 1999 ) . in the second school , commonly followed in solar system chemistry ( fegley & lodders 1994 ) condensed species `
rain out ' of the atmosphere and are no longer are available to react with surrounding gas at lower temperatures . in the pure equilibrium school for example gaseous na and k
react with previously condensed aluminum and silicate compounds to form feldspar minerals such as sanadine and albite ( see burrows et al .
2001 for a discussion ) .
meanwhile in the rain out conception the aluminum oxides and silicates rain out of the atmosphere after they condense and the feldspars do not form .
instead the gaseous na and k form chloride salts at much lower temperatures ( lodders 1999 ) . a number of lines of evidence ( marley et al .
2002 , morley et al .
2012 ) , have pointed towards rainout chemistry as being the most applicable to brown dwarf atmospheres , but conclusive evidence of na and k depletion in the atmosphere over the expected temperature range has been somewhat lacking . given our observed trend , and its correlation with the condensate base pressure , we have provided definitive evidence that this rain out paradigm is indeed correct . as demonstrated in
[ sec : element_abund ] and [ sec : trends ] , condensates are likely responsible for the depletion of oxygen and the alkali metal trend .
we might anticipate the resulting clouds to impact the spectrum .
as such , we have included the impact of a gray non - scattering cloud in our retrieval scheme ( equation [ eq : cloud_eq ] ) . to investigate the importance of clouds , from the three retrieved cloud parameters in equation [ eq : cloud_eq ]
we can compute the column integrated optical depth histograms over the photospheres for each object , shown in the last column ( @xmath12 ) of figure [ fig : figure3 ] .
the cloud optical depth histograms for each object all show a dramatic decline in probability near optical depths of unity ( vertical dotted lines ) .
this suggests that for these objects , larger cloud optical are rejected by the data indicating that optically thick clouds are not present .
this is consistent with our findings in part i where we applied bayesian hypothesis testing to show that the inclusion of clouds for gl570d and hd3651b was not justified given the data nor did their inclusion impact the retrieved values for any of the other parameters .
we note , that for some objects there is a small trickle of probability towards higher optical depths , though not significant relative to the total integrated probability .
the lack of large cloud optical depth is in reasonable agreement with expectations for late t - dwarfs from forward modeling investigations ( lodders 1999 ; lodders & fegley 2002 ; visscher et al .
2006;2010 ; morley et al .
2012 ) . based upon the intersection of the enstatite condensation curve with the tp - profiles ( figure [ fig : figure4 ] ) ,
we find that the silicate clouds largely reside below the photosphere , e.g. , the gaseous opacities become optically thick at pressure levels above where the enstatite cloud should form and thus should have no impact on the emergent spectrum .
this is consistent with the findings of morley et al .
( 2012 ) , where forward models of objects covering a similar temperature range suggest that the silicate clouds are are largely below the photosphere except for the objects approaching @xmath5900k .
figure [ fig : figure4 ] and the alkali metal trend suggests that na@xmath0s and kcl clouds should form within the photosphere .
morley et al . ( 2012 ) showed that the largely neglected cloud species na@xmath0s , kcl , mns , and zns can form within cool dwarf atmospheres .
they showed that , depending on gravity and the degree of droplet sedimentation ( higher gravity and more vertically extended clouds result in higher cloud optical depths ) , the optical depths of na@xmath0s ( more so than kcl ) can range from negligible to a few within the 1 - 6@xmath13 m photospheres of late t - dwarfs ( their figure 4 and 8) .
our photospheric optical depth upper limits suggest that the cloud optical depths can not be significantly greater than unity ) , which would allow for higher optical depths by a factor of @xmath5(1-@xmath14 ) ] .
this may potentially place constraints on the vertical extent of the clouds , largely favoring more vertically compact clouds ( higher degree of sedimentation within their modeling framework ) .
continued work on determining more precise cloud optical depth upper limits , and or even bounded constraints on cloud optical depths , will be invaluable in constraining cloud physics in substellar atmospheres .
the elemental abundances of a star or substellar object are diagnostic of their formation history and evolution as well as atmospheric chemical and physical processes ( bond et al .
2010 ; burrows 2001 ; marley & robinson 2015 ) . detailed stellar / substellar abundances are also critical to the understanding of galactic chemical evolution ( timmes et al .
1995 ; mcwilliam 1997 ; venn et al .
there has been a recent push to obtain high precision stellar abundance measurements of many elements around solar type stars in order to better understand planet formation ( petigura & marcy 2011 ; teske et al .
2014 ; fortney 2012 ) .
much work has gone into determining stellar abundances , but it is by no means a solved problem as different methods and stellar parameter assumptions can result in different elemental abundance determinations ( e.g. , hinkel et al .
2016 ) .
while solar - type abundance determinations have received much attention , little work has been done in determining brown dwarf elemental abundances beyond their metallicities ( e.g. , cushing et al .
2008 ; stephens et al . 2009 ) .
planets have been hypothesized ( payne & lodato 2007 ) , and some detected ( han et al .
2013 ) , around brown dwarfs .
brown dwarfs will eventually provide a unique opportunity to detect / explore habitable planets via the transit method due to their favorable separation and radius ratios ( belu et al .
we would therefore expect that in the not too distant future , brown dwarf elemental abundances determinations will become increasingly more important .
the challenging aspect of elemental abundance determinations in cool substellar objects ( m - dwarfs included ) is the presence of molecular species and condensates .
molecular species and condensates sequester what would otherwise be liberated elements available to form atomic absorption lines .
furthermore , the nature of their broadband absorption obscures the precious continuum so often required in classic stellar elemental abundance analysis ( e.g. , bean et al .
2006 ) . therefore
, novel methods need to be employed in order to leverage the molecular absorption to obtain the intrinsic elemental abundances . even then
, molecular abundances can be readily perturbed by atmospheric processes like vertical mixing and equilibrium cloud condensation .
these processes need to be taken into account when inferring elemental abundances from molecular absorption . while there has been effort to infer metallicities ( cushing et al .
2008 ; stephens et al .
2009 ) and other properties ( cloud thickness , gravity , effective temperature ) , many of these types of investigations of rely upon grids of pre - computed grid models over just a few parameters ( for instance non - solar c / o is often not accounted for ) .
often times these grid models do nt adequately fit the data ( e.g. , patience et al . 2012 ) and/or do not account for uncertainties in missing model physics as that physics is a - priori prescribed .
our retrieval methodology relaxes these assumptions and allows for more satisfying fits to the data .
the danger , again being , the possibility for unphysical " solutions , e.g. , solutions that do not conform to reasonable chemical / physical assumptions .
however , as demonstrated in @xmath15[sec : tp ] and [ sec : chemistry ] we are indeed deriving physically / chemically plausible atmospheric properties .
we proceed to derive the elemental abundances from the molecular abundances using the thermochemical model described in @xmath15[sec : chemistry ] as follows .
we first parameterize elemental abundances with the c / o and metallicity .
we then focus on ch@xmath1 and h@xmath0o as the metallicity and c / o diagnostics since their mixing ratios are not strongly influenced by chemical quenching at these temperatures ( in contrast to say , nh@xmath2 and co ) .
these are the molecular abundance data " that we will fit .
we execute a more rigorous ( in contrast to @xmath15[sec : chemistry ] ) fit to the molecular abundances by applying the chemical retrieval - on - retrieval " method described in line et al .
( 2016 ) .
the basic method is to fit the retrieved water and methane mixing ratio posteriors with the thermochemical model with the free parameters being the c / o and metallicity using emcee .
the exact data " being fit here are the water and methane histograms from figure [ fig : figure3 ] . at each mcmc step a tp - profile
is randomly drawn from the retrieval posterior in order to propagate the tp - profile uncertainty into the c / o - metallicity results . given the randomly selected tp - profile , the c / o and metallicity for that mcmc step ,
the thermochemical abundances of methane and water are computed at a representative photospheric pressure ( 3 bars , and corresponding temperature at that pressure level ) , and evaluate their probability of occurrence from the methane and water histograms .
these two probabilities are then multiplied and logged to compute the equivalent of chi - square " .
note that we could have a - priori fit the water and methane histograms with a gaussian ahead of time to derive a data point " with an error bar of which we could then use the chi - square formula .
however , the fit the histogram " approach is more rigorous because it can account for any arbitrarily shaped distribution ( see line et al .
2016 for details ) and is fundamentally what minimizing chi - square tries to do anyway .
we apply this procedure under two different condensate rainout assumptions .
the first assumes no rainout , e.g. , pure equilibrium .
all condensates remain in the atmospheric layer at which the formed .
the second is the silicate rainout paradigm as it affects oxygen whereby we remove 3.28 oxygen atoms per silicon atom ( as was done in @xmath15[sec : chemistry ] ) .
silicate condensation , and subsequent rainout , occur below the photosphere where the tp - profiles intersect the silicate condensate curves ( figure [ fig : figure4 ] show s enstatite ) essentially , we are simplistically exploring the effect that different condensation assumptions can have on the derived elemental abundances . recall that only photons emanating from a relatively thin slab of atmosphere are accessible by the observations .
unseen processes occurring outside of the photosphere can impact the molecular abundances that we retrieve within the photosphere .
we must indirectly account for these processes
. figure [ fig : figure6 ] summarizes the results of this analysis .
we find that our sample tends to have an overall sub - solar metallicity and super - solar c / o .
given that the retrieved water abundances for all objects is greater than or equal to the methane abundance , it is unsurprising that we in turn derive a super solar c / o .
we also find that the rainout assumption can have a fairly significant effect on the derived c / o and metallicity . increasing the amount of oxygen lost because of rainout shifts " the population towards more solar - like abundances .
one could imagine that if an even more efficient mechanism for oxygen sequestration existed ( e.g. , hydrated mineral formation via something like montmorillonite ) or some other unaccounted for condensation mechanism ) , these objects would further shift towards the solar values .
we are rather confident in robustness of our methods and assumptions used to derive these quantities given the good metallicity and c / o agreement in the benchmark systems , gl570 and hd3651 from part i. next , we compare the brown dwarf metallicities and carbon - to - oxygen ratios to those of the local f - g - k stellar population in figure [ fig : figure6 ] derived from the hypatia catalogue ( hinkel et al . 2014 ) .
we do not intend to delve into the details of stellar abundance determinations , suffice it to say , that it is indeed no trivial task and often different datasets and methodologies can lead to inconsistent results ( e.g. , hinkel et al .
2016 ) , resulting in relatively large systematic abundance uncertainties per any given star . in general , we find that the objects in our sample tend to lie towards lower metallicities and higher carbon - to - oxygen ratios than the bulk of the stellar population .
the reason for this as of now , is unknown , but we again , hypothesize that it may have to due with how efficient oxygen is sequestered into condensates , preventing us from determining the true oxygen inventory . more detailed thermochemical condensate modeling is required to identify additional sinks of oxygen .
it is also possible that the si / o ratio need not be in solar ( e.g. , hinkel et al .
2014 ) . with higher si abundances , more oxygen could be depleted through silicate rainout
. both [ si / fe ] and [ o / fe ] are known to decrease within decreasing [ fe / h ] in the stellar population ( figure 5 & 7 in hinkel et al .
2014 ) because they are @xmath16-elements ( e.g. , gratton & ortolani 1986 ; marcolini et al .
however , oxygen tends to decrease more than silicon with decreasing [ fe / h ] , suggesting that lower metallicity objects should have a _
greater _ proportion of o relative to si than at higher metallicities .
why the intrinsic si / o would be preferentially higher amongst brown dwarfs is unknown .
bulk metallicity ( sum of all species heavier than hydrogen and helium ) and carbon - to - oxygen ratios derived from the retrieved methane and water abundances .
the red points show those quantities derived from pure equilibrium .
the dark blue points show those quantities derived assuming silicate rainout condensate chemistry which removed 3.28 oxygen atoms per silicon atom .
this rainout effectively depletes " the oxygen that builds water in the photosphere , resulting in a reduced water abidance relative to methane . in this regime , in order to reproduce the retrieved methane and water molecular abundances , the bulk carbon to oxygen ratios have to be lower and the metallicity higher .
the data in purple was determined from the hypatia catalog ( hinkel et al .
2014 ) , an amalgamate dataset of f - g - k stellar abundances , where all literature source measurements are solar normalized to lodders et al .
( 2009 ) . in the instance
where multiple groups measure the same element in the same star ( hinkel et al . 2016 ) , the median value is used for the stellar abundance and the range or _
spread _ of those measurements is averaged to determine how well different determinations agree , shown as the representative error bar in the top right .
our results tend to sit towards lower metallicities and higher c / o s than the stellar population .
, scaledwidth=50.0% ] the radius , gravity , and effective temperature of a substellar object is diagnostic of its evolutionary history ( burrows et al .
2001 ; saumon & marley 2008 ) .
late field t - dwarfs are typically expected to be high gravity ( log@xmath175 ) and low effective temperatures , typically associated with higher ages .
we find the gravities of these objects falling between 4.75 and 5.3 .
gravities are not typically expected to exceed @xmath55.3 even for the oldest , most massive objects ( e.g. , saumon & marley 2008 ) .
figure [ fig : figure9 ] summarizes the effective temperatures and gravities of the 11 objects in our sample compared with the evolution models of saumon & marley ( 2008 ) .
the retrieved gravity and radii are consistent with ages between 1 and 5 gyr and masses between @xmath520 and 50 jupiter radii . as an additional , more quantiative check
, we use the approximate evolutionary relationships ( as a function of gravity and effective temperature ) from burrows et al .
( 2001 ) ( their equation 4 ) to derive the possible ages for each object ( last column in figure [ fig : figure3 ] ) .
again , we find that the median ages for each object are typically greater than @xmath51gyr , but some span a wide range of possible ages due to the larger uncertainties on gravity . the gravity and effective temperatures of this sample are also in line with expectations from population synthesis evolutionary models ( figure 9 in saumon & marley 2008 ) for late t - dwarfs
. the photometric radius of each object can be determined by the radius - to - distance dilution factor and photometry .
this factor effectively scales the spectra uniformly in wavelength . from this scale factor , the distance , and photometry
we can directly determine the radii of these objects .
we find that the radii for most of the objects in our sample fall between @xmath50.7 and 1.3 jupiter radii ( figure [ fig : figure3 ] ) .
we can test for the self - consistency of the spectroscopically derived radii using the analytic evolutionary relationship of radius with effective temperature and gravity ( equation 5 in burrows et al .
2001)-log@xmath18 pairs .
we found that for the same t@xmath19-log@xmath18 pair , radii differences between the two models are no larger than 0.02r@xmath20 .
this is small compared to the relative to the retrieved photometric radii errors of @xmath50.08 . ] . propagating the uncertainties in gravities and effective temperatures for each object into the evolutionary relationship for radius we find that our photometrically derived radii are systematically higher than expected from the evolution models .
the largest discrepancies ( in terms of differences in the medians of the radii distributions ) come from 2mass j1553 , 2mass j1217 , and to a lesser degree 2mass j0939 . as discussed in
[ sec : sample ] , from hst imagery , 2mass j1553 has been shown to be a binary , where - as the evidence for 2mass j1217 and 2mass j0939 being binaries is more circumstantial , based on somewhat higher fluxes than expected based on their spectral types . in future
work we if we a priori identify a binary system , or like in previous works ( e.g. , burgasser et al .
2008 ) , we identify an overly large radius , we may try fitting for the atmospheric properties of two separate objects and combining them into one spectrum ( as done in leggett et al .
2009 ) . at this time it is unclear why discrepancies in the evolution derived radii , and
the photometric radii are inconsistent .
perhaps some of these objects are indeed like 2mass j1217 and 2mass j0939 in they are unresolved binary systems .
we encourage further imaging campaigns of discrepant systems to determine if they are indeed binaries .
we note that the recent detection of a highly irradiated ( t@xmath212400k ) transiting brown dwarf , kelt-1b ( siverd et al . 2012 , see also bouchy et al .
2011 ) , also shows a larger radius , relative to evolution models , of @xmath22r@xmath23 given its mass of @xmath527@xmath24 and age ( @xmath51.5 - 2 gyr ) .
however , they attribute this anomalously large radius to its high degree of stellar irradiation , driving some as - of - yet unknown inflation mechanism operating on many hot jupiter s ( e.g. , laughlin et al .
it is currently unclear how stellar irradiation driven inflation mechanisms would impact higher gravity objects ( log@xmath18=4 - 5 , transiting planets have log@xmath18 @xmath25 3.5 with most less than 3 . ) .
bouchy et al .
( 2011 ) argue that large stellar insolation is likely to play a negligable role in increasing the radii of irradiated brown dwarf - mass objects .
these measured radii are typically @xmath510% larger than expectations from evolution models given their masses and a reasonable range of ages .
the large radii for the non - binary systems in our sample are in good agreement with these transiting systems . however , some transiting brown dwarf mass objects , e.g. , lhs 6343abb not highly irradiated ( johnson et al . 2011 ; montet et al .
2016 ) , do not show evidence for inflated radii ( @xmath26 r@xmath23 ) and are inline with evolution model expectations . continuing to identify these ultimate benchmark systems over a range of irradiation levels will further test evolution models and provide context for our radius determinations .
comparison of retrieved gravities and effective temperatures with evolution models ( saumon & marley 2008 , adapted from part i ) . the inferred ages ( blue dashed curves ) and masses ( black curves ) are consistent with expectations for field t - dwarfs .
, scaledwidth=50.0% ]
table [ tab : table3 ] summarizes our results compared with the literature .
there are a wide range of literature values for effective temperatures , gravities , and radii for any individual object . in a broad sense ,
our results are largely in - line with what has been found before .
however , we do find that in most cases our effective temperatures are systematically lower by up to @xmath5100k , which is many sigma , when compared to previous determinations .
however , the differences in the estimates from previous works can also be on the order of one to several hundred k due to different wavelength regions explored and/or different model grid .
overall , we find that our derived effective temperatures are systematically the coolest .
our gravity determinations agree very well with previous estimates , with most previous estimates for each object falling within our 1 sigma ( typically @xmath5 0.2 dex ) .
the inferred metalliciites are also typically in good agreement with previous determination , for select objects . for instance ,
for objects previously reported to have low metallicities ( 2mass j0939 and ulas j1416our two lowest metallicity objects ) , we too find significantly low metallicity .
we find a slightly higher , though , not significantly , metallicity for 2mass j0415 , consistent with the 0 - 0.3 dex determination from saumon et al .
we also find that the photometric radii are consistent with previous determinations where available ( 2mass j0415 with yamamura et al .
( 2010 ) reporting 1.14r@xmath23 based on akari data is with 2-sigma our value of 1.06@xmath27r@xmath23 ) and that binaries stand out as objects with overly large photometric radii ( e.g. , 1.26r@xmath23 for 2mass j0939 from burgasser et al .
2008 , compared with our determination of 1.22@xmath28r@xmath23 ) .
we have applied the benchmark - validated retrieval methodologies of line et al .
2015 ( part i ) to low resolution near infrared spectra of a small ensemble of 11 late t - dwarfs .
a systematic , uniform analysis of objects observed with the same instrument , and processed using the same modeling analysis provides a solid foundation for atmospheric property comparison and identification of diagnostic trends .
we focused on late t - dwarfs in this investigation for several reasons : firstly , their spectra have a large dynamic range , directly translating into a wide range of atmospheric pressures probed ; this in turn provides strong leverage on temperature - pressure profile constraints . secondly ,
late t - dwarfs are largely consistent with cloud free atmospheres , mitigating a significant temperature - cloud - gas abundance degeneracy , and allowing for a vertically wide photosphere .
thirdly , the cool " temperatures ( @xmath5600 - 800 k ) occupied by late t - dwarfs thermochemically result in large methane and water abundances .
these species are both spectrally prominent with multiple bands across the near infrared , and are the dominant carbon and oxygen reservoirs , un - tainted by disequilibrium chemistry , making for robust metallicity and carbon - to - oxygen ratio determinations .
finally , this temperature range encompasses the peak of the equilibrium temperatures spanned by most transiting exoplanets ( bathala et al .
understanding the physical and chemical processes in these t - dwarf laboratories , with data of similar spectral quality as the james webb space telescope , will undoubtably provide a solid foundation and context with which we can base our planetary atmosphere inferences in the not too distant future . from this analysis ,
we have arrived at the following conclusions : + 1 .
our retrieval modeling approach can fit the data very well .
the residuals are quite small and appear to be random ( figure [ fig : figure2 ] ) .
only water , methane , ammonia , and the alkalis have strong , bounded constraints ( mole fractions typically constrained at 68% confidence to within @xmath290.2 dex , or a factor of 1.6 ) .
we only obtain upper limits on the mixing ratios of the other include species ( co , co@xmath0 , and h@xmath0s ) ( figure [ fig : figure3 ] ) .
ammonia is present and well constrained in all objects , consistent with strong detections presented via 3 independent methods in part i. 3 .
the retrieved temperature - pressure profiles are consistent with expectations from radiative - convective equilibrium temperature profiles over the photosphere ( figure [ fig : figure4 ] ) .
deviations from grid model expectations , potentially diagnostic of additional physics , that occur outside of the photospheres should not be too heavily interpreted .
the retrieved molecular abundances ( for the constrained species ) are largely consistent with thermochemical equilibrium predictions for given temperature - pressure profile , metallicity , and carbon - to - oxygen ratio .
this suggests chemical plausibility ( figure [ fig : figure5 ] ) .
there is no justifiable trend in the mixing ratios of methane , water , or ammonia with effective temperature .
there is however , a very strong trend in the alkali metal mixing ratios with effective temperature ( figure [ fig : figure7 ] ) .
this alkali trend is consistent with expectations from condensate rainout chemistry , which results in a depletion of the column integrated mixing ratios of elemental sodium and potassium at cooler effective temperatures as the temperature - condensate intersection pressures for na@xmath0s and kcl move towards the deeper atmosphere ( figure [ fig : figure8 ] ) .
we only retrieve cloud optical depth upper limits .
these upper limits are typically near unit optical depth ( figure [ fig : figure3 ] ) suggesting that clouds do not substantially influence the spectra of these 11 late t - dwarfs .
this is consistent with our findings from part i where the inclusion of cloud parameters was not justified given the data , nor did their inclusion , or lack there - of , influence in any way the constraints for the other parameters .
. the elemental abundance analysis on the retrieved methane and water mixing ratios suggests sub - solar metallicities and somewhat super - solar carbon - to - oxygen ratios , and are somewhat lower and higher , respectively , when compared to the field f - g - k stellar abundances .
one caveat being that the late t - dwarf metallicities and carbon - to - oxygen ratios are sensitive to the degree of oxygen depletion in condensates that form below the photosphere ( figure [ fig : figure6 ] ) .
more efficient oxygen depletion due to additional rainout mechanisms could further shift the metallicity and carbon - to - oxygen ratios towards solar .
the retrieved gravities and effective temperatures are largely consistent with evolutionary models ( figure [ fig : figure9 ] ) and previous works , though our retrieved effective temperatures are systematically cooler than previous grid - model derivations .
we also find that the spectroscopically retrieved radii of several of the objects are larger than those expected from the evolution models derived from the retrieved effective temperatures and gravities .
some of these large radii are indicative un - resolved binary systems , consistent with the literature .
the other object s larger radii are difficult to explain , but those larger radii are in line with what has been found for some transiting brown dwarfs .
we emphasize , that all of these constraints have been made with relatively low resolution spectra ( spex prism , at r=80 - 120 ) .
hi - resolution is not necessarily a requirement for obtaining useful abundance information .
broad spectral coverage spanning multiple molecular bands goes a long way towards constraining molecular abundances , even with a large number of highly degenerate parameters .
while we have learned a lot from just this small sample , there is still very much unknown and much work to be done .
expanding our sample to more objects over a wider range of temperatures will allow us to obtain better statistics on the preliminary trends presented here .
we would like to first expand our sample to more low resolution spectra of late t - dwarfs and y - dwarfs .
we have demonstrated that our methodologies work within this cool temperature regime and for low - resolution data .
we should understand these simpler objects before we can move onto more complex , warmer , cloudier objects ( burningham et al . in prep ) . as we move towards cooler temperatures
, we would like to definitively identify the transition effective temperature at which we can no longer determine bounded constraints on the alkali abundances .
color - spectral type analysis suggest that this transition occurs around t8 , near the y - t transition ( e.g. , schneider et al .
2015 , figure 13 ) . measuring the transition abundance and
effective temperature would place further constraints on the rainout mechanism , and can be directly compared to equilibrium chemistry predictions . moving towards slightly higher temperatures , we would like to identify when the strong ammonia detection disappears ( becomes an upper limit ) , or if we can retrieve a decline in its abundance with increasing temperature ( e.g. , figure 5 , burrows & sharp 1998 ) .
the rate of decline , and where this transition occurs relative to equilibrium chemistry predictions , will place constraints on the role of disequilibrium nitrogen chemistry in late t - dwarf atmospheres ( e.g. , the dispersion in vertical mixing strengths ) . continued validation of the model against star - dwarf benchmark systems ( like gl570 and hd3651 from part 1 ) and also dwarf - dwarf benchmark systems is also necessary to identify improvements and/or pitfalls in our current modeling . in that
regards , there are several modeling improvements .
firstly , more testing and identification of robust temperature - pressure profile parameterizations are needed . our current
p - spline smoothing prior " parameterization introduced in part i works well for t - dwarfs , and is in agreement with expectations from self - consistent cloud - free grid models . however
, cloudy objects with dampened photospheres will require less flexible parameterizations with fewer free parameters .
understanding the influence of these temperature profile assumptions on the abundances , gravities , cloud optical depths , etc .
, will become more important as we move towards hotter objects or objects with less spectral coverage and signal - to - noise , like self - luminous directly imaged planets .
the role that various cloud parameterizations play in influencing the other retrieved properties will also need to be explored ( e.g. , gray non - scattering , vs. wavelength dependent scattering , cloud vertical profiles / particle sizes / compositions / patchiness , etc . ) , as has been done within grid modeling frameworks ( e.g. , burrows et al .
2006 ; cushing et al .
2008 ; morley et al .
finally , more sophisticated data - model comparison techniques should be explored ( e.g. , czekala et al . 2015 ) especially when retrieving on higher resolution spectra and stitching together spectra from different instruments , orders , and epochs .
failure to account for these direct data - model comparison subtleties may result in biased answers a motivation for why we choose a single instrument ( and mode ) , low resolution set of data .
despite these challenges , we are optimistic that continued improvement and application of these retrieval methodologies to large samples of hundreds of objects spanning a wide array of temperatures , gravities , and ages , will undoubtably provide new insights and surprises in physics and chemistry of brown dwarf atmospheres .
we thank roxana lupu and richard freedman for continued development and support of their extensive opacity database of which makes much of this work possible .
we also thank michael cushing , jackie fahrety , trent dupuy , brenden bowler , didier saumon , ryan garland , ty robinson & adam schneider for useful discussions .
m.r.l . acknowledges support provided by nasa through hubble fellowship grant 51362 awarded by the space telescope science institute , which is operated by the association of universities for research in astronomy , inc .
, for nasa , under the contract nas 5 - 26555 .
bb acknowledges financial support from the european commission in the form of a marie curie international outgoing fellowship ( piof - ga-2013- 629435 ) .
nrh acknowledges the support of the vanderbilt office of the provost through the vanderbilt initiative in data - intensive astrophysics ( vida ) fellowship .
allard , f. , hauschildt , p. h. , baraffe , i. , & chabrier , g. 1996 , , 465 , l123 ackerman , a. s. , & marley , m. s. 2001 , , 556 , 872 aitchison , j. 1986 , the statistical analysis of compositional data , chapman & hall london barnes , s. a. 2007 , , 669 , 1167 barstow , j. k. , aigrain , s. , irwin , p. g. j. , fletcher , l. n. , & lee , j .- m .
2013 , , 434 , 2616 belu , a. r. , selsis , f. , raymond , s. n. , et al .
2013 , , 768 , 125 benneke , b. , & seager , s. 2012 , , 753 , 100 bond , j. c. , obrien , d. p. , & lauretta , d. s. 2010 , , 715 , 1050 bonfanti , a. , ortolani , s. , & nascimbeni , v. 2016 , , 585 , a5 bowler , b. p. , liu , m. c. , & dupuy , t. j. 2010 , , 710 , 45 , a. j. 2007 , , 658 , 617 burgasser , a. j. , burrows , a. , & kirkpatrick , j. d. 2006 , , 639 , 1095 burgasser , a. j. , geballe , t. r. , leggett , s. k. , kirkpatrick , j. d. , & golimowski , d. a. 2006 , , 637 , 1067 burgasser , a. j. , et al .
2002 , , 564 , 421 burgasser , a. j. , kirkpatrick , j. d. , cruz , k. l. , reid , i. n. , leggett , s. k. , liebert , j. , burrows , a. , & brown , m. e. 2006 , , 166 , 585 burgasser , a. j. , kirkpatrick , j. d. , reid , i. n. , brown , m. e. , miskey , c. l. , & gizis , j. e. 2003 , , 586 , 512 burgasser , a. j. , looper , d. , & rayner , j. t. 2010 , , 139 , 2448 burgasser , a. j. , tinney , c. g. , cushing , m. c. , saumon , d. , marley , m. s. , bennett , c. s. , & kirkpatrick , j. d. 2008 , , 689 , l53 burgasser , a. j. , et al .
1999 , , 522 , l65 . 2000 , , 531 , l57 . 2010 , , 725 , 1405 burningham , b. , et al .
2010 , , 404 , 1952 burrows , a. , sudarsky , d. , & hubeny , i. 2006 , , 640 , 1063 burrows , a. , hubbard , w. b. , lunine , j. i. , & liebert , j. 2001 , reviews of modern physics , 73 , 719 canty , j. i. , lucas , p. w. , yurchenko , s. n. , et al .
2015 , , 450 , 454 casagrande , l. , flynn , c. , portinari , l. , girardi , l. , & jimenez , r. 2007 , , 382 , 1516 casagrande , l. , schnrich , r. , asplund , m. , cassisi , s. , ramrez , i. , melndez , j. , bensby , t. , & feltzing , s. 2011 , , 530 , a138 conrath , b. j. , gierasch , p. j. , & ustinov , e. a. 1998 , icarus , 135 , 501 crisp , d. , atlas , r. m. , breon , f. m. , brown , l. r. , burrows , j. p. , ciais , p. , schroll , s. 2004 , adv . in space res . , 34 , 700 - 709 .
cushing , m. c. , marley , m. s. , saumon , d. , et al .
2008 , , 678 , 1372 czekala , i. , andrews , s. m. , mandel , k. s. , hogg , d. w. , & green , g. m. 2015 , , 812 , 128 del burgo , c. , martn , e. l. , zapatero osorio , m. r. , & hauschildt , p. h. 2009 , , 501 , 1059 dupuy , t. j. , & liu , m. c. 2011 , , 733 , 122 . 2012 , , 201 , 19 faherty , j. k. , et al . 2012 , , 752 , 56 fegley , b. , jr . , & lodders , k. 1994 , icarus , 110 , 117 fernandes , j. m. , vaz , a. i. f. , & vicente , l. n. 2011 , , 532 , a20 fischer , d. a. , butler , r. p. , marcy , g. w. , vogt , s. s. , & henry , g. w. 2003 , , 590 , 1081 foreman - mackey , d. , hogg , d. w. , lang , d. , & goodman , j. 2013 , , 125 , 306 fletcher , l. n. , irwin , p. g. j. , teanby , n. a. , et al .
2007 , icarus , 189 , 457 freedman , r. s. , marley , m. s. , & lodders , k. 2008 , , 174 , 504 - 513 freedman , r. s. , lustig - yaeger , j. , fortney , j. j. , et al .
2014 , , 214 , 25 fortney , j. j. 2012 , , 747 , l27 geballe , t. r. , saumon , d. , golimowski , d. a. , leggett , s. k. , marley , m. s. , & noll , k. s. 2009 , , 695 , 844 greathouse , t. k. , richter , m. , lacy , j. , et al . 2011 , icarus , 214 , 606 grimm , s. l. , & heng , k. 2015 , , 808 , 182 gordon , s. , & mcbride , b. j. 1994 , nasa reference publ . 1311
han , c. , jung , y. k. , udalski , a. , et al .
2013 , , 778 , 38 heng , k. , & lyons , j. r. 2016 , , 817 , 149 hinkel , n. r. , timmes , f. x. , young , p. a. , pagano , m. d. , & turnbull , m. c. 2014 , , 148 , 54 hinkel , n. r. , young , p. a. , pagano , m. d. , et al . 2016 , , 226 , 4 hubeny , i. , & burrows , a. 2007 , , 669 , 1248 irwin , p. g. j. , teanby , n. a. , de kok , r. , et al .
2008 , , 109 , 1136 khandrika , h. , burgasser , a. j. , melis , c. , luk , c. , bowsher , e. , & swift , b. 2013 , , 145 , 71 kirkpatrick , j. d. 2005 , , 43 , 195 knapp , g. r. , et al .
2004 , , 127 , 3553 lee , j .- m . , fletcher , l. n. , & irwin , p. g. j. 2012 , , 420 , 170 , s. k. , et al .
2002 , , 564 , 452 leggett , s. k. , marley , m. s. , freedman , r. , saumon , d. , liu , m. c. , geballe , t. r. , golimowski , d. a. , & stephens , d. c. 2007 , , 667 , 537 leggett , s. k. , saumon , d. , marley , m. s. , geballe , t. r. , golimowski , d. a. , stephens , d. , & fan , x. 2007 , , 655 , 1079 leggett , s. k. , et al .
2009 , , 695 , 1517 . 2010 , , 710 , 1627 . 2012 , , 748 , 74 ligi , r. , et al .
2016 , , 586 , a94 line , m. r. , teske , j. , burningham , b. , fortney , j. j. , & marley , m. s. 2015 , , 807 , 183 line , m. r. , zhang , x. , vasisht , g. , et al .
2012 , , 749 , 93 line , m. r. , wolf , a. s. , zhang , x. , et al .
2013 , , 775 , 137 line , m. r. , knutson , h. , wolf , a. s. , & yung , y. l. 2014a , , 783 , 70 line , m. r. , fortney , j. j. , marley , m. s. , & sorahana , s. 2014b , , 793 , 33 line , m. r. , stevenson , k. b. , bean , j. , et al .
2016 , , 152 , 6 liu , m. c. , delorme , p. , dupuy , t. j. , et al .
2011 , , 740 , 108 liu , m. c. , leggett , s. k. , & chiu , k. 2007 , , 660 , 1507 lodders , k. 1999 , , 519 , 793 lodders , k. , & fegley , b. 2002 , icarus , 155 , 393 lodders , k. , & palme , h. 2009 , meteoritics and planetary science supplement , 72 , 5154 looper , d. l. , kirkpatrick , j. d. , & burgasser , a. j. 2007 , , 134 , 1162 lucas , p. w. , et al .
2010 , , 408 , l56 luhman , k. l. , et al .
2007 , , 654 , 570 mamajek , e. e. , & hillenbrand , l. a. 2008 , , 687 , 1264 marley , m. s. , saumon , d. , guillot , t. , et al .
1996 , science , 272 , 1919 marley , m. s. , & robinson , t. d. 2015 , , 53 , 279 mcwilliam , a. 1997 , , 35 , 503 metchev , s. a. , et al .
2015 , , 799 , 154 morley , c. v. , fortney , j. j. , marley , m. s. , et al .
2012 , , 756 , 172 mugrauer , m. , seifahrt , a. , neuhuser , r. , & mazeh , t. 2006 , , 373 , l31 payne , m. j. , & lodato , g. 2007 , , 381 , 1597 petigura , e. a. , & marcy , g. w. 2011 , , 735 , 41 pineda , j. s. , hallinan , g. , kirkpatrick , j. d. , cotter , g. , kao , m. m. , & mooley , k. 2016 , , 826 , 73 patience , j. , king , r. r. , de rosa , r. j. , et al .
2012 , , 540 , a85 radigan , j. 2014 , , 797 , 120 radigan , j. , lafrenire , d. , jayawardhana , r. , & artigau , e. 2014 , , 793 , 75 rice , e. l. , barman , t. , mclean , i. s. , prato , l. , & kirkpatrick , j. d. 2010 , , 186 , 63 rodgers , c. d. 1976 , reviews of geophysics and space physics , 14 , 609 rodgers , c. d. 2000 , inverse methods for atmospheric sounding - theory and practice .
series : series on atmospheric oceanic and planetary physics , isbn : < isbn>9789812813718</isbn>. world scientific publishing co. pte . ltd .
, edited by clive d. rodgers , vol . 2 , 2 saumon , d. , & marley , m. s. 2008 , , 689 , 1327 saumon , d. , marley , m. s. , cushing , m. c. , leggett , s. k. , roellig , t. l. , lodders , k. , & freedman , r. s. 2006 , , 647 , 552 saumon , d. , et al .
2007 , , 656 , 1136 schmidt , s. j. , west , a. a. , burgasser , a. j. , bochanski , j. j. , & hawley , s. l. 2010 , , 139 , 1045 scholz , r. 2010 , , 510 , l8 sharp , c. m. , & burrows , a. 2007 , , 168 , 140 sorahana , s. , & yamamura , i. 2012 , , 760 , 151 sorahana , s. , suzuki , t. k. , & yamamura , i. 2014 , , 440 , 3675 stephens , d. c. , leggett , s. k. , cushing , m. c. , et al .
2009 , , 702 , 154 stevenson , k. b. , bean , j. l. , madhusudhan , n. , & harrington , j. 2014 , , 791 , 36 teske , j. k. , cunha , k. , smith , v. v. , schuler , s. c. , & griffith , c. a. 2014 , , 788 , 39 testi , l. 2009 , , 503 , 639 timmes , f. x. , woosley , s. e. , & weaver , t. a. 1995 , , 98 , 617 tinney , c. g. , burgasser , a. j. , & kirkpatrick , j. d. 2003 , , 126 , 975 tinney , c. g. , burgasser , a. j. , kirkpatrick , j. d. , & mcelwain , m. w. 2005 , , 130 , 2326 tremblin , p. , amundsen , d. s. , mourier , p. , et al .
2015 , , 804 , l17 tsuji , t. , yamamura , i. , & sorahana , s. 2011 , , 734 , 73 tsuji , t. , ohnaka , k. , aoki , w. , & nakajima , t. 1996 , , 308 , l29 twomey , s. , herman , b. , & rabinoff , r. 1977 , jas , 34 , 1085 valenti , j. a. , & fischer , d. a. 2005 , , 159 , 141 van leeuwen , f. 2007 , , 474 , 653 vican , l. 2012 , , 143 , 135 visscher , c. , lodders , k. , & fegley , b. , jr .
2006 , , 648 , 1181 visscher , c. , lodders , k. , & fegley , b. , jr .
2010 , , 716 , 1060 - 1075 venn , k. a. , irwin , m. , shetrone , m. d. , et al .
2004 , , 128 , 1177 wilson , p. a. , rajan , a. , & patience , j. 2014 , , 566 , a111 yamamura , i. , tsuji , t. , & tanab , t. 2010 , , 722 , 682 | brown dwarf spectra are rich in information revealing of the chemical and physical processes operating in their atmospheres .
we apply a recently developed atmospheric retrieval tool to an ensemble of late t - dwarf ( 600 - 800k ) near infrared spectra . with these spectra
we are able to place direct constraints the molecular abundances of h@xmath0o , ch@xmath1 , co , co@xmath0 , nh@xmath2 , h@xmath0s , and na+k , gravity , thermal structure ( and effective temperature ) , photometric radius , and cloud optical depths .
we find that ammonia , water , methane , and the alkali metals are present and well constrained in all 11 objects . from the abundance constraints we find no significant trend in the water , methane , or ammonia abundances with temperature , but find a very strong ( @xmath325@xmath4 ) increasing trend in the alkali metal abundances with effective temperature , indicative of alkali rainout .
we also find little evidence for optically thick clouds . with the methane and water abundances ,
we derive the intrinsic atmospheric metallicity and carbon - to - oxygen ratios .
we find in our sample , that metallicities are typically sub solar and carbon - to - oxygen ratios are somewhat super solar , different than expectations from the local stellar population .
we also find that the retrieved vertical thermal profiles are consistent with radiative equilibrium over the photospheric regions .
finally , we find that our retrieved effective temperatures are lower than previous inferences for some objects and that our radii are larger than expectations from evolutionary models , possibly indicative of un - resolved binaries .
this investigation and methodology represents a paradigm in linking spectra to the determination of the fundamental chemical and physical processes governing cool brown dwarf atmospheres . |
following the first discovery of the presence of heavy elements in the photospheres of h - rich ( da ) white dwarfs with _ iue _
( bruhweiler & kondo 1981 , 1983 ) , it is now well established that they are ubiquitous in the group of hottest objects , with effective temperatures in excess of 55000k ( e.g. marsh _
et al . _
1997b ; barstow _ et al . _
extensive further studies in the far uv have eventually revealed the presence of absorption lines from c , n , o , si , s , p , fe and ni in a number of objects ( e.g. vennes _
et al . _ 1992
et al . _ 1992
; holberg _ et al . _ 1994 ;
et al . _
1996 ) .
apart from the detectable far uv absorption lines , these heavy elements have important effects on other spectral regions . for example
, they systematically alter the flux level and shape of the optical balmer line profiles , from which and log g can be determined .
the effect is to lower the measured value of by several thousand degrees , compared to that determined under the assumption that the star has a pure h envelope ( barstow _ et al . _ 1998 ) . in the extreme ultraviolet ( euv )
, the heavy element opacity dramatically blocks the emergent flux , yielding a much steeper short wavelength cutoff than is seen in a star with a pure h atmosphere . while , the general shape of the euv spectrum of metal - containing hot da white dwarfs has been understood qualitatively , since vennes _
et al . _ ( 1988 ) first successfully interpreted the _ exosat _ spectrum of feige 24 with an arbitrary mixture of elements , quantitative agreement between the observations and the predictions of theoretical model atmospheres has proved much more elusive .
for example , initial attempts failed to match either the flux level or even the general shape of the continuum of one of the best studied white dwarfs , g191@xmath0b2b ( e.g. barstow _ et al . _
this was eventually perceived to be due to inclusion of an insufficiently large number of heavy element lines ( mainly fe and ni ) .
improved non - lte calculations , containing some 9 million predicted fe and ni lines , rather than just the 300,000 or so observed experimentally , were able to provide a self - consistent model which could accurately reproduce the euv , uv and optical spectra ( lanz _ et al . _
subsequently , other authors have obtained similar results with both lte and non - lte models ( e.g. wolff _ et al . _ 1998
; koester _
et al . _ 1997
; chayer _ et al . _
interestingly , wolff _
et al . _ ( 1998 ) obtained a lower fe abundance from their far uv analysis than did lanz _
et al . _ ( 1996 ; @xmath2 and @xmath3 respectively ) , while agreeing with the amount of fe ( @xmath3 ) required by the euv spectra . however , we note that wolff _ et al . _ ( 1998 ) only study the restricted wavelength range available from the ghrs spectra , confining their analysis to only a few , relatively weak fev lines .
in contrast , lanz _
et al . _ ( 1996 ) considered the best match to all the fe lines visible in the _ iue _ spectrum .
furthermore , the far uv lines are much less sensitive to the fe abundance than the euv continuum .
consequently , it is likely that the apparent discrepancy is not significant . despite these advances
important problems remain .
first , the good agreement between the observed euv spectrum of g191@xmath0b2b and latest model predictions can only be obtained by inclusion of a significant quantity of helium , either in the photosphere or in the form of an interstellar / circumstellar component ( see lanz _ et al . _
1996 ) . at the comparatively limited @xmath4 resolution of _ euve _ in the region of the lyman series , and the consequent blending of these lines with features from heavier elements , the inferred he contribution can not be directly detected .
if he is really present , two alternative interpretations arise .
either there is an interstellar / circumstellar component or the material resides predominantly in the stellar photosphere .
if the former explanation holds , the amount of required implies an extremely high ionization fraction ( 80% ) when compared with the measured column density . this is a much higher value than appears to be typical of the local interstellar medium ( ism ) in general ( barstow _ et al . _
on the other hand , if there is a significant photospheric component the implied abundance of he / h=@xmath5 is in disagreement with the upper limit of @xmath6 imposed by the absence of a detectable 1640 feature in the uv .
a partially successful attempt has been made to resolve these issues by adopting a physically more realistic model , where the helium is gravitationally stratified rather than homogeneously distributed within the atmosphere ( barstow & hubeny 1998 ) .
the required interstellar he ionization fraction remains high ( 59% ) , compared to that of the local ism , but the predicted strength of the 1640 line becomes consistent with observation .
unfortunately , with the stratified model , the predicted lyman series lines are somewhat stronger than can be accommodated by the _ euve _ spectrum . in the end , the issue of the component will only be solved when a much higher resolution spectrum is obtained , capable of resolving any lines from those of heavier elements .
the j - pex sounding rocket spectrometer , with an expected flight in early 1999 , should provide such data ( bannister _ et al . _
1999 ) .
the second major difficulty in understanding the spectrum of g191@xmath0b2b has received considerably less attention .
indeed , to some extent , with the recent successes in dealing with the optical , uv and medium - long wavelength _ euve _
simultaneously , it has been specifically ignored .
all the results of lanz _
et al . _
1996 , barstow & hubeny ( 1998 ) and others , discussed above , only consider the euv spectral data longward of @xmath7 .
nevertheless , there is significant flux detected shortward of this , in the _ euve _ short wavelength channel and in the soft x - ray .
furthermore , the flux level predicted by the most successful current models is between five and ten times that observed .
a complete understanding of the atmosphere of g191@xmath0b2b requires that we also explain the short wavelength spectrum as well as the longer wavelength spectral ranges .
we show here that the _ entire _ spectrum of g191@xmath0b2b , from the short wavelength euv to the optical , can be explained if the fe known to be present in the atmosphere is not homogeneously mixed but stratified , with a decreasing abundance towards the outer layers of the stellar envelope .
such a heavy element distribution may be an indication of ongoing mass loss from the star , which has important consequence for the spectral evolution of g191@xmath0b2b and hot da white dwarfs in general .
as in our earlier papers on g191@xmath0b2b , we utilised the ` dithered ' _ euve _ spectrum obtained on 1993 december 7 - 8 ( see lanz _ et al . _ 1996 ) and the coadded _ iue _ echelle spectrum of holberg
_ et al . _ ( 1994 ) .
the _ euve _ ` dither ' mode consists of a series of pointings slightly offset in different directions from the nominal source position to average out flat field variations . as these data have been extensively described elsewhere
, we just give a brief summary here .
the _ euve _
exposure times were 58,815s , 49239s and 60,816s in the sw , mw and lw ranges respectively .
we assume that the residual efficiency variation , after the effect of the ` dither ' , is 5% , as observed for hz43 ( e.g. barstow holberg and koester 1995 ; dupuis _ et al . _
1995 ) , quadratically adding a systematic error of this magnitude to the statistical errors of the data . as one of the brightest euv sources in the sky , the spectrum of g191@xmath0b2b is well - exposed across most of the wavelength range , achieving the maximum signal - to - noise possible with _ euve
_ ( limited by the residual fixed pattern efficiency variation ) at the maximum spectral resolution . since the raw spectra oversample the true resolution by a factor @xmath8 , we have generally chosen to bin the data by this factor during the spectral analysis
. however , in the sw range , the heavy element opacity causes a dramatic reduction in the observed stellar brightness .
consequently , in an exposure optimised for the mw and lw ranges , the signal - to - noise at shorter wavelengths is very low . to approach the signal - to - noise achieved at longer wavelengths and produce
a data set with no bins containing zero counts , it was necessary to rebin the data below @xmath1 by a further factor 8 ( 32 in total ) . inevitably , detailed spectral line information is lost but the data can otherwise constrain the model spectra .
figure 1 shows the complete _ euve _ spectrum of g191@xmath0b2b , comparing it to best fit model of lanz _
et al . _ ( 1996 ) .
recent work ( barstow hubeny & holberg 1998 ) has shown that the value of determined from an analysis of the optical balmer lines is sensitive to a combination of non - lte effects and heavy element line blanketing . in their analysis ,
et al . _ ( 1996 ) adopted an effective temperature of 56000k , after taking these effects into account .
however , their work was carried out using a model grid with a fixed value of the surface gravity ( log g=7.5 ) . the more complete analysis of barstow _
_ ( 1998 ) spans the log g range from 7.0 to 8.0 and yields a slightly lower temperature of 53720k , from a combined balmer and lyman line analysis . for this study
we adopt = 54000k and log g=7.5 in the model calculations .
the homogeneous non - lte heavy element rich calculations used here originate in the work of lanz _
et al . _ ( 1996 ) and barstow _
et al . _ ( 1998 ) and have been described extensively in those papers .
briefly , the models include a total of 26 ions of h , he , c , n , o , si , fe and ni in calculations with the programme ( hubeny 1988 ; hubeny & lanz 1992 , 1995 ) .
radiative data for the light elements have been extracted from topbase , the database of the opacity project ( cunto _ et al . _
1993 ) , except for extended models of carbon atoms ( k. werner , private communication ) . for iron and nickel , all the levels predicted by kurucz ( 1988 ) are included , taking into account the effect of over 9.4 million lines .
barstow _ et al .
_ ( 1998 ) computed a model grid over a range of from 52000k to 68000k and log g from 7.0 to 8.0 but only considered a single representative value of the fe abundance .
these calculations have also been extended to deal with helium stratification by barstow & hubeny ( 1998 ) . as part of our continuing study of g191@xmath0b2b - like hot da white dwarfs
we have extended the stratified computations to match the grid of barstow _
et al . _ ( 1998 ) in and log g while enlarging the range of fe abundances considered ( see table [ mgrid ] ) .
.physical parameters adopted for model calculations with a homogeneous mixture of fe [ cols= " < , < " , ] [ mgrid ] in addition to the calculation of the intrinsic stellar euv spectrum , it is necessary to take account of the effect of the intervening interstellar medium .
the basic model that deals with , and opacity is now well - established ( rumph bowyer & vennes 1994 ) and we apply the modifications described by dupuis _
et al . _ ( 1995 ) to treat the converging line series near the and edges .
the analysis technique used to compare models and data has been described extensively in several earlier papers ( e.g. lanz _ et al . _ 1996
; barstow _ et al . _ 1997a , b etc . ) .
hence , we just give a brief resum here .
we utilise the programme to fold model spectra through the _ euve _ instrument response , taking into account the overlapping higher spectral orders in the instrumental effective area and applying the long wavelength corrections described by dupuis _
et al . _ ( 1995 ) .
goodness of fit is determined using a @xmath9 statistic and the best agreement between model and data is achieved by seeking to minimise the value of that parameter . while visually , good agreement can be achieved between model and data disagreements in some details of the line strengths often lead to high values of the reduced @xmath9 ( @xmath10 , where @xmath11 is the number of degrees of freedom ) . formally , a good fit should have @xmath12 or less and if it is much greater than this , estimates on the parameter uncertainties can not be evaluated from the change in @xmath9 according the usual values of @xmath13 ( e.g. @xmath14 for @xmath15 uncertainty and 5 degrees of freedom , see press _ et al . _
1992 ) . an alternative way to estimate such uncertainties is to make use of the f test , which calculates the significance of differences in the values of @xmath9 determined from separate fits .
the f parameter is simply the ratio of the two values of @xmath16 .
the significance of its value depends on the number of degrees of freedom and can be determined from standard tables .
this can be used to determine whether or not one model might provide a significantly better fit than another .
uncertainties are then estimated by tracking the value of f as an individual parameter is varied until it reaches a predetermined value corresponding to a particular significance . the work of lanz _
et al . _ ( 1996 ) and barstow & hubeny ( 1998 ) has been very successful in explaining the spectrum of g191@xmath0b2b at wavelengths longward of @xmath17 . however , taking the models that best match that spectral region and extending the comparison to shorter wavelengths reveals a significant discrepancy , with the observed flux falling well below that predicted ( figure [ f1 ] ) .
the largest difference is nearly an order of magnitude .
closer inspection of figure [ f1 ] reveals that , while the onset of the disagreement is quite sudden , at @xmath1 , the fit is not perfect at the longer wavelengths .
agreement is very good above 260 , but there are some significant differences between model and data between 190 and 260 .
interestingly , the match of the mw and lw data to the model can be much improved by restricting further the wavelength range under consideration .
for example , ignoring the data shortward of 210 delivers a large improvement in the fit , changing the value of @xmath9 by more than a factor 2 , from 4827 ( 530 degrees of freedom ) to 2064 ( 491 dof ) .
applying the f - test to these values shows that the improvement is hugely significant .
furthermore , this is coupled with a reduction in the required fe abundance to fe / h@xmath18 .
the quality of agreement with the long wavelength data , already very good with the current generation of models , remains more or less unchanged .
therefore , nearly all the improvement in the fit lies between 210 and @xmath19 , as illustrated in figure [ f2 ] . here
, we highlight the most interesting spectral range , between 100 and 300 , below which the source is barely detectable with _ euve_. as expected from the above discussion , the agreement between model and data is excellent above 210 , but there is an increasing discrepancy to shorter wavelengths .
the difference between the model prediction and observation ( see figure [ f2 ] ) is characterised mainly by a difference in slope .
a clue as to the possible explanation of the short wavelength flux discrepancy can be found by considering the region below 180 separately from the rest of the spectrum .
an improved match to the observed flux level is obtained by allowing the fe abundance to vary freely , also shown in figure [ f2 ] , after fixing the interstellar columns at the values determined from the longer wavelength fit . however , this exercise yields a fe / h ratio of @xmath20 , much higher than either the values obtained in the analyses of lanz _
et al . _ ( 1996 ) and barstow & hubeny ( 1998 ) or with the more restricted short wavelength limit considered here .
figure [ f3 ] shows the mass depth ( @xmath21 , total mass above the point of interest ) of the euv line and continuum at monochromatic optical depth @xmath22 as a function of wavelength computed for the model that gives the best overall fit to the data .
it can be seen that there is a steep change in the depth at which both the continuum and lines are formed at @xmath17 , just where short wavelength discrepancy begins ( see figures [ f1 ] and [ f2 ] ) .
the sharp decrease in the depth of the continuum formation between 190 and 160 is a combined result of a number of intervening continuum edges , mostly of fev ( @xmath23 165.5 , 173.1 , 176.1 , 180.2 , 184.3 ) , and partly niv ( @xmath23 163.2 , 171.3 , 174.9 , 177.0 , 180.0 ) .
there are also edges of light ions , namely civ ground - state ( 192.2 ) , niv ( 160.1 , 179.4 ) and oiv ( 160.3 , 181.1 ) , which are less important than the fev features .
coupling this with the larger fe abundance required to match the short wavelength data leads us to suggest that the fe is not homogeneously mixed in the atmosphere but has a depth dependent abundance .
we examine this possibility in the rest of this paper .
barstow & hubeny ( 1998 ) have already reported on the effects of helium stratification in heavy element - rich model atmospheres .
they utilised the programme ( hubeny 1988 ; hubeny & lanz 1992 , 1995 ; lanz _ et al . _
1996 ) including modifications to allow the abundance of any element at any depth point within an atmosphere to be a completely free parameter .
however , to generate a physically realistic model it would be necessary to carry out diffusion calculations for every element . in the case of helium , this is comparatively straightforward . as pointed out by vennes _
et al . _ ( 1988 ) , the radiation pressure on helium is insufficient to counteract the downward force of gravity to a degree yielding a photospheric helium abundance large enough to explain the observed euv / soft x - ray flux deficiency , due to the comparatively small number of lines in the euv waveband .
hence , vennes _
et al . _ ( 1988 ) proposed a model where the the relative abundance of he is determined by the equilibrium between ordinary diffusion and gravitational settling and depends on effective temperature , surface gravity and the mass of the overlying h layer .
barstow & hubeny ( 1998 ) adopted the approach of vennes _
et al . _ ( 1988 ) to calculate the depth dependent abundance profile of he for their stratified models and this is also the case in the models used in this work . for heavier elements , the number of lines found in the euv is much greater than for helium and radiative forces become important . in these circumstances ,
simple diffusive equilibrium is no longer an adequate treatment of the relative abundances of the elements .
detailed models dealing with the effects of radiative levitation have been constructed by several workers .
most recently , chayer , fontaine & wesmael ( 1995a ) and chayer _
et al . _ ( 1994 ) have carried out calculations that include all the heavy elements so far found in the atmosphere of g191@xmath0b2b .
unfortunately , for our purposes , the range of mass depth considered by them , running from @xmath24 upto @xmath25 , correponds to a region of the envelope below that of the line / continuum formation depths in the models .
we note that chayer _
et al . _ ( 1994 ) consider the fractional mass depth @xmath26 , whereas we deal with the total mass @xmath21 , independent of the mass of the star .
however , it is a simple matter to convert our scale to theirs by dividing @xmath21 by the known mass of g191@xmath0b2b ( @xmath27 , marsh _
et al . _
hence , with line / continuum formation depths above @xmath28 ( corresponding to @xmath29 ) , there is little information in the chayer _
et al . _ ( 1994 , 1995a ) work that could be usefully incorporated into the models .
since , we have no a priori information on the possible depth dependent abundance of fe , either from observation or a detailed physical model , we have taken two somewhat arbitrary approaches in an attempt to produce one or more atmosphere models that can match the complete _ euve _ spectrum of g191@xmath0b2b .
first , we looked at a series of simple ` slab ' models , where we divide the atmosphere into two or three discrete regions and fix the fe abundance at a constant value within these depth ranges but allow it to vary from region to region .
as starting points we used the nominal fe abundances determined from the best fit models to the short wavelength and medium wavelength _ euve _
spectra and placed the slab divisions near the continuum formation depth .
we consider models with two and three layers , but it is important to note that the resulting discontinuities are unphysical and that this choice is only a rough attempt to mimic the true depth dependence of the fe abundance that might be expected from diffusion equilibrium .
table [ slg ] lists the details of the slab models , giving fe abundances and depth ranges , the quoted depth representing the lower limit of the given abundance for each layer .
we note that the abundances of all other elements were constant , as specified in barstow & hubeny ( 1998 ; c / h@xmath30 , n / h@xmath31 , o / h@xmath32 , si / h@xmath33 , ni / h@xmath34 ) .
lcccccc model index & depth top zone & fe / h & depth ( optional ) mid zone & fe / h & depth deep zone & fe / h + fe1 & @xmath35 & @xmath6 & @xmath36 & @xmath3 & @xmath37 & @xmath38 + fe2 & @xmath35 & @xmath6 & & & @xmath37 & @xmath39 + fe3 & @xmath35 & @xmath40 & & & @xmath37 & @xmath6 + fe4 & @xmath35 & @xmath40 & & & @xmath37 & @xmath20 + fe5 & @xmath35 & @xmath39 & & & @xmath37 & @xmath6 + fe6 & @xmath35 & @xmath39 & & & @xmath37 & @xmath20 + fe9 & @xmath35 & @xmath41 & & & @xmath37 & @xmath42 + fe10 & @xmath35 & @xmath43 & & & @xmath37 & @xmath20 + fe11 & @xmath35 & @xmath43 & & & @xmath37 & @xmath44 + fe12 & @xmath45 & @xmath43 & @xmath46 & @xmath40 & @xmath37 & @xmath42 + fe13 & @xmath45 & @xmath43 & @xmath36 & @xmath40 & @xmath37 & @xmath42 + fe14 & @xmath45 & @xmath43 & @xmath47 & @xmath40 & @xmath37 & @xmath42 + + diffusion / levitation & @xmath48 & log @xmath49 & reservoir fe / h + fe7 & @xmath50 & 7.4 & @xmath51 + fe8 & @xmath52 & 7.45 & @xmath51 + [ slg ] a second approach to modelling the fe abundance was to modify the diffusion calculation for gravitational settling of helium to deal with fe ( see vennes _
et al . _ 1988 ;
barstow & hubeny 1998 ) .
this is straightforward , since the atomic mass and effective charge are free parameters in the calulation and can be adjusted .
however , in dealing with helium , it is realistic to assume that there is an infinite reservoir overlaid by a thin h shell , a useful boundary condition .
as fe is not a product of nuclear burning in a white dwarf of normal mass , such as g191@xmath53b2b , it can only exist as a trace element and the fe reservoir can not be specified in the same way . in this case
it is necessary to assume that there is some limiting fe abundance in the deeper layers of the atmosphere .
irrespective of any assumptions about this limiting abundance , what is clear from a range of test diffusion calulations for fe is that the rate of change of the abundance profile is very steep .
that is , for a sensible abundance of fe at the continuum formation depth , the outer layers are completely depleted of fe while the fe abundance in the deeper regions is so high that the emergent euv flux is negligible .
this is not too suprising , as the atomic mass of fe is fourteen times that of he , and highlights the already well documented fact that radiative levitation effects are necessary to explain the presence of photospheric fe .
we have not developed a complete radiative levitation calculation for this work but the possible effects can be examined , to first order , by applying a reverse acceleration term in the diffusive equilibrium calculation .
examination of radiative acceleration predictions for fe ( see figure 12 of chayer _
et al . _
1995a ) shows that the value of @xmath49 reaches a plateau - like maximum near the stellar surface .
consequently , we can adopt a constant value for the radiative acceleration term in our calculations .
it is important to stress that our choice of log @xmath49 and the reservoir fe abundance are entirely empirical , based on achieving fe abundances that approximately correspond to those included in the slab models and which also match the range observed in the homogeneous analysis of section 3 ( see table [ slg ] ) .
in principle , a detailed study of any individual star requires a grid of models to be calculated spanning the possible range of values of all parameters . in practice , this is not feasible for a star like g191@xmath0b2b because of the number of free parameters that must be considered .
a heavy element - rich atmospheric model is specified by and log g , plus the abundance of each element included in addition to hydrogen - seven in the models used here . for g191@xmath0b2b and similar stars , determination of the values of all the parameters is a multi - stage process .
temperature and gravity are estimated from the balmer lines while abundances can be determined from an analysis of the far uv absorption line strengths .
abundance determinations are typically iterative , with initial estimates made using a first guess at the composition refined by recalculation of a fully converged model .
we now know that , to achieve a completely consistent interpretation of the data , it is also necessary to redetermine and log g in the light of the measured heavy element abundances ( barstow _ et al . _ 1998 ) .
most analyses performed have made the reasonable simplifying assumption that the stellar envelopes are homogeneous .
once it is necessary to consider depth dependent abundances , the number of possible variables increases dramatically , since it is then necessary to specify the element abundances at each depth point . since , the models constructed for this work have 70 depth points , this could mean that it is necessary to deal with @xmath54 times the number of variables handled in the homogeneous work unless the problem is restricted in some sensible way . the dominant opacity source in the euv
is fe , therefore , we have chosen to confine the analysis to the study of fe stratification and assume that all other heavy elements are homogeneously mixed , with the exception of helium which is also stratified as described above and by barstow & hubeny ( 1998 ) . even dealing with a single element it is necessary to define the abundance at 70 depth points and the problem can be reduced further by dividing the atmosphere into a smaller number of discrete regions or slabs or trying to specify a smooth abundance profile with a smaller number of diffusion / levitation parameters , as described in section 3 above .
even so , there is no single variable that can be adjusted to yield the desired result in terms of matching both short and longer wavelength _ euve _ spectra at the same time .
however , from the discussion of the short wavelength problem ( section 2 ) , it is at least possible to express the goal as one of steepening the short wavelength region of the spectrum below @xmath55 compared to the longward flux level . taking as a starting point the fe abundances determined from separate fits to short and long wavelength euv spectra ,
the model grid listed in table [ slg ] has been constructed in an incremental way , in response to the results of a fit to an earlier model . in each new model , only one or two small changes were made from the previous example , devised to bring the predicted spectrum into closer agreement with that observed ( but not always successfully ) .
we consider the results of all these analyses together in table [ fits ] , together with the best - fit homogeneous model , for reference .
the probability that the best fit model ( fe6 ) is a significant improvement on each of the other models is calculated and listed in table [ fits ] . in each case , the entire useful euv spectrum of g191@xmath0b2b from 120 to 600 was considered .
the interstellar column densities were allowed to vary completely freely while h layer mass ( for h and he stratification ) , and log g were fixed at their nominal values of @xmath56 , 54000k and 7.5 respectively .
lccccc model index & ( @xmath57 ) & ( @xmath58 ) & ( @xmath59 ) & @xmath16 & f test probability ( % ) + homogeneous & 2.15 & 2.18 & 1.18 & 21.0 & @xmath60 + fe1 & 2.02 & 1.76 & 7.78 & 28.2 & @xmath60 + fe2 & 2.09 & 1.84 & 5.20 & 45.4 & @xmath60 + fe3 & 1.96 & 1.76 & 4.92 & 15.8 ( 1.62 ) & 89.5 ( 95 ) + fe4 & 2.09 & 2.06 & 2.21 & 14.3 ( 1.75 ) & 31.5 ( 97 ) + fe5 & 1.92 & 1.67 & 5.52 & 17.0 ( 1.02 ) & 98.6 ( 42 ) + fe6 & 2.05 & 2.00 & 2.57 & 13.8 ( 1.00 ) & ( 38 ) + fe9 & 2.08 & 2.11 & 1.22 & 15.7 ( 0.89 ) & 87.6 ( 10 ) + fe10 & 2.06 & 2.11 & 1.51 & 15.6 ( 0.85 ) & 84.8 ( ) + fe11 & 2.14 & 2.30 & 0.00 & 20.3 & @xmath60 + fe12 & 1.73 & 1.39 & 16.2 & 34.0 & @xmath60 + fe13 & 1.76 & 1.44 & 7.86 & 23.8 & @xmath60 + fe14 & 1.87 & 1.65 & 4.46 & 17.0 ( 1.32 ) & 98.7 ( 82 ) + + fe7 & 1.53 & 1.16 & 15.1 & 28.3 & @xmath60 + fe8 & 2.40 & 2.76 & 0.00 & 69.8 & @xmath60 + [ fits ] most of the stratified models offer a significant improvement over that homogeneous model which gives the best match to the euv spectrum of g191@xmath0b2b .
however , some are rather more successful than others .
we examined two different types of slab model , first where the upper layers of the photosphere have the greater fe abundance and second with this situation reversed .
model fe1 , which falls into the former category gives a worse agreement overall , in comparison with the homogeneous case ( figure [ f4 ] ) .
the predicted mw flux clearly falls below the observed level while there is an opposite discrepancy in the sw range .
those slab models where the deeper fe abundance is greater than that in the outer layers give the best agreement in most cases .
the very best of these , model fe6 , is a good match to the observed spectrum throughout the complete euv range ( figure [ f5 ] ) .
any residual differences are of similar magnitude to those obtained with the homogeneous models when those deal with the more restricted wavelength range above @xmath7 .
neither of the two models which have fe abundance profiles determined from the balance of radiative levitation and gravitational settling are in good agreement with the observations .
for example , the better of these , fe7 , requires a very high column density to force agreement with the short wavelength flux level .
this leaves an unacceptably high flux decrement at and below the 228 lyman series limit ( figure [ f6 ] ) .
thus far , the analysis has only addressed the level of agreement between model predictions and observations in the euv spectral range .
however , as the fe abundances are adjusted at different depths to force agreement with the euv observations , it is important to consider the effect this has on the predicted fe line strengths in the far uv . at this point
, we can discard those stratified models which do not work and limit the analysis to the selected few that do .
using the f - test , we define this group by assessing which model fits have values of @xmath16 which indicate that the probability the fit is significantly worse than the best model ( fe6 ) lies below 99% ( see table [ fits ] ) .
the models included are fe3 , fe4 , fe5 , fe6 , fe9 , fe10 and fe14 .
figure [ f7 ] shows the a region of the _ iue _ newsips coadded spectrum of g191@xmath0b2b spanning the range @xmath61 to 1380 and including several of the strongest fev lines .
also shown , in decreasing order of predicted line strength are synthetic spectra computed for models fe4 , fe3 , fe6 , fe9 and fe14 .
models fe4 and fe3 are a very good match , with the fe6 and fe9 fev line strengths being slightly weaker than observed . to assess quantitatively the level of agreement , we have carried out a further analysis , fitting the models predictions to the two strongest fev lines seen in figure [ f7 ] ( at 1373.8 and 1376.5 ) .
the resulting values of @xmath16 are listed ( in brackets ) in table [ fits ] , showing that all models are in good agreement with the data and that none can be excluded on the basis of the f test . while model fe6 is the best in the euv range , fe10 , which is not significantly different gives the best match to the far - uv fev lines
we have been able to provide the first consistent explanation of the complete spectral flux distribution of the hot da white dwarf g191@xmath0b2b , including the short wavelength euv region ( @xmath62 ) which has previously been problematic , using a grid of atmosphere models with a depth dependent abundance of fe .
potentially , this is an important breakthrough in our understanding of this and related das with significant photospheric heavy element abundances .
however , the choice of model structures is somewhat arbitrary , even artificial , in the absence of any a priori physical constraints that we might apply .
hence , it is necessary to re - examine the justification for this approach , question its physical reality and assess the uniqueness of the models before consideration of the possible implications of the results . in comparison with the homogeneous atmosphere models used in the earlier studies of g191@xmath0b2b ( see figure [ f1 ] ) , the stratified models described here seek to suppress the level of the short wavelength ( @xmath62 ) flux compared to the longer wavelengths , also steepening the spectral slope towards short wavelengths in this region .
a number of the models tried were successful , but there may be other mechanisms that might yield a similar effect .
probably the most important question concerns the atomic data included in the models .
it is well - known that there are large uncertainties in the continuum and line opacities of the fe group elements . as the dominant euv opacity source , the fe data will be the most critical
. however , there are two arguments againts this possibility .
first , it seems unreasonable to suppose that any errors should be concentrated in a particular wavelength range .
second , where the greatest uncertainties lie , below @xmath63 , there are very few fe lines ( or of any other species ) .
this might suggest that , in fact , there should be less problem in the short wavelength region .
however , we can not ignore this potential problem completely .
bautista ( 1996 ) and bautista and pradhan ( 1997 ) have computed photoionization cross - section and oscillator strengths for feiv and fev , expanding the opacity project database by and appreciable number of new transitions .
their calculations take account of transitions between states lying below the first ionization threshold and states lying above . as bautista and pradhan note
, these transitions can make an important contribution to the total photoabsorption , even though they do not appear as resonances in the photoionisation cross - sections .
an alternative explanation for the short wavelength flux decrement is that extra euv opacity may be present in the form of species , different elements or ionization stages , that have not been taken into account in the models .
again , this seems unlikely , as the few detected elements that we have not included ( s and p ) are only present in traces too small to have any noticeable effet in the euv ( vennes _ et al . _
any elements not yet detected will have still lower abundances .
the choice of stratified models is really determined by the level of complexity that can be accomodated in the modelling process and by the availability of information to determine the abundance depth profile of a given element .
the two layer slab models calculated , therefore , represent the simplest possible development of the idea of a depth dependent abundance . extending this to deal with three layers ( models fe1 , fe12 , fe13 , fe14 )
is a modest increase in complexity , adding two further unknowns ( position of the lower layer boundary and the layer abundance ) to the three required for two layers . in this work ,
the additional , narrow layer ( 2 - 4 depth points ) is used to make a smoother transition between the two main slabs .
generation of a smooth abundance profile is also a result of the diffusion / levitation calculations included in models fe7 and fe8 .
it is rather striking that it is the simplest two region slab models that give the best agreement with the data .
of the three layer models , only fe14 , where the intermediate layer is reduced to one depth point is in reasonable agreement and the diffusion / levitation models are the worst as a group .
one conclusion we could draw from this is that the transition region for the change in fe abundance is indeed narrow .
in addition , the evidence seems to indicate that the fe abundance is greatest in the deeper layers of the atmosphere but with the abundance in the outer regions being finite and the contrast between the two values being limited to a factor @xmath64 .
for example , the best fit fe6 model ( to the euv data ) has a layer abundance ratio of 40 , and predicts far uv fe line strengths that are consistent with the observed spectrum . in comparison ,
the fe14 model ( layer abundance ratio 60 ) , which also gives a good match in the euv , yields far uv fe line strengths that are much weaker than observed .
this also indicates that the abundance of fe in the outer layer can not be less than @xmath65 .
it is interesting to note that models fe9 and fe10 , which are not significantly worse than the best model ( fe6 ) , have a lower column than any of the other successful models .
the resulting he ionization fractions , 37% and 42% respectively , are much closer to the 27% mean value typical of the local ism ( barstow _ et al . _
1997 ) , compared with either the homogeneous analysis of lanz _
et al . _ ( 1996 ; @xmath66% ) , the stratified h / he work of barstow & hubeny ( 1998 ; @xmath64% ) or the good models considered here ( 52% for fe6 ) . on the basis of achieving the lowest he ionization fraction we might favour model fe9 over the others
. however , we must be cautious in taking any of these results too literally , since there are an enormous number of possible , more complex , abundance profiles that we have not yet tested and which might give an equally good or better result .
furthermore , we have only examined the stratification of fe .
there is no reason to assume that this is the only element that might be stratified .
indeed , there is good evidence to show that nitrogen is stratified in the slightly cooler , less heavy element - rich , da white dwarf rej1032 + 532 ( holberg _ et al . _ in preparation ) .
ideally , we should investigate all these possibilities with new calculations but there is a major problem in producing the large number of models needed .
we must seek ways of confining the problem .
it may be possible to provide some constraints by measuring abundances using lines that are formed at different depths within the envelope .
however , this approach will be particularly sensitive to any uncertainties in the oscillator strengths .
furthermore , as any effects are likely to be quite subtle and the range of uncertainty in determining abundances from far uv line strengths is typically a factor 2 , it will be necessary to obtain data of considerably higher signal - to - noise than that currently available . even then , as examination of the euv and far uv line formation depths shows ( figure [ f3 ] ) , it will only be possible to investigate a narrow region of the photosphere , occupying 1 dex in mass depth .
interestingly , there is much more contrast in line formation depth within the euv and between the euv and far uv than in the far uv range alone .
if we accept at face value the evidence this work presents , that the fe in g191@xmath0b2b is stratified in two main layers with abundances of @xmath67 ( lower layer ) and @xmath68 ( upper layer ) , it is interesting to explore the possible implications .
a relative depletion of fe in the outer layers of the envelope may be an indication of ongoing mass - loss in the star .
the effects of mass - loss in white dwarf atmospheres has hardly been examined , although there is evidence for this occuring in the hot do white dwarf rej0503@xmath0289 ( barstow & sion , 1994 ) . a preliminary study of this problem by chayer _
et al . _ ( 1993 ) shows that the outer layers of the envelope will become depleted over time .
however , the mass loss rate used in the calculation ( @xmath69/yr ) was sufficient to eliminate the reservoir of the heavy elements completely within a few thousand years . on this basis , to see any heavy elements at all , the mass loss rate in g191@xmath0b2b must be considerably lower .
there is clearly a need for new radiative levitation calculations coupled with mass - loss to evaluate this problem properly .
we have demonstrated that the complete spectrum of g191@xmath0b2b can be explained by a model atmosphere where fe is stratified , with increasing abundance at greater depth .
the abundance profile appears to be sharply stepped and may explain the difficulties in matching observed photospheric abundances , usually obtained by analyses utilising homogeneous model atmospheres , to the detailed radiative levitation predictions . particularly as the latter are only strictly valid for regions deeper than where the euv / far uv lines and continua are formed .
et al . _ ( 1993 ) show that the outer layers of the envelope will become depleted over time if a weak wind is present . hence ,
if found to be the only explanation of the observed spectrum , the relative depletion of fe in the outer layers of the atmosphere could be the first evidence for radiatively driven mass loss in the star .
in addition , the work presented here may contribute to the resolution of the issue of the possible presence of along the line of sight to the star , discussed by lanz _
et al . _ ( 1996 ) and barstow & hubeny ( 1998 ) , and its likely location , in the photosphere or ism .
we find that two of our best stratified models yield an column density and he ionization fraction closer to the local ism values than results obtained in the earlier studies .
however , this particular problem will only be completely solved when a much higher resolution spectrum is obtained , capable of separating lines from those of heavier elements .
we anticipate that the j - pex spectrometer will provide such data in early 1999 ( bannister _ et al . _ 1999 ) . through its ability to study individual lines , this instrument may also be able to deliver new information on the stratification of fe and other elements .
the work of mab was supported by pparc , uk , through an advanced fellowship .
jbh wishes to acknowledge nasa grants nag 5 - 2738 and nag 5 - 3472 .
data analysis and interpretation were performed using noao , nasa heasarc and starlink software . bannister n.p . , cruddace , r.g .
, kowalski , m. , fritz , g. , barstow m.a . , fraser g.w . ,
barbee , t. , lapington j.s .
, tandy j. , 1999 , in the proceedings of the 11th european workshop on white dwarfs , ed .
solheim , asp conference series , in press barstow m.a .
_ et al . _ , 1993 ,
mnras , 264 , 16 barstow m.a . , dobbie , p.d .
, holberg j.b . , hubeny i. , lanz t. , 1997a , mnras , 286 , 58 barstow m.a . ,
holberg j.b . , hubeny i. , lanz t. , 1997b , in white dwarfs , eds .
j. isern , m. hernanz and e. garcia - berro , kluwer , in press barstow m.a . , holberg j.b . , koester d. , 1995 , mnras , 274 , l31 barstow m.a . , hubeny i. , 1998 , mnras , 299 , 379 barstow m.a . , hubeny i. , holberg j.b . , 1998 , mnras , in press barstow m.a
, hubeny i. , lanz t. , holberg j.b .
, sion e.m .
, 1996 , in astrophysics in the extreme ultraviolet , eds .
s. bowyer and r.f .
malina , kluwer , 203 barstow m.a .
sion e.m . , 1994 , mnras , 271 , l52 bautista m.a . , 1996 , a & ass , 199 , 105 bautista m.a . , pradhan a.k . , 1997 , a & ass , 126 , 365 bruhweiler f.c . , kondo y. 1981 , apj , 248 , l123 bruhweiler f.c .
, kondo y. 1983 , apj , 269 , 657 chayer p. , leblanc f. , fontaine g. , wesemael f. , michaud g. , vennes s. , 1994 , apj , 436 , l161 chayer p. , fontaine g. , wesemael f. , 1995a , apjs , 99 , 189 chayer p. , pelletier c. , fontaine g. , wesemael f. , 1993 , in ` white dwarfs : advances in observation and theory ' , ed m.a .
barstow , kluwer , dordrecht , 261 .
chayer p. , vennes s. , dupuis j. , pradhan a.k .
, 1997 , in ` white dwarfs ' , eds . j. isern , m. hernanz and e. garcia - berro , kluwer , dordrecht , 273 chayer p. , vennes s. , pradhan a.k . , thejll p. , beauchamp a. , fontaine g. , wesemael f. , 1995b , apj , 454 , 429 cunto w. , mendoza c. , ochsenbein f. , zeippen c.j .
, 1993 , a & a , 275 , l5 dupuis j. , vennes s. , bowyer s. , pradhan a.k . , thejll p. , 1995 ,
apj , 455 , 574 koester d. , wolff b. , jordan s. , dreizler s. , 1997 , in proceedings of the 3rd conference on faint blue stars " , in press .
holberg j.b .
, hubeny i. , barstow m.a .
, lanz t. , sion e.m .
, tweedy r.w . , 1994 ,
apj , 425 , l105 hubeny i. , 1988 , comp.phys.comm . , 52 , 103 hubeny i. , lanz t. , 1992 , a & a , 262 , 501 hubeny i. , lanz t. , 1995 , apj , 439 , 875 kurucz r.l . , 1988 , in iau trans . , ed .
m. mcnally , volxxb , kluwer , dordrecht , 168 lanz t. , barstow m.a . , hubeny i. , holberg j.b . , 1996 ,
apj , 473 , 1089 marsh , m.c .
_ et al . _ , 1997a , mnras , 286 , 369 marsh , m.c .
_ et al . _ , 1997b , mnras , 287 , 705 press w.h . , teulosky s.a . , vetterling w.t . , flannery b.p . , 1992 , numerical recipes ( 2nd edition ) , p687ff , cambridge rumph t. , bowyer s. , vennes s. , 1994 , aj , 107 , 2108 sion , e.m . ,
bohlin , r.c . ,
tweedy , r.w . , vauclair , g.p .
1992 , apj , 391 , l29 vennes s. , chayer p. , hurwitz m. , bowyer s. , 1996 , apj , 468 , 989 vennes s. , chayer p. , thorstensen j.r . , bowyer s. , shipman h.l .
, 1992 , apj , 392 , l27 vennes s. , pelletier c. , fontaine g. , wesmael f. , 1988 , apj , 331 , 876 wolff b. , koester d. , jordan s. , haas s. , 1998 , a & a , 329 , 1045 | the presence of heavy elements in the atmospheres of the hottest h - rich da white dwarfs has been the subject of considerable interest .
while theoretical calculations can demonstrate that radiative forces , counteracting the effects of gravitational settling , can explain the detections of individual species , the predicted abundances do not accord well with observation .
however , accurate abundance measurements can only be based on a thorough understanding of the physical structure of the white dwarf photospheres , which has proved elusive .
recently , the availability of new non - lte model atmospheres with improved atomic data has allowed self - consistent analysis of the euv , far uv and optical spectra of the prototypical object g191@xmath0b2b .
even so , the predicted and observed stellar fluxes remain in serious disagreement at the shortest wavelengths ( below @xmath1 ) , while the inferred abundances remain largely unaltered .
we show here that the complete spectrum of g191@xmath0b2b can be explained by a model atmosphere where fe is stratified , with increasing abundance at greater depth .
this abundance profile may explain the difficulties in matching observed photospheric abundances , usually obtained by analyses using homogeneous model atmospheres , to the detailed radiative levitation predictions . particularly as the latter are only strictly valid for regions deeper than where the euv / far uv lines and continua are formed .
furthermore , the relative depletion of fe in the outer layers of the atmosphere may be evidence for radiatively driven mass loss in g191@xmath0b2b .
[ firstpage ] stars : abundances stars : atmospheres stars : white dwarfs ultraviolet : stars x - rays : stars . |
although radiation represents a negligible fraction of the energy in core - collapse supernova ( sn ) explosions , most of what we know about these events is inferred from the analysis of spectra and light curves .
the gravitational binding - energy of the newly - formed protoneutron star is on the order of 10@xmath1erg . within milliseconds after
core bounce , a few @xmath2 10@xmath3erg of this energy is used to photodissociate the infalling outer iron core . starting with an electron - neutrino burst when the core reaches nuclear densities , the radiation of neutrinos of all flavors operates over a few tens of seconds after the bounce as the protoneutron star cools , and carries away a few @xmath2 10@xmath3erg .
these neutrinos are believed to deposit @xmath410@xmath5erg ( 1% of the total ) into the infalling progenitor mantle , reversing the accretion , and to provide the internal and kinetic energy for the sn ejecta ( woosley & janka 2005 ) . only @xmath410@xmath6erg ( 0.01% of the total ) is eventually processed into light , that gets radiated with a rate equivalent to a few @xmath2 10@xmath7 sustained for three months .
depending on the progenitor radius , the shock emerges at the surface between a few hours and a day after core bounce , with a radiative precursor that is expected to heat and accelerate the surface layers , producing a soft x - ray flash ( chevalier 1982 ; ensman & burrows 1992 ; matzner & mckee 1999 ; blinnikov et al .
2000 ) . cooling , due to adiabatic expansion and radiative losses , is moderated as time progresses by recombination of hydrogen during the photospheric phase of type ii sn and by non - thermal excitation by unstable isotopes .
although uv emission is significant for the first few days after shock breakout , the spectral energy distribution ( sed ) subsequently peaks further and further into the red , first in the visual and then in the near - ir , a few months after explosion ( kirshner et al .
1973,1975 ; mitchell et al .
2002 ; leonard et al . 2002a
; brown et al . 2007 ) .
core - collapse sn explosions constitute an excellent laboratory for inferring the properties of massive stars at the end of their lives and , indirectly , their pre - sn evolution . as the material expands , the photosphere recedes to deeper and deeper layers of the star s former envelope , allowing the observer to probe all the way from the surface at shock breakout to the innermost layers just outside the compact remnant .
the nebular phase , the epoch when the ejecta become optically thin throughout ( @xmath43 - 4 months after explosion for type ii sne ) , permits the inspection of the regions where the blast originated , offering a means to constrain the mechanism and the morphology of the explosion as well as the nucleosynthetic yields .
detailed quantitative analyzes focus on the early , photospheric , evolution ( hoflich 1988 ; eastman & kirshner 1989 ; schmutz et al .
1990 ; baron et al 1996 , 2000 , 2003 , 2004 , 2007 ; mitchell et al .
2002 ; dessart & hillier 2005a,2006ab ) , because the ejecta are best suited for `` standard '' model atmosphere computations : there is an optically - thick base where the radiation thermalizes ; there is no contribution from non - thermal excitation by isotopes like @xmath8ni and its daughter product @xmath8co ; electron scattering dominates the opacity ; and line and continuum formation is spatially confined .
advances in computer technology permit better handling of non - local - thermodynamic - equilibrium ( non - lte ) effects and the critical role played by metals is now systematically accounted for ( baron et al .
2004 ; dessart & hillier 2005a ) .
non - lte effects are important even at early times , for example , to reproduce the hei lines observed in optical spectra . here
, we present a quantitative spectroscopic analysis of the early photospheric - phase evolution of the type ii sne 2005cs in ngc 5194 and sn 2006bp in ngc 3953 , both with plateau - type light curves as shown in fig .
[ fig_lc_obs ] .
we employ the non - lte model atmosphere code cmfgen ( hillier & miller 1998 ; dessart & hillier 2005a ) , in its steady - state configuration , following a similar approach to dessart & hillier ( 2006a ; hereafter dh06 ) and their analysis of sn 1999em .
a novel component of our study is the use of _ swift _ ( gehrels et al . 2004 ) ultra - violet / optical telescope ( uvot ; roming et al . 2005 ) ultraviolet photometry that places important constraints on the sed in the uv , the early cooling of the photosphere , and the reddening .
we infer from our models that sn 2005cs and sn 2006bp , were discovered only @xmath42 days after explosion , the earliest detections after explosion since sn 1987a ( type ii peculiar ) and sn 1993j ( type iib ; see , e.g. , matheson et al .
2000 ) . as a result of this prompt discovery
, we have the opportunity to observe a bright uv spectrum and to see high - excitation lines that have not previously been observed in sn ii - p .
finally , the fine time sampling of our spectra during the first month , along with dense photometric monitoring allows us to make an accurate distance determination .
this is a good demonstration of the potential of type ii - p sne , when coupled with detailed models , to provide accurate extragalactic distances . unlike the standard approach that depend on the cepheid period - luminosity relation in the lmc and other nearby galaxies ,
our method is based on physical models fitted directly to observations , and the distances it provides are independent of all the lower rungs of the distance ladder .
we aim to assess the accuracy of distance determinations based on type ii - p sne , and to refine the value of the hubble constant @xmath9 , in the spirit of the earlier efforts of schmidt et al .
measurements of the total matter and baryonic content of the universe via cmb experiments or large - scale structure correlations depend on the square of @xmath9 ( spergel et al .
2007 ; eisenstein et al . 2005 ) . in concordance solutions ,
these uncertainties affect measurements of the dark energy density and its associated equation - of - state parameter @xmath10 ( garnavich et al .
1998 ; riess et al 2005 ; see macri et al .
2006 for an illustration of the effect an improved accuracy on @xmath9 has on @xmath10 ) .
a small improvement in our knowledge of @xmath9 will lead to a significant improvement in the understanding the energy content of the universe .
this paper is structured as follows . in the next section
, we present the photometric and spectroscopic datasets . in [ sect_mod_pres ] , we present the model atmosphere code cmfgen and describe our methods . in
[ sect_05cs ] , we present a quantitative spectroscopic analysis at 13 photospheric - phase epochs for sn 2005cs , and in [ sect_06bp ] , analyze 18 epochs for sn 2006bp . in [ sect_dist ] , we apply the expanding photosphere method ( epm ) as well as a direct - fitting method ( in the spirit of the spectral - fitting expanding atmosphere method , seam , of baron et al . 2000 ) to infer the distance and the explosion date for both sn events . in
[ sect_discussion ] , we discuss the implications of our results , and in [ sect_conclusion ] , we conclude .
lccccccccc 2005 - 06 - 30 & 552.4 & 13.48 @xmath0 0.05 & 14.36 @xmath0 0.05 & 14.48 @xmath0 0.02 & 14.46 @xmath0 0.04 & 14.44 @xmath0 0.04 & .. & .. & p06 + 2005 - 07 - 01 & 553.2 & .. & 14.42 @xmath0 0.02 & 14.42 @xmath0 0.02 & .. & .. & 14.68 @xmath0 0.03 & 14.34 @xmath0 0.03 & cfa + 2005 - 07 - 01 & 553.3 & 13.53 @xmath0 0.06 & 14.36 @xmath0 0.05 & 14.46 @xmath0 0.05 & 14.40 @xmath0 0.02 & 14.34 @xmath0 0.06 & .. & .. & p06 + 2005 - 07 - 02 & 554.2 & 13.50 @xmath0 0.02 & 14.47 @xmath0 0.01 & 14.41 @xmath0 0.01 & .. & .. & 14.64 @xmath0 0.02 & 14.32 @xmath0 0.02 & cfa + 2005 - 07 - 02 & 554.5 & 13.69 @xmath0 0.10 & 14.45 @xmath0 0.04 & 14.51 @xmath0 0.03 & 14.47 @xmath0 0.03 & 14.29 @xmath0 0.10 & .. & .. & p06 + 2005 - 07 - 03 & 555.2 & .. & 14.52 @xmath0 0.02 & 14.46 @xmath0 0.02 & .. & .. & .. & 14.35 @xmath0 0.02 & cfa + 2005 - 07 - 03 & 555.3 & .. & 14.45 @xmath0 0.02 & 14.45 @xmath0 0.02 & 14.30 @xmath0 0.02 & 14.37 @xmath0 0.05 & .. & .. & t06 + 2005 - 07 - 04 & 556.2 & .. & 14.55 @xmath0 0.01 & 14.52 @xmath0 0.02 & .. & .. & 14.66 @xmath0 0.02 & 14.36 @xmath0 0.02 & cfa + 2005 - 07 - 05 & 557.4 & .. & 14.52 @xmath0 0.04 & 14.54 @xmath0 0.04 & 14.42 @xmath0 0.05 & 14.29 @xmath0 0.08 & .. & .. & p06 + 2005 - 07 - 06 & 557.8 & 14.02 @xmath0 0.06 & 14.60 @xmath0 0.05 & 14.55 @xmath0 0.03 & 14.36 @xmath0 0.03 & 14.40 @xmath0 0.03 & .. & .. & p06 + 2005 - 07 - 06 & 558.2 & 13.87 @xmath0 0.02 & 14.61 @xmath0 0.02 & 14.52 @xmath0 0.02 & .. & .. & 14.60 @xmath0 0.03 & 14.33 @xmath0 0.03 & cfa + 2005 - 07 - 06 & 558.3 & .. & 14.57 @xmath0 0.02 & 14.54 @xmath0 0.02 & 14.34 @xmath0 0.02 & 14.29 @xmath0 0.03 & .. & .. & t06 + 2005 - 07 - 07 & 559.2 & .. & 14.58 @xmath0 0.03 & 14.53 @xmath0 0.01 & .. & .. & 14.59 @xmath0 0.02 & 14.33 @xmath0 0.01 & cfa + 2005 - 07 - 07 & 559.4 & .. & 14.65 @xmath0 0.04 & 14.56 @xmath0 0.09 & 14.41 @xmath0 0.09 & 14.35 @xmath0 0.11 & .. & .. & p06 + 2005 - 07 - 08 & 560.2 & .. & .. & 14.54 @xmath0 0.01 & .. & .. & .. & 14.23 @xmath0 0.01 & cfa + 2005 - 07 - 08 & 560.3 & .. & 14.72 @xmath0 0.01 & .. & .. & .. & .. & .. & cfa + 2005 - 07 - 09 & 561.3 & .. & 14.74 @xmath0 0.02 & 14.50 @xmath0 0.02 & 14.26 @xmath0 0.02 & 14.25 @xmath0 0.03 & .. & .. & t06 + 2005 - 07 - 10 & 562.2 & 14.64 @xmath0 0.03 & 14.84 @xmath0 0.02 & 14.52 @xmath0 0.02 & .. & .. & .. & .. & cfa + 2005 - 07 - 11 & 563.1 & .. & 14.93 @xmath0 0.01 & 14.50 @xmath0 0.02 & .. & .. & .. & .. & cfa + 2005 - 07 - 11 & 563.3 & .. & .. & 14.50 @xmath0 0.02 & .. & .. & .. & .. & t06 + 2005 - 07 - 11 & 563.4 & 14.84 @xmath0 0.05 & 14.90 @xmath0 0.04 & 14.54 @xmath0 0.04 & 14.37 @xmath0 0.02 & 14.24 @xmath0 0.05 & .. & .. & p06 + 2005 - 07 - 12 & 564.3 & .. & 14.92 @xmath0 0.02 & 14.50 @xmath0 0.02 & 14.32 @xmath0 0.02 & 14.20 @xmath0 0.03 & .. & .. & t06 + 2005 - 07 - 13 & 565.4 & .. & 15.08 @xmath0 0.05 & 14.65 @xmath0 0.04 & 14.42 @xmath0 0.04 & 14.29 @xmath0 0.06 & .. & .. & t06 + 2005 - 07 - 13 & 565.4 & 15.27 @xmath0 0.06 & 15.04 @xmath0 0.05 & 14.64 @xmath0 0.02 & 14.37 @xmath0 0.03 & 14.29 @xmath0 0.05 & .. & .. & p06 + 2005 - 07 - 14 & 566.3 & 15.36 @xmath0 0.06 & 15.19 @xmath0 0.02 & 14.63 @xmath0 0.02 & 14.39 @xmath0 0.02 & 14.32 @xmath0 0.03 & .. & .. & t06 + 2005 - 07 - 14 & 566.4 & 15.29 @xmath0 0.05 & 15.10 @xmath0 0.05 & 14.67 @xmath0 0.02 & 14.42 @xmath0 0.04 & 14.27 @xmath0 0.02 & .. & .. & p06 + 2005 - 07 - 15 & 567.3 & .. & 15.13 @xmath0 0.03 & 14.60 @xmath0 0.02 & 14.34 @xmath0 0.02 & 14.23 @xmath0 0.03 & .. & .. & t06 + 2005 - 07 - 17 & 569.4 & 15.92 @xmath0 0.07 & 15.31 @xmath0 0.05 & 14.67 @xmath0 0.03 & 14.40 @xmath0 0.03 & 14.27 @xmath0 0.04 & .. & .. & p06 + 2005 - 07 - 19 & 571.3 & .. & 15.40 @xmath0 0.02 & 14.72 @xmath0 0.02 & 14.43 @xmath0 0.02 & 14.27 @xmath0 0.03 & .. & .. & t06 + 2005 - 07 - 19 & 571.4 & .. & 15.34 @xmath0 0.02 & 14.72 @xmath0 0.02 & 14.38 @xmath0 0.04 & 14.22 @xmath0 0.04 & .. & .. & t06 + 2005 - 07 - 19 & 571.4 & .. & 15.36 @xmath0 0.12 & 14.72 @xmath0 0.05 & 14.45 @xmath0 0.09 & 14.26 @xmath0 0.07 & .. & .. & p06 + 2005 - 07 - 20 & 572.4 & .. & 15.39 @xmath0 0.07 & 14.71 @xmath0 0.03 & 14.45 @xmath0 0.03 & 14.27 @xmath0 0.03 & .. & .. & p06 + 2005 - 07 - 25 & 577.4 & .. & 15.46 @xmath0 0.09 & 14.72 @xmath0 0.04 & 14.36 @xmath0 0.06 & 14.24 @xmath0 0.06 & .. & .. & p06 + 2005 - 07 - 27 & 579.4 & .. & 15.60 @xmath0 0.07 & 14.73 @xmath0 0.04 & 14.39 @xmath0 0.04 & 14.16 @xmath0 0.04 & .. & .. & p06 + 2005 - 07 - 28 & 580.3 & .. & 15.69 @xmath0 0.05 & 14.73 @xmath0 0.03 & 14.42 @xmath0 0.03 & 14.20 @xmath0 0.04 & .. & .. & t06 + lcccccc 2006 - 04 - 10 & 836.27 & .. & 15.59 @xmath0 0.02 & 15.44 @xmath0 0.02 & 15.42 @xmath0 0.02 & 15.39 @xmath0 0.02 + 2006 - 04 - 11 & 837.20 & 14.97 @xmath0 0.03 & 15.47 @xmath0 0.02 & 15.31 @xmath0 0.02 & 15.24 @xmath0 0.02 & 15.23 @xmath0 0.02 + 2006 - 04 - 13 & 839.20 & 14.92 @xmath0 0.03 & 15.44 @xmath0 0.02 & 15.16 @xmath0 0.02 & 15.03 @xmath0 0.02 & 14.94 @xmath0 0.02 + 2006 - 04 - 14 & 840.23 & .. & 15.43 @xmath0 0.02 & 15.12 @xmath0 0.02 & 14.97 @xmath0 0.02 & 14.91 @xmath0 0.01 + 2006 - 04 - 17 & 843.24 & 15.04 @xmath0 0.03 & 15.46 @xmath0 0.02 & 15.11 @xmath0 0.02 & 14.94 @xmath0 0.02 & 14.87 @xmath0 0.02 + 2006 - 04 - 18 & 844.16 & 15.09 @xmath0 0.03 & 15.47 @xmath0 0.02 & 15.12 @xmath0 0.02 & 14.93 @xmath0 0.02 & 14.86 @xmath0 0.02 + 2006 - 04 - 20 & 846.17 & 15.20 @xmath0 0.02 & 15.52 @xmath0 0.02 & 15.15 @xmath0 0.02 & 14.93 @xmath0 0.02 & 14.87 @xmath0 0.02 + 2006 - 04 - 22 & 848.22 & 15.33 @xmath0 0.03 & 15.59 @xmath0 0.02 & 15.18 @xmath0 0.02 & 14.94 @xmath0 0.02 & 14.90 @xmath0 0.02 + 2006 - 04 - 23 & 849.23 & 15.42 @xmath0 0.03 & 15.63 @xmath0 0.02 & 15.19 @xmath0 0.02 & .. & 14.91 @xmath0 0.02 + 2006 - 05 - 02 & 858.27 & 16.44 @xmath0 0.03 & 16.01 @xmath0 0.02 & 15.22 @xmath0 0.02 & 14.93 @xmath0 0.02 & .. + 2006 - 05 - 03 & 859.19 & 16.55 @xmath0 0.04 & 16.07 @xmath0 0.02 & 15.24 @xmath0 0.02 & 14.93 @xmath0 0.02 & 14.89 @xmath0 0.02 + 2006 - 05 - 05 & 861.17 & 16.74 @xmath0 0.04 & 16.16 @xmath0 0.02 & 15.27 @xmath0 0.02 & 14.95 @xmath0 0.02 & 14.88 @xmath0 0.02 + 2006 - 05 - 06 & 862.16 & 16.79 @xmath0 0.05 & 16.21 @xmath0 0.02 & 15.26 @xmath0 0.02 & 14.96 @xmath0 0.02 & 14.88 @xmath0 0.02 + 2006 - 05 - 11 & 867.19 & 17.12 @xmath0 0.06 & 16.39 @xmath0 0.03 & 15.32 @xmath0 0.02 & 15.00 @xmath0 0.02 & 14.89 @xmath0 0.02 + 2006 - 05 - 14 & 870.20 & .. & 16.48 @xmath0 0.04 & 15.37 @xmath0 0.02 & 15.05 @xmath0 0.02 & 14.92 @xmath0 0.02 + 2006 - 05 - 30 & 886.23 & 18.36 @xmath0 0.07 & 16.85 @xmath0 0.02 & 15.52 @xmath0 0.02 & 15.11 @xmath0 0.02 & 14.98 @xmath0 0.02 + ccccccccc + 2005 - 06 - 30.23 & 551.73 & 4 & 1.1 & 1.31 & bd28 , bd17 & & 1200 & rh + 2005 - 07 - 01.27 & 552.77 & 5 & 5.0 & 1.53 & bd28 , bd17 & 4 & 900 & rh + 2005 - 07 - 02.23 & 553.73 & 6 & 3.1 & 1.34 & bd28 , bd26 & 2 & 1200 & rh + 2005 - 07 - 03.15 & 554.65 & 7 & 32.7 & 1.09 & f34 , bd33 & 2 & 720 & rh + 2005 - 07 - 04.16 & 555.66 & 8 & 85.4 & 1.11 & bd28,bd17 & 1 & 1200 & jg + 2005 - 07 - 04.18 & 555.68 & 8 & 78.4 & 1.15 & bd28,bd17 & 1 & 1200 & jg + 2005 - 07 - 05.18 & 556.68 & 9 & 8.9 & 1.16 & f34 , bd17 & 1 & 1200 & jg + 2005 - 07 - 06.17 & 557.67 & 10 & 5.2 & 1.15 & bd28 , bd17 & 1 & 1200 & jg + 2005 - 07 - 07.15 & 558.65 & 12 & 1.8 & 1.11 & bd28 , bd17 & 1 & 1200 & ef + 2005 - 07 - 09.19 & 560.69 & 13 & 0.7 & 1.24 & bd28 , bd17 & 1 - 2 & 1200 & pb + 2005 - 07 - 10.18 & 561.68 & 14 & 1.1 & 1.22 & bd28 , bd17 & 1 - 2 & 1200 & pb + 2005 - 07 - 11.18 & 562.68 & 15 & 2.9 & 1.23 & bd28 , bd17 & 1 - 2 & 1200 & mc + 2005 - 07 - 12.16 & 563.66 & 16 & 2.8 & 1.17 & bd28 , bd17 & 1 - 2 & 1200 & mc + 2005 - 07 - 28.17 & 579.67 & 32 & 1.0 & 1.38 & bd28 , bd17 & 1 - 2 & 1200 & pb + 2005 - 07 - 29.15 & 580.65 & 33 & 2.1 & 1.30 & bd28 , bd17 & 2 - 3 & 960 & mc + 2005 - 12 - 25.56 & 730.04 & 182 & 5.0 & 1.11 & bd28 , bd17 & 1 & 1800 & mc + + 2006 - 04 - 20.30 & 845.80 & 12 & 12.3 & 1.13 & f34 , hd84 & 1.5 & 1200 & wp + 2006 - 04 - 21.21 & 846.71 & 13 & 3.5 & 1.07 & f34 , hd84 & 1.5 & 1200 & wp + 2006 - 04 - 24.19 & 849.69 & 16 & 67.9 & 1.08 & f34 , bd26 & 1.5 & 1500 & jh + 2006 - 05 - 01.35 & 856.85 & 23 & 0.8 & 1.41 & f34 , hd84 & 1.5 & 780 & wb + 2006 - 05 - 02.14 & 857.64 & 24 & 3.8 & 1.10 & f34 , hd84 & 1.5 - 2 & 1200 & wb + 2006 - 05 - 03.22 & 858.72 & 25 & 64.4 & 1.08 & f34 , hd84 & 1.5 - 2 & 1200 & tg + 2006 - 05 - 04.15 & 859.65 & 26 & 2.8 & 1.09 & f34 , hd84 & 1.5 - 2 & 1200 & tg + 2006 - 05 - 07.19 & 862.69 & 29 & 12.2 & 1.07 & f34 , bd17 & 1.5 - 2 & 1200 & wp +
to sample the early photospheric phase with as many optical observations as possible , we use multiple data sources . for both objects , _ swift _ uvot photometry is used .
spectra are less frequently observed than the photometry , typically once every few days , and this sets the limit to our analysis . in the next sections , we review the ground - based optical photometry , the optical spectroscopy , the _ swift _ uvot data .
ubvri photometry was obtained for sn 2005cs and sn 2006bp at the flwo 1.2 m telescope on mt .
the minicam instrument was used for sn 2005cs while the keplercam was used for sn 2006bp ( for details on cameras and photometric reduction see hicken , 2007 in preparation ) .
all data were bias subtracted and flat fielded using standard procedures .
psf - fitting photometry was performed using dophot ( schechter et al .
host - galaxy subtraction was performed for sn 2006bp .
the host - subtracted photometry was approximately 0.02 mag fainter than the photometry from the unsubtracted images , across all bands .
a host - galaxy subtraction template was not available for sn 2005cs but the supernova was on a relatively faint and smooth background and observations were taken while sn 2005cs was bright so the absence of host subtraction is negligible .
comparison stars for each sn were calibrated on only one night . however , our confidence in their accuracy is high as other sne that had been observed on other photometric nights were observed on these nights , and the results are consistent .
the comparison stars of these other sne showed good agreement across all nights ( the standard deviation of a typical comparison star across these nights being about 0.015 mag in v ) .
we present the photometry obtained for sn 2005cs in table [ tab_phot_05cs ] and in fig .
[ fig_lc_obs ] ( left panel ) , and for sn 2006bp in table [ tab_phot_06bp ] and in fig . [ fig_lc_obs ] ( right panel ) , where , for completeness , we have also included , where appropriate , the observations obtained by pastorello et al .
( 2006 ; hereafter , p06 ) and tsvetkov et al . ( 2006 ) , and the _ swift _ observations for the three bluest filters . some of the spectra for sn 2005cs presented and analyzed in this paper have been previously published by p06 , while a subset for sn 2006bp have been presented by quimby et al .
we refer the reader to those papers for more information on those spectra . in the analysis section of sn 2005cs
, we indicate this source as p06 , wherever appropriate , while in the analysis section of sn 2006bp , we indicate this source as quimby .
all the other spectra were obtained with the 1.5 m tillinghast telescope at the fred lawrence whipple observatory ( flwo ) using the low - dispersion fast spectrograph ( fabricant et al . 1998 ) .
observational details of the spectra are listed in table [ table : speclog ] .
the fast data were all reduced in the same manner , using standard iraf routines for ccd processing and spectrum extraction .
all the spectra were extracted using the optimal algorithm of horne ( 1986 ) .
the spectra are then wavelength calibrated using calibration - lamp spectra ; the calibration is later checked against night sky lines and adjusted accordingly .
we used our own idl procedures for flux calibration .
these include the removal of telluric lines using the well - exposed continua of the spectrophotometric standard stars ( wade & horne 1988 ; matheson et al .
2000 ) .
the relative spectrophotometry is good to about 5% , when done at the parallactic angle ( matheson et al .
2007 , in prep . ) , but no attempt was made to determine an absolute flux calibration . to avoid second - order light contamination , we calibrated each spectrum using two standard stars of different colors ( see table [ table : speclog ] ) . in general ,
the spectra were observed at the parallactic angle to minimize the effects of atmospheric differential refraction ( filippenko 1982 ) .
when this is not the case , the spectra can suffer from a relative flux depletion at bluer wavelengths , especially when observed at high airmass ( @xmath11 , where @xmath12 is the zenith angle ) . in table
[ table : speclog ] we list the angle in degrees , @xmath13 , between the observed position angle and the parallactic angle for each spectrum , along with the airmass . when @xmath14 , the relative spectrophotometry is less accurate . around maximum light ,
both sne 2005cs and 2006bp were bright with respect to their host galaxy .
nevertheless , the earliest ( june 30@xmath15ut ) and latest ( dec 25@xmath16 ut ) fast spectra of sn 2005cs have substantial galaxy background , and the relative flux calibration is affected by this contamination .
finally , all sn 2005cs spectra presented here have been de - redshifted , adopting our own measurement of 466@xmath035of the recession velocity of ngc5194 , while for sn 2006bp and ngc 3953 , we adopt 1280 ( quimby et al .
2007 ; verheijen & sancisi 2001 ; a comparable value of 1053is given by ho et al .
1997 ) .
_ swift _ uvot photometry allows us to track the ultraviolet color evolution of sne from a few days to a few weeks after shock breakout .
observations of both sn 2005cs and sn 2006bp were obtained with _
swift _ uvot ( gehrels et al . 2004 ; roming et al .
2005 ; poole et al .
2007 ) at multiple epochs covering from just hours after discovery until the uv light had faded to the background emission level of the sn environment . the swift uvot data for sne 2005cs and 2006bp , originally presented in brown et al .
( 2007 ) and immler et al .
( 2007 ) , have been reanalyzed using the improved calibration of poole et al .
the magnitudes in table [ table_swift_05cs ] and in table [ table_swift_06bp ] are in the uvot system , distinguished by lower case letters .
the uvot @xmath17 , @xmath18 , and @xmath19 are close to the johnson @xmath20 ; however , the uvot @xmath17 band is bluer than the johnson @xmath21 so there can be significant differences at early times when the sne are blue .
the @xmath18 and @xmath19 agree well with their johnson counterparts .
the magnitudes and flux densities were measured using a @xmath22 aperture source region , corresponding to that used to calibrate the zeropoints and coincidence loss corrections , using @xmath23 . here , we subtract the background light by means of late - time template images , selecting a background region with the same average count rate per pixel as in the @xmath22 source region . when the background subtracted count rate drops below that of the background ( as measured in the template images ) we report the magnitude of the background as an upper limit .
the errors reported include the systematic errors in the zeropoints .
we use the flux equivalent of the _ swift _ uvot magnitudes given in table [ table_swift_05cs ] & [ table_swift_06bp ] , accounting for the different ( and varying ) energy distribution of the supernova by adding an additional systematic error of 10% in quadrature with the statistical error .
we plot these fluxes at the effective wavelengths of the uvot filters along with appropriately reddened synthetic flux distributions computed with the non - lte model atmosphere code cmfgen ( see [ sect_mod_pres ] , [ sect_05cs ] , and [ sect_06bp ] ) .
_ swift _ uvot filters @xmath24 , @xmath25 , @xmath26 , @xmath17 , @xmath18 , and @xmath19 have their effective wavelength at 2030 , 2231 , 2634 , 3501 , 4329 , and 5402 ( poole et al .
2007 ) .
cccccccc 2005 - 06 - 30 & 2453552.4 & 13.07 @xmath0 0.03 & 12.64 @xmath0 0.03 & 12.78 @xmath0 0.03 & & & 14.63 @xmath0 0.03 + 2005 - 07 - 03 & 2453555.1 & 14.21 @xmath0 0.03 & & & & & 14.63 @xmath0 0.06 + 2005 - 07 - 05 & 2453557.3 & 15.05 @xmath0 0.04 & 14.87 @xmath0 0.04 & 14.39 @xmath0 0.04 & 13.64 @xmath0 0.03 & 14.75 @xmath0 0.03 & 14.68 @xmath0 0.03 + 2005 - 07 - 06 & 2453557.9 & 15.26 @xmath0 0.04 & 15.15 @xmath0 0.04 & 14.60 @xmath0 0.04 & 13.75 @xmath0 0.03 & 14.76 @xmath0 0.03 & 14.66 @xmath0 0.03 + 2005 - 07 - 07 & 2453558.8 & @xmath2715.7 & @xmath2715.7 & 14.98 @xmath0 0.04 & 13.85 @xmath0 0.03 & 14.76 @xmath0 0.02 & 14.64 @xmath0 0.03 + 2005 - 07 - 08 & 2453560.0 & @xmath2715.7 & @xmath2715.7 & 15.63 @xmath0 0.05 & 14.14 @xmath0 0.03 & 14.78 @xmath0 0.03 & 14.59 @xmath0 0.03 + 2005 - 07 - 09 & 2453560.6 & @xmath2715.7 & @xmath2715.7 & @xmath2715.7 & 14.32 @xmath0 0.04 & 14.78 @xmath0 0.04 & 14.58 @xmath0 0.04 + 2005 - 07 - 10 & 2453562.0 & @xmath2715.7 & @xmath2715.7 & @xmath2715.7 & 14.67 @xmath0 0.04 & 14.87 @xmath0 0.05 & 14.63 @xmath0 0.05 + 2005 - 07 - 11 & 2453562.8 & @xmath2715.7 & @xmath2715.7 & @xmath2715.7 & 14.92 @xmath0 0.05 & 14.98 @xmath0 0.05 & 14.70 @xmath0 0.05 + 2005 - 07 - 13 & 2453564.5 & @xmath2715.7 & @xmath2715.7 & @xmath2715.7 & & & 14.67 @xmath0 0.05 + 2005 - 07 - 14 & 2453565.5 & @xmath2715.7 & @xmath2715.7 & @xmath2715.7 & 15.43 @xmath0 0.05 & 15.21 @xmath0 0.04 & 14.74 @xmath0 0.04 + 2005 - 07 - 17 & 2453568.5 & @xmath2715.7 & @xmath2715.7 & @xmath2715.7 & @xmath2715.8 & 15.48 @xmath0 0.05 & 14.76 @xmath0 0.05 + 2005 - 07 - 22 & 2453573.5 & @xmath2715.7 & @xmath2715.7 & @xmath2715.7 & @xmath2715.8 & 15.62 @xmath0 0.06 & 14.90 @xmath0 0.05 + 2005 - 09 - 14 & 2453628.5 & @xmath2715.7 & @xmath2715.7 & @xmath2715.7 & @xmath2715.8 & 16.33 @xmath0 0.06 & 14.95 @xmath0 0.04 + cccccccc 2006 - 04 - 10 & 2453836.0 & 14.49 @xmath0 0.03 & 14.74 @xmath0 0.03 & 14.62 @xmath0 0.03 & 14.61 @xmath0 0.02 & 15.75 @xmath0 0.02 & 15.63 @xmath0 0.03 + 2006 - 04 - 11 & 2453836.6 & 14.57 @xmath0 0.04 & 14.83 @xmath0 0.05 & 14.66 @xmath0 0.04 & 14.60 @xmath0 0.04 & 15.60 @xmath0 0.04 & 15.52 @xmath0 0.05 + 2006 - 04 - 11 & 2453836.9 & 14.62 @xmath0 0.03 & 14.83 @xmath0 0.04 & 14.63 @xmath0 0.03 & 14.55 @xmath0 0.03 & 15.59 @xmath0 0.03 & 15.52 @xmath0 0.03 + 2006 - 04 - 11 & 2453837.1 & 14.68 @xmath0 0.03 & 14.83 @xmath0 0.04 & 14.64 @xmath0 0.04 & 14.58 @xmath0 0.03 & 15.53 @xmath0 0.03 & 15.44 @xmath0 0.04 + 2006 - 04 - 12 & 2453837.9 & 14.95 @xmath0 0.03 & 14.95 @xmath0 0.04 & 14.68 @xmath0 0.03 & 14.49 @xmath0 0.03 & 15.52 @xmath0 0.02 & 15.31 @xmath0 0.03 + 2006 - 04 - 13 & 2453838.7 & 15.16 @xmath0 0.03 & 15.05 @xmath0 0.03 & 14.74 @xmath0 0.03 & 14.51 @xmath0 0.02 & 15.47 @xmath0 0.02 & 15.25 @xmath0 0.03 + 2006 - 04 - 13 & 2453839.0 & 15.27 @xmath0 0.03 & 15.17 @xmath0 0.04 & 14.80 @xmath0 0.03 & 14.52 @xmath0 0.03 & 15.43 @xmath0 0.02 & 15.18 @xmath0 0.03 + 2006 - 04 - 15 & 2453840.8 & 15.77 @xmath0 0.03 & 15.53 @xmath0 0.04 & 15.08 @xmath0 0.03 & 14.58 @xmath0 0.03 & 15.41 @xmath0 0.02 & 15.20 @xmath0 0.03 + 2006 - 04 - 15 & 2453841.1 & 15.87 @xmath0 0.04 & 15.69 @xmath0 0.05 & 15.14 @xmath0 0.04 & 14.60 @xmath0 0.03 & 15.47 @xmath0 0.03 & 15.13 @xmath0 0.03 + 2006 - 04 - 17 & 2453842.7 & 16.23 @xmath0 0.04 & 16.04 @xmath0 0.04 & 15.35 @xmath0 0.04 & 14.64 @xmath0 0.03 & 15.42 @xmath0 0.02 & 15.14 @xmath0 0.03 + 2006 - 04 - 17 & 2453842.9 & 16.30 @xmath0 0.05 & 16.22 @xmath0 0.05 & 15.40 @xmath0 0.05 & 14.66 @xmath0 0.04 & 15.41 @xmath0 0.04 & 15.19 @xmath0 0.05 + 2006 - 04 - 17 & 2453843.3 & 16.37 @xmath0 0.04 & & 15.42 @xmath0 0.04 & 14.67 @xmath0 0.03 & 15.41 @xmath0 0.03 & 15.19 @xmath0 0.04 + 2006 - 04 - 20 & 2453846.1 & 17.01 @xmath0 0.04 & 16.91 @xmath0 0.05 & 15.91 @xmath0 0.04 & 14.94 @xmath0 0.03 & 15.50 @xmath0 0.03 & 15.16 @xmath0 0.03 + 2006 - 04 - 22 & 2453847.6 & 17.30 @xmath0 0.05 & 17.34 @xmath0 0.07 & 16.17 @xmath0 0.04 & 15.00 @xmath0 0.03 & 15.56 @xmath0 0.02 & 15.19 @xmath0 0.03 + 2006 - 04 - 24 & 2453849.6 & 17.73 @xmath0 0.08 & 18.13 @xmath0 0.10 & 16.65 @xmath0 0.05 & 15.18 @xmath0 0.03 & 15.60 @xmath0 0.03 & 15.27 @xmath0 0.03 + 2006 - 04 - 26 & 2453851.6 & 18.22 @xmath0 0.09 & @xmath2718.8 & 17.07 @xmath0 0.07 & 15.46 @xmath0 0.03 & 15.65 @xmath0 0.03 & 15.23 @xmath0 0.03 + 2006 - 04 - 28 & 2453853.5 & 18.78 @xmath0 0.10 & @xmath2718.8 & 17.54 @xmath0 0.07 & 15.83 @xmath0 0.03 & 15.76 @xmath0 0.02 & 15.23 @xmath0 0.03 + 2006 - 05 - 01 & 2453857.0 & @xmath2718.8 & @xmath2718.8 & 18.29 @xmath0 0.07 & 16.31 @xmath0 0.03 & 15.91 @xmath0 0.02 & 15.22 @xmath0 0.02 + 2006 - 05 - 04 & 2453859.7 & & & 18.48 @xmath0 0.11 & 16.64 @xmath0 0.04 & 15.98 @xmath0 0.02 & 15.22 @xmath0 0.03 + 2006 - 05 - 16 & 2453871.8 & & & @xmath2718.6 & 17.43 @xmath0 0.07 & & 15.38 @xmath0 0.02 + 2006 - 05 - 26 & 2453882.0 & & & @xmath2718.6 & 17.91 @xmath0 0.09 & 16.64 @xmath0 0.04 & 15.50 @xmath0 0.03 + 2006 - 05 - 30 & 2453885.9 & & & @xmath2718.6 & @xmath2718.2 & 16.72 @xmath0 0.02 & 15.56 @xmath0 0.02 +
the analysis of sne 2005cs and sn 2006bp is analogous to that presented in dh06 , which analyzed sn 1999em at 8 epochs in the first 45 days after explosion . that paper determined its distance by means of various epm - based approaches ( kirshner & kwan 1974 ; eastman & kirshner 1989 ; schmidt et al .
1994 ; eastman et al . 1996 ; hamuy et al . 2001 ; leonard et al .
2002a ; elmhamdi et al .
2003 ; dessart & hillier 2005b ) and another method akin to the seam ( baron et al . 2000 ) . here
, we have two advantages : _ swift _ observations monitor the uv light and we have more densely sampled data . _
swift _ provides additional constraints on the reddening , the temperature , and the ionization state of the ejecta at early epochs .
our analysis covers 13 epochs in ( @xmath430 days ) for sn 2005cs and 18 epochs in ( @xmath445 days ) for sn 2006bp , providing better tracking of the spectral and photometric evolution of each .
time - dependent terms in the statistical- and radiative - equilibrium equations become important a few weeks after explosion ( utrobin & chugai 2005 ; dessart & hillier 2007a , b ) .
we avoid that complication by focusing on the early - time photospheric phase . we know
that assuming a steady state in the statistical and radiative equilibrium equations produces a prediction of the ejecta ionization state that is somewhat below its true level , that hei lines persist over a longer period than we predict , and that strong balmer lines at late times are a clear sign of the epoch when this assumption is no longer valid .
cmfgen solves the radiative transfer equation in the comoving frame , subject to the constraints of radiative and statistical equilibrium ( hillier & miller 1998 ) .
the ejecta are assumed to be spherically symmetric and chemically homogeneous , a good approximation given the very spatially confined regions of continuum and line formation ( dessart & hillier 2005a ) , and low polarization observed in the ejecta of photospheric phase type ii sn ejecta ( leonard & filippenko 2001 ) . the expansion becomes truly homologous about a week after the explosion , but , for simplicity , is assumed homologous at all times .
we adopt a density distribution characterized by a base density @xmath28 ( at the base radius @xmath29 ) and a density exponent @xmath30 , with @xmath31 .
the value of @xmath28 is chosen so that the rosseland - mean optical depth @xmath32 at @xmath29 is @xmath450 .
the hydrogen - rich ejecta are always ionized at least once at the base and electron scattering is the primary source of continuum opacity , while the large hi bound - free cross sections ensure the thermalization of the radiation below the photosphere ( see 4.2 and figs7 - 8 - 9 in dessart & hillier 2005b ) .
the value of @xmath30 is usually quite large , i.e. , @xmath410 , but could be much larger at early times , and can be constrained from line profile shapes , through the magnitude of the blueshift of peak emission of optical p - cygni profiles ( dessart & hillier 2005a ; blondin et al . 2006 ; aldering et al .
2005 ) , or through the strength of p - cygni profiles with respect to the continuum .
for example , lower values of n lead to p - cygni profiles having stronger and broader absorption and emission components , and with a peak emission that is closer to line center .
the early - time spectra of sne 2005cs and 2006bp , show weak p - cygni profiles with a strongly blueshifted peak emission , and point towards high values of n. we quote properties of models at the photosphere , which corresponds to the location where the continuum optical depth , integrated inwards from the outer grid radius , is 2/3 .
the luminosity at depth is not known in advance unless one uses a consistent hydrodynamical input of the ejecta properties , following simultaneously the time - dependent evolution of the radiation field , internal energy , and level populations .
instead , we adjust the base ( comoving - frame ) luminosity so that the synthetic emergent ( observer - frame ) luminosity matches approximately the observed luminosity at each epoch , for an adopted distance and reddening .
usually , with our inferred distance and reddening , the emergent reddened synthetic flux is within a factor of 3 of the observed flux .
we find that radial / luminosity scaling of that magnitude , preserving the ratio @xmath33 produce the same color and relative line strengths , line to continuum flux ratios , and other essential features of the spectrum .
( see dessart & hillier 2005a for a discussion ) .
reddening is constrained using the cardelli et al .
( 1988 ) law for both sne , assuming @xmath34 .
we adopt a single value of @xmath35 to obtain satisfactory fits to the observed spectra at all epochs .
this works best when the relative flux calibration is accurate . for sn 2005cs , we find that the reddening is very small , i.e. , @xmath36 ( baron et al .
2007 infer a value of 0.035 from a quantitative spectroscopic analysis with phoenix ; hauschildt & baron 1999 ) , while for sn 2006bp , it is quite large , i.e. , @xmath37 ( ho et al .
1997 infer a value of 0.4 to the galaxy host ngc3953 from the h@xmath38/h@xmath39 ratio and assuming case b recombination ) . _
swift _ uvot observations allow more leverage on the estimate of reddening than is permitted by optical observations alone by lengthening blueward the observed wavelength range .
cccccrcccc 2453552.25 & 2005 - 06 - 30 & 1.85 & 3.0 & 2.69 & 15750 & 2.00 & 6880 & 24.6 & 20 + 2453553.25 & 2005 - 07 - 01 & 2.85 & 2.3 & 1.99 & 13350 & 2.55 & 6950 & 10.8 & 12 + 2453554.50 & 2005 - 07 - 02 & 4.10 & 2.3 & 2.01 & 13420 & 2.54 & 6370 & 10.6 & 12 + 2453555.25 & 2005 - 07 - 03 & 4.85 & 2.0 & 1.74 & 10850 & 3.26 & 6080 & 7.6 & 10 + 2453556.00 & 2005 - 07 - 04 & 5.60 & 1.8 & 1.64 & 9300 & 3.91 & 5450 & 6.8 & 10 + 2453557.75 & 2005 - 07 - 05 & 7.35 & 1.8 & 1.56 & 8620 & 4.93 & 4780 & 5.6 & 10 + 2453558.25 & 2005 - 07 - 06 & 7.85 & 1.5 & 0.97 & 8250 & 4.18 & 5230 & 6.0 & 10 + 2453561.25 & 2005 - 07 - 09 & 10.85 & 1.5 & 1.21 & 7770 & 5.97 & 4710 & 4.7 & 10 + 2453562.50 & 2005 - 07 - 10 & 12.10 & 1.5 & 1.19 & 7120 & 6.38 & 4580 & 6.0 & 10 + 2453563.50 & 2005 - 07 - 11 & 13.10 & 1.5 & 1.10 & 7300 & 5.88 & 4440 & 8.9 & 10 + 2453564.25 & 2005 - 07 - 12 & 13.85 & 1.5 & 1.08 & 6980 & 6.01 & 4320 & 10.9 & 10 + 2453566.00 & 2005 - 07 - 14 & 15.60 & 1.1 & 0.79 & 6700 & 5.47 & 3930 & 28.2 & 10 + 2453580.25 & 2005 - 07 - 28 & 29.85 & 1.0 & 1.17 & 6140 & 7.84 & 2240 & 125.6 & 10 + to guide our reasoning when interpreting type ii sn spectra , in fig .
[ fig_s15 ] , we illustrate the chemical stratification and velocity distribution versus interior mass for a stellar model that is characteristic of the ejecta of type ii - p sne , resulting from the explosion of a red supergiant ( rsg ) progenitor .
the figure shows the ejecta of a 15(on the main sequence ) progenitor star model ( ignoring rotation ) , exploded with a kinetic energy of @xmath41b ( @xmath4010@xmath5erg ) and with modest ( artificial ) nickel mixing ( woosley & heger 2007 ) . at core collapse
, this star has a mass of @xmath413(@xmath42have been lost through a wind during the pre - sn evolution ) , 1.83is left behind in a protoneutron star and the rest escapes to form the sn ejecta .
starting from the surface , there is @xmath47of essentially homogeneous material , made up primarily of hydrogen ( @xmath41 ) and helium ( @xmath42 ) just slightly evolved from the primordial , solar metallicity , mixture ( grevesse & sauval 1998 ) .
the velocity at the base of this hydrogen shell is about 2000 .
just below is a @xmath43shell dominated by helium ( @xmath43 ) and oxygen ( @xmath44 ) , between velocities 1000 to 2000 . finally , just above the inner mass cut at 1.83
resides an oxygen - rich shell ( @xmath45 ) , with a @xmath40.1 mass fraction of helium , carbon , and nickel .
the radioactive nickel is deep inside , at velocities below 2000 , and separated by @xmath410of material from the photosphere at breakout . in such ( 1d ) radiation - hydrodynamics simulations ,
explosion parameters are adjusted to obtain the right amount of nickel production and mixing , which influence the light curve shape or the potential appearance of lines from radioactive isotopes at the photosphere ( from @xmath8co , for example ) . for the peculiar type ii sn 1987a ,
this required allowing mixing out to @xmath44000 ( pinto & woosley 1988 ; kumagai et al .
1989 ; blinnikov et al .
2000 ) , while for the type ii - p sn 1999em , nickel mixing out to @xmath4650seems optimal ( utrobin 2007 ) .
more detailed 2d hydrodynamical simulations of core - collapse supernovae explosions , assuming the energy demands are fulfilled , do not naturally predict nickel at large velocities , placing it at @xmath42500 ( hachisu et al . 1990 , 1991 ;
kifonidis et al . 2000 , 2003 ) .
observationally , there is a range of inferred values and a conservative assumption for type ii - p sne may be @xmath41000 - 2000 , with 4000representing an extreme circumstance . in the above 15model
, @xmath8ni is buried under about 10of material , mostly composed of hydrogen and helium , so that any non - thermal excitation and ionization of the photosphere by radioactive isotopes is delayed until the end of the plateau phase , when the photosphere has receded to these deep and slow - moving ejecta layers . in our analysis of the early photospheric phase of type ii - p sn , we neglect any contribution from radioactive decay .
we expect slow and small variations in chemistry during the plateau phase , and , thus , adopt chemical homogeneity at a given epoch .
this is consistent with our finding that c and o are only modestly enhanced at the photosphere one month after explosion .
finally , in such explosions , the density profile follows closely a power law distribution with exponent @xmath410 at depth in the hydrogen envelope , but is characterized by a much steeper drop - off near the surface layers , that may result from the strong radiation pressure in these layers at shock breakout ( blinnikov et al .
2000 , shigeyama & nomoto 1990 ) .
such a steepening is not well - modeled by lagrangean simulations , in particular for the shock - breakout phase , but our interpretation of our earliest observations supports this trend .
we adopt the surface composition of blue supergiants analyzed by crowther et al .
( 2006 ) , which is in very close agreement with the outer composition of the 15model described above . by number ,
we adopt h / he = 5 , c / he=0.0004 , n / he=0.0013 , o / he = 0.0016 , with all metals taken at their solar value ( grevesse & sauval 1998 ) .
given as mass fractions , these abundances are : @xmath46=0.55 , @xmath47=0.44 , @xmath48=0.0005 , @xmath49=0.002 , @xmath50=0.003 , and @xmath51=0.0013 . at later times , when in particular oi7770 gains in strength , we need to enhance both the carbon and the oxygen abundances , with c / he=0.001 , o / he = 0.01 , and to reduce the nitrogen abundance , n / he=0.001 to match the data . a key asset of cmfgen is the explicit treatment of line blanketing using a detailed description of the atomic structure for a large number of species .
the iterative march to solving many non - lte level populations together with the radiation field comes at great cpu expense .
we optimize the problem by including only the relevant ions of the most important species , although the ejecta ionization stratification still requires that at least three ionization stages of iron , for example , be included to capture the sources of bound - bound and bound - free opacity at all depths .
typically , any model crafted for very early conditions ( to match the first few , nearly featureless spectra of sne 2005cs and 2006bp , with weak hei5875 ) will contain the following species and levels ( the suffixed parenthesis contains the number of full- and super - levels ; see hillier & miller 1998 for details ) : hi ( 30,20 ) , hei(51,40 ) , heii ( 30,13 ) , cii ( 26,14 ) , ciii ( 112 , 62 ) , civ ( 64,59 ) , nii ( 3,3 ) , niii ( 26,26 ) , niv ( 60,34 ) , nv ( 67,45 ) , oii ( 3,3 ) , oiii ( 115,79 ) , oiv ( 72,53 ) , ov ( 152,73 ) , siiv ( 48,37 ) , siv ( 71,33 ) , feii ( 309,116 ) , feiii ( 563 , 191 ) , feiv ( 787,120 ) , fev ( 191,47 ) , fevi ( 433,44 ) , niiii ( 67,15 ) , and niiv ( 242,36 ) . at one week after explosion , when hei5875 is strong , we take out niv , ov , siiv , siv , fevi , and niiv , and add ni ( 104,54 ) , oi ( 75,23 ) , mgii ( 65,22 ) , caii ( 77,21 ) , siii ( 59 , 31 ) , tiii ( 152,37 ) , tiiii ( 206,33 ) , niii ( 93,19 ) , and niiii ( 67,15 ) .
when hydrogen starts recombining , we put more emphasis on low ionization states and adopt the following model atoms : hi ( 30,20 ) , hei(51,40 ) , ci ( 63,33 ) , cii ( 59,32 ) , ni ( 104,54 ) , nai ( 71,22 ) , mgii ( 65,22 ) , siii ( 59 , 31 ) , oi ( 75,23 ) , caii ( 77,21 ) , fei ( 136,44 ) , feii ( 309,116 ) , feiii ( 477,61 ) , niii ( 93,19 ) , and tiii ( 152,37 ) . references to the corresponding atomic data sources
are given in dessart & hillier ( 2005a ) .
to summarize , inputs to a cmfgen model are global characteristics of the ejecta such as base ( comoving - frame ) luminosity , radius , chemical composition , velocity and density .
these parameters are iterated ( see dessart & hillier 2005a for details ) until the synthetic spectrum matches the observations .
assessment of suitability is done by eye .
spectroscopic observations used here are accurate in relative flux , while only accurate to within 1020% in absolute flux . for spectral fitting , relative flux calibration is essential while absolute flux calibration is not .
when inferring distances , we use the photometry to set the absolute flux scale .
one important result of our study is that the cmfgen model parameters needed to fit the optical data of sne 2005cs and 2006bp during the first 2 - 3 weeks after explosion also naturally explain the _ swift _ uvot data _
simultaneously_. there appears to be a common , photospheric , origin for the uv and optical light for both objects : circumstellar interaction contributes very little to the observed light in these spectral regions ( see also brown et al .
note that sn 2006bp was , however , detected in the x - rays for up to 12 days after explosion , suggesting that interaction between the ejecta and the circumstellar medium ( csm ) does occur ( immler et al .
2007 ) , but without any sizable effect on the uv and the optical flux .
the first month after explosion , when the photosphere resides in the ionized hydrogen - rich layers of the ejecta , is not the best epoch to probe the innermost details of the explosion mechanism .
however , it is ideal for distance determinations based on variants of the expanding photosphere method : a true continuum exists and the effects of line - blanketing are weak in the optical ; the properties of the ejecta are well determined by the model , using both the continuum energy distribution and lines from the most abundant ions .
the homogeneous hydrogen - rich envelope ensures that a well defined photosphere exists and that radiation thermalizes at depth , allowing the adequate use of the diffusion approximation at the model base .
there , our non - lte approach naturally predicts lte conditions , the departure coefficients of all levels tending to unity , and at the thermalization depth , the mean intensity is well represented by a blackbody ( lte is , in other words , never enforced ) .
the analysis presented here allows the inference of distances to type ii - p sne ( see [ sect_dist_05cs ] and [ sect_dist_06bp ] ) . to be more concise , we present all model parameters in table [ tab_model_05cs ] and [ tab_model_06bp ] , but mention in the text only the parameters that are most directly related to the spectroscopic evolution , i.e. , @xmath52 , @xmath53 , and @xmath30 .
sn 2005cs was discovered in the whirlpool galaxy ( m51a ; ngc5194 ) by kloehr et al .
( 2005 ) on 28.905 june 2005 ( jd 2453550.4 ) .
it was not detected on 26.89 june below the limits @xmath54 17.3 , @xmath55 17.7 , and @xmath5617.6 ( p06 ) , which places a narrow window of two days on the explosion date . in this work ,
we employ observations from various sources at 13 epochs , starting on 30 june 2005 , which could be within 3 days of the explosion .
note that m51a , an sa(s)bc galaxy , also hosted the type ic sn 1994i ( wheeler et al .
1994 ; filippenko et al .
1995 ; baron et al .
importantly , maund et al .
( 2005 ) and li et al .
( 2006 ) associate the sn 2005cs event with a 9@xmath57k0-m4 rsg progenitor star , based on the analysis of pre - explosion hst images , compatible with the observed plateau evolution of the light curve in the optical bands .
sn 2005cs was not detected in the x - rays and the upper limit to the x - ray luminosity yields a limit to the progenitor mass loss rate of @xmath58@xmath59 , where @xmath60 is the asymptotic velocity of the progenitor wind ( brown et al .
model properties are given in table [ tab_model_05cs ] and corresponding synthetic spectral fits to observations are presented in figs .
[ 05cs_0630][05cs_0728 ] .
we show the first observations ( blues crosses : _ swift _ uvot photometry ; black curve : optical spectrum of 30 june 2005 from cfa ) of sn 2005cs in fig .
[ 05cs_0630 ] , together with synthetic spectra ( red : full ; blue : continuum only ) , reddened with @xmath61 and scaled by a factor 1.17 .
observations reveal a very hot spectrum , with a steep decline of the observed flux longward of @xmath42300 .
we identify optical lines of the balmer series ( h@xmath38 , h@xmath39 , and h@xmath62 ) , hei5875 , the nii multiplets at @xmath44600 and @xmath45400 , and , as suggested by baron et al .
( 2007 ) , an oii multiplet at @xmath44600 .
note that the lower ionization inferred for the earliest spectra of sn 1999em suggested that nii alone was sufficient to reproduce adequately the observed features at both 4600 and 5400 .
it is the higher ionization in the first spectrum of sn 2005cs that calls for this novel oii identification , more consistent with solar cno chemistry , and slightly different from the abundances used in dh06 ( see our discussion in [ sect_mod_pres ] and [ sect_discussion ] ) .
the 43005000 region also contains numerous hei lines .
the observed dip shortward of 2000 stems from feiii line - blanketing , rather than dust extinction ( the continuum - only synthetic spectrum , shown in blue , does not show this dip ) .
the individual contributions from all important ions on that date are presented in fig . 5 , left panel , of brown et al .
( 2007 ) . on that date
, we have @xmath63k , @xmath64 , and @xmath65 .
such a steep density distribution is required to bring optical lines down to their ( weak ) observed strength , by reducing their emission volumes .
this is higher than the density exponent adopted in the past , but these are also the first analyses of such early - time observations of type ii - p sn spectra .
a steep density gradient is also used to fit the earliest observations of sn 2006bp .
we discuss this choice and comment on its implications in [ sect_ion ] .
we present p06 s observations of sn 2005cs on the 1st july 2005 in fig .
[ 05cs_early ] , top left panel .
the slope of the spectrum in the blue appears steeper than predicted by the _ swift _
uvot data points , which are quite closely matched by our model ( the observed flux in the spectrum , inaccurate below @xmath44000 , is omitted ; pastorello , priv . comm . ) .
the synthetic flux is scaled by a factor 1.1 .
the sn ejecta is getting cooler ( @xmath66k ) at the photosphere , but with a comparable velocity ( within the uncertainties ; @xmath67 ) .
the density exponent @xmath30 has been reduced to a value of 12 ( the density distribution flattens ) , motivated by the strengthening of optical line features , whose identity is the same as that on 30 june .
the uv flux is weaker , resulting primarily from the cooling of the ejecta - the line blanketing is still mostly due to feiii .
little changes as we move onto the 2nd of july 2005 , as shown in fig .
[ 05cs_early ] , top right panel , with a model characterized by @xmath68k , @xmath69 , and @xmath70 , or a lower ejecta velocity ( dessart & hillier 2005a ) .
this 10% range in expansion velocity ( @xmath4500 ) may in fact be physical , reflecting the potential presence of fluid instabilities or turbulence .
the _ swift _ uvot photometry would be better fit by a faster decline in the uv than our cmfgen model provides .
this situation improves on july 3rd , the following day ( fig .
[ 05cs_early ] , bottom left panel ) , where a single model fits both the _ swift _ uvot photometry and the optical spectrum , with parameters @xmath71k , @xmath72 , and @xmath73 .
the photosphere is both cooler and slower , while the density exponent is now set to 10 at all remaining epochs .
the observations on july 4th resemble those of july 3rd in the optical , but the uvot photometry indicates a fading of the uv flux . in the bottom right panel of fig .
[ 05cs_early ] , we show the model fit to the july 4th observations is cooler , with @xmath74k , @xmath75 , and @xmath73 .
these early observations show a strong uv flux that dominates over the optical flux , with low extinction .
optical lines are hi , hei , oii and nii , with blanketing in the uv from feiii .
the density distribution is steep , flattening over a few days to the canonical @xmath76 exponent predicted for blast wave solutions ( imshennik & nadezhin 1988 ; shigeyama & nomoto 1990 , chevalier 1982 ; ensman & burrows 1992 , blinnikov et al .
we surmise that the steep density profile may reflect work done by the intense uv radiation at and shortly - after shock breakout ( see [ sect_ion ] ) .
our models support a common origin for both the uv and the optical light observed , with emission from the regions at and above the photosphere .
no additional contribution to the uv and optical fluxes is present from an external source , such as an interaction between the sn ejecta and the csm . in this epoch ,
the flux level decreases dramatically in the uv and metal lines , mostly of feii , appear in the optical .
this is the onset of the hydrogen - recombination epoch leading to the `` plateau '' phase of such type ii sne . at the same time , hei5875 weaken , so that the feature at @xmath45900 is due only to naid at late times .
these spectral properties are visible in fig .
[ 05cs_int ] , top left panel , in which observations on the 5th of july are fitted with a model characterized by @xmath77k , @xmath78 , and @xmath73 .
the oi7770 line is now clearly seen , following the cooling and the recession of the photosphere .
the sn 2005cs ejecta have lower velocities than sn 1999em : this allows the clear identification of this line profile , which was severely corrupted by the atmospheric extinction in the spectra of sn 1999em ( dh06 ) .
similarly , while siii6355 was predicted to overlap with h@xmath38 in the spectrum of sn 1999em ( figs . 3 and 5 in dessart & hillier 2005a ) , we clearly observe it here as a separate line blueward of the h@xmath38 trough ( see also p06 ) .
this line , in combination with unidentified nii / oii lines at similar doppler velocities in the blue wing of h@xmath39 and hei5875 , led to speculations on the presence of inhomogeneities above the photosphere ( leonard et al 2002ab ) . at early epochs
, it seems these subtle spectral features can be explained quite naturally by numerous lines of oii ( see above ) , nii , hei , and siii . at later times , the presence of a cool dense shell at the outer edge of type ii - p sn ejecta is expected to produce real absorption dips in the trough of h@xmath38 ( and hei10830 ; see chugai et al .
the fit to observations on july 6th is shown on the top right panel of fig .
[ 05cs_int ] , with little change from the previous day except for a modest decline in temperature ( e.g. , @xmath79k ) . by the 9th of july ( fig .
[ 05cs_int ] , middle left panel ) , the uv flux is now an order of magnitude lower than the optical flux .
the caiih&k doublet appears as a strong p - cygni line profile at the blue edge of the optical spectrum , strengthening with time , together with the feii lines present throughout the uv range ( as a forest of overlapping lines that give an apparently continuous light blocking effect - see fig . 5 in dessart & hillier 2005a ) and up to @xmath45200 .
niii ( shortward of 2500 ) and tiii ( in the 3000 region ) also contribute significant line blanketing .
the model properties for the 9th of july are @xmath80k , @xmath81 , and @xmath73 .
subsequent evolution is quite slow from the 10th ( fig .
[ 05cs_int ] , middle right panel ) , 11th ( fig .
[ 05cs_int ] , bottom left panel ) , 12th ( fig .
[ 05cs_int ] , bottom right panel ) , and 14th ( fig .
[ 05cs_0714 ] ) . over this 5-day period ,
the photospheric layers cool slowly and modestly ( 7120k , 7300k , 6980k , 6700k , respectively ) , and slow down ( 4580 , 4440 , 4320 , 3930 ) . on the 14th , the p06 spectrum extends beyond 9000 and reveals the presence of the caii8500 multiplet , providing a good check that the ionization state of the sn ejecta is well modeled . fitting the oi7770 line requires a higher oxygen abundance ( o / he=0.01 ) , which we adopt for sn 2005cs starting on the 11th of july .
we show a comparison of the enhanced and standard oxygen abundance cmfgen models in the left and right panels of fig .
[ 05cs_0714 ] , to show the effect on the predicted strength of the oi7770 line .
note finally that the h@xmath38 line strength ( as well as h@xmath39 , although less discernible ) is underestimated by a factor of two in both absorption and emission , the more so for higher carbon / oxygen abundances ( corresponding to lower hydrogen mass fractions ) .
the observation of the 28th of july 2005 is the last in our sn 2005cs sample .
this is about one month after the first observation taken and well into the plateau phase .
we show our fits to these observations in fig .
[ 05cs_0728 ] for which the model parameters are @xmath82k , @xmath83 , and @xmath73 .
the photosphere has now receded to slow - moving ejecta layers close to the bottom of the hydrogen shell . in practice ,
given the low luminosity and low velocity of this sn envelope , the ( outer ) hydrogen shell of the progenitor may stretch even further down to the slower velocity and more c / o enriched layers .
unfortunately , the cfa observations do not cover the oi and ci lines longward of 7500 , which would provide a measure of the oxygen and carbon mass fractions . in dh06 , and by comparison with the surface chemistry of the progenitor , we inferred significant enhancements in carbon and oxygen abundances in the photospheric layers of sn 1999em at 45 days after explosion .
the main line diagnostics are oi7775 and oi7984 , while redder transitions at 8446 and 9264 overlap with caii , mgii , and ci lines , and are therefore less reliable .
useful ci diagnostics are the features at @xmath49100 , 9400 , and 9600 , due to transitions of the ( 3p-3s ) series at 9061 , 9062 , 9078 , 9088 , 9112 , 9405 , 9603 , 9621 , and 9658 . in the near - ir ,
the ci1.07@xmath84 m is an additional diagnostic , although it overlaps with p@xmath62 , and potentially with hei1.083@xmath84 m .
observations in this wavelength range would be especially valuable .
observing the c / o abundance change would be important and warrants thorough late time spectroscopic observations in the photospheric phase of type ii sne , extending at least up to 1@xmath84 m .
a census of lines contributing in the optical is shown in fig .
8 of dh06 .
feii blankets completely whatever uv flux emerges from the thermalization layers below the photosphere .
tiii and feii leave a complex set of features in the optical , most noticeably resulting in a pronounced dip in the b - band region .
naid is reproduced here by enhancing its abundance by a factor of four above its solar metallicity value ( grevesse & sauval 1998 ) .
however , the predicted line remains narrower than observed and time - dependent effects were found to lead to a strong and broad line profile without any change in sodium abundance .
the underestimate of the h@xmath38 line strength suggests that this is the epoch when our steady - state approach departs significantly from the more correctly - computed time - dependent conditions in the sn ejecta ( dessart & hillier 2007a , b ) .
cccccrcccc 2453836.6 & 2006 - 04 - 11 & 1.5 & 15 & 16.3 & 20600 & 2.386 & 14100 & 51.7 & 50 + 2453837.6 & 2006 - 04 - 12 & 2.5 & 15 & 15.1 & 18120 & 3.165 & 12550 & 31.6 & 50 + 2453842.6 & 2006 - 04 - 17 & 7.5 & 9 & 6.9 & 11800 & 5.541 & 11300 & 8.3 & 20 + 2453844.6 & 2006 - 04 - 19 & 9.5 & 9 & 6.6 & 11400 & 5.868 & 11050 & 6.5 & 16 + 2453845.8 & 2006 - 04 - 20 & 10.7 & 9 & 6.2 & 10700 & 6.485 & 10300 & 4.5 & 12 + 2453846.7 & 2006 - 04 - 21 & 11.6 & 9 & 6.2 & 10700 & 6.485 & 10300 & 4.5 & 12 + 2453849.6 & 2006 - 04 - 24 & 14.5 & 5 & 3.6 & 9350 & 6.345 & 7920 & 4.2 & 10 + 2453850.6 & 2006 - 04 - 25 & 15.5 & 2.7 & 1.8 & 8150 & 6.580 & 8750 & 4.0 & 10 + 2453855.7 & 2006 - 04 - 30 & 20.6 & 1.5 & 1.1 & 7320 & 6.000 & 6450 & 7.3 & 10 + 2453856.8 & 2006 - 05 - 01 & 21.7 & 1.5 & 1.1 & 7100 & 6.147 & 6330 & 8.8 & 10 + 2453857.6 & 2006 - 05 - 02 & 22.5 & 1.5 & 1.1 & 7100 & 6.147 & 6330 & 8.8 & 10 + 2453858.7 & 2006 - 05 - 03 & 23.6 & 2.0 & 1.4 & 6800 & 7.26 & 5810 & 10.1 & 10 + 2453859.6 & 2006 - 05 - 04 & 24.5 & 2.0 & 1.4 & 6800 & 7.26 & 5810 & 10.1 & 10 + 2453862.7 & 2006 - 05 - 07 & 27.6 & 2.0 & 1.7 & 6420 & 8.90 & 4550 & 17.9 & 10 + 2453867.7 & 2006 - 05 - 12 & 32.6 & 2.0 & 1.7 & 6420 & 8.90 & 4550 & 17.9 & 10 + 2453876.7 & 2006 - 05 - 21 & 41.6 & 2.0 & 1.8 & 6300 & 9.28 & 4320 & 30.3 & 10 + 2453891.7 & 2006 - 06 - 05 & 56.6 & 2.0 & 1.8 & 6160 & 9.73 & 3760 & 50.7 & 10 + 2453907.6 & 2006 - 06 - 21 & 72.5 & 2.0 & 1.9 & 6050 & 10.18 & 3160 & 79.4 & 10 +
sn 2006bp was discovered in ngc3953 , an sb(r)bc galaxy , by koichi itagaki ( teppo - cho , yamagata , japan ) around 9.60 april 2006 ( jd 2453835.1 ) , and located at r.a .
= 11h53mn55s.74 , decl .
= + 52@xmath852109``.4 ( equinox 2000.0 ) , which is 62 '' east and 93 " north of the center of ngc3953 .
quimby et al .
( 2006 ) obtained an optical spectrum on march 11.11 ut with the 9.2-m hobby - eberly telescope ( het ) , reporting the observation of a blue continuum , a narrow absorption at 592 nm ( associated with naid ) , and line features at 583 nm ( associated with hei5875 ) , and 627 nm ( associated with h@xmath38 ) .
sn 2006bp was not detected in unfiltered rotse - iiib images taken on apr .
9.15 ( limiting mag about 16.9 ) , but it was found at about 15.3 mag on images taken on apr . 10.15 .
sn 2006bp was caught promptly after the explosion , as confirmed from our analysis of the spectra .
immler et al . (
2007 ) report the 4.5@xmath86 detection of x - ray emission at the level of 1.8 @xmath0 0.4 @xmath2 10@xmath87ergs@xmath88 up until 12 days after discovery , indicating circumstellar interaction with a pre - sn rsg wind characterized by @xmath89 .
this x - ray luminosity is less than 1% of the uv and optical luminosity in the combined uv and optical ranges and differs from the strongly interacting type iin sne .
model properties are given in table [ tab_model_06bp ] and corresponding synthetic spectral fits to observations are presented in figs.[06bp_0411][06bp_late ] .
the first spectral observation of sn 2006bp , shown in fig .
[ 06bp_0411 ] , is impressive in several regards .
it has a nearly featureless spectrum , with a very steep slope .
the dip at @xmath42200 is not caused by line blanketing , as in the first spectrum of sn 2005cs ( feiv is the main blocking ion in the uv , and it does so blueward of 1800 ; its effect on the synthetic continuum sed is shown ) .
the dip stems from interstellar extinction , with an inferred value @xmath35=0.4 .
we also see , at all epochs , naid as a narrow absorption feature , indicating significant column density in the line of sight to sn 2006bp .
hydrogen balmer lines are present , but very weak and very broad , and limited to h@xmath38 , h@xmath39 , and h@xmath62 .
hei5875 is just barely identifiable .
it is surprising that lines of nii and oii identified in the first spectra taken of sn 2005cs are not seen here .
given the high reddening , fitting the steep slope of the sed requires much higher temperatures and ionization conditions ( fig .
[ fig_gammas ] ) .
consequently , cmfgen predicts , although somewhat too strong , the presence of heii4686 ( together with a small , @xmath420% contribution from ciii4650 ) , a line that is more typical of the photosphere / wind of early - type o and wolf - rayet stars ( hillier & miller 1999 ; dessart et al . 2000 ; crowther et al .
2002 ; hillier et al .
2003 ) , as well as features arising from niii40014099 , niii4638 , and ciii2290 ( see bottom panel of fig . [ 06bp_0411 ] ) . in practice
, we used @xmath90k , @xmath91 , and @xmath92 .
numerous directions in the parameter space were explored to reproduce the steepness of the sed , the small strength of p - cygni profiles , and the strong blueshift of p - cygni peak emission .
these included varying the reddening and temperature , extent of the photosphere , ejecta chemistry , and density exponent .
the most critical was the density exponent .
quimby et al .
( 2007 ) discuss the presence of weak and narrow lines in this optical spectrum , associated with heii4200 , heii4686 , and civ5808 .
they suggest that they emanate from material located just outside the sn ejecta that is flash ionized at shock breakout .
this gives additional support to our identification here of heii4686 emission from the sn ejecta itself .
note the photospheric velocity of sn 2005cs inferred from the first observed spectrum at 5 days after explosion is only 6880 , compared to 10430 at a comparable epoch for sn 1999em ; see [ sect_05cs ] and dh06 ) .
we discuss further the properties of sn 2006bp at that age in [ sect_ion ] .
we could only reproduce this spectrum with cmfgen by increasing the density exponent to 50 ( see also [ sect_ion ] ) . because the density distribution is so steep , line and continuum formation coincide and occupy a very confined region of space at such early times .
as time progresses , this region recedes , in the co - moving frame , to deeper layers where the density profile is flatter , but the spectrum formation region remains confined , which prevents the observer from simultaneously probing the outer regions .
changes in the exponent that describes the density distribution are best inferred through consecutive , rather than isolated , spectroscopic observations ( dessart & hillier 2005a ) .
the peak emission of , e.g. , the h@xmath38 line is blueward of its rest wavelength , corresponding to a velocity shift of @xmath48000initially , but decreases with time to reach @xmath41000on the 21st of june .
this is consistent with a large density exponent ( dessart & hillier 2005a ) , although our overestimate of the observed h@xmath38-peak blueshift on the 11th of april ( fig .
[ 06bp_0411 ] ) suggests that @xmath30 may not be as high as 50 , and that the weakness of optical lines may not be caused entirely by the steep density fall - off .
we show , in the top left panel of fig .
[ 06bp_early ] , cmfgen fits to observations of sn 2006bp on the 12th of april 2006 .
the spectrum looks very similar to that on the 11th , the fit quality is superior , and the model differs in having a lower photospheric temperature and velocity .
the model parameters are @xmath93k , @xmath94 , and @xmath92 .
the synthetic uv flux also closely matches the flux - equivalent _ swift _ uvot magnitudes , with a nearly featureless spectrum strongly extinguished by the high reddening in the line of sight of @xmath95 . along a low extinction line of sight
, the flux at @xmath42000 would have been about ten times as large , reaching a level of @xmath410@xmath96erg@xmath97s@xmath88@xmath88 , i.e. , about a hundred times greater than the optical flux at that time !
the het observations on the 15th of april 2006 were of very low signal and were excluded from our sample . for the 17th of april 2006 , we show cmfgen fits in the top right panel of fig
. [ 06bp_early ]
. the optical range shows hydrogen balmer lines and hei5875 , weak nii features in the blue edge of hei5875 and h@xmath39 , as well as a weak feature that we attribute to oii lines in the range 43004600 ( see above discussion for sn 2005cs ; baron et al .
the uv flux no longer dominates the optical and decreases rapidly from the combined effects of cooling ( due to radiative losses and expansion ) and increased line - blanketing .
the optical lines now appear stronger and are well fitted with a density exponent of 20 .
the model parameters for the 17th of april 2006 are @xmath98k , @xmath99 , and @xmath65 .
two days later , on the 19th of april 2006 , the spectrum looks very similar ( bottom left panel of fig . [ 06bp_early ] ) , with optical lines somewhat stronger and a diminishing uv flux .
the model parameters are @xmath100k , @xmath101 , and @xmath102 .
note the decreasing density exponent from 50 to 16 in just one week , and the decreasing photospheric velocity and temperature . because the density distribution flattens ,
the outer ionized regions contribute more to the photospheric optical depth , shifting it outward to a lower density ( table [ tab_model_06bp ] ) .
the same holds on the following day , the 20th of april 2006 ( fig .
[ 06bp_early ] , bottom right panel ) , for which the model parameters are @xmath103k , @xmath104 , and @xmath70 .
the best - fit density exponent has a value of 12 , and changes little after this time to reach the canonical value of 10 .
the spectral evolution is very slow and no new lines are identified in the spectrum .
nii and oii lines have , however , disappeared , following the recombination of the ejecta . on the 21st of april 2006
, we have two sources of observations . the data of quimby et al .
( 2007 ) starts only at @xmath44200but extends out to 10700 , revealing a weak oi7770 and the absence of the caii multiplet at 8500 ( left panel of fig .
[ 06bp_0421 ] ) .
the _ swift _ uvot data points are well matched by the synthetic uv flux distribution , although we overestimate the u - band flux somewhat .
the model parameters are @xmath103k , @xmath104 , and @xmath70 . in the right panel of fig .
[ 06bp_0421 ] , we present a similar model with a photospheric temperature @xmath41000k lower , allowing a better match to the uv range , but a poorer match to the optical lines . this cooler model seems to have the right ionization for the uv flux , but has too low an ionization to explain the optical .
we surmise that this stems from time - dependent effects , present at all times , but more visible in spectral lines of species ( here , helium ) that are about to change their ionization state in the ejecta ( dessart & hillier 2007a , b ) .
this cooler model fits much better the weak feii5169 that indicates the start of the recombination of feiii to feii , together with the recombination of hydrogen .
so this date closes the early epoch of sn 2006bp corresponding to fully ionized conditions in the photosphere region . from this epoch
onwards , the ejecta in the photospheric layers is too cool to radiate a substantial uv flux , and line blanketing , due in particular to feii lines , blocks whatever uv light is produced , re - processing it to longer wavelengths .
extinction plays only a secondary role . on the 24th of april 2006 ( fig .
[ 06bp_int ] , top left panel ) , in addition to the strengthening hydrogen balmer lines and the weakening hei5875 , we see the appearance of spectral features due to feii lines , around 4400 and 5200 ( the main contributor there is feii5169 ) and mgii4480 .
the model parameters are @xmath105k , @xmath106 , and @xmath73 .
note that we set the density exponent to 10 for this and all following dates . on the 25th of april 2006 ( fig .
[ 06bp_int ] , top right panel ) , the lines of feii strengthen and we see the appearance of siii6355 , although not as clearly as for sn 2005cs due to the higher photospheric velocities .
it is not clear if the observed dip at around 5800 is due to weakening hei5875 ( as supported by time - dependent calculations which predict a higher ejecta ionization sustained over a longer time ) or due to the strengthening naid . in the red part of the spectrum ,
we also see the unambiguous presence of oi7770 and the weak appearance of the caii8500 multiplets , which are not predicted by cmfgen at this epoch .
h@xmath38 is somewhat overestimated both in emission and absorption , which may be related to the assumption of steady state or to the constant density exponent adopted at a given epoch .
the model parameters are @xmath107k , @xmath108 , and @xmath73 .
the spectrum of sn 2006bp on the 30th of april 2006 ( middle left panel of fig . [ 06bp_int ] ) shows evidence of recombination in the ejecta . in particular , feii lines are now clearly visible throughout the @xmath109-band , affecting the h@xmath38 trough ( with a small contribution from siii6355 ) .
caii lines at 3800 and 8500 are also strong and broad .
naid is the likely contributor to the dip at 5800 . at this epoch ,
we employ enhanced oxygen and carbon abundance compared to the starting conditions ( c / he=0.0017 and o / he=0.01 by number ) , while nitrogen is depleted ( n / he=0.0068 ) and nitrogen lines are not seen . in the red part of the spectrum , we observe numerous ci lines due to multiplets around 9100 , 9400 , and 9630 .
we also predict additional contributions from oi at 7990 , 8450 , and 9260 , but all much weaker than the strong oi7770 .
the model parameters are @xmath110k , @xmath111 , and @xmath73 . for the cfa observations of the 1st and 2nd of may 2006bp ( middle right and bottom left panels in fig .
[ 06bp_int ] ) , we employ the same model with @xmath112k , @xmath113 , and @xmath73 .
the spectral range misses the red part of the spectrum where the oxygen , carbon , and calcium lines of interest would be found .
the spectral evolution is slow : we see a slight fading of the flux in the blue , with a modest recession of the photosphere , as shown by the narrower line profiles .
we now discuss the late - time observations of sn 2006bp , an epoch that corresponds , in our steady - state modeling with cmfgen , to conditions where hydrogen has recombined at and above the photosphere , and where the density of free electrons is low .
we present in fig .
[ 06bp_late ] a montage of synthetic spectra fitted to observations of sn 2006bp on may 4th ( top left ) , 7th ( top right ) , 12th ( middle left ) , 21st ( middle right ) , and june 5th ( bottom left ) and 21st ( bottom right ) , 2006 . in this elapsed time , the model photospheric temperature ( velocity ) decreases steadily from 6800k ( 5810 ) to 6050k ( 3160 ) , while the luminosity and the density exponent remain constant ( a full set of model parameters is given in table [ tab_model_06bp ] ) . throughout this period
, the uv flux fades and the sed peaks in the optical .
the continuum energy distribution is very well fitted at all times , becoming more and more affected by metal line blanketing , whose effect appears first in the optical with feii lines .
contributions from tiii lines in the 4200 region strengthen with time and cause the @xmath114-band magnitude to plummet . by the 21st of june
, feii lines even affect the flux in the red part of the spectrum , around 7300 , while oi and ci lines continue to contribute numerous narrow features in the red .
caiih&k is overestimated early on , while the caii multiplet at 8500 is generally well fitted .
we note the striking evolution of the h@xmath38 strength , observed with a fixed strength over this 6-week period , while steady - state cmfgen models predict an ever decreasing strength and width ( always underestimated ) for that line ( fig . [ 06bp_late ] ) .
this is caused by time - dependent effects , as proposed by utrobin & chugai ( 2005 ) and dessart & hillier ( 2007a , b ) , and can be remedied by accounting for the time - dependent terms in the statistical and radiative equilibrium equations .
other lines tend to be too narrow , in particular naid .
the good fit to the central part of the line profile indicates that the photospheric velocity is well matched , but the extent of the corresponding emission line regions is systematically underestimated .
so , while we provide fits to observations at those late epochs , time dependent models would yield different parameters and synthetic spectra more compatible with the observations .
since time dependence does not profoundly alter the continuum energy distribution , our synthetic fits may still be useful for distance determinations .
it remains to be checked to what extent the effects introduced by time dependence affect the correction factors used in the epm , the true magnitude of the flux dilution , etc ( dessart & hillier 2005b ) . in practice
, we exclude the observations of sn 2006bp on the 5th and on the 21st of june for the determination of its distance .
in this section , we compute distances and explosion times of sne 2005cs and 2006bp using our photometric observations and synthetic spectra . we use two methods , which emerged as the most suitable from the detailed discussion in dessart & hillier ( 2005b ) and dh06 .
the first approach is to apply the expanding photosphere method ( epm ; kirshner & kwan 1974 ; eastman & kirshner 1989 ; schmidt et al .
1994 ; eastman et al . 1996 ; hamuy et al . 2001 ; leonard et al .
2002a ; elmhamdi et al .
2003 ) , but using correction factors and color temperatures calculated directly from our tailored models at each epoch ( as in dh06 ) .
this approach is superior to using analytical formulae describing the dependence of correction factors on color temperature , since those are computed for models that may not match the observed spectrum .
this approach avoids the physical scatter that stems from such a diversity among sn ii .
the second approach is to use our synthetic spectra to directly fit observations , as described in dh06 , in a similar fashion to the seam ( baron et al .
2004 ) . for both ,
the observed magnitudes shown in table [ tab_phot_05cs ] and [ tab_phot_06bp ] are used , either directly or interpolated linearly between adjacent dates to correspond to the date of the spectroscopic observation .
this does not introduce any sizable error given the very slow change ( nearly constant from day to day within the errors ) of @xmath114 , @xmath109 , and @xmath115 magnitudes in the early photospheric phase of type ii - p sne .
we have described at length the epm in dessart & hillier ( 2005b ) , complementing the previous work of eastman et al .
( 1996 ) , providing some insights into the method and its limitations , as well as a new and independently determined set of correction factors based on a large database of cmfgen models .
correction factors determined with cmfgen are comparable to , albeit systematically 10 - 20% larger than , those of eastman et al . ,
although the origin of this small , but important , difference remains unclear . in dh06 ,
we applied the epm to sn 1999em and found that , unless the correction factors and associated color temperatures are determined with tailored models , the epm distance and the explosion time were inconsistent when different bandpass - sets were used ( amongst @xmath116 , @xmath117 , and @xmath118 ) . since the photospheric radius is nearly constant across the optical range ( dessart & hillier 2005b ) the inferred distance ought to be the same whatever optical set is used .
another source of uncertainty , although smaller , is the inference of the photospheric velocity , @xmath52 , usually approximated by adopting the doppler velocity at maximum absorption in weak and seemingly isolated p - cygni profiles ( note that sphericity and projection effects , subtleties in the line transfer problem , and diversity in atomic properties lead to a non - trivial relation between the photospheric velocity and the velocity at maximum absorption in a given line ; see dessart & hillier 2005b for a discussion on this issue ) . in our approach
, we adopt the value from our cmfgen model .
although our model fits are satisfactory , they sometimes overestimate or underestimate profile widths , matching some profiles very well while doing a poorer job on others at the same time .
there is likely a genuine range of velocities at the photosphere due to fluid instabilities and density variations , so that a 10% inaccuracy in the photospheric velocity at a given time may in fact exist .
however , we expect the sense and the magnitude of the uncertainty in our inferences to behave randomly
. we do not consistently overestimate or underestimate v@xmath119 , or the flux in a given spectral region , so that observations at many times should reduce the uncertainty .
reddening is another ingredient requiring special care in distance estimates . in our approach
, we use synthetic seds to constrain the reddening @xmath35 within @xmath120 .
the dust properties in other galaxies may not be well reproduced by the cardelli law , which we employ , and this introduces an additional uncertainty , which is significant for the strongly extinguished sn 2006bp ( for such a discussion , see , e.g. , elias - rosa et al .
2006 ) . for both objects
, we have a good constraint on the explosion time from prior optical detection / non - detection .
however , compact progenitor stars like blue supergiants brighten in the optical much more quickly than rsg stars with their extended envelopes , whose sed may peak in the uv a few days after explosion ( the difference comes from the energy penalty associated with the ejecta expansion ) . hence , optical non - detection / detection may not constrain the explosion date to better than one day . searching for
the uv and x - ray signatures at shock breakout would be more valuable in this respect .
the high quality spectral fits to the 13 photospheric - phase observations of sn 2005cs presented in the preceding sections are well suited to determine the distance to the sn and its host galaxy m51a ( ngc5194 ) .
ccrrrrrrrrr 2453552.25 & 2005 - 06 - 30 & 3.006 & 3.028 & 3.028 & 0.562 & 0.597 & 0.617 & 17835 & 16813 & 16122 + 2453553.25 & 2005 - 07 - 01 & 3.778 & 3.822 & 3.822 & 0.496 & 0.532 & 0.552 & 16272 & 15280 & 14649 + 2453554.50 & 2005 - 07 - 02 & 3.645 & 3.778 & 3.822 & 0.496 & 0.527 & 0.552 & 16242 & 15400 & 14619 + 2453555.25 & 2005 - 07 - 03 & 4.505 & 4.527 & 4.527 & 0.466 & 0.501 & 0.527 & 14438 & 13597 & 12935 + 2453556.00 & 2005 - 07 - 04 & 5.139 & 5.139 & 5.104 & 0.431 & 0.486 & 0.526 & 13498 & 12327 & 11456 + 2453557.75 & 2005 - 07 - 05 & 5.783 & 5.748 & 5.692 & 0.403 & 0.471 & 0.518 & 12792 & 11441 & 10525 + 2453558.25 & 2005 - 07 - 06 & 5.982 & 5.960 & 5.872 & 0.416 & 0.486 & 0.542 & 12184 & 10951 & 10020 + 2453561.25 & 2005 - 07 - 09 & 6.599 & 6.643 & 6.709 & 0.471 & 0.486 & 0.491 & 10200 & 9989 & 9899 + 2453562.50 & 2005 - 07 - 10 & 6.776 & 6.820 & 6.864 & 0.552 & 0.506 & 0.491 & 8997 & 9448 & 9689 + 2453563.50 & 2005 - 07 - 11 & 6.291 & 6.423 & 6.555 & 0.567 & 0.516 & 0.486 & 9118 & 9599 & 10080 + 2453564.25 & 2005 - 07 - 12 & 6.445 & 6.577 & 6.754 & 0.622 & 0.537 & 0.491 & 8547 & 9268 & 9929 + 2453566.00 & 2005 - 07 - 14 & 6.291 & 6.379 & 6.423 & 0.888 & 0.657 & 0.542 & 6923 & 7975 & 9178 + 2453580.25 & 2005 - 07 - 28 & 7.084 & 7.084 & 7.216 & 1.359 & 0.858 & 0.642 & 5240 & 6262 & 7525
+ & & & + + & & & + & & & + + & & & + & & & + lcccccc 2453547.8 & 8.9 & 0.04 & 1.0 & @xmath116 & 13 + 2453547.8 & 8.8 & 0.04 & 1.0 & @xmath117 & 13 + 2453547.9 & 8.7 & 0.04 & 1.0 & @xmath118 & 13 + 2453547.8 & 8.8 & 0.04 & 1.0 & @xmath117 & 12 + 2453548.0 & 8.5 & 0.04 & 1.0 & @xmath117 & 11 + 2453548.3 & 8.2 & 0.04 & 1.0 & @xmath117 & 10 + 2453548.6 & 7.7 & 0.04 & 1.0 & @xmath117 & 9 + 2453548.7 & 7.5 & 0.04 & 1.0 & @xmath117 & 8 + 2453547.8 & 8.0 & 0.04 & 0.9 & @xmath117 & 13 + 2453547.8 & 8.8 & 0.04 & 1.0 & @xmath117 & 13 + 2453547.8 & 9.7 & 0.04 & 1.1 & @xmath117 & 13 + 2453547.8 & 9.3 & 0.00 & 1.0 & @xmath117 & 13 + 2453547.8 & 8.8 & 0.04 & 1.0 & @xmath117 & 13 + 2453547.7 & 8.2 & 0.10 & 1.0 & @xmath117 & 13 + 2453547.2@xmath121 & 9.4 & 0.04 & 1.0 & @xmath116 & 13 + 2453547.8@xmath121 & 8.9 & 0.04 & 1.0 & @xmath116 & 13 + 2453548.4@xmath121 & 8.3 & 0.04 & 1.0 & @xmath116 & 13 + focusing first on the epm , we calculated the correction factors and color temperatures for all dates in our sn 2005cs sample .
as in dh06 , we employ three bandpass sets , using \{b , v } , \{b , v , i } , and \{v , i}. we log in table [ tab_epm_05cs ] the color temperature and associated correction factors for each model .
both quantities follow the same trend discussed in dessart & hillier ( 2005b ) or eastman et al .
( 1996 ) , with comparable values to those found at comparable epochs for sn 1999em ( dh06 ) . at late times corresponding to the recombination epoch
, the correction factors rise for \{b , v } and \{b , v , i } , driven by the strong feii and tiii line blanketing in the @xmath114-band .
the resulting fading of the @xmath114-band magnitude is , thus , not related to a variation in the flux - dilution magnitude . also given in table [ tab_epm_05cs ] are the angular sizes resulting from the epm , at each date .
note that these are determined with an estimated accuracy of @xmath12210% , which is a minimum error that accounts for the mismatch between synthetic and observed spectra .
the sn 2005cs photosphere approximately doubles in radius within one month of explosion , a comparable increase to that inferred for sn 1999em . using the model photospheric velocity ( table [ tab_model_05cs ] ) together with these angular sizes
, we use the method of least squares to fit the equation @xmath123 , taking time as the independent variable , and list the resulting distance and explosion dates in table [ tab_epm_05cs ] . using either the first 12 dates ( to test the effect of a shorter time baseline of two weeks
, we exclude the last date , corresponding to the recombination epoch ) or all 13 dates ( encompassing four weeks ) , we find a distance of [email protected] and an explosion that occurred on jd [email protected] , i.e. , 2.8 days prior to discovery ( @xmath45 days prior to the first observation ) .
the agreement between the three bandpasses is good .
lengthening the time baseline by two weeks by adding the thirteenth date yields the same results , which provides further credence to the reliability of this result . a plot of these epm data points for the @xmath117 passband , together with a linear fit to the data ,
are shown in fig .
[ fig_dist ] ( top curve ) .
the second approach is to fit directly the observations with our models . assuming homologous expansion and adopting the model @xmath52 ( as in the epm ) , we select an explosion time to obtain a photospheric radius at each epoch . each model photospheric radius
is then scaled to match that corresponding value , together with the synthetic magnitude .
we then calculate the distance modulus and the associated standard deviation .
we repeat this calculation for a range of explosion times , using a fine time increment , and select the explosion date that corresponds to the smallest scatter in distance modulus among all dates .
the results are given in table [ tab_seam_05cs ] .
using all 13 dates , a reddening of 0.04 , and our model photospheric velocity , we find a distance of [email protected] and an explosion date of jd 2453547.8 , in agreement with the results obtained with the epm ( here , the quoted error on the distance corresponds to the minimum distance - modulus scatter between all observations for the chosen time of explosion , and is typically 510% ) .
we investigate , as in dh06 the effect of changing some of the parameters , limiting ourselves to the @xmath117 set .
varying the number of observations from 13 to 8 moves forward the explosion date by about one day while the distance decreases by @xmath41.4mpc . reducing the time baseline to less than two weeks does not give a satisfactory solution : observations every three days over the first month is optimal for distance determinations based on the seam or the epm .
increasing the photospheric velocity at all times by 10% ( to mimic a systematic overestimate by 10% ) places the sn further away by 0.9mpc , the opposite being true for a 10% systematic underestimate of photospheric velocity .
these uncertainties are gross overestimates since our models would not give good fits to the observations at all times with such systematic over- or under - estimates .
our multi - wavelength multi - epoch modeling to spectroscopic observations of sn 2005cs yields a reddening of 0.04 .
a value greater than 0.1 would be impossible to accommodate with our models , but values of 0 or 0.1 would yield acceptable fits , albeit poorer . keeping the same models at all dates , changing the reddening from 0 to 0.1 makes the distance vary from 9.4mpc down to 8.2mpc . in dh06 , we discussed how allowing instead for changes in model parameters simultaneously , to adjust to the change in reddening , actually reduces the effect of the reddening uncertainty .
overall , the agreement between the epm and the seam - like distances to sn 2005cs give us confidence that our inference is accurate ( further confidence on such an accuracy could come from a cepheid distance to m 51 , in analogy to the case of sn 1999em and ngc 1637 ; see leonard et al .
2003 , baron et al .
2004 , dh06 ) .
we thus adopt a distance of [email protected] to sn 2005cs and m51 , and an explosion date of jd [email protected] .
this inferred explosion date is compatible within errors with the constraint of 2453549@xmath01 imposed by prior non - detection ( p06 ) , since even after the explosion has occurred , optical brightening is not instant . adopting a 10% uncertainty in the inferred angular size ( which corresponds to a minimum error associated with fitting mismatches )
leads to uncertainties of 0.5mpc and 0.5day for the distance and explosion date determined through the epm .
the additional uncertainties in photospheric velocity , reddening , and photometry , are small , so that , through our use of many observations , our inferred distance should be accurate to within @xmath12210% .
our inferred distance is slightly larger than two recent epm - like distance measurements to sn 2005cs .
using phoenix version 14 ( hauschildt & baron 1999 ) , baron et al .
( 2007 ) obtained a seam distance to sn 2005cs of 7.9@xmath124mpc using all @xmath125 magnitudes , but at only two epochs . in some sense
the level of agreement is satisfying : phoenix and cmfgen are two very distinct codes , and use different assumptions in the modeling .
takts & vink ( 2006 ) used the epm ( with analytical correction factors and an estimate of the reddening @xmath126 ) and the standard candle method of hamuy & pinto ( 2002 ) to determine a distance of [email protected] and [email protected] , respectively .
their epm distance was actually an average of two determinations one in which the explosion time was fixed ( d=8.34@xmath127 ) and one in which the explosion time was a free parameter ( d=@xmath128 ) .
the later gave a better fit to the data , but yielded an explosion time 4 days later than that inferred directly from the observations .
takts & vink ( 2006 ) find a challenging enhancement in the photospheric angular size of about 1020 over the same period , which , in light of our cmfgen model results , would suggest an outward migration , in the lagrangean ( or co - moving ) frame , of the photosphere with time . at later times our results , and those of takts & vink , appear to be consistent ; at early times takts & vink estimate much hotter photospheric temperatures , and smaller photospheric radii .
in general , our inferred distance is larger than most other estimates in the literature .
previously , using light - curve or spectroscopic modeling of the type ic sn 1994i , iwamoto et al .
( 1994 ) and baron et al .
( 1996 ) obtained a distance of 6.9@xmath01mpc and [email protected] , respectively , although the reddening may be uncertain in the line of sight to this object .
georgiev et al .
( 1990 ) obtained a distance of [email protected] using surface photometry of stellar associations .
feldmeier et al .
( 1997 ) found a distance of [email protected] using planetary nebulae , revised downward by ciardullo et al .
( 2002 ) to [email protected] with an improved reddening of 0.04 .
note that this is the reddening we infer in the line of sight to sn 2005cs .
tonry et al ( 2001 ) infer a distance to ngc5195 ( m 51b , a companion galaxy to m 51 ) of 7.66@xmath01mpc from surface brightness fluctuations .
our estimate is compatible , within the errors , with that of tonry et al .
( 2001 ) , but mildly in disagreement with the other studies .
obtaining a cepheid distance to m 51 would help in pinning down the distance to that galaxy and in providing an independent test for the accuracy of our distance estimate .
ccrrrrrrrrr 2453836.6 & 2006 - 04 - 11 & 2.345 & 2.373 & 2.352 & 0.586 & 0.663 & 0.788 & 24899 & 21736 & 17632 + 2453837.6 & 2006 - 04 - 12 & 2.856 & 2.884 & 2.842 & 0.563 & 0.657 & 0.734 & 21916 & 18753 & 16511 + 2453842.6 & 2006 - 04 - 17 & 4.838 & 4.929 & 4.873 & 0.499 & 0.557 & 0.608 & 15330 & 14009 & 12847 + 2453844.6 & 2006 - 04 - 19 & 5.111 & 5.216 & 5.160 & 0.480 & 0.538 & 0.589 & 15030 & 13708 & 12547 + 2453845.8 & 2006 - 04 - 20 & 5.580 & 5.685 & 5.615 & 0.454 & 0.512 & 0.560 & 14489 & 13188 & 12107 + 2453846.7 & 2006 - 04 - 21 & 5.524 & 5.643 & 5.580 & 0.454 & 0.512 & 0.560 & 14489 & 13188 & 12107 + 2453849.6 & 2006 - 04 - 24 & 6.239 & 6.351 & 6.365 & 0.422 & 0.490 & 0.544 & 13548 & 12127 & 11026 + 2453850.6 & 2006 - 04 - 25 & 7.527 & 7.569 & 7.597 & 0.451 & 0.496 & 0.531 & 11126 & 10445 & 9864 + 2453855.7 & 2006 - 04 - 30 & 7.772 & 7.884 & 7.940 & 0.583 & 0.525 & 0.486 & 8963 & 9524 & 10125 + 2453856.8 & 2006 - 05 - 01 & 7.891 & 8.003 & 8.039 & 0.634 & 0.544 & 0.486 & 8423 & 9164 & 10025 + 2453857.6 & 2006 - 05 - 02 & 7.828 & 7.961 & 8.032 & 0.634 & 0.544 & 0.486 & 8423 & 9164 & 10025 + 2453858.7 & 2006 - 05 - 03 & 8.165 & 8.249 & 8.333 & 0.692 & 0.576 & 0.502 & 7802 & 8583 & 9524 + 2453859.6 & 2006 - 05 - 04 & 8.039 & 8.165 & 8.263 & 0.692 & 0.576 & 0.502 & 7802 & 8583 & 9524 + 2453862.7 & 2006 - 05 - 07 & 8.627 & 8.732 & 8.823 & 0.862 & 0.653 & 0.531 & 6681 & 7562 & 8743 + 2453867.7 & 2006 - 05 - 12 & 8.172 & 8.396 & 8.571 & 0.862 & 0.653 & 0.531 & 6681 & 7562 & 8743 + 2453876.7 & 2006 - 05 - 21 & 7.969 & 8.200 & 8.375 & 1.058 & 0.724 & 0.560 & 6001 & 7022 & 8343 + & & & + + & & & + & & & + + & & & + & & & + lcccccc 2453833.4 & 17.4 & 0.4 & 1.0 & @xmath116 & 16 + 2453833.4 & 17.1 & 0.4 & 1.0 & @xmath129 & 16 + 2453833.5 & 16.7 & 0.4 & 1.0 & @xmath130 & 16 + 2453833.5 & 16.8 & 0.4 & 1.0 & @xmath129 & 15 + 2453833.5 & 16.8 & 0.4 & 1.0 & @xmath129 & 14 + 2453833.3 & 17.4 & 0.4 & 1.0 & @xmath129 & 13 + 2453833.3 & 17.6 & 0.4 & 1.0 & @xmath129 & 12 + 2453833.1 & 18.0 & 0.4 & 1.0 & @xmath129 & 11 + 2453833.1 & 18.3 & 0.4 & 1.0 & @xmath129 & 10 + 2453832.9 & 18.9 & 0.4 & 1.0 & @xmath129 & 9 + 2453833.4 & 17.1 & 0.4 & 1.0 & @xmath129 & 16 + 2453833.4 & 18.8 & 0.4 & 1.1 & @xmath129 & 16 + 2453833.4 & 15.4 & 0.4 & 0.9 & @xmath129 & 16 + 2453833.4 & 15.9 & 0.45 & 1.0 & @xmath129 & 16 + 2453833.4 & 17.1 & 0.40 & 1.0 & @xmath129 & 16 + 2453833.4 & 18.3 & 0.35 & 1.0 & @xmath129 & 16
+ 2453833.0@xmath121 & 17.6 & 0.4 & 1.0 & @xmath129 & 16 + 2453833.5@xmath121 & 16.9 & 0.4 & 1.0 & @xmath129 & 16 + 2453834.0@xmath121 & 16.2 & 0.4 & 1.0 & @xmath129 & 16 + we now apply the methods of the previous section to sn 2006bp .
to avoid the hydrogen - recombination phase , when steady - state cmfgen models struggle to reproduce the hydrogen balmer and naid lines , we exclude the two extreme observations in our sample , those of the 5th and of the 21st of june 2006 , leaving us with 16 observations and a time coverage of 4 weeks .
we present the epm - inferred quantities in table [ tab_epm_06bp ] , giving the angular size , correction factors and color temperatures for the three sets of bandpasses @xmath116 , @xmath129 , and @xmath130 . because of the faster expansion of sn 2006bp compared to sn 2005cs , the photospheric angular size more than triples over the first month of observation . at the recombination epoch ,
the associated front , which coincides closely with the photosphere , recedes in the co - moving frame at a rate that is comparable to the expansion rate of the ejecta . at late times , the angular size of the photosphere is , thus , roughly constant . using all 16 observations in our sample and our epm - based technique leads to a distance of [email protected] for the @xmath129 set , only slightly varying with bandpass set ( [email protected] for the @xmath116 set and [email protected] for the @xmath130 set ; table [ tab_epm_06bp ] ) . including only the first 13 dates ( to assess the effect of a shorter time baseline of one month ) leads to a higher distance by 0.20.3mpc in each band and a larger error of @xmath00.8mpc .
the explosion date is also in good agreement for either 13 or 16 dates in our sample , with a julian date of [email protected] . a plot of these epm data points for the @xmath129 passband , together with a linear fit to the data , are shown in fig . [ fig_dist ] ( bottom curve ) .
we also use our synthetic spectra and magnitudes to constrain directly the distance , in a fashion similar to the seam .
we find a distance in agreement with the epm inference , with 17.1mpc for the @xmath129 set ( table [ tab_seam_06bp ] ) .
uncertainties are larger than the scatter of values across bandpasses , but should remain at the 510% level , in particular due to the large number of observations .
errors are unlikely to be purely statistical , but they have some level of randomness as the deviations from a perfect match of synthetic to observed spectra are very different for individual epochs ( blue featureless spectrum at early time to strongly blanketed spectrum at the recombination epoch ) , different spectral regions ( the blue and red wavelength regions reflect different contributions from different elements that we model with variable levels of success ) . reducing the number of observations to 9 dates increases the distance estimate by 1.7mpc .
uncertainties in reddening and photospheric velocity have a similar effect to that described for sn 2005cs ( [ sect_05cs ] ) and sn 1999em ( dh06 ) . to conclude , we estimate a distance to sn 2006bp and its galaxy host ngc3953 of [email protected] together with an explosion date of [email protected] only 1.7 days after discovery ( @xmath43 days prior to the first observation ) .
this makes it the earliest - discovered type ii - p sn and gives further support to the novel properties of the sn ejecta inferred from the first observed spectra ( steep density distribution , highly ionized ejecta with heii4686 line emission ) .
this distance estimate is in close match to the value of 17mpc obtained by tully ( 1988 ) .
this explosion date is also in agreement with the date of april 9th , inferred by quimby et al .
( 2007 ) from fits to optical rotse - iii light curve ( note however that the sn may remain dim in the optical for many hours after breakout ) .
we have presented quantitative spectroscopic analyses of two type ii - p sne , using observations triggered at an exceptionally early time after explosion ( 35 days ) , providing a glimpse into the ejecta structure relatively soon after shock breakout .
we infer a very high - temperature / very high ionization in the corresponding photospheric layers and in particular identify the presence of broad , ejecta - related , heii4686 in the first two observations of sn 2006bp ( narrow emission in heii4686 is also identified by quimby et al .
( 2007 ) , related to material outside of the sn ejecta , which also testifies for the high temperature associated with shock breakout ) . in fig .
[ fig_gammas ] , we show the ionization level for h , he , c , n , o , and fe in the ejecta computed with cmfgen and for models that fit the first spectroscopic and photometric observations of sn 2006bp ( left panel ) , sn 2005cs ( middle panel ) , and sn 1999em ( right panel ) . these correspond , in the same sequence , to 3 , 5 , and 6 days after explosion , as determined in this work and in dh06 .
one clearly sees in this figure the evolution to lower ionization through this sequence , following directly the elapsed time since explosion .
oxygen changes from being twice- to singly - ionized at mid - heights , from sn 2006bp to sn 1999em .
helium is singly ionized throughout the sn 1999em ejecta , but twice ionized in the inner and outer regions of the sn 2006bp ejecta .
this trend towards much higher ionization at earlier times supports the identification of heii4686 emission in the sn 2006bp ejecta , as well as oii lines for both sn 2006bp ( at about 5 days after explosion ) and sn 2005cs ( in the discovery spectrum ) .
oii lines were not detected unambiguously in the first spectrum taken of sn 1999em because it was taken too late , about 12 days later than sn 2005cs relative to the explosion time .
moreover , at such early times , adopting the nitrogen abundance of dh06 now overestimates the strength of nii lines and underestimates the strength of oii lines , justifying the adoption in this work of cno abundances closer to solar , i.e. c / he=0.0004 , n / he=0.0013 , o / he = 0.0016 , in agreement with the surface chemistry of blue supergiants analyzed by crowther et al . (
2006 ) . by comparison , dh06 used c / he=0.00017 , n / he=0.0068 , o / he = 0.0001 ( all given by number ; for both works we use h / he=5 ) . we thus concur with the findings of baron et al .
( 2007 ) on the necessary cno chemistry to reproduce the observations of oii and nii lines in the early spectra of sn 2005cs ( as well as in sn 2006bp ) , with a depletion of carbon and oxygen compared to solar that is modest rather than severe . in general , an accurate determination of cno abundances in the fast - expanding photospheres of type ii - p sne is difficult since the corresponding lines are weak and overlap with lines of other species such as hei .
having recourse to quantitative spectroscopy of the emergent light from their supergiant progenitors , which can be found in number and nearby , seems a more viable route to constraining the surface chemistry of such pre - sn massive stars . in both objects , the near featureless spectral appearance over the first few days requires the choice of a very high power - law density exponent of 2050 to dwarf the emission volume of all optical lines , as well as very high ejecta velocities to reproduce the large widths of emission line features .
this is admittedly well above the value @xmath73 for the propagation of a shock wave through the envelope of a massive star progenitor , both obtained theoretically ( imshennik & nadezhin 1988 ; shigeyama & nomoto 1990 , chevalier 1982 ; ensman & burrows 1992 , blinnikov et al .
2000 ) and inferred observationally ( see , e.g. , eastman & kirshner 1989 ; schmutz et al . 1990 ; baron et al .
2000 ; dessart & hillier 2005a,2006a ) .
nonetheless , for the outermost layers of the progenitor envelope , a much steeper density profile is expected , with values in the vicinity of a few tens ( blinnikov et al .
2000 , and in particular their fig . 1 ) , and , under certain conditions , a small amount of mass at the progenitor surface may even reach relativistic speeds as the shock breaks out ( matzner & mckee 1999 ) .
unfortunately , the radiation - hydrodynamics codes used to compute the ejecta structures of core - collapse sn are lagrangean , and are thus not well designed to describe the dynamics and the structure of the low density surface layers . in those fully - ionized regions , the luminosity is super - eddington by up to a few orders of magnitude , even when accounting only for the electron - scattering opacity .
when accounting for the additional contribution of optically thick and thin lines , we find that the radiation pressure can be very substantial , following the combination of a huge luminosity , a low density , and a large velocity gradient ( castor , abbott , & klein 1975 ) . in the cmfgen model of sn 2006bp on 30 june 2006 , the radiative acceleration in the layers above the photosphere reaches a few thousand cms@xmath131 , which combined with a velocity of a few 10000and a length scale of 10@xmath132 cm lead to a velocity gain of few tens of percent ( assuming the acceleration remains constant ) .
sooner after shock breakout , the radiative acceleration could be even larger .
overall , we expect that the outer optically - thin layers would adopt a steeper density distribution and a faster - than - linear velocity distribution with radius .
we plan to investigate these issues more quantitatively in the future , but at present , it seems that the earliest observations of sne 2005cs and 2006bp support the very steep density distribution obtained in radiation - hydrodynamics simulations of core - collapse sn ejecta soon after shock breakout . before closing this section
, we note that sn 1999gi is another type ii - p sn discovered very early , perhaps only @xmath44@xmath133 days after explosion ( leonard et al .
interestingly , leonard et al .
report the presence of a full p - cygni profile shifted from the rest wavelength of h@xmath39 by @xmath134 in the first spectrum of sn 1999gi , and speculate on the potential association with ejecta inhomogeneities . in the context of a density inhomogeneity far above the photosphere
, a blueshifted absorption is a possibility , but a p - cygni profile entirely blueshifted , and by that amount , is not ( see , however , dessart & hillier 2005a ) .
we reproduce their observation in fig .
[ fig_he2 ] , and overplot the first spectrum of sn 2006bp ( red line ) .
given the close correspondence between the two spectra and the likely early detection of sn 1999gi , we propose that , in fact , the broad line feature just blueward of h@xmath39 in the spectrum of sn 1999gi is due to heii4686 , as for sn 2006bp .
having inferred the reddening and the distance to sne 2005cs and 2006bp , we can deduce the absolute magnitude for each event at all epochs studied .
together with results presented for sn 1999em in dh06 , we show in the top - left panel of fig .
[ fig_mag_vphot ] the evolution with time since explosion of the absolute magnitude in @xmath114 ( black ) , @xmath109 ( blue ) , and @xmath115 ( red ) , for sn 2005cs ( solid line ) , sn 2006bp ( dashed line ; @xmath135 is shown instead of @xmath115 ) , and sn 1999em ( dotted line ) .
note that the time of explosion is computed together with the distance , rather than estimated based on prior non - detection of the sn on pre - explosion images of the corresponding host - galaxy .
while the spectroscopic and photometric evolution of three events are comparable in all spectral regions at a given time after explosion , they display a sizable range of intrinsic brightness .
sn 2005cs is about two magnitudes fainter than sn 2006bp , about a factor of six in ( bolometric ) luminosity , with sn 1999em somewhat less luminous than sn 2006bp .
sn 2006bp was detected earliest of all three and is in fact the first type ii - p to be caught so soon after explosion , with a @xmath42-day delay at most ( note that 1987a and 1993j are not type ii - p sne ) . in the top - right panel of fig .
[ fig_mag_vphot ] , we show the corresponding evolution of the model photospheric velocity , for sne 2005cs , 1999em , and 2006bp . the above trend is maintained , i.e. , larger intrinsic brightness is associated with larger ejecta velocity .
note that if the bolometric luminosity scaled with the square of the ejecta velocity , a brightening by 1.5mag would result from a factor of 2 enhancement in velocity , which is compatible with our inferences for sne 2005cs and 2006bp ( top row of fig .
[ fig_mag_vphot ] ) .
since we adopt similar ejecta structures ( after one week past explosion ) , this suggests that the bolometric luminosity correlates with the ejecta kinetic energy in the hydrogen shell , as pointed out in the past by , for example , falk & arnett ( 1977 ) , litvinova & nadezhin ( 1983 ) , hamuy ( 2003 ) , zampieri et al .
( 2003 ) , pastorello et al .
the late time brightness is not controlled by the energy deposited at shock passage , but rather stems from the amount and location of the unstable nickel at the start of homologous expansion , so the above trend may not carry over until much later times , although it might , but for different reasons ( higher energy explosions may systematically produce more @xmath8ni ) . similarly , and for completeness
, we include in the bottom panels of fig .
[ fig_mag_vphot ] the temporal evolution of the photospheric radius ( left ) and temperature ( right ) .
note the smaller extent of the slowly - expanding ejecta of sn2005cs , the converging photospheric temperatures , and the near - constant photospheric radii at the recombination epoch for all three sne , tightly linked to the plateau appearance of the visual light - curves at such times .
our modeling approach has been first to fit the optical range , using both the overall shape of the sed and spectral lines formed in the sn ejecta . as a consistency check , as well as an additional constraint on the resulting model parameters and the choice of reddening
, we compared the synthetic spectrum to the observed _ swift _ uvot values . for both sn 2005cs and sn 2006bp
, we obtain very satisfactory fits and we thus conclude that the observed uv and optical light emerges from the photospheric layers of the sn ejecta , over the first 1 - 2 months after explosion ( starting 3 - 5 days afterwards ) .
this result is unambiguous , and , thus , no sizable contribution to the uv and optical fluxes stems from an interaction with the csm .
a further evidence for the photospheric origin of the uv flux is the identification of the mgii2800line as a broad p - cygni profile in _ swift _ uvot grism spectra ( brown et al .
2007 ) , with a width compatible with the photospheric velocity at the corresponding epoch .
so , while circumstellar interaction may be important in some type ii sne ( the iin events ) , it seems to contribute negligibly to the uv and optical light we received from sn 2005cs and sn 2006bp , and more generally in agreement with the weak radio and x - ray emission from type ii - p sne ( see , e.g. , pooley et al .
2002 ; chevalier et al . 2006 ) .
the recent spectroscopic analysis by baron et al . ( 2007 ) of sn 2005cs at two epochs in the photospheric phase agrees in many respects with the findings of the present study for that object and , thus , gives some credence to the type ii sn spectroscopic analyzes performed with phoenix ( hauschildt & baron 1999 ) and cmfgen ( hillier & miller 1998 ; dessart & hillier 2005a ) .
this is particularly satisfying since phoenix and cmfgen are two very distinct codes , and use different assumptions in the modeling .
a potential important difference in assumptions is our neglect of energy contribution from radioactive decay throughout this work .
by contrast with baron et al .
( 2007 ) , we find no strong evidence for nickel mixing and non - thermal energy deposition at the photosphere over the first month after explosion ( in this work , we obtain good fits to observations prior to the hydrogen - recombination epoch without invoking this additional source of energy ) .
the ejecta ionization structure computed by the two codes may thus differ . in this work ,
we infer a reddening @xmath36 , in agreement with baron et al .
( 2007 ) who obtain a value of 0.035 .
we also concur with the findings of baron et al .
( 2007 ) on the necessary cno chemistry to reproduce the observations of oii and nii lines in the early spectra of sn 2005cs ( as well as in sn 2006bp ) , with a depletion of carbon and oxygen compared to solar that is modest rather than severe . as discussed in [ sect_ion ] , this constraint can be placed for sn 2005cs , by contrast with , e.g. , sn 1999em , mostly because it was observed at an earlier time when the ionization conditions at the photosphere are ripe for oii line emission .
the identification of oii and nii lines can explain most , if not all , the mysterious kinks seen in the blue wing of h@xmath39 or hei5875 ( see , e.g. , leonard et al .
2002a , b who reported such kinks in the spectra of sne 1999em and 1999gi ) . here ,
the abundances adopted at early times for both sne 2005cs and 2006bp are , given as mass fractions , @xmath46=0.55 , @xmath47=0.44 , @xmath48=0.0005 , @xmath49=0.002 , @xmath50=0.003 , and @xmath51=0.0013 . at later times , when in particular oi7770 gains in strength , we enhance both the carbon and the oxygen abundances , with c / he=0.001 , o / he = 0.01 , and reduce the nitrogen abundance , n / he=0.001 .
again , observations support the same ejecta chemistry at the photosphere at corresponding epochs after explosion for both sne .
we are however reserved concerning the sodium abundance , which we predicted , based on steady - state cmfgen models , to be enhanced by a factor of four over its primordial value in sn 1999em ( dh06 ) .
comparable enhancements would be needed here for 2005cs and 2006bp , and , indeed , baron et al .
( 2007 ) predict a sodium enhancement of a factor of 10 .
_ time - dependent _ cmfgen models performed with a _ primordial _ sodium abundance , however , suggest that naid may remain as strong and broad as observed ( dessart & hillier 2007a , b ) ; further work is needed to quantify the effect of time - dependence on abundance determinations . for the observations of the 14th of july 2005 ,
baron et al . ( 2007 ) find at the photosphere a temperature of 6000k , a velocity of 4000kms@xmath88 and a density exponent of 8 . in the same order and for that date
, we find 6700k , 3930kms@xmath88 , and 10 , in agreement with their values .
for the observations of the 31st of july 2005 , baron et al .
find at the photosphere a temperature of 5500k , a velocity of 6000kms@xmath88 and a density exponent of 12 ( see their 6 ) . in this work , for the observations of the 28th of july 2005 , just three days before ( note that observations do not change visibly over just a few days at such late times ) , we find in the same order 6140k , 2240kms@xmath88 and a density exponent of 10 . while we fit the cfa observations reasonably well with this parameter set on that date ( fig.[05cs_0728 ] ) , baron et al .
require a photospheric velocity nearly three times as large to do so , and also larger than the value on day 17 ( in fact nearly as large as on the first spectrum taken , for which we use 6880kms@xmath88 ) .
upon inspection of their figs .
12 - 13 , it appears that h@xmath38 peaks at @xmath46480 , while the observations they use , as originally published in p06 , and our ( restframe ) cfa observations show a peak at 6550 .
this difference of 70 corresponds to a velocity shift of 3200kms@xmath88 at this wavelength and suggests that their extravagant proposition of an outward moving photosphere between day 17 and day 34 ( see their 6 ) may stem solely from a problem with the wavelength of their flux datapoints .
baron et al .
( 2007 ) obtained a seam distance to sn 2005cs of 7.9@xmath124mpc using all @xmath125 magnitudes , but at only two epochs , which is in agreement with our estimate of [email protected] using @xmath117 and 12 or 13 epochs .
although the observed spectra of sne 2005cs and 2006bp presented in this work are different in many ways , these reflect mostly the difference in the ejecta expansion rate ( small for 2005cs ) , the reddening in the line of sight ( large for 2006bp ) , and the elapsed time between explosion and discovery ( very short for 2006bp ) . as shown in fig .
[ fig_mag_vphot ] , both sne evolve in a similar fashion , 2005cs being merely underluminous and boasting a lower ejecta kinetic energy .
we have presented a multi - epoch multi - wavelength quantitative spectroscopic analysis of the two type ii - p sn events 2005cs and 2006bp , over the first two months after explosion , and have reached similar conclusions on the ejecta properties to those for sn 1999em ( dh06 ) .
the progenitor stars have a composition compatible with a bsg / rsg progenitor .
the spectral evolution shows a reddening of the sed with time , with a flux peaking in the uv at very early times , but progressively shifting to longer wavelengths . the photosphere recedes to deeper layers with time , reaching mass shells moving at about 2000 - 3000about 6 weeks after explosion and having enhanced
carbon and oxygen , and reduced nitrogen , abundances , compared to the surface layers .
we find a density distribution that is very steep at early times , and we surmise that this could be an imprint of the shock breakout epoch . from about a week until six weeks after explosion , we find a more standard and constant density exponent of ten for the corresponding power law distribution , compared to a value of 50 ( for sn 2006bp ) and 20 ( for sn 2005cs ) used to model the first observation .
_ swift _ uvot photometry and ground - based optical spectroscopy and photometry support a common origin for the uv and the optical light , so that we anticipate that if csm interaction does occur , it does not contribute significantly to the observed uv and optical fluxes . using our analysis , we infer the distance and the explosion date of each sn .
we find , in that order , jd [email protected] days ( 2.8 days prior to discovery ) and [email protected] for sn 2005cs , and jd [email protected] ( 1.7 days prior to discovery ) and [email protected] for sn 2006bp .
the large number of observations , the quality of our fits , the agreement between the two methods followed suggests a very high level of internal consistency of the distances determined here with type ii sne . the actual level of accuracy on such determined distances , which we estimate to be on the order of 10% ( related to the spectral fitting accuracy )
, would be best assessed by confrontation with alternate methods ( cepheids for example , although all alternate methods have their own inaccuracies , claimed and otherwise ) , the determination of distances to a galaxy host with well - recorded multiple type ii sn events ( the typical type ii sn rate of 2 per century per galaxy makes this proposition somewhat unrealistic ) , a galactic sn that could also be resolved with an interferometer or a large - aperture telescope .
our perspective on the uncertainty of our distance measurements is that , were we to increase the number of type ii sne suitable for analysis in m 51 to 10 or 100 , the scatter in the distance determined with our technique for that sample would not be larger than 10% .
but this would still be no proof that the resulting average distance is indeed the right one since enlarging the sample would only allow for an internal consistency check on our method . in the short term , a cepheid distance to the galaxy host of 2005cs and 2006bp is highly desirable , and would provide an independent check on the distance computed with type ii - p sne , as was done with success in the past with sn 1999em ( leonard et al . 2003 ; baron et al . 2004 ; dh06 ) , giving further support to the potential use of such objects for distance determinations in the universe .
this work fits within an ambitious project to build a general understanding of sn ejecta , to infer properties of the progenitor star ( with ramifications for massive star evolution ) and the explosion mechanisms at the origin of the event .
a major ongoing theoretical development is the treatment of time - dependent effects both in the radiation field and in the level populations to allow for a direct modeling of the evolution over months after shock breakout , through the photospheric and into the nebular phases , and based on a range of hydrodynamical models of the explosion . in parallel ,
high - quality multi - epoch multi - wavelength observations for a substantial number of nearby sne , irrespective of the type , galaxy host properties , covering from as soon as possible after explosion until the nebular phase , is eagerly sought .
we thank stan woosley for providing the ejecta structure of a 15model .
we also thank andrea pastorello , robert quimby , and doug leonard for providing their optical spectra of sn 2005cs , sn 2006bp , and sn 1999gi , respectively .
l.d . acknowledges support for this work from the scientific discovery through advanced computing ( scidac ) program of the doe , under grant numbers de - fc02 - 01er41184 and de - fc02 - 06er41452 , and from the nsf under grant number ast-0504947 .
this research was supported in part by the national science foundation under grant no .
phy05 - 51164 to the kavli institute for theoretical physics at uc santa barbara and ast-0606772 to harvard university .
this work is sponsored at psu by nasa contract nas5 - 00136 . | we analyze the type ii plateau supernovae ( sn ii - p ) 2005cs and 2006bp with the non - lte model atmosphere code cmfgen .
we fit 13 spectra in the first month for sn 2005cs and 18 for sn 2006bp .
_ swift _ ultraviolet photometry and ground - based optical photometry calibrate each spectrum .
our analysis shows both objects were discovered less than 3 days after they exploded , making these the earliest sn ii - p spectra ever studied .
they reveal broad and very weak lines from highly - ionized fast ejecta with an extremely steep density profile .
we identify heii4686 emission in the sn 2006bp ejecta .
days later , the spectra resemble the prototypical type ii - p sn 1999em , which had a supergiant - like photospheric composition . despite the association of sn 2005cs with possible x - ray emission ,
the emergent uv and optical light comes from the photosphere , not from circumstellar emission .
we surmise that the very steep density fall - off we infer at early times may be a fossil of the combined actions of the shock wave passage and radiation driving at shock breakout .
based on tailored cmfgen models , the direct - fitting technique and the expanding photosphere method both yield distances and explosion times that agree within a few percent .
we derive a distance to ngc 5194 , the host of sn 2005cs , of [email protected] and [email protected] for sn 2006bp in ngc 3953 .
the luminosity of sn 2006bp is 1.5 times that of sn 1999em , and 6 times that of sn 2005cs .
reliable distances to type ii - p supernovae that do not depend on a small range in luminosity provide an independent route to the hubble constant and improved constraints on other cosmological parameters . |
( heron formula ) let @xmath0 , @xmath1 and @xmath2 be the lengths of the sides of a triangle , @xmath3 the half of the perimeter and @xmath4 the area . then holds @xmath5 @xmath6 now we define : a _ heron triangle _ is a triangle with integer sides and area .
[ prop : peri ] the perimeter of a heron triangle is even .
assume the perimeter is odd .
then all factors on the right hand side of equation ( [ eq : heron ] ) are halves of odd integers , so the product is an odd integer divided by @xmath7 , which can not be an integer .
@xmath6 if the sides of a heron triangle have a common factor @xmath8 , then the area is divisible by @xmath9 ,
i.e. the triangle scaled down by factor @xmath8 is also heron .
furthermore the area of a heron triangle is always divisible by @xmath10 .
it is sufficient to show this for @xmath8 prime .
for @xmath11 it is obvious , for @xmath12 consider the term @xmath13 modulo 4 and one gets that it ca nt be congruent @xmath14 .
the second part can in a similar way be checked modulo @xmath15 and @xmath16 .
@xmath6 now i will give some examples for heron triangles .
there are two main classifications for triangles : \{acute , right - angled and obtuse } and \{regular , isosceles and generic}. so we first search for right - angled and isosceles heron triangles .
right - angled triangles have to be pythagorean but the reverse is also true : any pythagorean triangle is a heron triangle .
the lengths of the sides are integer , so we only have to proof that the area is integer . for the area @xmath4 holds @xmath17 , where @xmath0 and @xmath1 are the lengths of the catheti .
hence if @xmath4 is not an integer then @xmath0 and @xmath1 have to be odd . but then follows @xmath18 and @xmath19 and that is not possible.@xmath6 any heron isosceles triangle is divided by the center line into two congruent pythagorean triangles .
using proposition [ prop : peri ] we get that the basis must be an even number , say @xmath20 .
let @xmath1 the length of the arms and @xmath21 be the length of the center line .
now we only have to proof that @xmath21 is an integer .
we get @xmath22 and from the heron formula @xmath23 . from @xmath4 is an integer follows that @xmath24 is a square number , but that is @xmath25 , so @xmath21 is integer.@xmath6 there is not intersection of this two classes right - angled isosceles triangles do nt have integer sides .
the examples in the previous section show that some classes of heron triangles can trivially be represented as lattice triangles .
first we show the following weaker result .
any heron triangle can be represented with rational coordinates .
we only have to show that the height is rational and the length of the height - tiles . but
this follows from the equations @xmath26 and @xmath27.@xmath6 there are heron triangles without integer heights as the example @xmath28 with area @xmath29 shows . any heron triangle can be represented with integer coordinates .
we know that it is representable with rational coordinates
. then we can apply proposition [ prop : em2].@xmath6
a tetrahedron with given edges exists if and only if all faces exist ( i.e. in all triangles the triangle inequality holds , which is equivalent to : the right hand side of equation ( [ eq : heron ] ) is positive ) and the right hand side of equation ( [ eq : v3 ] ) is positive.@xmath6 if the sides of a heron tetrahedron have a common factor @xmath8 , then the volume is divisible by @xmath33 , i.e. the tetrahedron scaled down by factor @xmath8 is also heron .
furthermore the volume of a heron tetrahedron is always divisible by @xmath34 . as for triangles
it is sufficient to show this for @xmath8 prime .
for @xmath35 it is obvious . for @xmath36 one
have to proof that if the determinant on the right hand side of equation ( [ eq : v3 ] ) is the double of a square and all areas of the faces are integer ( i.e. the right hand side of equation ( [ eq : heron ] ) is a square ) then it is divisible by @xmath37 . because i have no elegant proof for that this was done by a `` brute force '' calculation modulo @xmath38 and modulo @xmath7 using a computer program . 1 .
[ sr : opp ] the opposite edges have the same length .
[ sr : cong ] all faces are congruent ( acute ) triangles .
[ sr : peri ] the perimeters of the faces coincide .
[ sr : area ] all faces have the same area .
[ sr : hei ] all body heights have the same length
[ sr : box ] the tetrahedron can be inscribed in a box ( see figure [ fig : srinabox ] ) .
[ sr:3p ] two of the following three points coincide : the mass center , the center of the circumscribed and inscribed sphere .
[ sr : circ ] the circumcircle radii of the faces coincide .
[ sr : angle ] in all vertices the sum of the face angles equals @xmath42 .
[ sr : net ] the net of the tetrahedron is a triangle with connected center points . ( 30,40 ) ( 25,5)(0,0)[cc][sr : circ ] ( 25,15)(0,0)[cc][sr : angle ] ( 25,25)(0,0)[cc][sr : net ] ( 25,35)(0,0)[cc][sr : peri ] ( 15,5)(0,0)[cc][sr:3p]ci ( 15,15)(0,0)[cc][sr:3p]mc ( 15,25)(0,0)[cc][sr : opp ] ( 15,35)(0,0)[cc][sr : cong ] ( 5,5)(0,0)[cc][sr : box ] ( 5,15)(0,0)[cc][sr:3p]mi ( 5,25)(0,0)[cc][sr : hei ] ( 5,35)(0,0)[cc][sr : area ] ( 15,27)(0,1)6 ( 5,17)(0,1)6 ( 15,17)(0,1)6 ( 25,17)(0,1)6 ( 5,7)(0,1)6 ( 25,7)(0,1)6 ( 5,30)(0,1)3 ( 5,30)(0,-1)3 ( 16,35)(1,0)8 ( 6,5)(1,0)7 ( 17,5)(1,0)7 ( 14,35)(-1,0)8 ( 7,7)(1,1)6 ( 7,33)(1,-1)6 ( 23,27)(-1,1)6 ( 23,33)(-1,-1)6 ( 10,15)(1 , 2)4 ( 10,15)(-1 , -2)4 and much more . for
a scheme of the proofs see figure [ fig : srproof ] .
if one of these properties is fulfilled we call the tetrahedron _ semi - regular_. i also found the name _
isosceles _ for that ( e.g. on @xcite , there are a lot of references ) , but i prefer the name semi - regular because these tetrahedra are `` more regular than isosceles '' , at least i will show in section [ sec : if ] a generalization of isosceles triangles which fits better in this context . hence to get a heron semi - regular tetrahedron one has only to take an acute heron triangle and verify whether the volume is integer , and the volume can be calculated by @xmath43 .
so the smallest examples of this type can be easy computed as @xmath44 , @xmath45 and @xmath46 its multiples .
these were the first known heron tetrahedra .
see appendix [ ap : sr ] for more examples of heron semi - regular tetrahedra .
the right - angled triangle can be generalized in two ways . the first type i will call _ right - angled - vertex tetrahedron _ : a tetrahedron where in one vertex @xmath47 all angles are right angles ( see left picture of figure [ fig : ra ] ) .
let @xmath48 , @xmath49 and @xmath50 be the lengths of these edges which ends in @xmath47 .
then the area of the faces which contains @xmath47 are automatically integer , and also the volume is integer .
so there s only left to examine the lengths @xmath51 , @xmath52 and @xmath53 , and the area of the face opposite to @xmath47 .
this area can easy be computed by formula ( [ eq : heron ] ) as @xmath54 .
let @xmath55 , @xmath56 and @xmath57 be the lengths of the box .
then the lengths of the face diagonals are @xmath58 , @xmath59 and @xmath60 , and the length of the body diagonal is @xmath61 .
so we have an integer box .
lets @xmath62 , @xmath63 and @xmath64 the lengths of an integer box .
set @xmath65 , @xmath66 and @xmath67 .
then the lengths of the other three edges are @xmath68 , @xmath69 and @xmath70 , which are integers .
the area of the to @xmath47 opposite face is @xmath71 which is also an integer.@xmath6 the second type i will call _ right - angled - face tetrahedron _ : a tetrahedron whose vertices @xmath72 , @xmath73 , @xmath74 , @xmath75 can be represented by the cartesian coordinates @xmath76 , @xmath77 , @xmath78 , @xmath79 ( see right picture of figure [ fig : ra ] ) .
note that this condition is equivalent to all faces being right - angled triangles .
let s examine under which conditions this will generate a heron tetrahedron .
first all edges have to be integer , i.e. @xmath80 , @xmath81 , @xmath82 , @xmath83 , @xmath84 and @xmath85 .
the faces are pythagorean triangles , so their area is integer and the volume is @xmath86 which is an integer ( the product of the catheti is always divisible by @xmath10 ) .
now take a look to @xcite problem d18 and you get to know that this problem is equivalent to find a nearly integer box , where only one edge is not integer . the smallest solutions @xmath87 and
@xmath88 were known to euler .
hence these solutions gives us the two smallest right - angled - face tetrahedra .
see appendix [ ap : raf ] for more . 1 .
@xmath89 ( such heron tetrahedra seem not to exist ) , 2 .
@xmath90 and @xmath91 ( such heron tetrahedra do nt exist : assume there is one , then from the classification of heron isosceles triangles follows , that the two lengths are doubles of integer @xmath0 and @xmath1 with @xmath92 and @xmath93 . adding this equations one gets @xmath94 which implies @xmath95 ) and 3 .
@xmath96 .
figure [ fig : ra2if ] shows how to construct an isosceles - face tetrahedron from a right - angled - face tetrahedron and the decomposition of an isosceles - face tetrahedron into four right - angled - face tetrahedron ( picture on the lower right ) .
@xmath97 and @xmath98 are bases of isosceles triangles , so they have lengths @xmath20 and @xmath99 respectively .
let @xmath48 be the length of the other four edges .
we have to show that the length @xmath100 of the center line is integer . from @xmath101
follows that @xmath100 is rational and from @xmath102 follows that @xmath103 is integer .
hence @xmath100 is integer.@xmath6 | in @xcite problem d22 richard guy asked for the existence of simplices with integer lengths , areas , volumes . in dimension two this is well known , these triangles are called _
heron triangles_. here i will present my results on heron tetrahedra , their connection to the existence of an integer box ( problem d18 ) , the tools for the search for higher dimensional heron simplices and my nice embedding conjecture about heron simplices , which i can only proof in dimension two , but i verified it for a large range in dimension three . |
time - reversal breaking phases are predicted to occur in the so called _ pseudogap _ region of the phase diagram of the cuprates.@xcite these phases do not change the translational symmetry of the lattice and therefore do not produce new bragg spots in ordinary diffraction experiments . since the spontaneous magnetic field generated averages to zero in each unit cell , no simple magnetic dichroism effects arise .
however , it was shown that time - reversal breaking phases produce angle dependent dichroism in photoemission with circularly polarized photons.@xcite one such experiment has indeed detected such a phase.@xcite in this phase both time - reversal symmetry and inversion are broken but their product is preserved .
( such a symmetry is called _
magnetoelectric_.@xcite ) if this experiment is correct , the principal features of the theory of cuprates may be said to have been found .
it is therefore important to have independent experiments of quite a different kind verifying the existence of such phases .
here we propose some experimental possibilities , using x - ray resonant diffraction and absorption techniques .
another possibility using second harmonic generation has also been suggested .
@xcite magnetoelectric phases are known to produce non - reciprocal dichroism in resonant x - ray absorption in certain specified geometries .
we present here the symmetry considerations to find the geometries for specific crystals of the cuprate family . in certain crystals ( body - centered tetragonal or orthorhombic , for example ) certain bragg spots
are forbidden due to interference between scattering from different atoms in the unit cell .
were such samples to have a transition to magnetoelectric phases , such bragg spots would be allowed for some pattern of currents ( which enlarge the _ primitive _ unit cell ) due to transition amplitude tensors for x - ray scattering which also break time - reversal and inversion . even in the case
that the current pattern does not enlarge the primitive unit cell , it is possible to detect such a phase , either by means of x - ray non - reciprocal or magnetochiral dichroism , or by means of a polarization analysis of the scattering signal . in section
ii we present an analysis for this processes valid for the x - ray scattering . in section iii
we show what can be done with the different kind of dichroisms and finally , in section iv , consider some specific crystal symmetries in the cuprate family and illustrate the geometrical conditions to show the effect on these materials . as the effect of time - reversal and inversion breaking
is purely orbital , we shall deal in the following with resonant scattering and absorption at the k - edge , where only orbital quantities are involved.@xcite the generalization to any other edge is straightforward .
x - ray elastic resonant scattering has become a well - known technique to detect magnetic signals in single crystals.@xcite in the resonant regime , the structure factor for anomalous diffraction is given by @xmath0 where @xmath1 , @xmath2 are the wave vectors of the incoming and outgoing photon , respectively , @xmath3 and @xmath4 is the usual bragg factor .
@xmath5 is the energy - dependent atomic scattering amplitude for the atom at site @xmath6 .
it can be obtained by expanding the hamiltonian for electrons in an electromangetic field up to the second order in the electron - radiation coupling . a complete description about how to derive it can be found in blume.@xcite neglecting relativistic effects
, the interaction term is the usual @xmath7 , where @xmath8 is the vector potential for the electromagnetic field and @xmath9 is the linear momentum of the @xmath10 atom .
the scattering amplitude close to a resonance is : @xmath11 the operator @xmath12 is the usual matter - radiation interaction operator expanded up to the quadrupolar term , and @xmath13 is the photon polarization .
primed quantities refer to outgoing objects .
we neglect , as usual in the x - ray range , the magnetic dipole contribution . at thermal equilibrium at a temperature @xmath14 , @xmath15 where @xmath16 is the energy of a state with wavefunction @xmath17 , and
@xmath18 is the partition function .
the sum extends over all the excited states @xmath19 with energy @xmath20 .
finally , @xmath21 , the index @xmath22 indicates the lattice site of the photoabsorbing atom and @xmath23 is the energy of the incoming photon . in the dipole - dipole ( e1-e1 ) approximation ,
( [ resonance ] ) becomes : @xmath24 where @xmath25 . in the following we introduce the notation : @xmath26 .
we notice that the average in eq .
( [ e1e1 ] ) can be taken also on @xmath27 and @xmath28 which are time - reversed with respect to @xmath29(the corresponding energies are @xmath30 ) , provided we consider also the time - reversed distribution function @xmath31 .
( [ e1e1 ] ) can be re - written as : @xmath32 eq .
( [ e1e1bis ] ) can be rewritten as the sum of a time - reversal even and a time - reversal odd part , @xmath33 and @xmath34 , respectively , with : @xmath35 and @xmath36 @xmath37 @xmath38 stand for the real and imaginary parts respectively of the arguments . when time - reversal is a symmetry of the problem @xmath39 and @xmath40 . therefore ,
under time - reversal @xmath41 . equation ( [ e1e1minus ] ) shows that @xmath42 unless @xmath43 .
this latter condition is realized if the states @xmath44 and @xmath27 are not degenerate , ie , if time - reversal is broken . for the mixed dipole - quadrupole channel ( e1-e2 ) ,
( [ resonance ] ) gives : @xmath45 again , rewriting eq .
( [ e1e2 ] ) in terms of the time - reversed states and taking the mean gives @xmath46 where , we used the notation @xmath47 .
it is possible to write a time - reversal even and a time - reversal odd contribution in the form : @xmath48 where @xmath49 @xmath50 and @xmath51 @xmath52 what is measured experimentally is of course the intensity , ie , @xmath53 to calculate @xmath54 , one must first add the contributions @xmath55 and @xmath56 on the right side of eq .
( 1 ) and then take the square of the absolute magnitude . note that the time - reversal even part of the scattering amplitude is real in the e1-e1 channel and imaginary in the e1-e2 channel , while the opposite is true for the time - reversal odd parts .
this means that in the intensity for scattering of linearly polarized waves there is no interference term between the e1-e1 and e1-e2 channels when time - reversal is preserved but an interference exists between the time - reversal even part of the e1-e1 and the time - reversal odd part of the e1-e2 channel .
this has important consequences for the detection of the magnetoelectric phase in cuprates materials , as we shall see later .
( for circular polarizations an interference term exists between the time - reversal even and the time - reversal odd parts in the e1-e1 channel alone . ) for the case of the cuprates no cartesian second - rank polar time - reversal breaking tensors exist for the symmetries predicted state@xcite but third rank polar time - reversal breaking tensors do exist ( see the appendix ) .
therefore one must rely on the time - reversal odd part of the e1-e2 term and its interference with the usual e1-e1 ( i.e. its time - reversal even part ) for the detection of the predicted states .
the polarization selection rules in the e1-e2 channels given by eqs .
( [ e1e2minusplus ] ) , ( [ e1e2minusminus ] ) are used in sec .
the expected order of magnitude for the time - reversal breaking signal may be estimated as follows .
a general time - reversal breaking wavefunction in a metal is of the form @xmath57 with @xmath58 where @xmath59 may be taken to be pure real in an appropriate gauge for the case that the _ charge - density _ has inversion symmetry and where the complete wavefunction @xmath60 can not be transformed to a real function by any gauge transformation . in order to obtain a finite time - reversal odd part , @xmath61 , the product of the two matrix element in eq .
( [ resonance ] ) must have an imaginary part.@xcite thus , either the first or the second matrix element in the product must be imaginary and since the core wavefunction is a real function , it then follows that the time - reversal odd scattering amplitude is of order @xmath62 with respect to the corresponding time - reversal even part . in the case the signal to be detected is in the e1-e2 channel , there is an extra factor of about 0.1 - 0.2 with respect to the e1-e1 main edge , due to the relative weight of the radial density of states . from refs .
[ ] one can get a rough estimate for @xmath63 . to get an interference term @xmath64
, one needs to have an interference with the main e1-e1 term , which is real ( its coefficient , @xmath65 is real for both circular and linear polarizations ) .
since the quadrupole matrix elements are are usually of @xmath66 of the dipole matrix elements , the magnitude of the effect is of @xmath67 of the typical resonant bragg peaks .
the atomic x - ray absorption cross section can be written as the imaginary part of the forward scattering amplitude or by the fermi golden rule : @xmath68 where @xmath69 is the fine - structure constant .
the total absorption cross - section is then given by the sum over all the atoms in the unit cell . in the following two subsections
we describe two dichroism experiments to detect the time - reversal broken phase in the pseudogap region of the cuprates .
the geometrical conditions for either are determined by the magnetic symmetries of the material as well as the experimental geometry .
the intensity of the effect is then simply determined on the basis of the so - called `` sum rules '' @xcite and of the explicit form of the conduction - band wave function , as determined for the cuprates.@xcite linear dichroism in the x - ray range is the differential spectrum obtained by subtracting two core absorption spectra taken with orthogonal linear polarizations .
if the difference is of dipole - dipole ( e1-e1 ) origin , it represents normal x - ray linear dichroism ( xld ) and if of dipole - quadrupole origin , it represents nonreciprocal linear dichroism ( nrxld ) . from ref .
[ ] the sum rules for xld and nrxld at the k - edge are : @xmath70 @xmath71 where @xmath72 and @xmath73 are the two arbitrary orthogonal directions of the two polarizations and @xmath74 is the direction of the photon wavevector @xmath75 .
@xmath76 is the usual angular momentum operator and the anapole operator @xmath77 , where @xmath78 is the unit radial vector operator.@xcite the averages of these operators are taken integrating over all the scattering ( excited ) states corresponding to the energy @xmath79 at which the spectrum is recorded . the main difference between eq .
( [ lindice1e1 ] ) and ( [ lindice1e2 ] ) is in the time - reversal and inversion properties of the two operators .
the first operator is time - reversal even and inversion even , being the symmetric traceless part of a cartesian second rank polar tensor . on the other hand
the second operator is time - reversal odd and inversion odd : its mean value over the states @xmath80 is different from zero only if both these symmetries are broken , ie , if the system under consideration is _
magnetoelectric_. this situation is fulfilled in one of the time - reversal broken phases found in refs .
[ ] . thus the detection of nrxld would be a clear evidence of the time - reversal broken phase .
the geometrical conditions to detect the effect are discussed in detail in the next section .
the order of magnitude of the effect is readily estimated as in the previous section for the scattering amplitude .
the nrxld is due to an electric - dipole electric - quadrupole transition . as such
, it mixes states of different parity : considering a k - edge process , these states are the @xmath81 and @xmath82 conduction bands . as shown in refs .
[ ] , the generic time - reversal broken wave function for the cuprates is such that @xmath83 is of @xmath81 type and in @xmath84 there is a part with @xmath82 symmetry . thus the average in eq .
( [ lindice1e2 ] ) is proportional to @xmath62 and the ratio of the total dichroic signal with respect to the absorption edge is proportional to @xmath62 times the ratio of radial density of states in the e1-e2 to the e1-e1 channels .
this of course gives the same estimate obtained for the scattering amplitude , about 0.5@xmath85 , which is detectable with modern synchrotron radiation sources.@xcite another technique that can be adopted to detect the time - reversal broken phase in the cuprates is the magnetochiral dichroism ( mcd ) .
circular dichroism in e1-e1 channel is excluded , being sensitive to the expectation value of the orbital angular momentum , which is zero for the patterns found in refs .
inversion - odd circular dichroism is time - reversal even , giving rise to x - ray natural circular dichroism , and thus it is excluded as well .
but , as shown recently by goulon et al.@xcite for cr@xmath86o@xmath87 , the sum of two right and left circularly polarized spectra can give rise to a dichroism in magnetochiral systems , ie , magnetoelectric systems whose magnetoelectric tensor has an antisymmetric off - diagonal part . note
that this is just the case of the magnetic symmetry predicted by simon and varma.@xcite in this case the dichroism is defined as the difference between the sum of two orthogonally polarized spectra in one magnetoelectric domain and those in the opposite magnetoelectric domain .
note that for cuprates the choice of the magnetoelectric domain could be made difficult by the fact that they are not insulators . as shown in ref .
[ ] , the magnetochiral dichroism is proportional to the expectation value , in the conduction states , of the operator @xmath88 , where @xmath74 is the direction of the wavevector of the absorbed photon : @xmath89 in section iv we analyze the geometrical conditions to detect the effect .
an eventual magnetochiral signal is a definite indication of the time - reversal broken phase . as for nrxld ,
the intensity of the signal should be of order 0.5@xmath85 , within the present experimental capabilities .
we have considered the following materials : hgba@xmath86cuo@xmath90 , tl@xmath86ba@xmath86cuo@xmath91 , la@xmath92sr@xmath93cuo@xmath94 , bi@xmath86sr@xmath86cacu@xmath86o@xmath95 . except the orthorombic phase of la@xmath92sr@xmath93cuo@xmath94 , that must be studied separately , these cuprates all have _ crystalline _ space group symmetry 4/mmm ( the crystal class is primitive for hgba@xmath86cuo@xmath90 , body - centered for all the others ) .
the copper ion occupies a site of maximal symmetry ( 4/mmm ) in all the cases except for bi@xmath86sr@xmath86cacu@xmath86o@xmath95 , where the symmetry is 4 mm : the mirror plane orthogonal to the four - fold axis is lost because in the bilayered compounds one of the apical oxygens is missing .
two types of time - reversal breaking states can be proposed , according to the results of ref .
[ ] , one with magnetic - group symmetry @xmath96 mm and the other with magnetic- group symmetry m@xmath97 m ( we use birss notation @xmath98 for a time - reversed symmetry operator @xmath99 ) .
the magnetic group @xmath96 mm , ( which remains tetragonal ) is such that the inversion symmetry is preserved , while both time - reversal and the product of the time - reversal and inversion are broken .
since inversion is preserved , no e1-e2 signals ( that are inversion - odd ) are possible . at the same time , the simultaneous presence of three mirror planes as symmetry elements ( even if one is time - reversed ) forbids the possibility of a magnetic moment as a local invariant ( because the magnetic moment is an axial quantity and can not be invariant under the action of two orthogonal mirror planes ) .
this implies that with this pattern time - reversal odd quantities can be detected only in the e2-e2 channel .
more interesting is the case when the symmetry is reduced to m@xmath97 m .
this is the symmetry consistent with the results of the arpes experiment in bi-2212 .
first of all , in this case the system is no longer tetragonal .
this can in principle be detected by an azimuthal scan around the original four - fold axis as the new periodicity is determined by the mirror and is no longer four - fold .
since inversion symmetry is also lost , in this phase the detection of some phenomena linked to magnetoelectricity becomes possible .
apart from a direct observation of magnetoelectric susceptibility , x - ray diffraction , non - reciprocal linear dichroism and magnetochirality in the e1-e2 channel are all possible experiments to observe this phase .
as we already saw the interesting effects arise in the e1-e2 channel through finite third - rank polar time - reversal odd tensors specified , for example , in eqs .
( [ e1e2minusplus ] ) , ( [ e1e2minusminus ] ) .
we specify the tensors using the following co - ordinates : the c axis of the crystal coincides with the @xmath100 direction , a and b crystalline axis ( those along the nearest neighbor cu - cu vector ) are along the @xmath101 and @xmath102 directions , as is clear from fig .
the point group m@xmath97 m has the following elements : the identity , two - fold axis rotation around @xmath103 ( @xmath104 ) , the reflections in the @xmath105 and @xmath106 planes ( @xmath107 and @xmath108 ) , @xmath109 , @xmath110 , @xmath111 and @xmath112 ( where @xmath113 is the time - reversal operator ) .
the distinct non - zero elements of the third - rank polar tensors , odd under time - reversal in this point group are seven@xcite : @xmath114 .
if we write the seven allowed tensors by symmetrizing the first two indices , we get the symmetric @xmath115 , @xmath116 , @xmath117 , @xmath118 and @xmath119 and the antisymmetric @xmath120 and @xmath121 . the formers can be detected only with the symmetrized part of the polarizations , ie , they correspond to @xmath122 ( eqs .
( [ e1e2minusplus ] ) and ( [ e1e2minusminus ] ) ) and they are proportional to @xmath123 .
they also correspond to tensors @xmath124 and @xmath125 in the appendix and as such they can be detected in both @xmath126 and @xmath127 polarizations ( @xmath124 only in @xmath126 ) .
the latters correspond to the tensor @xmath128 in the appendix .
even if proportional to @xmath129 , the signal is not maximized in a backscattering configuration , where it would be zero , because if @xmath1 and @xmath2 are both parallel to @xmath130 , then @xmath131 . this signal can be detected only in the @xmath127 channel .
thus , for the pattern depicted in fig .
1b , the preferred geometrical configuration should be when the wave vectors lie in the @xmath132 plane , with @xmath129 along @xmath130 and @xmath123 along @xmath133 , both about 45@xmath134 from the two axes , in order to optimize the projections of @xmath1 and of the outgoing polarization on the @xmath130-axis .
the incoming polarization in the @xmath127 channel is along @xmath135 and the outgoing polarization is in the @xmath132 plane , orthogonal to @xmath2 . in this situation
the only contributions come from the antisymmetric part of @xmath136 and from the symmetric part of @xmath137 .
an azimuthal scan about the @xmath130-axis could reveal the two - fold time - reversal breaking signal with respect to the four - fold time - reversal even one .
these considerations are , of course , subject to the requirement that @xmath138 , where @xmath139 is a bragg vector . another possibility to detect
the time - reversal breaking signal is to exploit the different polarization dependence of @xmath140 and @xmath141 .
the former is antisymmetric in the exchange of the two polarizations , while the latter is symmetric .
also the e1-e1 contribution antisymmetric in the polarizations ( ie , @xmath142 ) is zero , because it is proportional to the average magnetic moment on the cu site , while the time - reversal even e1-e2 term , antisymmetric in the polarizations ( ie , @xmath143 ) is imaginary .
therefore , subtracting the signals obtained with switched orthogonal incoming and outgoing polarizations leaves only @xmath144 .
observing the effect requires a sensitivity better than @xmath145 .
the orthorombic phase of la@xmath92sr@xmath93cuo@xmath94 has different symmetries , but the conclusions are similar , as well as the experimental techniques to detect the time - reversal breaking . in the more symmetric phase , due to the buckling of the oxygens , the symmetry around each copper atom is 2/m ( and the space crystal symmetry is cmce ) .
this symmetry is lowered , by the magnetic orbital patterns to @xmath146 or to one of the two magnetoelectric @xmath147/ m , @xmath148 .
for the first case we can apply , even with a lower symmetry , the same considerations as for the @xmath149/mm @xmath97 , while for the two magnetoelectric phases , being also magnetochiral , the same considerations as for m@xmath97 m apply .
we have already discussed the considerations regarding the magnitudes of the intensity of order @xmath62 due to time - reversal breaking .
there is of course a contribution of @xmath150 also .
apart from other considerations of domains noted below , time - reversed domains in the scattering volume are expected to affect adversely the effect of @xmath151 and not of @xmath150 . besides the estimate of the intensity of the signal , it is important to analyze how the geometry of the systems can influence the choice of the experimental technique .
the origin of the time - reversal breaking is due to orbital currents circulating around each copper ion with a well - defined pattern in each cuo@xmath86 plane , as described in ref . [ ] .
the arpes experiments probe only the surface cu - o layers and indicate a surprisingly large domains size , greater than about 10 microns .
if we accept this as correct , the next question concerns the relative order from plane to plane .
the inter - plane coupling is expected to be quite weak and answer to this question can not be reliably obtianed from theoretical considerations .
we shall examine the case of both `` ferromagnetic '' ( fm ) and `` antiferromagnetic '' ( afm ) ordering between the planes .
for all the cases we have examined , the ferromagnetic pattern preserves the size of the primitive cell and the antiferromagnetic pattern changes it .
therefore the `` forbidden '' diffraction spots ( at h+k+l = odd for body - centered tetragonal or body centered orthogonal lattices ) stay `` forbidden '' for the fm pattern but become allowed for the afm patterns
. if the ordering between the planes is random , as is not unlikely , bragg - rods are expected to arise . for the afm pattern
dichroic techniques are useless , as the signal averages to zero in the unit cell . in this case
only x - ray scattering can detect a signal , because the ordinarily forbidden bragg - spots can be revealed by means of linearly polarized light in the e1-e2 channel ( for pattern b of fig .
1 ) . in case
the alignment of successive layers is purely ferromagnetic , it is rather complicated to resolve the magnetic signal with respect to the non - magnetic by x - ray scattering , because of smallness of the former ( @xmath145 ) .
one possible way to perform the experiment in this case is linked to the different azimuthal scan between the normal ( 4-fold periodic ) and the magnetoelectric ( two - fold periodic ) phases , as shown in the following subsection .
probably more appropriate to the case of purely fm ordering is the use of dichroism . in the next subsection
we shall deal with the specific case of bi-2212 , that is specially note worthy in this respect , since the evidence for the symmetry breaking phase has first been discovered in it.@xcite the space group given is the tetragonal i4/mmm with two formula units in the cell .
the four cu ions are on the wyckoff sites 4e , whose symmetry is 4 mm , ie , not the full space symmetry of the cell ( full symmetry that is recovered only by the ca ions , on the 2a sites ) .
the reason for this is that the cu ions in a bilayered system like bi-2212 do not have one of the two apical oxygens and thus the /m plane , that is orthogonal to the 4-fold axis is missing . for this reason they have not even a centre of inversion and
the situation is quite different from the mono - layered compound , like hgba@xmath86cuo@xmath90 , where the inversion and the /m plane are present . from the experiment of kaminski _
et al._@xcite , we can only conclude that the bilayers on the surface cells have currents in phase ( in the examined region ) but nothing can be inferred for phases between different planes in the bulk region .
if we superimpose one of the two current structures , the _ local _ magnetic symmetry on the cu ion is reduced from 4mm@xmath152 to @xmath153 mm ( still tetragonal ) for pattern @xmath154 of fig . 1 , while it is reduced to @xmath155m@xmath156 for pattern @xmath157 . in this latter case
the magnetic symmetry is orthorhombic ( the four - fold axis is lost ) and the mirror planes are no longer orthogonal to the cu - o bonds , but to @xmath133 and @xmath135 of fig .
1b .. we already dealt with the afm case : let us now examine the case when all cuo@xmath86 layers are in phase ( fm ordering ) .
if all the four cu ions have a current pattern of type @xmath154 , the space magnetic symmetry is reduced to @xmath158 mm : the group is still tetragonal and , even if cu sites are not centrosymmetric , there is a global inversion centre in correspondance to the ca ions . because of the global centre of inversion , it is not possible to detect e1-e2 terms in an absorption experiment ( their average gives zero ) .
the only signal is in the e2-e2 channel and its intensity is expected to be very small ( there is an extra factor due to the radial matrix elements of the e2-e2 channel with respect of the e1-e2 , that can be estimated as an extra 0.1 ) .
if all the four cu ions have a current pattern of type @xmath157 , the space magnetic symmetry is reduced to m@xmath156 m , as already seen .
this group does not contain the time - reversal or the inversion as separated symmetries , but it contains their product .
it is a magnetoelectric group ( contrary to the original 4/mmm ) and , as the corresponding magnetoelectric tensor has an antisymmetric off - diagonal part , it admits also magnetochirality .
for this reason this configuration is even more interesting in that it can be revealed by linear ( non - reciprocal ) and magnetochiral dichroism .
here we determine the best geometrical conditions for such experiments .
we refer to fig .
1b and to eqs .
( [ lindice1e2 ] ) , ( [ cirdice1e1 ] ) .
it is clear that for both kind of dichroisms the signal is maximized when the light is shone in the same direction as the anapole , as in both eqs .
( [ lindice1e2 ] ) , ( [ cirdice1e1 ] ) @xmath159 stands for @xmath160 . the anapole , being a time - reversal odd polar vector , is invariant under a parallel mirror symmetry and changes sign if the mirror is orthogonal .
the contrary happens if the mirror is also time - reversed .
it follows that the only invariant direction for such a vector is the one drawn in fig . 1b , orthogonal to the time - reversed mirror plane .
thus , in order to have a non - zero time - reversal odd signal , the light must have a wave vector with a component along this direction . notice that , in this case , a linear dichroism of e1-e1 origin is present , because of the anisotropy of two orthogonal directions within the time - reversed mirror plane , and it is expected to be bigger than the time - reversal breaking one . in order to detect the latter , two experiments must be performed , with opposite magnetoelectric domains : the substraction of the two spectra gives then twice the e1-e2 nrxld , being the e1-e1 xld invariant in the two domains .
similar considerations also hold for the magnetochiral dichroism , ie , in order to maximize the signal , the light must be shone in the direction of the anapole : an useful description of the experiment is then given in ref . [ ] .
finally , the reduction of the group from tetragonal to orthorombic , with the pattern @xmath157 , allows , in principle , the detection of the signal also by means of x - ray resonant scattering .
the trick , as sketched before , is the following : even if all the allowed magnetic reflections correspond to non - magnetic allowed reflections , too , these latter have a four - fold symmetry around the axis orthogonal to the cuo@xmath86 plane .
consider a reflection , like the ( 002 ) , whose exchange vector is directed orthogonal to the cuo@xmath86 planes . in the normal phase , making an azimuthal scan around this vector one obtains a four - fold periodicity , ie , four minima .
the symmetry reduction due to the magnetic pattern ( two - fold ) reduces the number of minima to two .
thus deviations from a perfect four - fold azimuthal signal would be an indication of the magnetic effect .
this effect is however quite difficult to measure , because the deviation is again expected to be of the order @xmath161 .
cmv wishes to thank d.e .
moncton and g.sawatzky for asking the question of whether effects of time - reversal breaking in absorption experiments also have observable consequences in x - ray diffraction experiments , to p.carra for introducing him to the literature on x - ray dichroism and to m.e .
simon for useful discussions .
sdm would like to thank c.r .
natoli and i. marri for useful old discussions on x - ray scattering and dichroism
as in the text we dealt with both cartesian tensors and irreducible representations of these latters , it can be useful to list the way to express the ones in terms of the others . for the irreducible representations we shall adopt spherical coordinates , that turn out to be more versatile in this respect . for the e1-e1 channel
it is quite straightforward : the decomposition of a cartesian rank - two tensor into the irreducible spherical representations can be found in many textbooks.@xcite any rank - two polar tensor @xmath162 can be expressed in terms of its so(3 ) irreducible representations : \2 ) a vector , that is the dual of the antisymmetric part of @xmath162 : @xmath164 , where @xmath165 is the third - rank totally antisymmetric levi - civita tensor : it is proportional , at the k - edge , to the orbital angular momentum .
for example , referring to eq .
( 7 ) , @xmath166 things turn out to be slightly more complicated when we deal with the e1-e2 channel . in this case
there are @xmath168 , ie , 54 independent components ( the factor two comes out from the fact that in the e1-e2 channel also the complex conjugate matrix elements are present ) . the explicit form of eq .
( 8) shows that the tensor @xmath169 must be symmetric , thus reducing the numebr of independent components to @xmath170 .
they further reduce to @xmath171 because the scalar coupling of @xmath69 and @xmath172 in @xmath169 can not be detected as it is coupled to the scalar product @xmath173 . in order to classify these components properly , it is much more convenient to classify the tensors representing the properties of the x - ray beam , that are coupled scalarly to the ones representing the properties of the matter.@xcite this classification turns out to be material - independent , in that it only shows the tensorial propoerties to which the x - ray beam is sensitive in the e1-e2 channel .
it must then be combined with the explicit symmetries of the material , as done in section iv , to get more useful informations .
@xmath174^{(2 ) } \stackrel{(+)}{~~~~~ } \\
\hline \hline { \bf k } + { \bf k ' } \stackrel{(-)}{~~~~~ } & { \tilde { a}}^{(1 ) } & b^{(2 ) } & { \tilde { c}}^{(1 ) } , ~ { \tilde { c}}^{(2 ) } , ~ { \tilde { c}}^{(3 ) } \\ \hline { \bf k } - { \bf k ' } \stackrel{(+)}{~~~~~ } & a^{(1 ) } & { \tilde { b}}^{(2 ) } & c^{(1 ) } , ~ c^{(2 ) } , ~ c^{(3 ) } \\
\hline \end{array}$ ] the superscripts ( + ) and ( - ) refer to the time - reversal behaviour of the corresponding quantities and @xmath175^{(2)}$ ] is the @xmath176-@xmath177 coupling to a rank - two irreducible tensor .
non - tilded tensors are time - reversal even and tilded tensors are time - reversal odd .
their explicit expressions in spherical coordinates is ( apart from multiplicative prefactors ) : we just would like to note that , for example , one of the tensors , inversion and time - reversal odd , that can be detected in the cuprates , @xmath185 , is coupled to @xmath186 : it corresponds to what carra _
et al._@xcite identified as an indication of the tensorial magnetic and electric displacement : @xmath187^{(2)}_1 - [ { \bf l } , { \bf n}]^{(2)}_{-1 } $ ]
. c.m .
varma , phys .
b * 55 * , 14554 ( 1997 ) ; physical rev .
letters * 83 * , 3538 ( 1999 ) .
c.m.varma , phys .
b * 61 * , r3804 ( 2000 ) m.e .
simon and c.m .
varma , phys .
, accepted for publication ( 2002 ) ; cond - mat/0201036 a. kaminski , s. rosenkranz , h. m. fretwell , j. c. campuzano , z. li , h. raffy , w. g. cullen , h. you , c. g. olson , c. m. varma , h. hchst , nature * 416 * , 610 ( 2002 ) ; cond - mat/0203133 l.d .
landau , e.m .
lifshitz and l. p. pitaevskii , _ electrodyanmics of continuous media _ , 2nd edition , butterworth - heinemann , oxford ( 1984 ) . r.r .
birss , _ symmetry and magnetism _ , edited by e.p .
wohlfarth ( north - holland , amsterdam , 1964 ) , p. 137 , 104 and 60 . m.e .
simon and c.m.varma , cond - mat/0210672 .
p. carra , b.t .
thole , m. altarelli , x. wang , phys .
70 * , 694 ( 1993 ) j. luo , g.t .
trammell , j.p .
hannon , phys .
lett . * 71 * , 287 ( 1993 ) j.p . hannon , g.t .
trammell , m. blume and doon gibbs , phys .
* 61 * , 1245 ( 1988 ) m. blume , _ magnetic effect in anomalous dispersion _
, in resonant anomalous x - ray scattering , edited by g. materlik , c.j .
sparks and k. fischer ( elsevier science b.v . , 1994 ) , p. 495 m. blume , j. appl . phys .
* 57 * ( 1985 ) 3615 p. carra , r. benoist , phys .
b * 62 * , r7703 ( 2000 ) s. di matteo , c.r.natoli , j. synchr
9 * , 9 ( 2002 ) s. di matteo , c.r .
natoli , submitted to phys .
b ( 2002 ) d.a .
varshalovich , a.n .
moskalev , v.k .
kersonskii , _ quantum theory of angular momentum _ , world scientific ( 1988 ) j. goulon , a. rogalev , c. goulon - ginet , g. benayoun , l. paolasini , c. brouder , c. malgrange and p.a .
metcalf , phys .
lett * 85 * , 4385 ( 2000 ) j. goulon , a. rogalev , f. wilhelm , c. goulon - ginet , p. carra , d. cabaret , c. brouder , phys .
lett . * 88 * , 237401 ( 2002 ) s. di matteo , c.r .
natoli , accepted for publication in phys .
b ( 2002 ) p. carra , t. thole , rev .
phys . * 66 * , 1509 , ( 1994 ) p. carra , a. jerez , i. marri , pre - print : cond - mat/0210152 | time - reversal and inversion breaking in the pseudogap region of the cuprate compounds have been proposed and detected by means of arpes in the compound bi-2212 . given the implication of the effect in the physics of the cuprate superconductors , it is important to check the conclusions through quite independent experimental techniques . we show how time - reversal and inversion breaking may be revealed in the cuprates through resonant x - ray scattering and through non - reciprocal or magnetochiral dichroism . 2 |
in order to establish a sound understanding for any scientific phenomenon , one has from numerical data to obtain laws which can be next derived theoretically .
some difficulty arises in particular in nonlinear dynamical systems because of the problem to sort out noise from both chaos and deterministic components [ bhattacharya and kanjilal , 2000 ; provenzale et al . ,
moreover , algorithms for solving the inverse problem for nonlinear systems with underlying unknown dynamical processes are notoriously hardly reliable [ theiler , 1992 ] .
in fact , to extract dynamical equations for chaotic - like data is an enormous challenge [ rowlands , 1992 ] . practically one is often led to empirical relationships .
this is the case of the meteorology field where there is a widely mixed set of various ( sometimes ) unknown influences , over different time and space scales .
works of interest for sorting out deterministic ingredients in chaotic systems , as in [ hilborn , 1994 ; ott , 1993 ; schreiber , 1999 ; celluci , 1997 ; davis et al . ,
1996a ] , can be mentioned for general purpose though this is not an exhaustive list of references .
it is known that two equivalent @xmath3 @xmath4 govern the dynamics of a system , i.e. the fokker - planck equation and the langevin equation , the former for the probability distribution function of time and space signal increments , the latter for the increment themselves [ reichl , 1980 ; ernst et al . , 1969 ; risken , 1984 ; hggi and thomas , 1982 ; gardiner , 1983 ] .
they are both condensates of the huge set of ( 6n ) hamilton equations which should in practice describe the whole dynamics of the system of n particles by giving the time evolution of both coordinates and momenta of each individual particle .
this is a highly unrealistic scheme of work , and therefore conservation laws are used in order to derive the navier - stokes equations , from the averaging of basic quantities weighted by the above probability functions [ reichl , 1980 ; huang , 1967 ] .
such a `` micro - turned - macro - scopic description '' can be by passed when describing the _ evolution of the quasi - equilibrium probability distribution function _ of such particles , i.e. in writing a boltzmann equation [ huang , 1967 ] .
however this corresponds to a mean field description of the fokker - planck equation ( fpe ) [ reichl , 1980 ; ernst et al .
, 1969 ; risken , 1984 ; hggi and thomas , 1982 ; gardiner , 1983 ] , an approximation which might be too rough for this sort of nonlinear system dynamics .
recently much advance has been made in describing nonlinear phenomena in meteorology , after the canonical lorenz approximation [ lorenz , 1963 ; schuster , 1984 ] of the navier - stokes equations for sorting out some deterministic and stochastic components of turbulent fluid motion in the atmosphere .
in fact , turbulence [ frisch , 1995 ; parisi and frisch , 1985 ; hunt , 1999 ] seems now to be part of more general class of problems similar to those found when there are long range fluctuations , as in financial data [ ghashghaie et al .
, 1996 ; mantegna and stanley , 1995 ] , traffic [ chowdhury et al . ,
2000 ] , dielectric breakdown [ vandewalle et al . , 1999 ] , heart beats [ ivanov et al . ,
1999 ] , etc ... such observations corroborate the idea that scaling [ stanley , 1971 ] and fractal geometry [ schuster , 1984 ] principles could be useful in such studies on atmospheric turbulence [ marshak et al . , 1997 ] . thus to derive the fpe [ risken , 1984 ; hggi and thomas , 1982 ; gardiner , 1983 ] for some meteorological effect is of primary interest .
the aim of this paper is to derive such an fpe , in a first principle statistical approach , from given raw data , i.e. for the liquid water path ( lwp ) fluctuations in clouds , - in terms of a drift @xmath5 and a diffusion @xmath6 coefficient .
much following the lines of thought of [ friedrich et al . ,
2000 ] it will be shown that the fpe ( also known as the kolmogorov equation ) can be derived up to the _ first two _ moments of the ( _ measured _ ) conditional probability distribution functions @xmath7 , where @xmath8 is a signal increment and @xmath9 a time delay .
thus the method is not anymore based on the conventional phenomenological comparison between models and several stochastic aspects , but is a @xmath10 @xmath11 ( or first principle , _ from a statistical point of view _ ) approach .
this allows us to examine long and short time scales on the _ same footing _ , and suggests new features to be implemented in related weather forecasting .
indeed _ it is here below verified _ that the solution of the fpe yields the probability distributions with high accuracy , including the long and high tail @xmath8 and @xmath9 events . in some sense
the agreement proves that searching for more improvement with including further event of similar cases is not necessary .
furthermore the so found analytical form of both drift @xmath5 and diffusion @xmath6 coefficients has a simple physical interpretation .
it is also shown that a truncation of the fpe expansion in terms of high order joint probabilities to the first two terms is quite sufficient .
moreover the identification of the underlying process leading to the heavy tailed probability density functions ( pdf ) for the lwp fluctuation correlation with @xmath9 and the volatility clustering ( as seen in fig .
3 below ) seems to be a new observation leading to an interestingly new set of puzzles for this sort of clouds [ cahalan et al . ,
1995 ] , and more generally in atmospheric science .
the changes in a time series for a signal @xmath12 are commonly measured by the quantity @xmath13 , or increments @xmath8 = @xmath14 . results of the analysis of a data set @xmath15 , selecting among others lwp observations , are presented here .
the 25 772 data points are taken from microwave radiometer measurements at the southern great plains site of atmospheric radiation measurement ( arm ) ] of department of energy during the period january 9 - 14 , 1998 as used by [ ivanova et al . ,
the microwave radiometer measures the brightness temperature at two frequency channels , one at 23.8 ghz and the other at 31.4 ghz .
then both brightness temperature data series are used to retrieve the vertical columnar amount of water vapor and vertical columnar amount of liquid water in the cloud , i.e. the so - called liquid water path .
the raw data @xmath16 ( @xmath17 is equivalent to a time index ) of the liquid water path in stratus clouds is shown in fig . 1
. the first thing is to consider whether the distribution function of increments has short or long tails , and whether there is some scaling law involved .
next the central issue is the understanding of the statistics of fluctuations which determine the evolution of the cloud .
there are different estimators for the short and/or long range dependence of fluctuations correlations [ taqqu et al . ,
1995 ] . through a detrended fluctuation analysis ( dfa )
method [ ausloos and ivanova , 1999 ] we show first that the _ long range correlations _ are not brownian .
the method has been used previously in the meteorological field [ ivanova et al . , 2000 ; ausloos and ivanova , 1999 ; koscielny - bunde et al . , 1998
; ivanova and ausloos , 1999 ] and its concepts are not repeated here .
we show on fig .
2a the final result , for the @xmath18}^2 } } \sim \tau^{\alpha}\ ] ] indicating a scaling law characterized by an exponent @xmath19 , markedly different from @xmath20 , thus indicating antipersistence of the signal from about 3 to about 150 minutes . next the power spectrum @xmath21 with spectral exponent * @xmath22 * is also shown ( see fig .
2b ) [ davis et al . , 1996b ] .
a kolmogorov - smirnov test on surrogate data has indicated the statistical validity of the value and error bars .
thereafter we focus on the lwp changes measured as increments @xmath8 which are given in fig . 3
in units of the standard deviation @xmath23 of @xmath8 at @xmath24 s. ln order to characterize the statistics of these lwp changes , lwp increments @xmath25 , @xmath26 for delay times @xmath27 , @xmath28 at the same time @xmath29 are considered .
the corresponding joint probability density functions are evaluated for various time delays @xmath27 @xmath30 @xmath28 @xmath30 @xmath31 @xmath30 ... directly from the given data set .
one example of a contour plot of these functions is exhibited in fig .
4 . if two lwp changes , i.e. @xmath25 and @xmath26 are statistically independent , the joint pdf should factorize into a product of two probability density functions : @xmath32 leading to a camelback or an isotropic single hill landscape .
the tilted anisotropic form of the joint probability density ( fig .
4 ) clearly shows that such a factorization does not hold for small values of @xmath33log(@xmath34)@xmath33 , whence both lwp changes are _ statistically dependent_. the same is found to be true for other ( @xmath35 ) ratios .
this is in agreement with the hint taken from the above observations for the long range cross - correlation functions with the dfa .
this implies a ( fractal - like ) hierarchy of @xmath1 scales .
if fluctuations in lwp went up over a certain @xmath36 , then it is more likely that , on a shorter @xmath27 within the larger one , the lwp went down instead of up . to analyze these correlations in more detail , the question on what kind of statistical process underlies the lwp changes over a @xmath37 of nested time delays @xmath38 of decreasing duration should be raised .
a complete characterization of the statistical properties of the data set in general requires the evaluation of joint pdf s @xmath39(@xmath25,@xmath27; ... ;@xmath40,@xmath41 ) depending on n variables ( for arbitrarily large n ) . in the case of a markov process ( a process without memory ) [ schuster , 1984 ] ,
an important simplification arises : the n - point pdf @xmath39 is generated by a product of the conditional probabilities @xmath42(@xmath43,@xmath44,@xmath45 ) = @xmath42(@xmath43,@xmath46;@xmath47,@xmath48)/ @xmath42(@xmath47,@xmath45 ) for i = 1, ... ,n-1 .
the conditional probability is given by the probability of finding @xmath49 values for fixed @xmath47 .
as a necessary condition , the chapman - kolmogorov equation @xmath50 should hold for any value of @xmath51 , with @xmath27 @xmath30 @xmath51 @xmath30 @xmath36 .
we checked the validity of the chapman - kolmogorov equation for different @xmath45 triplets by comparing the directly evaluated conditional probability distributions @xmath42(@xmath52 ) with the ones calculated ( @xmath53 ) according to eq.(3 ) , in fig.5 .
the solutions of eq .
( 4 ) usually give rise to exponential laws , but initial conditions must be taken into account for adjusting the constants of the integration . starting with a stretched exponential ,
the latter is conserved after successive integrations , with a width varying with @xmath9 .
the power law probability density function _ which is observed _ is thus a signature of fractal properties of the signal increments because the tails of the pdf are not truly exponential ones . in fig .
5a , the contour lines of the two corresponding pdf s for all values of @xmath25 are shown .
there are mild deviations , probably resulting from a finite resolution of the statistics .
cuts for some exemplarily chosen values of @xmath25 are shown in addition in fig . 5 ( b - d ) .
as is well known , the chapman - kolmogorov equation yields an evolution equation for the change of the distribution functions @xmath42(@xmath54 ) and @xmath42(@xmath8,@xmath9 ) across the scales @xmath9 [ reichl , 1980 ; ernst et al .
, 1969 ; risken , 1984 ; hggi and thomas , 1982 ; gardiner , 1983 ]
. for the following it is convenient ( and without loss of generality ) to consider a normalized logarithmic time scale @xmath55= @xmath56 ( 640/@xmath9 ) .
the limiting case @xmath45 @xmath57 0 corresponds to @xmath55 @xmath57 @xmath58 .
the chapman - kolmogorov equation formulated in differential form yields a master equation , which can take the form of a fokker - planck equation ( for a detailed discussion , we refer the reader to [ reichl , 1980 ; ernst et al . , 1969
; risken , 1984 ; hggi and thomas , 1982 ; gardiner , 1983 ] ) to @xmath55 is non linear and might be debatable if the following equations ( 4 ) and ( 9 ) are integrated with respect to time .
this is not done here .
the integrations discussed in the text are made at fixed @xmath9 or @xmath55 . ] : @xmath59 p ( \delta x , \tau)\ ] ] in terms of a drift @xmath5(@xmath8,@xmath55 ) and a diffusion coefficient @xmath6(@xmath8,@xmath55 ) .
note that the fokker - p1anck equation results from a truncation of the master equation expanded in terms of high order joint probabilities . according to pawula s theorem [ risken , 1984 ] ,
such a truncation is valid provided that the fourth order coefficient @xmath60 vanishes .
we did check the @xmath60 coefficient for @xmath61 $ ] and obtained values that are two orders of magnitude smaller than @xmath6 and three decades smaller than @xmath5 , which we consider justifies the truncation .
the functional dependence of the drift and diffusion coefficients can be estimated directly from the moments @xmath62 of the conditional probability distributions ( cf .
5 ) : @xmath63 for different small @xmath64 s ( fig .
6 ) , such that @xmath65 for @xmath66 . the coefficient @xmath5 shows a linear dependence on @xmath8 , while @xmath6 can be approximated by a polynomial of degree two in @xmath8 .
this behavior was found for all scales @xmath55 and @xmath64 .
therefore the drift term @xmath5 is well approximated by a linear function of @xmath8 , whereas the diffusion term @xmath6 follows a function quadratic in @xmath8 . for large values of @xmath8
the statistics becomes poorer and the uncertainty increases . from a careful analysis of the data based on the functional dependences of @xmath5 and @xmath6 ( fig .
6 a - b ) , the following approximations hold true : @xmath67 @xmath68 it may be worthwhile to remark that the observed quadratic dependence of the diffusion term @xmath6 is essential for the logarithmic scaling of the intermittency parameter in previous studies on turbulence . finally , the fpe for the distribution function is known to be equivalent to a langevin equation for the variable , i.e. @xmath8 here , ( within the ito interpretation [ reichl , 1980 ; ernst et al . , 1969 ; risken , 1984 ; hggi and thomas , 1982 ; gardiner , 1983 ] ) @xmath69 where @xmath70 is a fluctuating @xmath71-correlated force with gaussian statistics , i.e. @xmath30 @xmath70 @xmath72@xmath73 = 2@xmath74 .
at least since the pioneering work of lorenz [ lorenz , 1963 ] stochastic problems in turbulence are commonly treated as processes running in time @xmath29 with long time correlations .
inspired by the idea of an existing energy cascade process [ cahalan , 1994 ] we present here a new approach , namely , we investigate among other related phenomena , how lwp changes are correlated on different time steps @xmath9 .
the pdf shape expresses an unexpected high probability ( compared to a gaussian pdf ) of large lwp changes which is of utmost importance for forecasting . in recent works [ friedrich et al . , 2000 ;
friedrich and peinke , 1997 ] this finding leads to postulating the existence of hierarchical features , i.e. an energy cascade process from large to small time scales .
the existence of finite non gausssian tails for large events is thought to be due or to imply drastic evolutions , as for earthquakes , financial crashes or heart attacks .
one unexpected result is the time dependence of the tails ( fig.3 ) which seems quite smooth for the lwp case .
this puzzle should initiate new research with a goal toward forecasting , in particular to find cases in which such a smoothness does not hold .
this evolution shows _ how the pdf s deviate more and more from a gaussian shape _ as @xmath9 increases or @xmath55 decreases .
this definitely is a new quality in describing the hierarchical structure of such data sets , - not seen in the dfa nor spectral result .
now it becomes clear that one must not require stationary probability distributions for lwp differences for different @xmath55 ; on the contrary the coupling between different scales @xmath55 via a markov process is essential .
thus also a proper modeling of the time evolution of lwp differences for a fixed time delay must take into account the coupling of these quantities to lwp differences on different time delays .
furthermore , it is stressed that in contrast to the use of phenomenological fitting functions , the above method provides the evolution process of pdf s _ from small time delays to larger ones_. interestingly this is through an analogy with two physically meaningful coefficients , a drift term @xmath5 and a diffusion term @xmath6 .
the first one linearly behaves , thus looks like a `` restoring force '' , the second behaving quadratically in @xmath8 , is obviously like an autocorrelation function as for _ bona fide _
( chemical ) diffusion .
of course further theoretical work is needed before understanding the numerical values in eqs .
( 7)-(8 ) .
finally , the present report presents a method on how to derive an underlying mathematical ( _ statistical or model free _ ) equation for a lwp cascade directly .
the method yields an effective stochastic equation in the form of a fpe in the variable @xmath9 .
the excellent agreement between the experimental and the statistical approach removes the need for much further proof based on examining many other cases for the same type of clouds .
the fpe provides the complete knowledge as to how the statistics of lwp distribution change correlations on different delay times .
since this includes an autocorrelation analysis in time @xmath29 for a scalar @xmath8 , it is suggesting that the findings could be implemented in atmospheric weather low dimensional @xmath75 [ grabowski and smolarkiewicz , 1999 ; ragwitz and kantz , 2000 ] .
consequently , even though for the first time in atmospheric science , a stochastic equation ( of lwp evolution ) is hereby introduced from first principles , several interesting statements can be presented and new questions opened .
we thank c. nicolis , t.p .
ackerman , h.n .
shirer and e.e .
clothiaux for stimulating discussions and comments .
email correspondence with j. peinke , ch . renner and r. friedrich
is specifically appreciated .
this investigation was partially supported by battelle grant number 327421-a - n4 .
davis , a. , a. marshak , w. j. wiscombe , and r. f. cahalan , 1996a : multifractal characterizations of intermittency in nonstationary geophysical signals and fields - a model - based perspective on ergodicity issues illustrated with cloud data in _ current topics in nonstationary analysis _ , eds . g. trevino , j. hardin , b. douglas , and e. andreas , ( world scientific , singapore , 1996 ) 97 - 158 .
davis , a. , a. marshak , w. wiscombe and r. cahalan , 1996b : scale - invariance of liquid water distributions in marine stratocumulus .
part i : spectral properties and stationarity issues , _ j. atmos .
_ , * 53 * , 1538 - 1558 .
ivanova , k. , and m. ausloos , 1999 : application of the detrended fluctuation analysis ( dfa ) method for describing cloud breaking , in _ applications of statistical mechanics _ , proc .
of a nato arw budapest 1999 , a. gadomski , j. kertesz , h.e .
stanley , and n. vandewalle , eds .
, _ physica a _ , * 274 * , 349 - 354 .
ivanova , k. , m. ausloos , e.e .
clothiaux , and t.p .
ackerman , 2000 : break - up of stratus cloud structure predicted from non - brownian motion liquid water and brightness temperature fluctuations , _ europhys .
, _ * 52 * ( 1 ) , 40 - 46 .
koscielny - bunde , e. , a. bunde , s. havlin , h. e. roman , y. goldreich , and h .- j .
schellnhuber , 1998 : indication of a universal persistence law governing atmospheric variability _ phys .
_ , * 81 * , 729 - 732 .
marshak , a. , a. davis , w. j. wiscombe , and r. f. cahalan , 1997 : scale - invariance of liquid water distributions in marine stratocumulus .
part ii : multifractal properties and intermittency issues , _ j. atmos .
_ , * 54 * , 1423 - 1444 .
vandewalle , n. , m. ausloos , m. houssa , p.w .
mertens and m.m .
heyns , 1999 : non - gaussian behavior and anticorrelations in ultra - thin gate oxides after soft breakdown_appl .
_ , * 74 * , 1579 - 1581 . | how to distinguish and quantify deterministic and random influences on the statistics of turbulence data in meteorology cases is discussed from first principles .
liquid water path ( lwp ) changes in clouds , as retrieved from radio signals , upon different delay times , can be regarded as a stochastic markov process .
a detrended fluctuation analysis method indicates the existence of long range time correlations .
the fokker - planck equation which models very precisely the lwp @xmath0 empirical probability distributions , in particular , their non - gaussian heavy tails is explicitly derived and written in terms of a drift and a diffusion coefficient .
furthermore , kramers - moyal coefficients , as estimated from the empirical data , are found to be in good agreement with their first principle derivation .
finally , the equivalent langevin equation is written for the lwp increments themselves .
thus rather than the existence of hierarchical structures , like an energy cascade process , _ strong correlations _ on different @xmath1 @xmath2 , from small to large ones , are considered to be proven as intrinsic ingredients of such cloud evolutions . 0.5 cm _ pacs numbers : _
05.45.tp , 05.45.gg , 93.30.fd , 89.69.+x ; 02.50.le , 05.40.-a , 47.27.ak , 87.23.ge 0.5 cm |
this paper considers the use of importance sampling for estimating hitting or exit probabilities for stochastic processes .
the process model is a @xmath0-dimensional diffusion @xmath1 satisfying the stochastic differential equation ( sde ) @xmath2 where @xmath3 and @xmath4 is a standard @xmath0-dimensional wiener process .
of particular interest is the case of gradient flows , @xmath5 , and constant diffusion coefficient , though many aspects of the analysis are more generally applicable .
let @xmath6 be an open set and denote by @xmath7 the exit time of @xmath8 from @xmath9 .
we are concerned with the estimation of quantities such as the probability that @xmath10 leaves @xmath9 before some time @xmath11 , or that it exits through a particular subset @xmath12 before @xmath13 , and related expected values .
the principal novel feature of this work is that the initial point is in the neighborhood of an equilibrium point of the noiseless dynamics .
the estimation of such probabilities has several mathematical and computational difficulties .
it is related to the estimation of transition probabilities between different metastable states within a given time horizon . as is well known , standard monte carlo sampling techniques lead to exponentially large relative errors as the noise coefficient @xmath14 tends to zero . when rest points are in the domain of interest the situation is even more complicated than usual .
this work will focus on this particularly difficult issue .
the performance of unbiased estimators for rare event problems is usually measured by the size of the second moment of the estimator based on a single simulation . for
a well designed scheme the ratio of this second moment to the quantity of interest will not grow too rapidly as @xmath15 .
one measure is the exponential rate of decay of the second moment .
if this rate of decay is exactly twice the decay rate for the probability of interest , then the scheme is called asymptotically efficient ( or weakly efficient ) .
the notion of _ strong _ efficiency requires that the ratio of the second moment to the square of the probability be bounded above uniformly for all small @xmath3 . while such performance is certainly desirable , it is not common when dealing with models such as ( [ eq : diffusion1 ] ) that involve state dependent dynamics and complicated geometries .
as we describe below , in some sense both these measures are inadequate for the situation considered here . a theory based on subsolutions to an associated hamilton - jacobi - bellman ( hjb ) equation has been developed for the design and performance analysis of importance sampling
( see for example @xcite ) . in this approach
the change of measure ( which for reasons made evident later on we will call the control ) used in the importance sampling is defined in terms of the gradient of a subsolution , and the performance , as measured by the decay rate of the second moment , is given by the value of the subsolution at the initial location @xmath16 .
this theory is the starting point of our analysis of ( [ eq : diffusion1 ] ) , though as mentioned previously the inclusion of rest points will motivate some further developments . for any particular class of process models and events ,
an essential step in the application of this approach is the construction of appropriate subsolutions . in this paper
we will exploit the fact that the freidlin - wentsell quasipotential @xcite can be used to construct various subsolutions for certain _ time independent _ problems related to ( [ eq : diffusion1 ] ) .
in addition , for particular but important classes of process models ( e.g. , gradient systems with constant diffusion matrix ) , the quasipotential and hence these subsolutions take explicit and simple forms . as we discuss in detail later on , these subsolutions also give subsolutions for the _ time dependent _ problems , and when @xmath13 is large the value of the subsolution at the starting point ( which now includes time @xmath17 as well as the location ) will be close to the maximal value .
it follows that if the final time @xmath13 is large enough , then existing theory implies that the estimator based on this subsolution should have a nearly optimal decay rate for its second moment .
while this is a valid statement , there is an important qualitative difference between problems which include a rest point in the domain of interest and those which do not . the distinction turns not on the decay rate , which behaves as expected in both situations , but rather depends on the pre - exponential terms not captured by the decay rate .
when the domain does not contain a rest point one has simultaneously good rates of decay and control over the pre - exponential terms . however , when a rest point is present schemes based only on this time independent subsolution keep the desired decay rate but lose the good control over the pre - exponential terms .
the qualitative difference is related to the fact that in the former case a subsolution designed on the basis of the @xmath18 problem can be shown to give useful bounds for the problem with @xmath3 , but in the latter case this is no longer true .
this qualitative distinction will be made precise later on , when we construct non - asymptotic bounds on the second moment for the two cases .
when @xmath3 is small but not too small , the loss of performance due to the large pre - exponential term can be significant , rendering the associated importance sampling scheme little better than ordinary monte carlo . as @xmath15 the exponential decay rate dominates , and importance sampling once again gives much greater performance than ordinary monte carlo .
however , the improvement is less than in the case where rest points are not included , and an approach which avoids this loss of performance would certainly be welcome .
one can show that in fact there is no time independent subsolution that will resolve this difficulty , and so one must bring in some form of time dependence .
it is not at all clear how one might incorporate time dependence throughout the entire domain @xmath19\times\mathcal{d}$ ] .
an alternative approach , and the one we follow in this paper , is to combine an explicit solution to an approximating time dependent problem in a neighborhood of the rest point with the time independent subsolution obtained via the quasipotential away from the rest point . as we will show , such an approach will maintain the high decay rate while at the same time properly controlling the pre - exponential term .
in the neighborhood of the rest point , one can approximate the dynamics of the diffusion process by a gauss - markov process , i.e. , a process with a constant diffusion matrix and drift that is affine in the state . for these dynamics and appropriate terminal conditions for the localizing problem
, the solution to the related pde can be constructed in terms to the famous linear / quadratic regulator problem from optimal control theory . as a consequence
an explicit and nearly optimal scheme for a surrogate problem can be identified in the neighborhood of the rest point , which is then merged with the explicit scheme based on the quasipotential in that part of the domain where it is particularly effective . in this paper
we analyze the difficulties caused by the presence of rest points in a general setting .
we describe and theoretically justify a resolution of these difficulties in the case of dimension one , and present computational data for this case .
the construction of the localizing problem is more elaborate in dimension greater than one , and will be presented in a companion paper along with the results of numerical experiments in higher dimensions .
the contents of this paper are as follows . in section [ s : ldpandis ]
we review the relevant large deviation theory and importance sampling . in section [ s : effectivenessofsubsolutions ] we discuss the effectiveness of time independent subsolutions . in particular , we show that if rest points are not part of the domain of interest then subsolutions lead to both good decay rates and non - asymptotic bounds for the second moment of the corresponding unbiased estimator .
however , when rest points are included in the domain of interest , the situation is more complicated and even if the decay rate is good , the prelimit bounds may not be as good as desired . in section [ s : linearproblem ] , we present a change of measure for the problem with a rest point for a quadratic potential function with provably good pre - asymptotic and asymptotic performance and which does not degrade as @xmath13 gets larger . in section [ s : nonlinearproblem ]
we extend the discussion to the nonlinear problem with rest points .
simulation data , demonstrating the discussions in sections [ s : linearproblem ] and [ s : nonlinearproblem ] are also presented in the corresponding sections .
the appendix has some auxiliary lemmas that are used in the main body of the manuscript .
in this section we recall well known large deviation results for probabilities of exit times ( subsection [ ss : ldp ] ) , review importance sampling in the context of small noise diffusions ( subsection [ ss : is ] ) , and also recall the notion of subsolutions to certain related hjb equations ( subsection [ ss : roleofsubsolutions ] ) . in most of this paper the following assumptions will be used .
the assumptions are stronger than necessary , but simplify the discussion considerably .
for example , the non - degeneracy of the diffusion matrix and regularity of the boundary of @xmath9 easily imply that a limit exists for ( [ eq : estimationtarget2 ] )
. they can be weakened , but the existence of the limit then requires conditions that are best addressed in a problem dependent fashion .
[ a : mainassumption ] 1 .
the drift @xmath20 is bounded and lipschitz continuous .
2 . the coefficient @xmath21 is bounded , lipschitz continuous and uniformly nondegenerate .
@xmath9 is an open and bounded subset of @xmath22 , and at all points on its boundary @xmath9 satisfies an interior and exterior cone condition , i.e. , there is @xmath23 such that if @xmath24 , then there exist unit vectors @xmath25 such that @xmath26 and @xmath27 we also assume that if @xmath28 and if @xmath29 solves @xmath30 , then @xmath31 for all @xmath32 .
fix @xmath11 and consider an initial point @xmath33 . consider a bounded and class @xmath34 function @xmath35 .
let @xmath36 denote expected value given @xmath37 , and define @xmath38 .
\label{eq : estimationtargetsmooth1}\ ] ] since @xmath39 scales exponentially it is also useful to define @xmath40 although for now we focus on the case where @xmath41 is bounded and continuous , one is also interested in cases where @xmath41 is discontinuous and takes the value @xmath42 .
an example is when for some set @xmath43 , @xmath44 if @xmath45 and @xmath46 if @xmath47
. in this case @xmath48 equals the probability of exiting @xmath9 through @xmath49 before time @xmath13 . for these cases and under mild regularity conditions on @xmath49 , statements analogous to theorem [ t : ldptheorem ] below hold .
let @xmath50:\mathbb{r}^{d}\right ) $ ] be the set of absolutely continuous functions on @xmath51 $ ] with values in @xmath22 .
we denote the local rate function by@xmath52 \right\rangle \label{eq : localratefunction}\ ] ] where @xmath53 , and the corresponding rate or action functional for @xmath54:\mathbb{r}^{d}\right ) $ ] by @xmath55 for all other @xmath56:\mathbb{r}^{d}\right ) $ ] set @xmath57 .
the following large deviations result is well known , e.g. , @xcite .
[ t : ldptheorem ] assume condition [ a : mainassumption ] .
then for each @xmath58 @xmath59 , \label{eq : ldpth}\ ] ] where @xmath60:\mathbb{r}^{d}):\phi ( t)=x,\phi(s)\in\mathcal{d}\text { for } s\in\lbrack t , t],\phi(t)\in \partial\mathcal{d}\right\ } .\ ] ] we briefly review the use of importance sampling for estimating @xmath48 for a given function @xmath41 .
let @xmath61 be any unbiased estimator of @xmath48 that is defined on some probability space with probability measure @xmath62 .
thus @xmath63 is a random variable such that @xmath64 where @xmath65 is the expectation operator associated with @xmath62 . in this paper
we will consider only unbiased estimators . in monte carlo simulation ,
one generates a number of independent copies of @xmath63 and the estimate is the sample mean .
the specific number of samples required depends on the desired accuracy , which is measured by the variance of the sample mean .
however , since the samples are independent it suffices to consider the variance of a single sample . because of unbiasedness , minimizing the variance is equivalent to minimizing the second moment . by jensen s inequality
@xmath66 it then follows from theorem [ t : ldptheorem ] that @xmath67 and thus @xmath68 is the best possible rate of decay of the second moment . if @xmath69 then @xmath63 achieves this best decay rate , and is said to be _ asymptotically optimal_. while asymptotic optimality or near asymptotic optimality is desirable , as noted in the introduction one may also desire good behavior of the pre - exponential term . to keep the terminology clear we will avoid the conventional usage of terms such as asymptotic optimality , and refer instead to properties of the decay rate and the pre - exponential term .
the unbiased estimators @xmath63 that we consider are all based on measure transformation .
consider @xmath70 , a sufficiently integrable and adapted function , such that @xmath71 defines a family of probability measures @xmath72 .
then by girsanov s theorem , for each @xmath3 @xmath73 is a brownian motion on @xmath51 $ ] under the probability measure @xmath74 , and @xmath10 satisfies @xmath75 and @xmath76 .\ ] ] for our purposes , @xmath70 is either given as a process that is progressively measurable with respect to a suitable filtration that measures the wiener process ( sometimes called an _
open loop _
control ) , or else it is of _ feedback _ form , in which case there is a suitably measurable function @xmath77\times\mathbb{r}^{d}\rightarrow \mathbb{r}^{d}$ ] such that @xmath78 .
of course when implementing importance sampling we consider only the latter form .
letting @xmath79 it follows easily that under @xmath72 , @xmath61 is an unbiased estimator for @xmath48 .
the performance of this estimator is characterized by its second moment @xmath80 .
\label{eq:2ndmoment1}\ ] ] the goal of this paper is to investigate the effect of rest points on @xmath81 and how one can choose controls that guarantee both good decay rates and pre - exponential bounds for @xmath81 .
we conclude this section with a review of subsolutions to related hjb equations .
such subsolutions are essential for constructing and analyzing good important sampling schemes .
let @xmath82 the construction of good importance sampling schemes for a quantity such as ( [ eq : estimationtargetsmooth1 ] ) is closely related to the hjb equation @xmath83@xmath84 and more precisely to its subsolutions .
it can be shown that @xmath85 defined in theorem [ t : ldptheorem ] is the unique continuous viscosity solution of ( [ eq : hjbequation1 ] ) and ( [ eq : hjbterm_cond ] ) , see @xcite .
[ def : classicalsubsolution ] a function @xmath86\times \mathbb{r}^{d}\mapsto\mathbb{r}$ ] is a classical subsolution to the hjb equation ( [ eq : hjbequation1 ] ) and ( [ eq : hjbterm_cond ] ) if 1 .
@xmath87 is continuously differentiable , 2 .
@xmath88 for every @xmath58 , 3 .
@xmath89 for @xmath90 and @xmath91 for @xmath92 .
the connection between subsolutions and the performance of importance sampling schemes has been established in several papers , such as @xcite .
these papers either consider classical subsolutions or , more generally , piecewise classical subsolutions . to simplify the discussion
, we consider here just classical subsolutions . in the present setting , we have the following theorem regarding asymptotic optimality ( theorem 4.1 in @xcite ) . [
t : uniformlylogefficientregime1 ] let @xmath93 be the unique strong solution to ( [ eq : diffusion1 ] ) . consider a bounded and continuous function @xmath35 and assume condition [ a : mainassumption ] .
let @xmath94 be a subsolution according to definition [ def : classicalsubsolution ] and define the control @xmath95
. then @xmath96 since @xmath87 is a subsolution it is automatic that @xmath97 for all @xmath98\times\mathcal{d}$ ] .
if @xmath99 then the scheme has the largest possible decay rate .
in this section we justify some of the claims made in the introduction regarding the differences in performance between importance sampling schemes when rest points are included in the domain of interest and when they are not .
we consider just the problem of estimating the probability of escape from a set before time @xmath13 , and even then consider a particular setup .
however , the example will illustrate the difference between the two cases , and also suggest how one might improve the performance of importance sampling when rest points are involved .
_ much of the prior application of subsolutions to importance sampling @xcite has involved the estimation of escape probabilities for classes of stochastic networks , in which case the origin is often the unique stable point for the law of large numbers dynamics [ the analogue of ( [ eq : diffusion1 ] ) with _
the event most often studied in this context is that of escape from a set ( i.e. , buffer overflow ) before reaching the origin , after starting near but not at the origin . the analogous event for the diffusion model ( [ eq : diffusion1 ] )
is one of the problems that are the focus of the present work .
however , the difficulties that will be described momentarily for the diffusion model do not arise when dealing with the analogous estimation problem for stochastic networks , and indeed in that setting the proximity of the rest point has little impact on either the rate of decay or the pre - exponential term .
this is related to the fact that the law of large numbers trajectories for stochastic networks reach the origin in finite time , as opposed to the infinite time it takes for the solution to ( [ eq : diffusion1 ] ) with _
_ @xmath18 _ to reach a stable equilibrium point when not starting at such a point . in turn , this property is responsible for the fact that minimizing trajectories in the definition of the quasipotential are achieved on bounded time intervals for stochastic network models , but take infinite time for processes such as ( [ eq : diffusion1 ] ) . _ for the remainder of this section we concentrate on the special case of @xmath5 and @xmath100 , and on a particular estimation problem .
we first argue that if the domain of interest does not include a rest point , then given a time - independent subsolution and associated control not only is a good decay rate obtained , but good bounds on the pre - exponential terms hold as well .
we then show why this is not possible when a rest point is included .
assume that @xmath101 is the global minimum for @xmath102 , so that @xmath103 , and that @xmath104 for all @xmath105 .
without loss we assume that @xmath106 .
let @xmath107 and define @xmath108 and @xmath109 .
then the problem is to estimate @xmath110 where the initial point @xmath111 is such that @xmath112 .
this corresponds to ( [ eq : estimationtargetsmooth1 ] ) , but here @xmath41 is not bounded and smooth , and instead @xmath46 if @xmath113 and @xmath44 if @xmath114 . for
this problem one can also identify the rate of decay @xmath115 via ( [ eq : ldpth ] ) .
a one - dimensional example is illustrated in figure [ fig : no_rest ] .
[ ptb ] no_rest.eps the quasipotential with respect to the equilibrium point @xmath116 is defined by @xmath117:\mathbb{r}^{d}),\phi(0)=0,\phi(t)=x , t\in(0,\infty)\}. \label{eq : quasipotential}\ ] ] it follows from the variational characterization of @xmath118 that @xmath119 is always a weak sense solution to @xmath120 , and therefore by adding an appropriate constant @xmath121 to satisfy any needed boundary and terminal conditions , @xmath122 will always define a weak sense subsolution . in the present case
@xmath123 takes the explicit form ( theorem 4.3.1 in @xcite ) @xmath124 and it is easy to check that @xmath125 is a subsolution according to definition [ def : classicalsubsolution ] . indeed , @xmath126 for @xmath113 , while the boundary condition @xmath127 for @xmath114 and terminal condition @xmath127 for @xmath92 hold vacuously
. see figure [ fig : sub_no_rest ] .
the control ( i.e. , change of measure ) suggested by this subsolution for the importance sampling scheme is @xmath128 .
[ ptb ] sub_no_rest.eps the following representation for @xmath129 follows essentially from the arguments of subsection 2.3 of @xcite , which is based on the representation for exponential integrals with respect to brownian motion given in @xcite .
the representation is given in terms of the value of a stochastic differential game , where the player corresponding to the importance sampling scheme has already selected their control ( i.e. , @xmath130 ) .
the characterization of performance for importance sampling in terms of games was first introduced in @xcite .
the only difference between the use here and in @xcite is that there the function @xmath41 is bounded , which is not true here since we consider an escape probability .
however , the bounds stated below can be obtained by first replacing @xmath131 by @xmath132 and then letting @xmath133 .
let @xmath134 be a filtration satisfying the usual conditions of completion and right continuity and which measures the wiener process .
let @xmath135 denote the set of all @xmath134-progressively measurable @xmath0-dimensional processes @xmath136 that satisfy @xmath137 let @xmath3 be fixed and let @xmath138 be the unique strong solution to@xmath139ds\right ] \label{eq : control_diffusion}\ ] ] with initial condition @xmath140 .
let @xmath141 denote the first time @xmath138 exits @xmath9 .
then @xmath142 .
\label{eq : game_rep}\ ] ] it is important to note that ( [ eq : game_rep ] ) provides a non - asymptotic representation for the performance measure @xmath143 .
however , to obtain a more concrete statement regarding the performance of the importance sampling scheme , we will want bounds on @xmath143 that are more explicit than the right hand side of ( [ eq : game_rep ] ) .
we do this by observing that when viewed as a function of an arbitrary starting point @xmath144 , @xmath145 also satisfies a nonlinear pde of the same general form as ( [ eq : hjbequation1 ] ) [ plus terminal and boundary conditions ] , and thus lower bounds can be obtained by constructing subsolutions for this pde . however , a key difference is that in contrast to ( [ eq : hjbequation1 ] ) , the pde for ( [ eq : game_rep ] ) involves a second derivative term .
one can not avoid this issue , in that second derivative information and @xmath14 dependence are needed if one is to obtain non - asymptotic bounds , even when the change of measure is based on a first order equation .
we next give the statement of the lower bound as it applies to the special case of this section .
a more general statement and the proof will be given in lemma [ l : generalbound ] .
however , the proof is an easy consequence of it s formula and the min / max representation @xmath146 .\ ] ] define @xmath147(t , x)=w_{t}(t , x)+\mathbb{h}(x , dw(t , x))+\frac { \varepsilon}{2}d^{2}w(t , x),\ ] ] and let @xmath148 be a subsolution to @xmath149=0 $ ] together with the boundary conditions @xmath150 for @xmath151 , @xmath152 for @xmath153 , and terminal condition @xmath154 for @xmath92 .
suppose @xmath130 is the control based on a given smooth function @xmath87 , i.e. , @xmath155 .
then @xmath156 \nonumber\\ & \qquad\geq2\bar{w}(0,y)\nonumber\\ & \qquad\qquad+\inf_{v\in\mathcal{a}:\hat{\tau}^{\varepsilon}\leq t\text { w.p.1}}\mathbb{e}\left [ \int_{0}^{\hat{\tau}^{\varepsilon}}2\mathcal{g}^{\varepsilon}[\bar{w}](s,\hat{x}^{\varepsilon}(s))ds-\int_{0}^{\hat{\tau } ^{\varepsilon}}\left\vert d\bar{w}(s,\hat{x}^{\varepsilon}(s))-d\bar { u}(s,\hat{x}^{\varepsilon}(s))\right\vert ^{2}ds\right ] .\nonumber\end{aligned}\ ] ] next we show how ( [ eq : lbq ] ) can be used to obtain bounds that are uniform in @xmath13 . for
@xmath157 define@xmath158 where @xmath159 is the subsolution based on the quasipotential for @xmath160 as above , and assume that @xmath159 is twice continuously differentiable .
then as with @xmath159 the appropriate boundary and terminal inequalities hold for @xmath161 .
we next evaluate the right side of ( [ eq : lbq ] ) when @xmath162 and @xmath163 .
a straightforward calculation gives @xmath164 and therefore @xmath165(x)-\frac{1}{2}\left\vert du^{\eta } ( x)-du(x)\right\vert ^{2}=2(\eta-2\eta^{2})\left\vert dv(x)\right\vert ^{2}-\varepsilon(1-\eta)d^{2}v(x).\ ] ] for @xmath3 but smaller than a constant that depends on @xmath166 and @xmath167 , there is @xmath168 with @xmath169 as @xmath170 such that the last display is non - negative .
we then obtain from ( [ eq : lbq ] ) the non - asymptotic upper bound _ _
@xmath171 note that this bound is independent of @xmath13 , and also that this argument is not possible when @xmath172 .
indeed , since @xmath173 for all @xmath174 , @xmath175(x)\geq0 $ ] is not possible for any choice of @xmath176 when @xmath172 .
the quality of the bound obtained by this method depends on the degree to which the subsolution obtained for the pde @xmath149=0 $ ] ( plus boundary and terminal conditions ) accurately approximates the solution to this equation . in this example , we have used quite crude method to produce such a subsolution , which is to simply reduce a given subsolution to the @xmath18 equation by a constant factor of @xmath177 .
an examination of the calculations suggest that the bound is not at all tight , which turns out to be true .
in fact , in this situation we can construct a better subsolution and hence a tighter bound .
for example , so long as the origin is not included @xmath178 can be used to obtain tighter bounds , though this function is not convenient for the time dependent problem .
note that the two functions @xmath159 and @xmath179 play very different roles here .
one is used to design an importance sampling scheme [ here @xmath159 ] , and one used for its analysis [ here @xmath179 ] .
indeed , @xmath179 with @xmath182 is used only for the analysis of the scheme that corresponds to @xmath159 and in particular to derive a bound that is independent of @xmath13 .
however , the design of the scheme and thus the simulation algorithm use the control @xmath155 .
next we consider the behavior of @xmath143 when @xmath172 . in this case
, we claim that @xmath143 grows without bound in @xmath13 for all @xmath3 , and therefore the performance of the control based on the quasipotential degrades as @xmath13 becomes large . to show this is true
, we use the game representation to establish a lower bound on @xmath143 .
we again examine a particular situation , which is to estimate the probability of escape from @xmath183 before time @xmath13 , after starting at @xmath111 at time @xmath116 .
see figure [ fig : with_rest ] .
the subsolution is still that of figure [ fig : sub_no_rest ] .
[ ptb ] with_rest.eps with the understanding that @xmath184 now represents the time of escape of @xmath138 from @xmath185 , the representation ( [ eq : game_rep ] ) is still valid .
suppose that @xmath13 is large and note that , while @xmath186 destabilizes the origin when used to construct the measure used for importance sampling , in the representation ( [ eq : game_rep ] ) it actually _ increases _ the stability of the origin , in the sense that @xmath187 . as a consequence ,
it is easy to construct a control @xmath188 which shows poor performance as @xmath189 .
the construction is suggested in figure [ fig : bad_large_t ] . with @xmath13 large we divide @xmath19 $ ] into an initial part @xmath190 and a final part @xmath191,@xmath192 $ ] , with @xmath193 fixed . during the first part we apply @xmath194 . because the resulting dynamics of @xmath138 are stable about the origin , with very high probability the process settles around @xmath116 for the entire interval @xmath190 . in the game representation there is then a running cost of @xmath195 , which one can check is of order @xmath3 . in the second portion
we apply a control which leads to escape prior to time @xmath13 , with a cost that may depend on @xmath193 but is independent of @xmath13 .
an example of such a control , at least away from the origin , is as illustrated in figure [ fig : bad_large_t ] .
the precise details of the construction in this second part are not important .
all that is needed is that such a control exists , which can easily be demonstrated .
[ ptb ] bad_large_t.eps when the two parts are combined , we have a control that provides an upper bound of the form @xmath196+c_{2},\ ] ] where @xmath197 and @xmath198 are positive constants .
this shows that @xmath199}e^{-\frac{1}{\varepsilon}c_{2 } } , \label{eq : bad_lower_bound}\ ] ] and thus for fixed @xmath3 and large @xmath13 the term we have called the pre - exponential term dominates and the scheme is very far from optimal .
we find that in this situation there are two exponential scalings , one in the noise strength and one in the length of the time interval , and the issue of which dominates depends on their relative sizes .
these effects are reflected in computational data . in tables
[ table1a ] and [ table1b ] we present both estimated values and relative errors for the problem of escape from an interval of the form @xmath200 $ ] , with @xmath201 . the process is a one dimensional gauss - markov model with drift @xmath202 and diffusion coefficient @xmath203 [ see ( [ eq : gm ] ) ] , with @xmath204 . in the tables ,
values of @xmath13 appear at the top and values of @xmath14 along the left side .
each computed value is based on @xmath205 samples .
a dash indicates that no samples escaped . to ease the presentation the relative errors are rounded to the nearest integer . owing to the fact that the subsolution based on the quasipotential is far from optimal in any sense when @xmath13 is small , the relative errors are large for small @xmath13 and decrease until approximately @xmath206 . for larger @xmath13
the errors grow rapidly with @xmath13 . note that the estimated relative errors in table [ table1b ] are not necessarily accurate for large @xmath13 , since they are subject to the same errors that can affect the probability estimates , but do indicate the qualitative worsening of estimation accuracy .
[ c]|c|c|c|c|c|c|c|c|c|@xmath207 & @xmath208 & @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath213 & @xmath214 & @xmath215 + @xmath216 & @xmath217 & @xmath218 & @xmath219 & @xmath220 & @xmath221 & @xmath222 & @xmath223 & @xmath224 + @xmath225 & @xmath226 & @xmath227 & @xmath228 & @xmath229 & @xmath230 & @xmath231 & @xmath232 & @xmath233 + @xmath234 & @xmath235 & @xmath236 & @xmath237 & @xmath238 & @xmath239 & @xmath240 & @xmath241 & @xmath242 + @xmath243 & @xmath244 & @xmath245 & @xmath246 & @xmath247 & @xmath248 & @xmath249 & @xmath250 & @xmath251 + @xmath252 & @xmath253 & @xmath254 & @xmath255 & @xmath256 & @xmath257 & @xmath258 & @xmath259 & @xmath260 + @xmath261 & @xmath253 & @xmath262 & @xmath263 & @xmath264 & @xmath265 & @xmath266 & @xmath267 & @xmath268 + @xmath269 & @xmath253 & @xmath270 & @xmath271 & @xmath272 & @xmath273 & @xmath274 & @xmath275 & @xmath276 + [ c]|c|c|c|c|c|c|c|c|c|@xmath207 & @xmath208 & @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath213 & @xmath214 & @xmath215 + @xmath216 & @xmath277 & @xmath278 & @xmath279 & @xmath210 & @xmath210 & @xmath213 & @xmath280 & @xmath281 + @xmath225 & @xmath282 & @xmath213 & @xmath279 & @xmath210 & @xmath210 & @xmath213 & @xmath283 & @xmath284 + @xmath234 & @xmath285 & @xmath286 & @xmath287 & @xmath210 & @xmath210 & @xmath288 & @xmath283 & @xmath289 + @xmath243 & @xmath290 & @xmath291 & @xmath287 & @xmath210 & @xmath210 & @xmath213 & @xmath292 & @xmath293 + @xmath252 & @xmath253 & @xmath294 & @xmath295 & @xmath279 & @xmath210 & @xmath288 & @xmath296 & @xmath297 + @xmath261 & @xmath253 & @xmath298 & @xmath299 & @xmath279 & @xmath210 & @xmath300 & @xmath301 & @xmath302 + @xmath269 & @xmath253 & @xmath303 & @xmath300 & @xmath279 & @xmath210 & @xmath300 & @xmath292 & @xmath304 + tables [ table1c ] and [ table1d ] present the approximated values and relative errors for the problem with the domain @xmath305 $ ] and escape possible therefore only at @xmath306 .
the results are of the same qualitative form as before , and carried out only to @xmath307 .
[ c]|c|c|c|c|c|c|c|c|@xmath207 & @xmath208 & @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath278 & @xmath213 + @xmath216 & @xmath308 & @xmath309 & @xmath310 & @xmath311 & @xmath240 & @xmath312 & @xmath313 + @xmath225 & @xmath314 & @xmath315 & @xmath316 & @xmath317 & @xmath219 & @xmath318 & @xmath319 + @xmath234 & @xmath320 & @xmath321 & @xmath322 & @xmath323 & @xmath324 & @xmath325 & @xmath326 + @xmath243 & @xmath327 & @xmath328 & @xmath256 & @xmath329 & @xmath218 & @xmath330 & @xmath331 + @xmath252 & @xmath253 & @xmath332 & @xmath321 & @xmath333 & @xmath334 & @xmath237 & @xmath218 + @xmath261 & @xmath253 & @xmath335 & @xmath336 & @xmath337 & @xmath321 & @xmath338 & @xmath339 + @xmath269 & @xmath253 & @xmath340 & @xmath341 & @xmath342 & @xmath343 & @xmath344 & @xmath345 + [ c]|c|c|c|c|c|c|c|c|@xmath207 & @xmath208 & @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath278 & @xmath213 + @xmath216 & @xmath346 & @xmath213 & @xmath287 & @xmath279 & @xmath279 & @xmath299 & @xmath347 + @xmath225 & @xmath348 & @xmath347 & @xmath287 & @xmath279 & @xmath279 & @xmath295 & @xmath214 + @xmath234 & @xmath349 & @xmath350 & @xmath295 & @xmath279 & @xmath279 & @xmath295 & @xmath214 + @xmath243 & @xmath351 & @xmath352 & @xmath295 & @xmath279 & @xmath279 & @xmath295 & @xmath214 + @xmath252 & @xmath253 & @xmath353 & @xmath354 & @xmath287 & @xmath279 & @xmath295 & @xmath355 + @xmath261 & @xmath253 & @xmath356 & @xmath278 & @xmath287 & @xmath279 & @xmath295 & @xmath357 + @xmath269 & @xmath253 & @xmath358 & @xmath357 & @xmath295 & @xmath279 & @xmath295 & @xmath357 + it is useful to compare the two situations , and identify why uniform control of pre - exponential terms was not possible when @xmath172 . in both cases
the control was based on the quasipotential , which is a valid subsolution to the @xmath18 problem . when using ito s formula to bound the second moment of the estimator , we must of course deal with the second derivative term , which is multiplied by @xmath3 .
it can happen that this term has a sign that degrades ( increases ) the second moment , and indeed this is always true in a neighborhood of the origin ( this is essentially due to the convexity of @xmath160 near the origin ) . for the case
where @xmath359 and for sufficiently small @xmath3 , this could be balanced by using that when @xmath105 @xmath360 and therefore @xmath361 are nonzero . however ,
this is not possible when the rest point is included in the domain of interest .
indeed , the running cost that is accumulated in the construction leading to ( [ eq : bad_lower_bound ] ) corresponds to this term , and as that argument shows it can not be removed .
the construction also suggests how the large variance comes about , which is that some trajectories generated under the change of measure defined by @xmath130 remain in a neighborhood of the origin for a long time , in spite of the fact that with such dynamics the origin is an unstable equilibrium point .
the likelihood ratios along these trajectories can vary greatly and , even though they are themselves relatively unlikely , they are likely enough to increase the variance of the estimator to the point where it will become worse than standard monte carlo . as such , they are reminiscent of the rogue trajectories which lead to poor performance of non - dynamic forms of importance sampling as discussed in @xcite
. it will turn out that to overcome the difficulties introduced by the rest point one must do a much better job of approximating the optimal change of measure than is possible based just on a time and @xmath14-independent subsolution .
however , it also turns out that the additional accuracy is needed only near the rest point , where in fact explicit time and @xmath362-dependent solutions can be found .
these are then combined with the simple time - independent subsolution based on the quasipotential to produce schemes that are nearly optimal and which protect against both sources of significant variance .
an overview of the construction of such schemes is the topic of the next section .
in this section we combine a local analysis that produces a time and @xmath14-dependent scheme near the rest point with a scheme based on the quasipotential elsewhere . there are of course few process models and problems for which the related hjb equation can be solved explicitly . however , a class of processes where this is possible are the gauss - markov models , i.e. , sdes with drift that is linear in the state and constant diffusion matrix
. for these processes and for terminal conditions of the appropriate form , both the limit ( @xmath18 ) pde and the prelimit ( @xmath3 ) pde have an explicit solution that can be expressed in terms of the value function of a linear - quadratic regulator ( lqr ) control problem . [ ptb ] part_for_sub2.eps our ultimate approach to the construction of importance sampling schemes
is suggested by figure [ fig : localization2 ] .
the problem of interest is of the form ( [ eq : estimationtargetsmooth1 ] ) or an analogous problem involving escape from @xmath9 prior to @xmath13 .
the particular problem will fix the boundary and terminal conditions on @xmath19\times\partial\mathcal{d}$ ] and @xmath363 . in the figure , we are interesting in estimating the exit probability @xmath364 where @xmath116 is the rest point and @xmath365 with @xmath366 . in most of the domain , which is the section outside the curves that terminate at @xmath367 and after time @xmath368 , the control is based on a subsolution @xmath159 constructed in terms of the quasipotential . within the curves the solution to ( [ eq : diffusion1 ] ) is well approximated by a gauss - markov process . hence within this region we will use a control that would be appropriate for a problem if the process were instead the approximating gauss - markov model .
the function @xmath369 that defines the pde for this region is therefore the one corresponding to the gauss - markov process . besides the dynamics in this region ( which are determined by the gauss - markov approximation )
, we must choose a terminal condition .
this will be given by the minimum of two quadratic functions , one centered at @xmath370 and one centered at @xmath371 .
the parameter @xmath370 is a nominal value that for purposes of the present discussion can be taken to be @xmath210 .
the parameter @xmath372 plays two roles .
one is to determine the size of the region on which the true dynamics are approximated by gauss - markov dynamics .
the second role is related to the fact that if the control based on the quadratics centered at @xmath373 is used all the way to @xmath13 then singularities will develop .
for this reason , we switch back to the control based on the quasipotential after @xmath368 , and @xmath372 must be chosen so that the subsolution property is preserved across the handoff time @xmath368 .
while the discussion above suggests the correct decomposition of the domain , the actual construction is more complex , since the control should transition nicely when moving between the regions , and the construction of a scheme for which rigorous bounds can be proved will require additional mollification and approximations
. however , the building blocks are always subsolutions to the indicated pdes .
the form of the surrogate problem for the gauss - markov model needs to be explained , as well as various boundary and terminal conditions . to simplify the discussion
, we consider a sequence of successively more general problems . in this paper
we complete the analysis for the one - dimensional problem .
the multi - dimensional problem will be addressed in a companion paper .
our first goal is to construct a subsolution for the @xmath18 problem for the processes that will be used in the localization .
this can be related to the problem of estimating the probability that the solution to @xmath374 escapes from the interval @xmath200 $ ] before time @xmath13 , a problem that was also used for the computational examples of the last subsection .
the parameters @xmath375 and @xmath376 are positive constants . in this subsection we will take @xmath377 , and because of this can postpone the issue regarding singularities in the control at @xmath378 .
we thus also take @xmath372 to be zero , and will return to the role of @xmath372 and its selection for a general problem in the next subsection .
the corresponding pde for the escape probability is @xmath379 plus the terminal and boundary conditions@xmath380{cc}0 & x=\pm a , t\in\lbrack0,t],\\ \infty & x\in(-a , a),t = t .
\end{array } \right.\ ] ] while simple in appearance , this equation does not have an explicit solution .
the equation obtained in the limit @xmath381 is more tractable , and the unique viscosity solution can be described as follows .
@xmath382 corresponds to the variational problem @xmath383\right\ } .\ ] ] depending on how and when the minimizing trajectory leaves @xmath200 $ ] , the solution takes a particular explicit form .
( in all cases the minimizer can be found by solving the appropriate euler - lagrange equation . )
if for the initial condition @xmath144 the minimizing trajectory leaves before time @xmath13 , then @xmath384 .\ ] ] this is the case when the minimal cost is the negative of the quasipotential , translated by a constant to satisfy the boundary condition at the exit location .
such initial conditions satisfy @xmath385 .
when @xmath386 @xmath387 the minimizer leaves through @xmath306 at exactly time @xmath13 , and the minimizing value is @xmath388}.\ ] ] one can also interpret @xmath389 as the minimal cost for a linear quadratic regulator with a singular terminal cost applied at time @xmath13 , i.e. , a cost that equals @xmath116 at @xmath306 and @xmath42 otherwise . by symmetry
it is clear that when @xmath390 the minimizing trajectory will exit at @xmath391 .
define @xmath392 setting @xmath393 , we have @xmath394 note that when @xmath395 @xmath396}=\frac{c}{\bar{\sigma}^{2}}\frac{(a - ae^{2c(t - t)})^{2}}{[1-e^{2c(t - t)}]}=\frac{c}{\bar{\sigma}^{2}}\left [ a^{2}-x^{2}\right ] = f_{1}(ae^{c(t - t)}),\ ] ] and therefore @xmath382 is continuous for all @xmath397 $ ] .
in fact more is true , and one can check that @xmath398 and thus @xmath382 has a continuous partial derivative in @xmath16 for @xmath399 .
the mapping @xmath400 is convex , and the graph of this mapping lies above that of @xmath401 , which is concave .
thus the two graphs intersect only at the point @xmath402/\bar{\sigma}^{2})$ ] , where both functions and their first derivatives in @xmath16 agree .
see figure [ fig : eps_zero_soln ] .
[ ptb ] eps_zero_soln.eps note also that @xmath403 for @xmath404 $ ] is the solution to the @xmath18 problem with escape from @xmath305 $ ] ( a one - sided version of the problem of escape from @xmath200 $ ] ) .
simulation data for the schemes based on these two value functions are presented below .
the approximated values are omitted since they are qualitatively similar to those in tables [ table1a ] and [ table1c ] , and only relative errors are presented .
table [ table2a ] gives data based on @xmath405 for the problem of two - sided exit , and should be compared to table [ table1b ] .
the use of the solution to hjb equation for @xmath18 drastically improves the performance for small @xmath13 , which is due to the fact that the subsolution based on the quasipotential is a very poor approximation to the solution for @xmath3 for such @xmath13 .
however , for large @xmath13 the two schemes are comparably bad .
[ c]|c|c|c|c|c|c|c|c|c|@xmath207 & @xmath208 & @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath213 & @xmath214 & @xmath215 + @xmath216 & @xmath279 & @xmath279 & @xmath210 & @xmath210 & @xmath210 & @xmath213 & @xmath406 & @xmath407 + @xmath225 & @xmath287 & @xmath287 & @xmath210 & @xmath210 & @xmath210 & @xmath213 & @xmath408 & @xmath409 + @xmath234 & @xmath295 & @xmath279 & @xmath210 & @xmath210 & @xmath210 & @xmath288 & @xmath280 & @xmath410 + @xmath243 & @xmath287 & @xmath279 & @xmath279 & @xmath210 & @xmath210 & @xmath288 & @xmath411 & @xmath412 + @xmath252 & @xmath287 & @xmath287 & @xmath279 & @xmath210 & @xmath210 & @xmath288 & @xmath296 & @xmath413 + @xmath261 & @xmath287 & @xmath287 & @xmath279 & @xmath279 & @xmath210 & @xmath288 & @xmath414 & @xmath415 + @xmath269 & @xmath295 & @xmath279 & @xmath279 & @xmath279 & @xmath210 & @xmath300 & @xmath416 & @xmath417 + the data for exit from one side are given in table [ table2b ] .
in contrast with the two - sided problem , the performance does not degrade so quickly with large @xmath13 .
this suggests that the decline in performance is not so much due to using the approximating @xmath18 pde , but rather the possible lack of regularity of the solution to this equation .
[ c]|c|c|c|c|c|c|c|c|@xmath207 & @xmath208 & @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath278 & @xmath213 + @xmath216 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath279 & @xmath287 + @xmath225 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath279 & @xmath295 + @xmath234 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath279 & @xmath287 + @xmath243 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath279 & @xmath287 + @xmath252 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath279 & @xmath287 + @xmath261 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath279 & @xmath287 + @xmath269 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath279 & @xmath287 + recall that the difficulty with the use of the subsolution based on the quasipotential alone ( here @xmath401 ) was that in a neighborhood of @xmath101 the second derivative was negative , and when @xmath13 is large this leads to poor control of the variance of the associated scheme for both the one - sided and two - sided problems .
when using the time - dependent @xmath18 pde as the basis for a scheme for the one - sided problem , the introduction of @xmath418 appears to have largely mitigated the problem due to the second derivative term .
note that this function determines the value of @xmath419 when @xmath101 and is convex rather than concave in @xmath16 .
in contrast , for the two - sided problem there is a concave singularity at the origin .
there the second derivative in @xmath16 is @xmath420 , and the subsolution property for the @xmath3 problem again fails .
what is needed is a subsolution that works across the point @xmath101 for the @xmath3 problem .
such a subsolution can be constructed using the mollification introduced in the next subsection .
simulations based on such a mollified version of the @xmath18 subsolution ( and with mollification parameter @xmath421 ) are presented in table [ table6b ] , and support the claim just made .
[ c]|c|c|c|c|c|c|c|c|c|c|@xmath207 & @xmath208 & @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath213 & @xmath214 & @xmath215 & @xmath350 + @xmath216 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath287 & @xmath299 & @xmath422 + @xmath225 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath287 & @xmath299 & @xmath423 + @xmath234 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath279 & @xmath295 & @xmath424 + @xmath243 & @xmath279 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath210 & @xmath279 & @xmath295 & @xmath425 + @xmath252 & @xmath279 & @xmath279 & @xmath279 & @xmath279 & @xmath210 & @xmath210 & @xmath279 & @xmath295 & @xmath426 + @xmath261 & @xmath279 & @xmath279 & @xmath279 & @xmath279 & @xmath210 & @xmath210 & @xmath279 & @xmath287 & @xmath408 + @xmath269 & @xmath279 & @xmath279 & @xmath279 & @xmath279 & @xmath210 & @xmath279 & @xmath279 & @xmath287 & @xmath427 + in spite of the shortcomings we have described in this section , the solution to the @xmath18 problem serves as the starting point for a construction that can be shown to perform well both in theory and in practice
. there are three important modifications that are needed . *
the first is that , in order to effectively deal with the @xmath3 dynamics and in particular to avoid the degradation still present in table [ table2b ] when @xmath13 is large , the region where the @xmath389 subsolution determines the dynamics must be enlarged .
the solution to the @xmath18 problem constructed above leads to a region that vanishes exponentially in @xmath428 [ see ( [ eq : basicfcn1 ] ) ] .
* the second modification will be the use a mollification to eliminate singularities such as the one at @xmath101 and help guarantee a global subsolution property for the @xmath3 pde .
the particular mollification we use is very convenient , and was first used for importance sampling in @xcite , though with a somewhat different intended use . *
the final modification was also alluded to previously , which is to revert to the quasipotential based control in an interval of the form @xmath429 $ ] .
all three modifications will be introduced in the next subsection in the context of the gauss - markov process . in this subsection
we generalize the construction of the last subsection .
as discussed there , the generalizations are needed to address issues that play a role in both the theoretical analysis of the scheme and its practical performance . after introducing these generalizations
, we will demonstrate ( numerically and theoretically ) that the suggested change of measure does not degrade in performance as @xmath13 gets larger and is close to optimality not only as @xmath15 , but for fixed @xmath3 as well .
we begin by introducing the first generalization , which is to replace the terminal condition @xmath430 , which was used to define @xmath418 , by a terminal condition of the form @xmath431 .
the parameter @xmath370 replaces @xmath306 and is a nominal value introduced to disconnect the localization from the boundary .
the solution to the lqr that corresponds to @xmath431 , which will be denoted @xmath432 , is automatically smaller than @xmath433 .
the motivation for replacing @xmath42 by @xmath434 is because we want the solution to the lqr problem [ i.e. , @xmath432 ] to determine the control near @xmath101 for a set whose width is uniformly ( in @xmath436 ) bounded below away from zero . as discussed in the last section ,
the second derivative term associated with @xmath437 is of the wrong sign and degrades performance .
the neighborhood where @xmath438 does not degenerate as @xmath439 , and its size is decreasing in @xmath440 .
the introduction of @xmath440 complicates the construction by also requiring mollification ( in addition to the one that will be needed at @xmath101 ) , since @xmath441 can no longer be smoothly merged with @xmath437 .
for @xmath442 the solution to this lqr takes the form @xmath443 recall that @xmath444/\bar{\sigma}^{2}$ ] so that @xmath445 .
. it will be important to know which of @xmath447 , @xmath448 , and @xmath437 is smallest , and we note here several properties .
let @xmath449 .
the first is that there are two real solutions to @xmath450 , and these take the form@xmath451 between these roots @xmath452 , and on the complement of the interval the reverse inequality holds .
the limit @xmath453 gives the asymptotic endpoints of the interval where @xmath452 , which are @xmath454 we point out here that a natural scaling for this problem , given that the size of the neighborhood of zero where the quasipotential based subsolution fails for the @xmath3 system scales like @xmath455 , is to ask that this width scale as @xmath456 $ ] . if for example the desired width is @xmath457 , then when @xmath3 is small then we should take @xmath458 .
the next adaptation is required so that singularities in the control associated with @xmath459 as @xmath460 do not cause a problem , as well as for purposes of localization . for a parameter @xmath461
we will want @xmath462 for all @xmath463 $ ] , where ( with an abuse of notation ) @xmath464 .
this is true if and only if the smaller solution to @xmath465 is less than or equal to zero .
this root was found to be @xmath466 and the restriction that this be non positive can be simplified to@xmath467 the inequality @xmath462 will ensure that the subsolution property is preserved if we switch from using @xmath405 for @xmath468 to using @xmath437 for @xmath469 . if @xmath470 this means that we always use @xmath437 , but our interest here is in large @xmath13 .
we assume that @xmath471 so that @xmath472 . to guarantee that the smaller root is strictly negative and to conveniently satisfy a bound used later on we assume @xmath473 besides enforcing the subsolution property across the handoff at time @xmath368
, the selection of @xmath372 plays a key role in determining the region used for the localization for the general nonlinear problem . owing to the exponential decay
, one can confine the localization to a small region with a modest value of @xmath372 .
suppose we consider confining to a region that scales as @xmath457 for all @xmath474 $ ] and arbitrarily large @xmath13 .
as discussed previously , for small @xmath436 and large @xmath368 this suggests that @xmath440 be approximately @xmath475 , which means that @xmath476 $ ] . recalling
that the loss in performance of the importance sampling scheme when the quasipotential based subsolution is used scales like @xmath477 , this gives a loss over such an interval of the form @xmath429 $ ] as scaling like @xmath478 $ ] .
we will return to such considerations in the next section .
finally there is the issue of mollification .
owing to the replacement of @xmath479 by @xmath459 , there are now several sources of discontinuity in the gradient .
the subsolution prior to mollification would in general take the form @xmath480 .
however , since we know that @xmath481 is dominated by @xmath401 near @xmath101 , we can also use@xmath482 see figure [ fig : various_functions ] .
[ ptb ] domains_for_bounds3.eps we next state a result that will be used to derive performance bounds for schemes based on the mollification .
the proof is deferred to the appendix .
we consider the general one dimensional process model @xmath483 where @xmath20 and @xmath21 are lipschitz continuous . letting @xmath484 , the relevant @xmath14-dependent pde is@xmath485 ( t , x)=u_{t}(t , x)+du(t , x)b(x)-\frac { 1}{2}\left\vert \sigma(x)du(t , x)\right\vert ^{2}+\frac{\varepsilon}{2}\alpha(x)d^{2}u(t , x)=0 .
\label{eq : eps}\ ] ] [ lem : mollification]suppose that the functions @xmath486\times\mathbb{r}$ ] , @xmath487 are twice continuously differentiable in @xmath16 , once continuously differentiable in @xmath436 , and satisfy @xmath488(t , x)\geq\gamma_{i}(x , t,\varepsilon).\ ] ] for @xmath23 define @xmath489 and define the weights@xmath490 then @xmath491 and for @xmath492@xmath493(t , x ) & \geq\frac{1}{2}\left ( 1-\frac{\varepsilon}{\delta}\right ) \left [ \sum_{i=1}^{n}\rho_{i}\left ( t , x;\delta\right ) \left\vert \sigma(x)d\tilde{u}_{i}(t , x)\right\vert ^{2}-\left\vert \sum_{i=1}^{n}\rho_{1}\left ( t , x;\delta\right ) \sigma(x)d\tilde{u}_{i}(t , x)\right\vert ^{2}\right ] \\ & \qquad+\sum_{i=1}^{n}\rho_{i}\left ( t , x;\delta\right ) \gamma _ { i}(x , t,\varepsilon)\\ & \geq\sum_{i=1}^{n}\rho_{i}\left ( t , x;\delta\right ) \gamma_{i}(x , t,\varepsilon).\end{aligned}\ ] ] based on this result we consider the mollification of @xmath494 , which is @xmath495 for reasons that will be made clear in the analysis ( see lemma [ l : region_y_hatx ] ) , we generally use @xmath421 . as discussed previously , we return to the control based on the quasipotential for the last @xmath372 units of time , and so the subsolution takes the form@xmath496{cc}f_{1}(x ) , & t > t - t^{\ast}\\ u^{\delta}(t , x ) , & t\leq t - t^{\ast}\end{array } \right . .
\label{eq : controldesignlinear}\ ] ] note that the mollification reduces values , in that @xmath497 , and so the requirement @xmath498 for @xmath463 $ ] holds , since @xmath462 .
we next present a rigorous and nonasymptotic bound for the second moment of the importance sampling scheme constructed in the last subsection . to derive a bound that is valid for @xmath3 and uniform in @xmath13
, we use the same representation as in ( [ eq : lbq ] ) .
in particular , we choose @xmath499 defined via ( [ eq : controldesignlinear ] ) for the design and @xmath500 for the analysis , where @xmath501{cc}f_{1}(x ) , & t > t - t^{\ast}\\ u^{\delta,\eta}(t , x ) , & t\leq t - t^{\ast}\end{array } \right . , \label{eq : controlanalysislinear}\ ] ] with @xmath502 as discussed in section 3 this choice is driven by the need for a subsolution for the @xmath3 dynamics with an explicit form , and this limitation of the technique leads to conservative bounds on the true performance . to simplify notation , for smooth functions @xmath503 we define @xmath504(t , x)=\mathcal{g}^{\varepsilon}[w](t , x)-\frac { 1}{2}\left\vert \bar{\sigma}\left ( dw(t , x)-du(t , x)\right ) \right\vert ^{2 } , \label{eq : ge_2}\ ] ] where @xmath149 $ ] is defined in ( [ eq : eps ] ) . using that @xmath505=\mathcal{g}^{0}[f_{2}^{m}]=0 $ ] , @xmath506(x)=\frac{\varepsilon}{2}\bar{\sigma}^{2}d^{2}f_{1}(x)=-\varepsilon c\text { and } \mathcal{g}^{\varepsilon}[f_{2}^{m}](t , x)=\frac{\varepsilon}{2}\bar{\sigma}^{2}d^{2}f_{2}^{m}(t , x)=\varepsilon\bar{\sigma}^{2}a^{m}(t ) .
\label{eq : values_for_pieces}\ ] ] since the problem is symmetric , it suffices to consider only @xmath507 $ ] .
a key role is played by the weights associated with the exponential mollification of the subsolution , which take the forms @xmath508 and @xmath509 to determine which of these dominate at any @xmath144 , the relative sizes of the functions @xmath459 and @xmath401 are required , with smaller functions corresponding to more dominant weights . for this reason the solutions to @xmath510 play an important role , and especially the larger one identified in ( [ eq : roots ] ) . see figure [ fig : various_functions ] .
the following bounds on this root will be used to partition the domain in the analysis .
let @xmath511 @xmath512 and let @xmath513 denote the larger root in ( [ eq : roots ] ) .
then we claim under ( [ eq : bound_on_tstar ] ) that @xmath514\ ] ] and uniformly in @xmath13 .
we recall the definition @xmath515 and that @xmath471 , so that @xmath516 . then the smallest possible value of @xmath513 satisfies @xmath517 while the largest satisfies @xmath518 where the first inequality uses @xmath519 .
we set @xmath520 . in order to obtain bounds on the performance under the corresponding scheme , we need to bound @xmath521(t , x)$ ] from below in various regions . by ( [ eq : ge_2 ] ) @xmath522(t , x ) & = \mathcal{g}^{\varepsilon}[u^{\delta,\eta}](t , x)-\frac{1}{2}\left\vert \bar{\sigma}\left ( du^{\delta,\eta}(t , x)-du^{\delta}(t , x)\right ) \right\vert ^{2}\nonumber\\ & = \mathcal{g}^{\varepsilon}[u^{\delta,\eta}](t , x)-\frac{1}{2}\eta ^{2}\left\vert \bar{\sigma}du^{\delta}(t , x)\right\vert ^{2}. \label{eq : combinedexpression}\ ] ] we will use the notation @xmath523(x)=-\varepsilon c \label{eq : def_gam0}\ ] ] and @xmath524(t , x)=\varepsilon \bar{\sigma}^{2}a^{m}(t ) .
\label{eq : def_gamm}\ ] ] straightforward calculations and some algebra give @xmath525(t , x)\geq(1-\eta)\mathcal{g}^{\varepsilon}[u^{\delta}](t , x)+\frac{1}{2}\left ( \eta-\eta^{2}\right ) \left\vert \bar{\sigma}du^{\delta}(t , x)\right\vert ^{2}.\ ] ] for notational convenience , define @xmath526 ( t , x).\end{aligned}\ ] ] note that by jensen s inequality @xmath527 .
we next apply lemma [ lem : mollification ] to @xmath528(t , x)$ ] ( while suppressing the dependence on @xmath529 in the notation for the @xmath530 s ) , and use ( [ eq : combinedexpression ] ) to get@xmath522(t , x ) & \geq ( 1-\eta)\frac{1}{2}\left ( 1-\frac{\varepsilon}{\delta}\right ) \beta _ { 0}(t , x)+(1-\eta)\left [ \rho_{2}^{m,+}(t , x)+\rho_{2}^{m,-}(t , x)\right ] \gamma_{2}^{m}(t)\nonumber\\ & \qquad+(1-\eta)\rho_{1}(t , x)\gamma_{1}+\frac{1}{2}\left ( \eta-2\eta ^{2}\right ) \left\vert \bar{\sigma}du^{\delta}(t , x)\right\vert ^{2 } \label{eq : bound_for_moll}\ ] ] for all @xmath463 $ ] and @xmath474 $ ] . we will partition the domain according @xmath511 @xmath531 and @xmath532 .
we consider three cases depending on whether @xmath533 $ ] , @xmath534 $ ] or @xmath535 $ ] if @xmath536 .
the case @xmath390 is symmetric . before proceeding with the analysis for each of the cases ,
we give the definition of exponential negligibility , a concept used frequently in the rest of the paper .
[ d : exponentialnegligibility ] a term is called exponential negligible if it is bounded above in absolute value by a quantity of the form @xmath537 , where @xmath538 and @xmath539 .
[ l : region_0_y ] assume that @xmath540\times \lbrack0,z]$ ] , @xmath541 and @xmath542 .
then , up to an exponentially negligible term @xmath543(t , x)\geq0.\ ] ] in this region @xmath544 , and we claim that the inequality is in fact strict .
we have @xmath545 -\frac{c}{k-\bar{\sigma}^{2}e^{2c(t - t)}}\left [ xe^{c(t - t)}-\hat{x}\right ] ^{2}. \label{eq : f1-f2}\ ] ] for each fixed @xmath436 this defines a concave function of @xmath16 . at @xmath101 the value is minimized when @xmath546 . using @xmath547^{4}\leq c / m\bar{\sigma}^{2}$ ] ( since @xmath548 ) and @xmath549 , we obtain the strictly positive lower bound @xmath550\right ] $ ] . since @xmath551 , by concavity there is @xmath552 such
that ( [ eq : f1-f2 ] ) is bounded below by @xmath553 for all @xmath540\times\lbrack0,z]$ ] .
thus the term in ( [ eq : bound_for_moll ] ) involving the weight @xmath554 is exponentially negligible . since @xmath527 , @xmath555 and @xmath556 , all other terms are non - negative , and the result follows .
[ l : region_hatx_a ] assume that @xmath540\times\lbrack h , a]$ ] , @xmath541 and @xmath557 .
then letting @xmath558 , we have that for all @xmath559 and with any @xmath560 $ ] , up to an exponentially negligible term@xmath543(t , x)\geq0 .
\label{eq : bound2_b}\ ] ] in this region @xmath561 , and it is straightforward that the terms associated with @xmath562 are exponentially negligible . note that @xmath563 is convex , recall that for each @xmath474 $ ] the largest value where the two functions agree is smaller than @xmath564 , and that @xmath565 inserting the largest root for the given @xmath436 gives the value@xmath566 a lower bound on the first term is @xmath567 . using @xmath568^{2}$ ] , the definition of @xmath193 , and @xmath569 to bound the second term from below produces the strictly positive lower bound@xmath570 for all @xmath474 $ ] and @xmath571 . since @xmath572
, this shows that there is @xmath573 such that @xmath574 for all @xmath540\times\lbrack h , a]$ ] .
it follows that terms involving @xmath575 are exponentially negligible . since @xmath527 and @xmath576 up to an exponentially negligible term , @xmath522(t , x ) & \geq ( 1-\eta)\rho_{1}(t , x)\gamma^{0}+\frac{1}{2}\left ( \eta-2\eta^{2}\right ) \bar{\sigma}^{2}\left\vert \rho_{1}(t , x)df_{1}(x)\right\vert ^{2}\\ & \geq-(1-\eta)\varepsilon c+2\left ( \eta-2\eta^{2}\right ) \frac{c^{2}}{\bar{\sigma}^{2}}x^{2}\ ] ] up to an exponentially negligible term .
choosing @xmath557 gives @xmath543(t , x)\geq-(1-\eta ) \varepsilon c+\eta\frac{c^{2}}{\bar{\sigma}^{2}}x^{2}\ ] ] and for @xmath14 small enough such that @xmath577 $ ] , the last display is non - negative . for this interval to be nonempty imposes the constraint @xmath578 .
hence in this region and up to an exponentially negligible term,@xmath543(t , x)\geq0.\ ] ] the final region is the most difficult , since @xmath579 can be either positive or negative .
[ l : region_y_hatx ] assume that @xmath540\times\lbrack z , h]$ ] , @xmath557 and set @xmath421 .
then up to an exponentially negligible term@xmath543(t , x)\geq\frac{1}{2}\left [ \frac{c^{2}\eta}{2\bar{\sigma}^{2}}\left ( z-\hat{x}e^{c(t - t)}\right ) ^{2}-2\varepsilon c\right ] \wedge0 .
\label{eq : bound2_a}\ ] ] while terms corresponding to @xmath562 are exponentially negligible in this region , since @xmath580 changes sign both @xmath581 and @xmath554 may be important . since they are negligible we omit terms corresponding to @xmath562 . by ( [ eq : bound_for_moll ] )
we have up to an exponentially negligible term@xmath522(t , x)\geq & ( 1-\eta)\frac{1}{4}\beta_{0}(t , x)+(1-\eta)\rho_{2}^{m,+}(t , x)\gamma_{2}^{m}(t)+(1-\eta)\rho_{1}(t , x)\gamma_{1}\nonumber\\ & + \frac{1}{2}\left ( \eta-2\eta^{2}\right ) \bar{\sigma}^{2}\left\vert \rho_{2}^{m,+}(t , x)df_{2}^{m}(t , x)+\rho_{1}(t , x)df_{1}(x)\right\vert ^{2 } \label{eq : boundforgreyarea1}\ ] ] as noted previously @xmath527 for all @xmath144 .
however , we will exploit the fact that @xmath582 only for points @xmath144 such that @xmath583 .
we distinguish two cases depending on whether @xmath584 or @xmath585 .
* case i : @xmath584 . *
we know that @xmath527 and @xmath586 and can ignore those terms .
using @xmath587 , the terms that remain are @xmath588 .\end{aligned}\ ] ] we claim that for @xmath540\times\lbrack z , h]$ ] @xmath589 first , we note that @xmath590 , and thus @xmath591
. therefore @xmath592 second , by ( [ eq : diff_of_deriv ] ) , the definition of @xmath593 and @xmath594 , we also have @xmath595 where the last inequality uses @xmath569 .
we conclude that @xmath596 since @xmath597 and @xmath557 , we obtain the bound@xmath543(t , x)\geq-(1-\eta ) \varepsilon c+\frac{1}{16}\eta\bar{\sigma}^{2}|df_{1}(x)-df_{2,+}^{m}(t , x)|^{2}. \label{eq : casei}\ ] ] this gives a bound for case i. * case ii : @xmath585 . * here we will have to use @xmath598 . dropping other terms on the right that are not possibly negative
, we obtain from ( [ eq : boundforgreyarea1 ] ) that @xmath543(t , x)\geq(1-\eta)\frac { 1}{4}\beta_{0}(t , x)+(1-\eta)\rho_{1}(t , x)\gamma_{1}.\ ] ] omitting exponentially negligible terms , we note that @xmath599 ( t , x)\\ & = \bar{\sigma}^{2}\rho_{1}\left [ \left\vert df_{1}\right\vert ^{2}-\left\vert df_{2,+}^{m}\right\vert ^{2}-2df_{2,+}^{m}(df_{1}-df_{2,+}^{m})-\rho_{1}(df_{1}-df_{2,+}^{m})^{2}\right ] ( t , x)\\ & = \bar{\sigma}^{2}\rho_{1}(df_{1}-df_{2,+}^{m})\left [ df_{1}+df_{2,+}^{m}-df_{2,+}^{m}-\rho_{1}(df_{1}-df_{2,+}^{m})\right ] ( t , x)\\ & = \bar{\sigma}^{2}\rho_{1}(1-\rho_{1})(df_{1}-df_{2,+}^{m})^{2}(t , x)\\ & \geq\frac{1}{2}\bar{\sigma}^{2}\rho_{1}(df_{1}-df_{2,+}^{m})^{2}(t , x),\end{aligned}\ ] ] where the last inequality uses @xmath585 , and so obtain @xmath543(t , x)\geq(1-\eta)\rho _ { 1}\left [ \frac{1}{8}\bar{\sigma}^{2}|df_{1}(x)-df_{2,+}^{m}(t , x)|^{2}-\varepsilon c\right ] .\ ] ] this gives a bound for case ii .
straightforward estimation gives , for all @xmath540\times\lbrack z , h]$ ] , the lower bound@xmath600 using this bound and @xmath557 , we get a lower bound from ( [ eq : casei ] ) in the form @xmath522(t , x ) & \geq\left [ \frac{1}{16}\eta\bar{\sigma}^{2}|df_{1}(x)-df_{2,+}^{m}(t , x)|^{2}-\varepsilon c\right ] \\ & \geq\left [ \frac{c^{2}\eta}{4\bar{\sigma}^{2}}\left ( z-\hat{x}e^{c(t - t)}\right ) ^{2}-\varepsilon c\right ] .\end{aligned}\ ] ] since this is less than the bound for case ii when both terms are negative , the conclusion of the lemma follows .
[ t : mainboundlinear ] assume @xmath601 , and that @xmath602 .
let @xmath130 be the control based on the function @xmath603 defined in ( [ eq : controldesignlinear ] ) , i.e. , @xmath604 . then up to an exponentially negligible term , we have@xmath605 where @xmath606 \wedge0\right ) \varepsilon.\ ] ] @xmath607 and @xmath608 .
although the bound provided by theorem [ t : mainboundlinear ] takes a complicated form , it is important to note that it does not degrade as @xmath189 , and this is also reflected in the simulation data . also , as noted previously there are natural scalings under which @xmath609 and @xmath610 as @xmath381 . using the bound from below given in lemma [ lem : mollification ] and the explicit form of @xmath611
, we obtain @xmath612 hence , we obtain the rate of decay@xmath613\ ] ] uniformly in @xmath13 as @xmath381 .
[ proof of theorem [ t : mainboundlinear]]the starting point is the representation ( [ eq : game_rep ] ) [ but rewritten for the more general process model and with time dependent @xmath614 , which is valid for every @xmath3 .
we can restrict to @xmath188 such that @xmath615 w.p.1 , obtaining @xmath616 .
\label{eq : rep_with_constraint}\ ] ] we can also assume @xmath617 since the bound is straightforward otherwise .
we recall that under ( [ eq : bound_on_tstar ] ) the subsolution property is preserved for @xmath618 at @xmath368 , i.e. , that@xmath619 .
\label{eq : subsolutionproperty_tstar}\ ] ] next consider any control in the representation such that @xmath620 w.p.1 .
we will apply it s formula separately over the intervals @xmath621 and @xmath622 and also use the boundary condition @xmath623 for @xmath624 $ ] . since @xmath625 we obtain @xmath626 1_{\left\ { \hat{\tau}^{\varepsilon}\geq t - t^{\ast } \right\ } } \nonumber\\ & \quad+\mathbb{e}_{0,0}\left [ \bar{u}^{\delta,\eta}((t - t^{\ast})\wedge \hat{\tau}^{\varepsilon},\hat{x}^{\varepsilon}((t - t^{\ast})\wedge\hat{\tau } ^{\varepsilon}))-\bar{u}^{\delta,\eta}(0,0)\right ] .
\label{eq : two_partsa}\ ] ] using lemma [ l : generalbound ] and recalling the definition of @xmath627 $ ] in ( [ eq : ge_2 ] ) , the contribution from @xmath621 gives@xmath628 \\ & = \mathbb{e}_{0,0}\int_{0}^{(t - t^{\ast})\wedge\hat{\tau}^{\varepsilon}}\left [ \mathcal{g}^{\varepsilon}[\bar{u}^{\delta,\eta},\bar{u}^{\delta } ] ( s,\hat{x}^{\varepsilon}(s))ds-\frac{1}{4}v(s)^{2}ds+\frac{1}{2}\bar { u}(s,\hat{x}^{\varepsilon}(s))^{2}\right ] ds.\end{aligned}\ ] ] an analogous formula holds for @xmath629 , save that since @xmath630 , the term @xmath631 $ ] simplifies to @xmath632 $ ] . rearranging and using ( [ eq : two_partsa ] ) , @xmath633 \geq2\bar{u}^{\delta,\eta}(0,0)\\ & + \mathbb{e}_{0,0}\int_{0}^{(t - t^{\ast})\wedge\hat{\tau}^{\varepsilon}}2\mathcal{g}^{\varepsilon}[\bar{u}^{\delta,\eta},\bar{u}^{\delta}](s,\hat { x}^{\varepsilon}(s))ds+\mathbb{e}_{0,0}\int_{(t - t^{\ast})\wedge\hat{\tau } ^{\varepsilon}}^{\hat{\tau}^{\varepsilon}}1_{\left\ { \hat{\tau}^{\varepsilon}\geq t - t^{\ast}\right\ } } 2\mathcal{g}^{\varepsilon}[f_{1}](s,\hat{x}^{\varepsilon}(s))ds.\end{aligned}\ ] ] we now replace each term by a lower bound , using lemmas [ l : region_0_y ] , [ l : region_hatx_a ] and [ l : region_y_hatx ] for @xmath634 $ ] .
since the bounds are independent of the control process @xmath188 , the representation ( [ eq : rep_with_constraint ] ) implies @xmath635 ds - t^{\ast}2\varepsilon c,\ ] ] where @xmath636 are the times in @xmath637 $ ] where the integrand is negative .
we next use the constraint @xmath638 , which guarantees that for @xmath639 sufficiently large the integrand is in fact positive .
let @xmath640 .\ ] ] then since the integrand is only negative for @xmath641 , we obtain the lower bound @xmath642 2\varepsilon c$ ] for the integral . adding the remaining @xmath643
then gives the result as stated . in this subsection
we present simulation data for the linear problem and make several comments on the application of the algorithm . for comparison purposes
, we consider the same two sided problem corresponding to the data from tables [ table1a ] , [ table1b ] , [ table2a ] and [ table6b ] .
thus we consider the small noise diffusion process with drift @xmath644 , where @xmath645 , diffusion coefficient @xmath455 , and starting from the stable equilibrium point @xmath101 .
the goal is to estimate the probability of exiting the set @xmath646 by a given time @xmath13 .
as discussed in subsection [ ss:1dimlinearmodel_algorithm ] , the change of measure for the importance sampling scheme is based on the subsolution ( [ eq : controldesignlinear ] ) . in order to apply it to a given pair @xmath647
, one needs to choose the parameters @xmath648 . before presenting simulation data , we comment on these choices .
the analysis in subsection [ s : analysislinearproblem ] assumes @xmath519 and @xmath421 , and we will take @xmath649 .
as noted before lemmas [ l : region_0_y]-[l : region_y_hatx ] , it is natural to allow quantities such as @xmath593 and @xmath650 , which characterize the region where the solution to the lqr replaces the subsolution based on the quasipotential , to depend on @xmath14 .
one would like the width of this region to scale like @xmath651 , with @xmath652 $ ] , which in turn suggests that @xmath440 scale like @xmath653 .
however , the exponential negligibility of certain terms that holds when parameters such as @xmath654 and @xmath440 are independent of @xmath14 need not hold when they depend on @xmath14 .
for example , the exponential negligibility of the term @xmath655 appearing in lemma [ l : region_0_y ] should be examined .
recall that the exponential rate of decay of terms like @xmath656 is bounded by the smallest value of @xmath579 .
a lower bound of the form @xmath550\right ] $ ] was obtained in the proof of lemma [ l : region_0_y ] .
inserting the given scaling and approximating for small @xmath14 gives @xmath657 , and upon dividing by @xmath421 gives the exponent @xmath658 .
hence exponential negligibility requires @xmath659 , with smaller values of @xmath660 giving a faster rate of decay .
note however that the analysis assumes @xmath471 and @xmath638 . with regard to the condition @xmath471 , inserting the given scaling for @xmath440
we get the constraint @xmath661 .
this is clearly satisfied for small @xmath3 if @xmath370 is of order @xmath210 .
one may also take here @xmath370 to be of order @xmath662 and the constraint will be satisfied for small @xmath14 if @xmath663 .
we also remark here that for the nonlinear problem , the condition @xmath471 needs to be strengthened to @xmath664 . with regard to the condition @xmath638 , inserting the given scaling for @xmath440 and recalling the definition @xmath665
, we obtain the constraint @xmath666 .
this constraint is satisfied for small enough @xmath14 when @xmath667 , and moreover one can allow @xmath609 as @xmath381 . below we present simulation data for various choices of the parameters as indicated in the corresponding tables . in table
[ table1a1 ] , estimated probabilities are reported when @xmath668 and @xmath669 , whereas the related relative error per sample estimates are reported in table [ table1b1 ] ( since the relative errors are consistently smaller we now round to the nearest @xmath670 ) . in tables
[ table4b1 ] and [ table5b1 ] relative errors per sample estimates are reported for combinations of @xmath671 that depend on @xmath14 .
the related probability estimates are almost identical to those in table [ table1a1 ] .
note that the degradation in performance as @xmath13 is gets larger observed previously , is no longer present .
this agrees with the theoretical performance bound appearing in theorem [ t : mainboundlinear ] .
[ c]|c|c|c|c|c|c|c|c|c|@xmath672 & @xmath211 & @xmath212 & @xmath299 & @xmath278 & @xmath213 & @xmath214 & @xmath215 & @xmath350 + @xmath216 & @xmath221 & @xmath673 & @xmath232 & @xmath674 & @xmath222 & @xmath223 & @xmath675 & @xmath676 + @xmath225 & @xmath229 & @xmath230 & @xmath677 & @xmath242 & @xmath231 & @xmath232 & @xmath678 & @xmath679 + @xmath234 & @xmath238 & @xmath239 & @xmath680 & @xmath681 & @xmath240 & @xmath241 & @xmath319 & @xmath682 + @xmath243 & @xmath247 & @xmath248 & @xmath317 & @xmath683 & @xmath249 & @xmath684 & @xmath685 & @xmath686 + @xmath252 & @xmath315 & @xmath257 & @xmath687 & @xmath688 & @xmath689 & @xmath690 & @xmath691 & @xmath692 + @xmath261 & @xmath264 & @xmath265 & @xmath693 & @xmath256 & @xmath266 & @xmath267 & @xmath268 & @xmath694 + @xmath269 & @xmath272 & @xmath273 & @xmath695 & @xmath696 & @xmath697 & @xmath275 & @xmath698 & @xmath699 + [ c]|c|c|c|c|c|c|c|c|c|@xmath207 & @xmath211 & @xmath212 & @xmath299 & @xmath278 & @xmath213 & @xmath214 & @xmath215 & @xmath350 + @xmath216 & @xmath700 & @xmath701 & @xmath702 & @xmath703 & @xmath704 & @xmath704 & @xmath705 & @xmath703 + @xmath225 & @xmath706 & @xmath707 & @xmath708 & @xmath702 & @xmath709 & @xmath705 & @xmath705 & @xmath709 + @xmath234 & @xmath710 & @xmath711 & @xmath712 & @xmath701 & @xmath713 & @xmath709 & @xmath709 & @xmath709 + @xmath243 & @xmath714 & @xmath715 & @xmath716 & @xmath712 & @xmath717 & @xmath713 & @xmath702 & @xmath703 + @xmath252 & @xmath718 & @xmath718 & @xmath715 & @xmath719 & @xmath707 & @xmath211 & @xmath717 & @xmath708 + @xmath261 & @xmath720 & @xmath721 & @xmath722 & @xmath723 & @xmath714 & @xmath710 & @xmath706 & @xmath724 + @xmath269 & @xmath725 & @xmath726 & @xmath727 & @xmath728 & @xmath729 & @xmath721 & @xmath730 & @xmath731 + [ c]|c|c|c|c|c|c|c|c|@xmath672 & @xmath212 & @xmath299 & @xmath278 & @xmath213 & @xmath214 & @xmath215 & @xmath350 + @xmath216 & @xmath717 & @xmath703 & @xmath705 & @xmath704 & @xmath704 & @xmath705 & @xmath702 + @xmath225 & @xmath211 & @xmath713 & @xmath709 & @xmath705 & @xmath705 & @xmath705 & @xmath703 + @xmath234 & @xmath700 & @xmath708 & @xmath702 & @xmath709 & @xmath705 & @xmath705 & @xmath709 + @xmath243 & @xmath707 & @xmath701 & @xmath708 & @xmath703 & @xmath709 & @xmath705 & @xmath709 + @xmath252 & @xmath716 & @xmath712 & @xmath717 & @xmath713 & @xmath703 & @xmath709 & @xmath709 + @xmath261 & @xmath711 & @xmath724 & @xmath712 & @xmath717 & @xmath713 & @xmath702 & @xmath703 + @xmath269 & @xmath710 & @xmath212 & @xmath706 & @xmath700 & @xmath211 & @xmath717 & @xmath713 + [ c]|c|c|c|c|c|c|c|c|@xmath672 & @xmath212 & @xmath299 & @xmath278 & @xmath213 & @xmath214 & @xmath215 & @xmath350 + @xmath216 & @xmath211 & @xmath703 & @xmath705 & @xmath705 & @xmath703 & @xmath713 & @xmath717 + @xmath225 & @xmath700 & @xmath702 & @xmath709 & @xmath705 & @xmath709 & @xmath702 & @xmath708 + @xmath234 & @xmath707 & @xmath713 & @xmath709 & @xmath709 & @xmath709 & @xmath702 & @xmath702 + @xmath243 & @xmath724 & @xmath713 & @xmath703 & @xmath705 & @xmath703 & @xmath703 & @xmath713 + @xmath252 & @xmath711 & @xmath708 & @xmath703 & @xmath709 & @xmath709 & @xmath703 & @xmath713 + @xmath261 & @xmath710 & @xmath717 & @xmath702 & @xmath709 & @xmath709 & @xmath703 & @xmath713 + @xmath269 & @xmath714 & @xmath211 & @xmath713 & @xmath703 & @xmath709 & @xmath703 & @xmath702 +
in this section , we extend the construction of section [ s : linearproblem ] to the general non - linear one dimensional setting .
we also generalize the notation and allow the stable equilibrium to be an arbitrary point @xmath732 .
consider the process model ( [ eq : diffusion1 ] ) and assume that @xmath733 and that @xmath734 , @xmath735 and @xmath736 for all @xmath737 .
thus we can write @xmath644 with unique local minimum at @xmath738 and @xmath739 .
it is easy to see that the quasipotential with respect to the equilibrium point @xmath732 takes the form@xmath740 the problem of interest is to estimate the exit probability@xmath741 where @xmath732 is the initial ( and rest ) point such that @xmath742 .
furthermore , we assume that @xmath743 for all @xmath744 $ ] and @xmath745 for all @xmath746 .
set @xmath747 $ ] .
the approach to the nonlinear problem is to merge the linearized dynamics around the equilibrium point with the subsolution based on the quasipotential .
this subsolution is @xmath748 observe that the second order approximation to this function around the equilibrium point @xmath732 is@xmath749 where @xmath750 and @xmath751 .
let @xmath752 be such that @xmath753 .
the appropriate translated version of @xmath447 is @xmath754 and @xmath755 .
the subsolution for times less than @xmath368 will be the mollification of @xmath756 .
note that since @xmath757 agrees with @xmath401 up to second order it is still the case that @xmath481 will be smallest near @xmath738 . letting @xmath758 and@xmath496{cc}\bar{f}_{1}(x ) ,
& t > t - t^{\ast}\\ u^{\delta}(t , x ) , & t\leq t - t^{\ast}\end{array } \right .
, \label{eq : nonlinearchangeofmeasure}\ ] ] the suggested importance sampling control that is used for the simulation is @xmath759 notice that this construction reduces to the construction of the linear case if the potential is indeed quadratic , since then @xmath760 . in subsection [ ss : simulationnonlinearproblem ] we present simulation data for the nonlinear problem , demonstrating the effectiveness of the suggested change of measure .
the analysis and the theoretical bound for the performance of this scheme are completely analogous to the linear problem , modulo the additional error coming from the linearization of the dynamics in the neighborhood of the stable equilibrium point . in subsection
[ ss : proofnonlinearproblem ] we rigorously analyze the performance of this algorithm . in this subsection , we present simulation data for the nonlinear problem .
we take the drift to be @xmath644 , where the potential function is @xmath761 .
this potential function has two stable points at @xmath762 and at @xmath763 , and an unstable equilibrium at @xmath116 .
we assume that the starting point is at the right equilibrium point @xmath764 and the exit set is the level set of the potential function @xmath765 , with @xmath766 .
thus exit occurs from either of the points @xmath767 or @xmath768 .
notice that the local quadratic approximation around the equilibrium point is @xmath769 with @xmath770 .
moreover , we have chosen , for simplicity , the diffusion coefficient to be constant @xmath771 .
@xmath772 independent trajectories were used for the simulations .
we first investigate the performance of a change of measure based on the quasipotential subsolution .
thus we change the measure via the control @xmath773 .
estimated values and the corresponding estimated relative errors per sample for several values of @xmath647 are in tables [ table1anl ] and [ table1bnl ] , respectively .
[ c]|c|c|c|c|c|c|c|c|c|c|@xmath672 & @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath295 & @xmath299 & @xmath354 & @xmath300 & @xmath213 + @xmath774 & @xmath775 & @xmath776 & @xmath777 & @xmath778 & @xmath779 & @xmath780 & @xmath781 & @xmath782 & @xmath783 + @xmath784 & @xmath785 & @xmath786 & @xmath787 & @xmath788 & @xmath789 & @xmath790 & @xmath791 & @xmath792 & @xmath793 + @xmath252 & @xmath794 & @xmath795 & @xmath796 & @xmath797 & @xmath798 & @xmath799 & @xmath800 & @xmath801 & @xmath802 + @xmath261 & @xmath803 & @xmath804 & @xmath805 & @xmath806 & @xmath807 & @xmath808 & @xmath809 & @xmath810 & @xmath811 + @xmath269 & @xmath812 & @xmath813 & @xmath814 & @xmath815 & @xmath816 & @xmath817 & @xmath818 & @xmath819 & @xmath820 + @xmath821 & @xmath822 & @xmath823 & @xmath824 & @xmath825 & @xmath826 & @xmath827 & @xmath828 & @xmath829 & @xmath830 + [ c]|c|c|c|c|c|c|c|c|c|c|@xmath672 & @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath295 & @xmath299 & @xmath354 & @xmath300 & @xmath213 + @xmath774 & @xmath279 & @xmath279 & @xmath295 & @xmath831 & @xmath832 & @xmath833 & @xmath834 & @xmath835 & @xmath836 + @xmath784 & @xmath279 & @xmath279 & @xmath295 & @xmath215 & @xmath837 & @xmath838 & @xmath839 & @xmath840 & @xmath841 + @xmath252 & @xmath279 & @xmath279 & @xmath295 & @xmath842 & @xmath843 & @xmath844 & @xmath845 & @xmath846 & @xmath847 + @xmath261 & @xmath279 & @xmath279 & @xmath295 & @xmath848 & @xmath423 & @xmath849 & @xmath850 & @xmath851 & @xmath293 + @xmath269 & @xmath279 & @xmath279 & @xmath295 & @xmath286 & @xmath852 & @xmath853 & @xmath854 & @xmath855 & @xmath856 + @xmath821 & @xmath279 & @xmath279 & @xmath287 & @xmath214 & @xmath857 & @xmath858 & @xmath859 & @xmath860 & @xmath861 + as we see from table [ table1bnl ] , even though the quasipotential subsolution performs relatively well for small values of @xmath13 , there is a clear degradation of performance as @xmath13 gets larger .
it is also interesting to note that the degradation is uniform across all values of @xmath14 for the same value of @xmath13 .
this behavior parallels what was observed for the linear problem .
as was mentioned there , the large per sample relative errors for @xmath862 should not be taken as being accurate , but just indicative of poor performance .
next we investigate how the suggested change of measure performs .
to apply the control , we choose values for the parameters @xmath863 according to the discussion in subsection [ s : simulationresultslinearproblem ] .
however , for reasons that will become clearer in the proof of lemma [ l : region_y_hatx_nonlinear ] , we need to strengthen the condition @xmath471 to @xmath664 , say @xmath864 . when we link the other parameters to @xmath14 the @xmath865 as @xmath381 , and because of this we do not explicitly take into account the error from the approximation around the neighborhood of the rest point of the true dynamics by its linearization when selecting the parameters for the implementation of the scheme . estimated values and corresponding estimated relative errors of the exit probabilities of interest for different pairs @xmath647 and different combinations values for @xmath593 are in tables [ table2anl ] through [ table5bnl ] .
estimated relative errors for @xmath866 with @xmath867 and @xmath868 are reported in table [ table2anl ] , whereas the related relative errors are reported in table [ table2bnl ] . in tables
[ table3bnl ] and [ table4bnl ] , we report only estimated relative errors for the same value of @xmath660 but for @xmath869 and @xmath669 respectively .
the related probability estimates are almost identical to those of table [ table2anl ] , so they are not repeated .
note that for table [ table3bnl ] , @xmath870 when @xmath871 and for @xmath872 .
similarly , for table [ table4bnl ] , @xmath870 when @xmath871 and for @xmath873 when @xmath874 . for such values
, the quasipotential subsolution is being used everywhere ( see ( [ eq : nonlinearchangeofmeasure ] ) ) and the numerical results for these values agree with those from table [ table1bnl ] . in order to illustrate the effect when the linear approximation is used over a relatively large region , the data in table [ table5bnl ]
are estimated relative errors when @xmath440 is considerably smaller than before , and thus @xmath593 is considerably larger .
in particular , we have taken @xmath875 and @xmath876 . comparing tables [ table4bnl ] and [ table5bnl ]
, we notice that if the error from the linearization is not confined to a small enough region then the algorithm degrades in @xmath14 though it appears stable in @xmath13 .
this is consistent with the theoretical results , which imply a uniformity in @xmath13 but only logarithmic optimality in @xmath14 .
that said , one would like to minimize errors associated with linearization as far as possible . as noted in subsection [
s : simulationresultslinearproblem ] , one should choose the scaling parameter @xmath667 .
however , with the nonlinear problem minimizing the region over which the approximation is used calls for larger @xmath660 , and so one want it close to but not equal to @xmath877 . for the problems considered here , @xmath878 worked well .
[ c]|c|c|c|c|c|c|c|c|c|c|c|@xmath672 & @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath295 & @xmath299 & @xmath354 & @xmath300 & @xmath213 & @xmath355 + @xmath774 & @xmath879 & @xmath880 & @xmath881 & @xmath882 & @xmath883 & @xmath884 & @xmath885 & @xmath886 & @xmath887 & @xmath888 + @xmath784 & @xmath889 & @xmath890 & @xmath891 & @xmath892 & @xmath893 & @xmath894 & @xmath895 & @xmath896 & @xmath897 & @xmath898 + @xmath252 & @xmath899 & @xmath900 & @xmath901 & @xmath902 & @xmath903 & @xmath904 & @xmath905 & @xmath906 & @xmath907 & @xmath908 + @xmath909 & @xmath910 & @xmath911 & @xmath912 & @xmath913 & @xmath914 & @xmath915 & @xmath916 & @xmath917 & @xmath918 & @xmath919 + @xmath261 & @xmath803 & @xmath804 & @xmath920 & @xmath921 & @xmath922 & @xmath923 & @xmath924 & @xmath925 & @xmath926 & @xmath927 + @xmath269 & @xmath928 & @xmath813 & @xmath929 & @xmath930 & @xmath931 & @xmath932 & @xmath933 & @xmath934 & @xmath935 & @xmath936 + @xmath937 & @xmath938 & @xmath939 & @xmath940 & @xmath941 & @xmath942 & @xmath943 & @xmath944 & @xmath945 & @xmath946 & @xmath947 + @xmath821 & @xmath948 & @xmath949 & @xmath824 & @xmath950 & @xmath951 & @xmath952 & @xmath953 & @xmath954 & @xmath955 & @xmath956 + [ c]|c|c|c|c|c|c|c|c|c|c|c|@xmath672 & @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath295 & @xmath299 & @xmath354 & @xmath300 & @xmath213 & @xmath355 + @xmath774 & @xmath957 & @xmath719 & @xmath707 & @xmath717 & @xmath713 & @xmath958 & @xmath702 & @xmath713 & @xmath717 & @xmath700 + @xmath784 & @xmath722 & @xmath959 & @xmath716 & @xmath211 & @xmath708 & @xmath960 & @xmath713 & @xmath713 & @xmath708 & @xmath211 + @xmath252 & @xmath961 & @xmath962 & @xmath959 & @xmath724 & @xmath211 & @xmath963 & @xmath708 & @xmath708 & @xmath708 & @xmath708 + @xmath909 & @xmath964 & @xmath957 & @xmath965 & @xmath706 & @xmath707 & @xmath966 & @xmath717 & @xmath708 & @xmath708 & @xmath708 + @xmath261 & @xmath967 & @xmath968 & @xmath969 & @xmath719 & @xmath707 & @xmath970 & @xmath211 & @xmath717 & @xmath708 & @xmath708 + @xmath269 & @xmath720 & @xmath971 & @xmath972 & @xmath969 & @xmath212 & @xmath973 & @xmath716 & @xmath700 & @xmath712 & @xmath717 + @xmath937 & @xmath969 & @xmath974 & @xmath971 & @xmath968 & @xmath969 & @xmath965 & @xmath959 & @xmath711 & @xmath716 & @xmath707 + @xmath821 & @xmath965 & @xmath975 & @xmath976 & @xmath977 & @xmath731 & @xmath978 & @xmath979 & @xmath980 & @xmath715 & @xmath710 + [ c]|c|c|c|c|c|c|c|c|c|c|c|@xmath672 & @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath295 & @xmath299 & @xmath354 & @xmath300 & @xmath213 & @xmath355 + @xmath774 & @xmath715 & @xmath212 & @xmath716 & @xmath211 & @xmath708 & @xmath958 & @xmath702 & @xmath702 & @xmath702 & @xmath713 + @xmath784 & @xmath959 & @xmath714 & @xmath710 & @xmath707 & @xmath701 & @xmath981 & @xmath713 & @xmath713 & @xmath713 & @xmath713 + @xmath252 & @xmath706 & @xmath957 & @xmath982 & @xmath710 & @xmath724 & @xmath983 & @xmath712 & @xmath701 & @xmath717 & @xmath708 + @xmath909 & @xmath724 & @xmath968 & @xmath979 & @xmath715 & @xmath706 & @xmath984 & @xmath707 & @xmath211 & @xmath701 & @xmath717 + @xmath261 & @xmath724 & @xmath972 & @xmath985 & @xmath980 & @xmath710 & @xmath986 & @xmath724 & @xmath700 & @xmath211 & @xmath701 + @xmath269 & @xmath707 & @xmath967 & @xmath987 & @xmath722 & @xmath969 & @xmath988 & @xmath715 & @xmath719 & @xmath716 & @xmath724 + @xmath937 & @xmath700 & @xmath989 & @xmath990 & @xmath987 & @xmath968 & @xmath957 & @xmath723 & @xmath714 & @xmath715 & @xmath719 + @xmath821 & @xmath716 & @xmath991 & @xmath992 & @xmath993 & @xmath977 & @xmath994 & @xmath972 & @xmath985 & @xmath718 & @xmath982 + [ c]|c|c|c|c|c|c|c|c|c|c|c|@xmath672 & @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath295 & @xmath299 & @xmath354 & @xmath300 & @xmath213 & @xmath355 + @xmath774 & @xmath211 & @xmath710 & @xmath959 & @xmath711 & @xmath707 & @xmath970 & @xmath211 & @xmath717 & @xmath708 & @xmath702 + @xmath784 & @xmath211 & @xmath719 & @xmath980 & @xmath715 & @xmath711 & @xmath995 & @xmath724 & @xmath712 & @xmath701 & @xmath717 + @xmath252 & @xmath712 & @xmath711 & @xmath985 & @xmath996 & @xmath969 & @xmath988 & @xmath715 & @xmath710 & @xmath706 & @xmath724 + @xmath909 & @xmath712 & @xmath706 & @xmath968 & @xmath722 & @xmath957 & @xmath997 & @xmath982 & @xmath715 & @xmath212 & @xmath706 + @xmath261 & @xmath707 & @xmath706 & @xmath722 & @xmath998 & @xmath999 & @xmath1000 & @xmath723 & @xmath980 & @xmath715 & @xmath710 + @xmath269 & @xmath707 & @xmath706 & @xmath731 & @xmath989 & @xmath964 & @xmath1001 & @xmath972 & @xmath985 & @xmath718 & @xmath969 + @xmath937 & @xmath724 & @xmath706 & @xmath1002 & @xmath1003 & @xmath1004 & @xmath1005 & @xmath967 & @xmath1002 & @xmath1006 & @xmath985 + @xmath821 & @xmath716 & @xmath706 & @xmath979 & @xmath1007 & @xmath992 & @xmath1008 & @xmath993 & @xmath1009 & @xmath977 & @xmath720 + [ c]|c|c|c|c|c|c|c|c|c|c|c|@xmath672 & @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath295 & @xmath299 & @xmath354 & @xmath300 & @xmath213 & @xmath355 + @xmath774 & @xmath211 & @xmath299 & @xmath299 & @xmath295 & @xmath287 & @xmath1010 & @xmath287 & @xmath279 & @xmath279 & @xmath279 + @xmath784 & @xmath211 & @xmath278 & @xmath278 & @xmath354 & @xmath299 & @xmath1011 & @xmath295 & @xmath295 & @xmath287 & @xmath287 + @xmath252 & @xmath712 & @xmath214 & @xmath1012 & @xmath848 & @xmath214 & @xmath1013 & @xmath357 & @xmath1014 & @xmath288 & @xmath300 + @xmath909 & @xmath712 & @xmath831 & @xmath1015 & @xmath1016 & @xmath350 & @xmath1017 & @xmath842 & @xmath848 & @xmath347 & @xmath214 + @xmath261 & @xmath707 & @xmath1018 & @xmath1019 & @xmath280 & @xmath292 & @xmath1020 & @xmath352 & @xmath1018 & @xmath1016 & @xmath1021 + @xmath269 & @xmath707 & @xmath352 & @xmath1022 & @xmath417 & @xmath1023 & @xmath1024 & @xmath1025 & @xmath1026 & @xmath855 & @xmath1027 + @xmath937 & @xmath724 & @xmath299 & @xmath1028 & @xmath1029 & @xmath1030 & @xmath1031 & @xmath1032 & @xmath1033 & @xmath1034 & @xmath1035 + @xmath821 & @xmath716 & @xmath287 & @xmath1036 & @xmath1037 & @xmath1038 & @xmath1020 & @xmath1039 & @xmath1040 & @xmath1041 & @xmath1042 + in this subsection , we present the theoretical analysis of the simulation scheme for general one - dimensional non - linear dynamics and provide rigorous bounds on performance . as for the linear case ,
the analysis is valid for @xmath3 without degradation as @xmath189 .
the analysis and the theoretical bound for the performance of this scheme are completely analogous to the linear problem , modulo the additional error coming from the linearization of the dynamics in the neighborhood of the stable equilibrium point . to distinguish between the linear and the nonlinear problem we need to introduce some notation . for a function
@xmath1043\times \mathbb{r})$ ] we define the operator @xmath1044(t , x)=w_{t}(t , x)+\bar{\mathbb{h}}(x , dw(t , x))+\frac{\varepsilon}{2}\sigma^{2}(x)d^{2}w(t , x),\quad \bar{\mathbb{h}}(x , p)=b(x)p-\frac{1}{2}\left\vert \sigma(x)p\right\vert ^{2}.\ ] ] in analogy to ( [ eq : ge_2 ] ) , for smooth functions @xmath503 , we define @xmath1045(t , x)=\bar{\mathcal{g}}^{\varepsilon } [ w](t , x)-\frac{1}{2}\left\vert \sigma(x)\left ( dw(t , x)-du(t , x)\right ) \right\vert ^{2}.\label{eq : ge_2nl}\ ] ] moreover , setting @xmath1046 and @xmath751 we recall that@xmath147(t , x)=w_{t}(t , x)+\mathbb{h}(x , dw(t , x))+\frac { \varepsilon}{2}\bar{\sigma}^{2}d^{2}w(t , x),\quad\mathbb{h}(x , p)=-cxp-\frac { 1}{2}\left\vert \bar{\sigma}p\right\vert ^{2}.\ ] ] the operators with bars correspond to the nonlinear problem , whereas the operators without bars give the corresponding first order approximations .
we define an operator measuring the error from the approximation by @xmath1047(t , x)=\bar{\mathcal{g}}^{\varepsilon}[w](t , x)-\mathcal{g}^{\varepsilon}[w](t , x ) \label{eq : errorofoperator}\ ] ] moreover , since @xmath1048 and @xmath1049 are @xmath1050 , we can write for any @xmath737 @xmath1051 where @xmath1052 and @xmath1053 are locally bounded . by assumption
we have that @xmath734 and @xmath1054 . as with the linear problem ,
the subsolution used for the analysis is based on the @xmath1055exponential mollification ( [ eq : second_moll ] ) reduced by the multiplicative factor @xmath177 .
we recall that this differs from the subsolution used for the design , which has @xmath1056 .
next we proceed with the mathematical analysis of the scheme .
the analysis is parallel to what was done for the linear problem , modulo adjustments due to the linearization of the dynamics in the neighborhood of the stable point , and therefore we mainly focus on the differences . in order to simplify the notation we assume without loss of generality ( as it was done in the linear problem ) that the stable equilibrium is @xmath1057 .
we write @xmath1058 , and for notational convenience assume that @xmath1059 .
as in the linear problem , @xmath1060 and @xmath1061 .
the following lemma bounds the error from the approximation .
[ l : boundforapproximationofoperatorg ] consider @xmath1062\times\lbrack0,z]$ ] .
then , for @xmath1063 @xmath1064(t , x)\right\vert & \leq2a^{m}(t)(z+\hat{x}e^{-c(t - t)})\sup_{x\in\lbrack0,z]}|b(x)+cx|\\ & \quad+\left [ 2\left ( a^{m}(t)(z+\hat{x}e^{-c(t - t)})\right ) ^{2}+\varepsilon a^{m}(t)\right ] \sup_{x\in\lbrack0,z]}|\sigma^{2}(x)-\bar{\sigma}^{2}|\\ & \leq c_{0}\left\ { a^{m}(t)(z+\hat{x}e^{-c(t - t)})z^{2}+\left ( a^{m}(t)(z+\hat{x}e^{-c(t - t)})\right ) ^{2}z+a^{m}(t)\varepsilon z\right\ } , \end{aligned}\ ] ] where @xmath1065}\left [ \frac{\left\vert r_{1}(x)\right\vert } { |x|^{2}}+\frac{\left\vert r_{2}(x)\right\vert } { |x|}\left\vert 2\bar{\sigma } + r_{2}(x)\right\vert \right ] .\ ] ] in addition @xmath1066}|\bar{f}_{1}(x)-f_{1}(x)|\leq c_{1}z^{3},\ ] ] where @xmath197 is a fixed constant .
since by assumption @xmath1067 and @xmath1054 , for @xmath1068\times\mathbb{r}$ ] @xmath1069(t , x ) & = \bar{\mathbb{h}}(x , dw(t , x))-\mathbb{h}(x , dw(t , x))+\frac{\varepsilon}{2}(\sigma^{2}(x)-\bar{\sigma}^{2})d^{2}w(t , x)\\ & = r_{1}(x)dw(t , x)-\frac{1}{2}\left ( \left\vert \sigma(x)dw(t , x)\right\vert ^{2}-\left\vert \bar{\sigma}dw(t , x)\right\vert ^{2}\right ) \\ & \quad+\frac{\varepsilon}{2}(\sigma^{2}(x)-\bar{\sigma}^{2})d^{2}w(t , x)\\ & = r_{1}(x)dw(t , x)-\frac{1}{2}r_{2}(x)\left ( 2\bar{\sigma}+r_{2}(x)\right ) \left\vert dw(t , x)\right\vert ^{2}\\ & \quad+\frac{\varepsilon}{2}r_{2}(x)\left ( 2\bar{\sigma}+r_{2}(x)\right ) d^{2}w(t , x).\end{aligned}\ ] ] for all @xmath1070 $ ] , we have @xmath1071 for a constant @xmath121 that depends on @xmath306 .
the bound follows by setting @xmath1072 and using @xmath1073 , and the proof of the first part of the lemma is concluded .
the proof of the second part goes as follows .
observe that@xmath1074 the last display and the bounds in ( [ eq : localizationbounds ] ) imply that @xmath1066}\left\vert \bar{f}_{1}(x)-f_{1}(x)\right\vert \leq c_{1}z^{3}\ ] ] which concludes the proof of the lemma . for notational convenience
we identify the quantity appearing in the upper bound for @xmath1075(t , x)\right\vert $ ] as given in lemma [ l : boundforapproximationofoperatorg]:@xmath1076 the following lemma shows that the error term induced by the local approximation of the dynamics in the neighborhood of the stable equilibrium point does not degrade as @xmath13 gets large .
[ l : integralofcorrectionterm ] we have that @xmath1077 where , letting @xmath1078 , @xmath1079
\nonumber\\ j_{2}(t^{\ast},t , m ) & = \frac{1}{\bar{\sigma}\sqrt{k}}\left ( 1-\frac{c}{\bar{\sigma}^{2}}\right ) \left [ \log\frac{1+\frac{\bar{\sigma}}{\sqrt{k}}e^{-ct^{\ast}}}{1-\frac{\bar{\sigma}}{\sqrt{k}}e^{-ct^{\ast}}}-\log \frac{1+\frac{\bar{\sigma}}{\sqrt{k}}e^{-ct}}{1-\frac{\bar{\sigma}}{\sqrt{k}}e^{-ct}}\right ] \nonumber\\ & + \frac{c}{2\bar{\sigma}^{4}}\left [ \frac{2\bar{\sigma}^{2}e^{-ct^{\ast}}}{k-\bar{\sigma}^{2}e^{-2ct^{\ast}}}-\frac{2\bar{\sigma}^{2}e^{-ct}}{k-\bar{\sigma}^{2}e^{-2ct}}\right ] \nonumber\\ j_{3}(t^{\ast},t , m ) & = \frac{c}{2\bar{\sigma}^{4}}\left [ \frac{\bar{\sigma } ^{2}}{k-\bar{\sigma}^{2}e^{-2ct^{\ast}}}-\frac{\bar{\sigma}^{2}}{k-\bar{\sigma}^{2}e^{-2ct}}\right ] \nonumber\\ j_{4}(t^{\ast},t , m ) & = \frac{1}{2\bar{\sigma}^{2}}\log\frac{k-\bar{\sigma } ^{2}e^{-2ct}}{k-\bar{\sigma}^{2}e^{-2ct^{\ast}}}.\nonumber\end{aligned}\ ] ] in particular , @xmath1080 .
the proof of this lemma follows by straightforward integration of @xmath1081 .
moreover , we note that @xmath1082 and that the definition of @xmath1060 implies @xmath1083 [ r : nonlinearremark ] _ the following three lemmas are analogous to lemmas [ l : region_0_y]-[l : region_y_hatx ] from the linear case .
the important difference between the nonlinear and the linear case is that the statements involve approximation errors and the statements hold if one confines the linearized dynamics to a small neighborhood of the equilibrium point as dictated by the sizes of @xmath372 and @xmath440 , or equivalently by @xmath372 and @xmath593 . due to the natural scaling of @xmath440 in terms of @xmath14
as indicated in subsection [ s : simulationresultslinearproblem ] , as @xmath14 gets smaller , @xmath593 will get smaller and be confined to a sufficiently small region that the statements of the lemmas are valid .
however , the lemmas below are stated for @xmath593 sufficiently small , without referencing to the natural scaling used in the simulation algorithm .
_ [ l : region_0_y_nonlinear ] assume that @xmath540\times\lbrack0,z]$ ] , @xmath541 and @xmath542 .
then , for sufficiently small @xmath593 we have up to an exponentially negligible term @xmath1084(t , x)\geq-(1-\eta ) c_{0 } r(\varepsilon,\hat{x},m , t).\ ] ] as in the proof of lemma [ l : region_0_y ] we need to show that in this region @xmath1085 is bounded by below away from zero .
we have @xmath1086 at @xmath101 , the value is minimized at @xmath546 and , as in lemma [ l : region_0_y ] , we obtain the lower bound @xmath1087>0 $ ] . for @xmath1088
we use the decomposition
@xmath1089 + \left [ \bar{f}_{1}(x)-f_{1}(x)\right ] .
\label{eq : nonlinearcasei}\ ] ] using the inequality @xmath1090 ^{4}\leq c / m\bar{\sigma}^{2}$ ] , the definition of @xmath593 and equation ( [ eq : f1-f2 ] ) , we obtain that for all @xmath540\times\lbrack0,z]$ ]
@xmath1091 -\frac{c}{k-\frac{c}{m}}\left ( ze^{-ct^{\ast}}-\hat { x}\right ) ^{2}\\ & \geq\frac{c}{\bar{\sigma}^{2}}\left [ \frac{\hat{x}^{2}}{m}\frac{c}{\bar{\sigma}^{2}+c / m}-z^{2}\right ] -z^{2}\frac{ce^{-2ct^{\ast}}}{\bar { \sigma}^{2}+c / m}\\ & \geq\frac{c}{\bar{\sigma}^{2}}z^{2}\left [ \frac{3\bar{\sigma}^{2}-c / m}{\bar{\sigma}^{2}+c / m}-\frac{c / m}{\bar{\sigma}^{2}+c / m}\right ] \\ & \geq\frac{c}{\bar{\sigma}^{2}}z^{2}\left [ \frac{2\bar{\sigma}^{2}+3c / m}{\bar{\sigma}^{2}+c / m}-\frac{c / m}{\bar{\sigma}^{2}+c / m}\right ] \\ & = 2\frac{c}{\bar{\sigma}^{2}}z^{2}.\end{aligned}\ ] ] hence , this term is of order @xmath1092 .
moreover , by lemma [ l : boundforapproximationofoperatorg ] , we also have that @xmath1066}|\bar{f}_{1}(x)-f_{1}(x)|\leq c_{1}z^{3}.\ ] ] thus , for sufficiently small @xmath593 the first term on the right hand side of ( [ eq : nonlinearcasei ] ) dominates the second term . hence the difference @xmath1085 is bounded from below away from zero if @xmath593 is sufficiently small , and the term involving the weight @xmath554 is exponentially negligible .
we next use that @xmath527 , @xmath555 and @xmath556 , lemma [ l : boundforapproximationofoperatorg ] , and also ( [ eq : errorofoperator ] ) and ( [ eq : defofr ] ) . these imply that up to an exponentially negligible term , @xmath1093(t , x ) & \geq(1-\eta)\bar{\mathcal{g}}^{\varepsilon}[f_{2}^{m}](t , x)\\ & = ( 1-\eta)\gamma_{2}^{m}(t)+(1-\eta)r^{\varepsilon}[f_{2}^{m}](t , x)\\ & \geq-(1-\mathbb{\eta})c_{0}r(\varepsilon,\hat{x},m , t),\end{aligned}\ ] ] which concludes the proof of the lemma .
[ l : region_hatx_a_nonlinear ] assume that @xmath1094\times\lbrack h , a]$ ] and assume @xmath541 .
define @xmath1095\cup[h , a]}\frac { -\varepsilon\sigma^{2}(x)d\left ( b(x)\sigma^{-2}(x)\right ) } { -\varepsilon \sigma^{2}(x)d\left ( b(x)\sigma^{-2}(x)\right ) + b^{2}(x)\sigma^{-2}(x)}\ ] ] let @xmath3 be sufficiently small such that @xmath1096 and consider @xmath1097 .
then , for sufficiently small @xmath593 and up to exponentially negligible terms , @xmath1098(t , x)\geq0.\ ] ] as with the linear problem we need to argue that there is a constant @xmath573 such that @xmath1099 for all @xmath1100\times\lbrack h , a]$ ] , which will then imply that the terms involving @xmath1101 are exponentially negligible .
we have the decomposition @xmath1102 + \left [ f_{1}(h)-\bar{f}_{1}(h)\right ] + \left [ \bar{f}_{1}(h)-\bar{f}_{1}(x)\right]\ ] ] focusing on @xmath1103 , lemma [ l : region_hatx_a ] guarantees that @xmath1104 > 0 $ ] uniformly in @xmath474 $ ] .
moreover , a straightforward computation shows that for @xmath650 sufficiently small , the lower bound of @xmath1105 $ ] is of the order @xmath1106 or equivalently @xmath1092 ( since @xmath1061 ) . by lemma [ l : boundforapproximationofoperatorg ] ,
the second term @xmath1107 $ ] is of order @xmath1108 .
thus , for sufficiently small @xmath650 , or equivalently for sufficiently small @xmath593 , the term @xmath1109 is bounded by below away from zero .
this statement , convexity of @xmath1110 and the fact that the third term of the last display is nonnegative for all @xmath535 $ ] , imply that there is a @xmath573 such that @xmath1111 for all @xmath540\times\lbrack h , a]$ ] .
hence , terms involving @xmath575 are exponentially negligible .
since @xmath527 , up to an exponentially negligible term@xmath1093(t , x ) & \geq(1-\eta)\rho_{1}(t , x)\bar{\mathcal{g}}^{\varepsilon}[\bar{f}_{1}](x)+\frac{1}{2}\left ( \eta-2\eta^{2}\right ) \left\vert \sigma(x)\rho _ { 1}(t , x)d\bar{f}_{1}(x)\right\vert ^{2}\\ & = \varepsilon(1-\eta)\sigma^{2}(x)d\left ( b(x)\sigma^{-2}(x)\right ) + 2\eta(1 - 2\eta)b^{2}(x)\sigma^{-2}(x).\end{aligned}\ ] ] thus for @xmath14 small enough that @xmath1112 we have , up to an exponentially negligible term@xmath1098(t , x)\geq0,\ ] ] concluding the proof of the lemma .
[ l : region_y_hatx_nonlinear ] assume that @xmath1094\times\lbrack z , h]$ ] and that @xmath864 .
set @xmath1113 and @xmath1114}\sigma^{2}(x)$ ] .
then , for sufficiently small @xmath593 we have up to an exponentially negligible term @xmath1093(t , x ) & \geq \frac{\sigma_{\ast}^{2}}{2}\left [ \frac{1}{\bar{\sigma}^{2}}\left ( \frac{c^{2}\eta}{2\bar{\sigma}^{2}}\left ( z-\hat{x}e^{c(t - t)}\right ) ^{2}-2\varepsilon c\right ) + \gamma(t , z , h,\hat{x},\varepsilon,\eta , t)\right ] \wedge0\\ & \qquad - c_{0}(1-\eta)r(\varepsilon,\hat{x},m , t).\end{aligned}\ ] ] where @xmath1115}\left [ \frac{c\eta}{2\bar{\sigma}^{2}}\left ( d\bar{f}_{1}(x)-df_{1}(x)\right ) \left ( z-\hat{x}e^{c(t - t)}\right ) \right .
\label{eq : defofgamma}\\ & \qquad\qquad\left .
+ \frac{1}{8}\eta\left ( d\bar{f}_{1}(x)-df_{1}(x)\right ) ^{2}+2\varepsilon\left ( \frac{c}{\bar{\sigma}^{2}}+d\left ( \frac{b(x)}{\sigma^{2}(x)}\right ) \right ) \right ] \nonumber\end{aligned}\ ] ] and @xmath736 for all @xmath737 . as with the analysis of the linear problem
, this region presents difficulties in its analysis , since the term @xmath1085 can be either positive or negative .
this means that both @xmath581 and @xmath554 may be important . as with the linear problem
, we ignore terms related to @xmath1116 due to exponential negligibility . up to an exponentially negligible term@xmath1093(t , x)\geq & ( 1-\eta)\frac{1}{4}\beta_{0}(t , x)+(1-\eta)\rho_{2}^{m,+}(t , x)\bar{\mathcal{g}}^{\varepsilon}[f_{2,+}^{m}](t , x)+(1-\eta)\rho_{1}(t , x)\bar{\mathcal{g}}^{\varepsilon}[\bar{f}_{1}](x)\\ & + \frac{1}{2}\left ( \eta-2\eta^{2}\right ) \sigma^{2}(x)\left\vert \rho _ { 2}^{m,+}(t , x)df_{2,+}^{m}(t , x)+\rho_{1}(t , x)d\bar{f}_{1}(x)\right\vert ^{2}.\end{aligned}\ ] ] we distinguish two cases depending on whether @xmath584 or @xmath1117 .
* case i : @xmath584 . *
we know that @xmath527 and due to the positivity of @xmath1118 we have by lemma [ l : boundforapproximationofoperatorg ] @xmath1119(t , x)=\mathcal{g}^{\varepsilon } [ f_{2,+}^{m}](t , x)+r^{\varepsilon}[f_{2,+}^{m}](t , x)\geq - c_{0}r(\varepsilon , \hat{x},m , t).\ ] ] next we need to argue that for @xmath540\times\lbrack z , h]$ ] @xmath1120 lemma [ l : region_y_hatx ] implies that @xmath1121 + \left [ df_{1}(x)-df_{2,+}^{m}(t , x)\right ] \nonumber\\ & \hspace{1cm}\leq\left [ d\bar{f}_{1}(x)-df_{1}(x)\right ] + \left [ \hat { x}\frac{2c}{\bar{\sigma}^{2}}\frac{k}{k-\bar{\sigma}^{2}e^{2c(t - t)}}\left ( \left ( \frac{2c}{m\bar{\sigma}^{2}}\right ) ^{2}-\frac{1}{2}\left ( \frac { c}{m\bar{\sigma}^{2}}\right ) ^{1/2}\right ) \right ] .
\label{eq : diff1}\ ] ] it follows easily from the proof of lemma [ l : boundforapproximationofoperatorg ] that @xmath1122}\left\vert d\bar{f}_{1}(x)-df_{1}(x)\right\vert \leq c_{1}h^{2}=100c_{1}z^{2}\ ] ] for some constant @xmath197 which is independent of @xmath593 . for the second term on the right hand side of ( [ eq : diff1 ] )
we have @xmath1123 where we used the definition @xmath1124 and the assumption @xmath1125 .
thus , this term is strictly negative and of order @xmath593 for @xmath593 sufficiently small .
thus for @xmath593 sufficiently small , the first term on the right hand side of ( [ eq : diff1 ] ) is dominated by the second term which is negative . together , with the non - positivity of @xmath1126 , we get that @xmath1127 for @xmath593 sufficiently small . since @xmath557
, we obtain @xmath1093(t , x ) & \geq(1-\eta)\rho_{1}(t , x)\bar{\mathcal{g}}^{\varepsilon}[\bar{f}_{1}](x)+\frac{1}{16}\eta\sigma^{2}(x)\left\vert d\bar{f}_{1}(x)-df_{2,+}^{m}(t , x)\right\vert ^{2}\\ & \quad-(1-\eta)c_{0}r(\varepsilon,\hat{x},m , t)\\ & \geq\frac{\sigma^{2}(x)}{2}\left [ \frac{1}{8}\eta\left ( d\bar{f}_{1}(x)-df_{2,+}^{m}(t , x)\right ) ^{2}+2\varepsilon d\left ( \frac { b(x)}{\sigma^{2}(x)}\right ) \right ] \\ & \quad-(1-\eta)c_{0}r(\varepsilon,\hat{x},m , t).\end{aligned}\ ] ] * case ii : @xmath585 . * similarly to the linear problem we have @xmath1098(t , x)\geq ( 1-\eta)\rho_{1}(t , x)\sigma^{2}(x)\left [ \frac{1}{8}|d\bar{f}_{1}(x)-df_{2,+}^{m}(t , x)|^{2}+\varepsilon d\left ( \frac{b(x)}{\sigma^{2}(x)}\right ) \right]\ ] ] writing @xmath1128 and using , as in the linear problem , the estimate @xmath1129 we obtain as in the proof of lemma [ l : region_y_hatx ] that @xmath1130(t , x)\\ & \qquad\geq\left\ { \frac{\sigma^{2}(x)}{2}\left [ \frac{1}{8}\eta\left ( d\bar{f}_{1}(x)-df_{1}(x)+\frac{2c}{\bar{\sigma}^{2}}\left ( z-\hat { x}e^{c(t - t)}\right ) \right ) ^{2}+2\varepsilon d\left ( \frac{b(x)}{\sigma^{2}(x)}\right ) \right ] \right\ } \wedge0\\ & \qquad\qquad - c_{0}(1-\eta)r(\varepsilon,\hat{x},m , t)\\ & \qquad\geq\frac{\sigma_{\ast}^{2}}{2}\left [ \frac{1}{\bar{\sigma}^{2}}\left ( \frac{c^{2}\eta}{2\bar{\sigma}^{2}}\left ( z-\hat{x}e^{c(t - t)}\right ) ^{2}-2\varepsilon c\right ) + \gamma(t , z , h,\hat{x},\varepsilon , \eta , t)\right ] \wedge0\\ & \qquad\qquad - c_{0}(1-\eta)r(\varepsilon,\hat{x},m , t).\end{aligned}\ ] ] where @xmath1131 was defined in ( [ eq : defofgamma ] ) .
this concludes the proof of the lemma .
the performance bound is then summarized in the following theorem .
the proof of theorem [ t : mainboundnonlinear ] is the same as the proof of theorem [ t : mainboundlinear ] for the linear case .
so , it will not be repeated here .
[ t : mainboundnonlinear ] assume @xmath421 , @xmath1132 , @xmath638 and that @xmath864 , where @xmath1133 is as in lemma [ l : region_hatx_a_nonlinear ] .
set @xmath1134}\sigma^{2}(x)$ ] .
let @xmath130 be the control based on the function @xmath603 defined via ( [ eq : nonlinearchangeofmeasure ] ) , i.e. , @xmath1135 .
then , up to an exponentially negligible term , for @xmath1136 such that @xmath1137 and for @xmath593 sufficiently small , we have@xmath1138 1_{\left\ { t\geq t^{\ast}\right\ } } \\ & \hspace{0.1cm}+2i_{2}(\varepsilon , t)1_{\left\ { t < t^{\ast}\right\ } } , \end{aligned}\ ] ] where @xmath1139 ds - t^{\ast}c^{\ast}\varepsilon,\end{aligned}\ ] ] where @xmath636 are the times in @xmath637 $ ] where the integrand is negative , @xmath1140 is as in ( [ eq : defofgamma ] ) , @xmath1141 and @xmath1142}\sigma^{2}(x)\left\vert d(b(x)/\sigma^{2}(x))\right\vert > 0.\ ] ] the bound of theorem [ t : mainboundnonlinear ] takes a complicated form , but as in the linear case , the performance does not degrade as @xmath1143 .
this was also reflected by the simulation data in subsection [ ss : simulationnonlinearproblem ] .
let us now justify this claim .
notice that , by lemma [ l : integralofcorrectionterm ] the term @xmath1144 is finite , uniformly in @xmath13 .
next we need to argue , similarly to the linear problem , that when @xmath639 is sufficiently large and @xmath593 is sufficiently small , the integrand of the second term in the definition of @xmath1145 is in fact positive .
let us denote the integrand of the second term by @xmath1146 where @xmath1147 .
the term @xmath1148 is composed by three terms and we shall argue below they are dominated ( even when they are negative ) , by the second term in the definition @xmath1149 , i.e. , by @xmath1150 when @xmath593 is small enough .
this means , as in the case of the linear problem , that when the integral will be finite uniformly in @xmath13 .
let us now support the claim just made .
it is easy to see that for @xmath1151 $ ] , the first term in the definition of @xmath1152 can be either positive or negative , but it is of order @xmath1153 .
the second term in the definition of @xmath1152 is positive and for @xmath1151 $ ] , it is of order @xmath1154 .
lastly , the third term in the definition of @xmath1152 , may be positive or negative , but in either cas !
e , it will be of order @xmath1155 for @xmath1151 $ ] .
therefore , for @xmath593 sufficiently small , @xmath1152 is dominated by the second term in the definition @xmath1156 , i.e. , by @xmath1157 .
hence , the argument that was used for the linear problem in order to show that the integrand of the second term in the definition of @xmath1145 is in fact positive when @xmath639 is large enough , allows us to reach to the same conclusion here as well , given that @xmath593 is chosen sufficiently small .
in this appendix we provide proofs of some auxiliary lemmas used in the main body of the manuscript .
[ proof of lemma [ lem : mollification]]without loss of generality , we can restrict attention to @xmath1158 .
we have @xmath1159@xmath1160 and@xmath1161 \\ & \quad-\rho_{2}(t , x;\delta)\left [ \frac{1}{\delta}d\tilde{u}_{2}(t , x)^{2}-d^{2}\tilde{u}_{2}(t , x)\right ] .\end{aligned}\ ] ] omitting function arguments for notational convenience , for @xmath1162@xmath1163 + \frac{\varepsilon}{2}\alpha d^{2}u^{\delta}\\ & = \rho_{1}\partial_{t}\tilde{u}_{1}+\rho_{2}\partial_{t}\tilde{u}_{2}+\rho_{1}d\tilde{u}_{1}b+\rho_{2}d\tilde{u}_{2}b-\frac{1}{2}\left\vert \sigma\left ( \rho_{1}d\tilde{u}_{1}+\rho_{2}d\tilde{u}_{2}\right ) \right\vert ^{2}\\ & \quad+\frac{\varepsilon}{2}\frac{1}{\delta}\left\vert \sigma\left ( \rho_{1}d\tilde{u}_{1}+\rho_{2}d\tilde{u}_{2}\right ) \right\vert ^{2}-\frac{\varepsilon}{2}\frac{1}{\delta}\left [ \rho_{1}\alpha(d\tilde{u}_{1})^{2}+\rho_{2}\alpha(d\tilde{u}_{2})^{2}\right ] \\ & \quad+\frac{\varepsilon}{2}\left [ \rho_{1}\alpha d^{2}\tilde{u}_{1}+\rho_{2}\alpha d^{2}\tilde{u}_{2}\right ] \\ & = \rho_{1}\left [ \partial_{t}\tilde{u}_{1}+d\tilde{u}_{1}b-\frac{1}{2}\left\vert \sigma d\tilde{u}_{1}\right\vert ^{2}+\frac{\varepsilon}{2}\alpha d^{2}\tilde{u}_{1}\right ] \\ & \quad+\rho_{2}\left [ \partial_{t}\tilde{u}_{2}+d\tilde{u}_{2}b-\frac{1}{2}\left\vert \sigma d\tilde{u}_{2}\right\vert ^{2}+\frac{\varepsilon}{2}\alpha d^{2}\tilde{u}_{2}\right ] \\ & \quad+\frac{1}{2}\left ( 1-\frac{\varepsilon}{\delta}\right ) \left [ \rho_{1}\left\vert \sigma d\tilde{u}_{1}\right\vert ^{2}+\rho_{2}\left\vert \sigma d\tilde{u}_{2}\right\vert ^{2}-\left\vert \sigma\left ( \rho_{1}d\tilde{u}_{1}+\rho_{2}d\tilde{u}_{2}\right ) \right\vert ^{2}\right ] \\ & \geq\frac{1}{2}\left ( 1-\frac{\varepsilon}{\delta}\right ) \left [ \rho_{1}\left\vert \sigma d\tilde{u}_{1}\right\vert ^{2}+\rho_{2}\left\vert \sigma d\tilde{u}_{2}\right\vert ^{2}-\left\vert \rho_{1}\sigma d\tilde{u}_{1}+\rho_{2}\sigma d\tilde{u}_{2}\right\vert ^{2}\right ] + \rho_{1}\gamma _ { 1}+\rho_{2}\gamma_{2}\\ & \geq\rho_{1}\gamma_{1}+\rho_{2}\gamma_{2},\end{aligned}\ ] ] where the last line is due to the convexity of @xmath1164 .
[ l : generalbound ] let @xmath1165 and @xmath1166 be two continuously differentiable functions from @xmath19\times\mathbb{r}\mapsto\mathbb{r}$ ] .
assume that @xmath20 and @xmath21 are lipschitz continuous .
set @xmath1167 , @xmath1168 and let @xmath1169 solve @xmath1170ds\right ] , \quad\hat{x}^{\varepsilon}(0)=y.\ ] ] then for every @xmath3 , @xmath1168 and stopping time @xmath1171 , we have , with probability @xmath210 , @xmath1172 ds\\ & \quad\geq2w(0,y)-2w(\hat{\tau}^{\varepsilon},\hat{x}^{\varepsilon}(\hat{\tau}^{\varepsilon}))+2\sqrt{\varepsilon}\int_{0}^{\hat{\tau } ^{\varepsilon}}dw(s,\hat{x}^{\varepsilon}(s))\sigma(\hat{x}^{\varepsilon } ( s))db(s)\\ & \quad+2\int_{0}^{\hat{\tau}^{\varepsilon}}\mathcal{g}^{\varepsilon } [ w](s,\hat{x}^{\varepsilon}(s))ds-\int_{0}^{\hat{\tau}^{\varepsilon}}\left\vert \sigma(\hat{x}^{\varepsilon}(s))\left ( dw(s,\hat{x}^{\varepsilon } ( s))-du(s,\hat{x}^{\varepsilon}(s))\right ) \right\vert ^{2}ds\end{aligned}\ ] ] assume we use the control @xmath1174 for the design of the scheme and choose @xmath1175
. then @xmath1176 \\ & \quad = dw(t , x)(b(x)+\sigma^{2}(x)du(t , x)-2\sigma^{2}(x)dw(t , x)\\ & \qquad-\frac{1}{2}\left\vert \sigma(x)du(t , x)\right\vert ^{2}+\left\vert \sigma(x)dw(t , x)\right\vert ^{2}\\ & \quad = dw(t , x)b(x)-\frac{1}{2}\left\vert \sigma(x)dw(t , x)\right\vert ^{2}-\frac{1}{2}\left\vert \sigma(x)\left
( dw(t , x)-du(t , x)\right ) \right\vert ^{2}\\ & \quad=\mathbb{h}(x , dw(t , x))-\frac{1}{2}\left\vert \sigma(x)\left ( dw(t , x)-du(t , x)\right ) \right\vert ^{2}.\end{aligned}\ ] ] applying it s formula to @xmath1177 then gives @xmath1178 \right ] ds\nonumber\\ & \quad+\int_{0}^{\hat{\tau}^{\varepsilon}}\frac{\varepsilon}{2}\sigma ^{2}(\hat{x}^{\varepsilon}(s))d^{2}w(s,\hat{x}^{\varepsilon}(s))ds+\int _
{ 0}^{\hat{\tau}^{\varepsilon}}\sqrt{\varepsilon}dw(s,\hat{x}^{\varepsilon } ( s))\sigma(\hat{x}^{\varepsilon}(s))db(s)\nonumber\\ & \geq\int_{0}^{\hat{\tau}^{\varepsilon}}\left [ \frac{1}{2}\bar{u}(s,\hat { x}^{\varepsilon}(s))^{2}-\frac{1}{4}v(s)^{2}\right ] ds+\sqrt{\varepsilon } \int_{0}^{\hat{\tau}^{\varepsilon}}dw(\hat{x}^{\varepsilon}(s))\sigma(\hat { x}^{\varepsilon}(s))db(s)\nonumber\\ & \quad+\int_{0}^{\hat{\tau}^{\varepsilon}}\left [ \partial_{s}w(s,\hat { x}^{\varepsilon}(s))+\mathbb{h}(\hat{x}^{\varepsilon}(s),dw(s,\hat { x}^{\varepsilon}(s)))+\frac{\varepsilon}{2}\sigma^{2}(\hat{x}^{\varepsilon } ( s))d^{2}w(s,\hat{x}^{\varepsilon}(s))\right ] ds\nonumber\\ & \quad-\frac{1}{2}\int_{0}^{\hat{\tau}^{\varepsilon}}\left\vert \sigma ( \hat{x}^{\varepsilon}(s))\left ( dw(s,\hat{x}^{\varepsilon}(s))-du(s,\hat { x}^{\varepsilon}(s))\right ) \right\vert ^{2}ds\nonumber\\ & = \int_{0}^{\hat{\tau}^{\varepsilon}}\left [ \frac{1}{2}\bar{u}(\hat { x}^{\varepsilon}(s))^{2}-\frac{1}{4}v(s)^{2}\right ] ds+\sqrt{\varepsilon } \int_{0}^{\hat{\tau}^{\varepsilon}}dw(s,\hat{x}^{\varepsilon}(s))\sigma ( \hat{x}^{\varepsilon}(s))db(s)\nonumber\\ & \quad+\int_{0}^{\hat{\tau}^{\varepsilon}}\mathcal{g}^{\varepsilon } [ w](s,\hat{x}^{\varepsilon}(s))ds-\frac{1}{2}\int_{0}^{\hat{\tau } ^{\varepsilon}}\left\vert \sigma(\hat{x}^{\varepsilon}(s))\left ( dw(s,\hat{x}^{\varepsilon}(s))-du(s,\hat{x}^{\varepsilon}(s))\right ) \right\vert ^{2}ds\nonumber\end{aligned}\ ] ] | we discuss importance sampling schemes for the estimation of finite time exit probabilities of small noise diffusions that involve escape from an equilibrium . a factor that complicates
the analysis is that rest points are included in the domain of interest .
we build importance sampling schemes with provably good performance both pre - asymptotically , i.e. , for fixed size of the noise , and asymptotically , i.e. , as the size of the noise goes to zero , and that do not degrade as the time horizon gets large .
simulation studies demonstrate the theoretical results . |
advances in thin film epitaxy growth techniques have made it possible to induce emergent electronic @xcite , magnetic @xcite and orbital @xcite states , which are not naturally occurring in bulk constituents , at atomically sharp transition metal oxide interfaces @xcite .
for example , the interface between the two nonmagnetic band insulators laalo@xmath1 and srtio@xmath1 @xcite has been reported to exhibit both conductance @xcite and magnetism @xcite ( see reviews @xcite and references therein ) . at the interface of mott insulators srmno@xmath1 and lamno@xmath1 , hole doping on the mn sites leads to rich phenomena , including metal - insulator transition , charge / spin / orbital ordering and magnetoresistance @xcite . in laalo@xmath1/srtio@xmath1 and related heterostructures , the interface electron gas is believed to be produced by the polar catastrophe mechanism , which leads to the transfer of charge from the sample surface to the interface . here , we consider a different mechanism for controlling the electronic properties of an interface : namely , electronegativity - driven charge transfer .
recently , we have shown that internal charge transfer in a latio@xmath1/lanio@xmath1 superlattice transforms metallic lanio@xmath1 into a @xmath2 mott insulator and mott insulating latio@xmath1 into a @xmath3 band insulator @xcite .
a natural question arises : can we reverse the process and utilize internal charge transfer to induce conductance via oxide interfaces ? in this regard , it is very tempting to explore mott interfaces ( one or both constituents are mott insulators ) due to the unusual phenomena ( colossal magnetoresistance and high temperature superconductivity ) exhibited in certain doped mott insulators .
in this paper we use density functional theory + dynamical mean field theory ( dft+dmft ) to theoretically design a superlattice with emergent metallic behavior .
we explore two different types of structure : the perovskite structure ( referred to as 113-type ) and the @xmath4 ruddlesden - popper structure ( referred to as 214-type ) . among the four bulk constituents ( srvo@xmath1 , srmno@xmath1 , sr@xmath0vo@xmath5 and sr@xmath0mno@xmath5 ) , all are correlation - driven insulators except srvo@xmath1 , which is a moderately correlated metal @xcite .
we show that the difference of electronegativity between the elements v and mn drives internal charge transfer from v to mn sites , leading to a `` self - doping '' at the interface and possibly inducing conductance as mn sites become weakly electron doped and v sites hole doped .
the rest of the paper is organized as follows : section ii presents the theoretical methods .
a schematic of band alignment is presented in section iii to illustrate the underlying mechanism of charge transfer .
all the bulk results from _ ab initio _ calculations are in section iv and the results of vanadate - manganite superlattices are in section v , both of which provide qualitative support and quantitative corrections to the schematic .
the conclusions are in section vi .
five appendices present technical details relating to the insulating gaps of sr@xmath0vo@xmath5 and sr@xmath0mno@xmath5 , alternative forms of the double counting correction , lda spectra of the superlattices and the possibility of two consecutive repeating layers ( i.e. 2/2 superlattices instead of 1/1 superlattices ) .
the dft @xcite component of our dft+dmft @xcite calculations is performed using a plane - wave basis @xcite , as implemented in the vienna ab - initio simulation package ( vasp ) @xcite using the projector augmented wave ( paw ) approach @xcite . both local density approximation ( lda ) @xcite and perdew - burke - ernzerhof generalied gradient approximation ( gga - pbe )
@xcite are employed .
the correlated subspace and the orbitals with which it mixes are constructed using maximally localized wannier functions @xcite defined over the full 10 ev range spanned by the @xmath6-@xmath7 band complex , resulting in an well localized set of @xmath7-like orbitals @xcite . to find the stationary solution for our dft+dmft functional , we first find the self - consistent charge density within dft .
subsequently we fix the charge density and converge the dmft equations .
a full charge self - consistency is not implemented in the present work . however
, this approximation procedure is found to yield reasonable results in calculations of bulk systems @xcite .
for the bulk materials , we consider two structures : the experimental one and the theoretical relaxed structure obtained by the use of dft . for the superlattice
, we use dft to obtain the relaxed structure .
we compare lda and gga calculations and both exchange correlation functions yield consistent results .
the simulation cell is illustrated in fig .
[ fig : structure ] .
the stacking direction of the superlattice is along [ 001 ] .
we use an energy cutoff 600 ev . a @xmath8 $ ] ( @xmath9 and @xmath10 are the lattice constants along the @xmath11 and @xmath12 directions , and [ @xmath11 ]
is the integer part of @xmath11 ) monkhorst - pack grid is used to sample the brillouin zone .
both cell and internal coordinates are fully relaxed until each force component is smaller than 10 mev / and the stress tensor is smaller than 10 kbar .
convergence of the key results are tested with a higher energy cutoff ( 800 ev ) and a denser @xmath14-point sampling @xmath15 $ ] and no significant changes are found . for the vanadates , we treat the empty @xmath16 orbitals with a static hartree - fock approximation ( recent work shows this approximation is adequate to describe the electronic structure of vanadates @xcite ) , while correlations in the v @xmath17 manifold are treated within single - site dmft including the slater - kanamori interactions using intra - orbital hubbard @xmath18 = 5 ev and @xmath19 = 0.65 ev @xcite . for manganites
, we treat the correlations on all the five mn @xmath7 orbitals within single - site dmft using the slater - kanamori interactions with intra - orbital hubbard @xmath20 = 5 ev and @xmath21 = 1 ev @xcite .
the dmft impurity problem is solved using the continuous time quantum monte carlo method @xcite . in order to use the `` segment '' algorithm @xcite
, we neglect the exchange and pairing terms in the slater - kanamori hamiltonian .
all the calculations are paramagnetic and the temperature is set to 232 k. long - range magnetic ordering ( in particular antiferromagnetism ) might be induced at low temperature on mn sites in manganites and in superlattices .
a thorough study of magnetic properties will be presented elsewhere @xcite . for the superlattice
, we solve the problem in the single - site dmft approximation , meaning that the self energy is site local and is one function on the v site and a different one on the mn site .
the self energies are determined from two quantum impurity models , which are solved independently but coupled at the level of the self consistency condition .
an important outstanding issue in the dft+dmft procedure is the `` double counting correction '' which accounts for the part of the slater - kanamori interactions already included in the underlying dft calculation and plays an important role by setting the mean energy difference between the @xmath7 and @xmath6 bands .
the @xmath6-@xmath7 separation plays a crucial role in determining the band alignment , which affects the charge transfer . however , currently there is no exact procedure for the double counting correction .
we use the @xmath22 double counting method recently introduced @xcite , where the parameter @xmath22 is the prefactor in the double - counting which determines the @xmath6-@xmath7 separation and equivalently the number of electrons in the @xmath7-manifold . in this study , @xmath22 is chosen to produce an energy separation between the o @xmath6 and transition metal @xmath7 bands which is consistent with photoemission experiments .
our main qualitative conclusions do not depend on the details of the double counting scheme ; in particular we show here they hold also for the conventional fully localized limit ( fll ) double counting @xcite which is the @xmath23 limit of the method of ref .
the reason behind that is because in the superlattice , it is the relative v@xmath7-mn@xmath7 energy separation that controls the charge transfer .
the fll double counting formula underestimates the @xmath6-@xmath7 separation in both srvo@xmath1 and srmno@xmath1 by about 1 ev .
however , such an error is cancelled in the calculation of v@xmath7-mn@xmath7 energy separation .
therefore the fll double counting does not change the charge transfer picture .
the spectral function presented throughout this work is defined as follows : @xmath24^{-1}\right)_{ii}\ ] ] where @xmath25 is the label of a wannier function , @xmath26 is the number of @xmath14-points , @xmath27 is an identity matrix , @xmath28 is the dft - lda band hamiltonian in the matrix form using the wannier basis .
@xmath29 is the total self - energy and is understood as a diagonal matrix only with nonzero entries on the correlated orbitals .
local tetragonal point symmetry of the v and mn sites ensures that @xmath30 is diagonal within the correlated orbital subspace .
@xmath31 is the chemical potential .
@xmath32 is the double counting potential , which is defined as @xcite : @xmath33 note that if @xmath23 , then we restore the standard fll double counting formula @xcite . for clarity
, all the spectra functions presented in this paper are obtained from lda+dmft calculations .
gga+dmft calculations yield qualitatively consistent results .
we consider the following materials as components of the superlattice : srvo@xmath1 , a moderately correlated metal with nominal @xmath7-valence @xmath34 ; sr@xmath0vo@xmath5 , a correlation - driven insulator also with nominal valence @xmath34 ; and srmno@xmath1 and sr@xmath0mno@xmath5 , both of which are @xmath35 correlation - driven ( mott ) insulators .
[ fig : structure ] shows the atomic structure of the bulk phases of the constituent materials and the corresponding superlattices .
[ fig : structure]*a * , * b * , and * c * are bulk srvo@xmath1 , bulk srmno@xmath1 , and srvo@xmath1/srmno@xmath1 superlattice , respectively .
[ fig : structure]*d * , * e * , and * f * are bulk sr@xmath0vo@xmath5 , bulk sr@xmath0mno@xmath5 , and sr@xmath0vo@xmath5/sr@xmath0mno@xmath5 superlattice , respectively . in both superlattices ,
the stacking direction is along the [ 001 ] axis . in the 214-type ,
the v atoms are shifted by a @xmath36 lattice constant in the @xmath37 plane relative to the mn atoms . fig . [
fig : band ] is a schematic of the band structure of bulk vanadates , bulk manganites , and the band alignments in the superlattice ( the small insulating gap of sr@xmath0vo@xmath5 is not relevant here .
there is a large energy separation ( around 2 ev ) between v @xmath7 and o @xmath6 states ( see fig .
[ fig : band]*a * ) . in the mn - based materials ( see fig . [
fig : band]*b * ) , the highest occupied states are mn @xmath17-derived and the lowest unoccupied states are mn @xmath16-derived . due to the electronegativity difference between v and mn , visible as the difference in the energy separation of the transition metal @xmath7 levels from the oxygen @xmath6 levels , if we align the o @xmath6 states between vanadates and manganites ( see fig . [
fig : band]*c * ) , the occupied v @xmath17 states overlap in energy with the unoccupied mn @xmath16 states .
the overlap drives electrons from v sites to mn sites .
as the superlattice is formed , a common fermi level appears across the interface and thus we expect that mn @xmath16 states become electron doped and v @xmath17 states hole doped .
we make two additional points : i ) though srvo@xmath1 is a metal and sr@xmath0vo@xmath5 is an insulator with a small energy gap ( around 0.2 ev ) @xcite , the near fermi level electronic structure does not affect the band alignment and therefore the internal charge transfer is expected to occur no matter whether there is a small energy gap in v @xmath17 states at the fermi level or not ; ii ) in our schematic , we assume that the main peak of o @xmath6 states are exactly aligned between the vanadates and manganites in the superlattices . of course , real material effects will spoil any exact alignment .
we will use _ ab initio _ calculations to provide quantitative information on how o @xmath6 states are aligned between the two materials .
.[tab : structure ] the in - plane and out - of - plane v - o and mn - o bond lengths @xmath38 of srvo@xmath1 , srmno@xmath1 , sr@xmath0vo@xmath5 and sr@xmath0mno@xmath5 .
the corresponding vo@xmath39 and mno@xmath39 octahedral volumes @xmath40 are also calculated .
the relaxed structures are obtained from dft - lda and dft - gga non - spin - polarized calculations .
the experimental values which are referenced in the main text are also provided for comparison . [
cols="^,^,^,^,^,^,^,^,^ " , ] | we perform density functional theory plus dynamical mean field theory calculations to investigate internal charge transfer in an artificial superlattice composed of alternating layers of vanadate and manganite perovskite and ruddlesden - popper structure materials .
we show that the electronegativity difference between vanadium and manganese causes moderate charge transfer from vo@xmath0 to mno@xmath0 layers in both perovskite and ruddlesden - popper based superlattices , leading to hole doping of the vo@xmath0 layer and electron doping of the mno@xmath0 layer .
comparison of the perovskite and ruddlesden - popper based heterostructures provides insights into the role of the apical oxygen .
our first principles simulations demonstrate that the combination of internal charge transfer and quantum confinement provided by heterostructuring is a powerful approach to engineering electronic structure and tailoring correlation effects in transition metal oxides . |
the metal - insulator transitions of the early transition - metal oxides have been attracting a lot of attention for decades . in particular , since the first report by morin , @xcite much interest has centered about the cornerstone materials @xmath3 and @xmath1 .
this is mainly due to the fact , that these compounds display first order phase transitions with very narrow hystereses of only few kelvin and strong changes of the conductivity of several orders of magnitude , making them promising candidates for applications .
interestingly , the transitions of both compounds are accompanied by characteristic structural changes , which initiated a discussion about the driving forces . dispute centered about the question whether the transitions are driven predominantly by structural instabilities of the parent corundum and rutile structure , respectively , or else by strong electronic correlations . nowadays , the importance of the latter is widely accepted .
however , despite intense work , the questions concerning the origins of the respective transitions have not yet been satisfactorily answered . in @xmath1 ,
the metal - insulator transition occurs at 340k and is connected with a structural change from the rutile structure to a distorted monoclinic structure .
the latter is characterized by ( i ) pairing of the metal atoms within chains parallel to the rutile @xmath4-axis and ( ii ) their lateral antiferroelectric zigzag - like displacement .
@xcite electronic states near the fermi energy are mainly of v @xmath5 @xmath6 character .
they fall into the @xmath7 band , which mediates v - v overlap along the metal chains , and the remaining , @xmath8 bands .
@xcite at the transition , splitting of the @xmath7 band and upshift of the @xmath8 bands due to increased metal - oxygen overlap produce a finite band gap .
controversial discussions addressed the origin of the @xmath7 splitting , which was assigned either to the metal - metal dimerization or to increased electronic correlations resulting from the reduced screening by the @xmath8 electrons .
@xcite state of the art electronic structure calculations gave strong hints at a structural instability but were not able to reproduce the insulating gap due to the shortcomings of the local density approximation ( lda ) .
@xcite only recently , lda+dmft calculations have demonstrated that the metal - insulator transition may be regarded as a correlation - assisted peierls - like transition .
@xcite remarkably , the structural distortions occurring at the transition of @xmath1 are characteristic of most of the neighbouring @xmath9 , @xmath10 , and @xmath11 transition - metal dioxides .
in particular , the metallic oxides @xmath0 , @xmath12 , @xmath13 , and @xmath14-@xmath15 all display the above - mentioned metal - metal pairing as well as the lateral displacement and crystallize in the same monoclinic structure as low - temperature @xmath1 .
@xcite this crystal structure is displayed in fig .
[ introfig1 ] , .
the dimerized metal pairs are highlighted by thick bonds . ] which allows to identify both the pairing and the lateral displacement of the metal atoms .
in contrast to the above - mentioned metallic @xmath10 and @xmath11 members , the @xmath9 compound @xmath2 , like @xmath1 , undergoes a metal - insulator transition ( at 1081 k ) and a simultaneous structural transition from rutile to a distorted variant having a body - centered tetragonal ( bct ) lattice .
@xcite however , despite the differences in long - range order , the local deviations from the rutile structure are the same as in @xmath1 , i.e. the niobium atoms dimerize and experience lateral displacements at the transition
. @xcite despite of the strong interest in the metal - insulator transition of @xmath1 , attempts at understanding the origin of the common structural characteristics are very rare .
indeed , any theory aiming at the metal - insulator transition of @xmath1 would be incomplete without proper account of the local structural distortions of all the rutile - related transition - metal dioxides . in order to close the apparent gap ,
we have started a systematic investigation especially on @xmath1 , @xmath2 , and @xmath0 already a decade ago , which by now provides a unified picture .
@xcite interest in @xmath0 was especially motivated by the metallic nature of this material , the reduced localization of the @xmath16 electrons as compared to the @xmath5 electrons of @xmath1 , and the rather small effective mass , which is similar to the free electron mass .
@xcite these facts point at a rather weak influence of electronic correlations . indeed ,
electronic structure calculations for @xmath0 show a large bonding - antibonding splitting of the @xmath17 bands .
@xcite since the occupied @xmath17 bands lie far below the fermi level , a possible influence of electronic correlations in these bands would not lead to an enhanced effective mass or susceptibility . as a consequence
, @xmath0 can be regarded as a `` model system '' for studying the origin of the structural characteristics of the early transition - metal dioxides .
in addition , an extensive examination of the electronic structure of @xmath0 could help clarifying open questions of the above - mentioned structurally related systems , in particular of the @xmath9 members @xmath2 and @xmath1 . in a previous work ,
we have presented a comprehensive theoretical study of the electronic properties of @xmath0 .
@xcite electronic structure calculations using the asw method , which is based on density functional theory and the local density approximation ( lda ) , were used to analyze the electronic states involved in low - energy excitations .
in addition , emphasis was put on investigating the origin of the small but distinct deviations of the monoclinic crystal structure from the parent rutile structure , which for this material mainly consist of a metal - metal dimerization parallel to the rutile @xmath18-axis . in short , we were able to show that the monoclinic structure results from what we called an embedded peierls - like instability .
this instability affects the quasi one - dimensional mo @xmath19 bands , which mediate the metal - metal overlap along chains parallel to the rutile @xmath20-axis and which are embedded in a background of the remaining three - dimensionally dispersing mo @xmath16 @xmath8 , electrons , which are also designated as the @xmath21 electrons .
the validity of this scenario had been demonstrated by comparing the electronic properties calculated for the monoclinic structure with those of a properly symmetrized artificial rutile structure .
in addition , we found nearly perfect agreement of our calculated results with the outcome of ultraviolet photoemission spectroscopy and x - ray absorption spectroscopy apart from an underestimation of the @xmath7 band splitting by about 1ev due to the well known shortcomings of the lda .
@xcite this scenario has also been shown to explain the metal - insulator transitions of the @xmath9 members @xmath1 and @xmath2 , which likewise are subject to peierls - like instabilities of the quasi one - dimensional v @xmath22 bands in the embedding backgrounds of the @xmath21 electrons .
@xcite as in @xmath0 , these instabilities cause distortions of the rutile structure by the characteristic metal - metal dimerization
. due to the particular position of the fermi energy in the @xmath9 compounds , the peierls - like instabilities are then accompanied by the observed metal - insulator transitions . at the same time , the opening of the optical band gaps is supported by the lateral antiferroelectric displacements of the metal atoms , which cause an upshift of the @xmath21 states . continuing our study of @xmath0 ,
we concentrate in the present work especially on the fermi surface and the band dispersions . for the first time , a comprehensive investigation of the full fermi body of @xmath0 , as determined both experimentally and theoretically , will be presented .
in particular , we turn to a comparison of the theoretical results with angle - resolved photoemission spectroscopy ( arpes ) and de haas - van alphen data ( dhva ) . in doing so , we apply a new implementation of the augmented spherical wave method , which goes beyond the version used in our previous work by taking into account the full potential rather than the approximate muffin - tin potential .
the resulting changes of the electronic states are small but important and lead to a much improved agreement of the theoretical and experimental data . while published photoemission studies on @xmath0 focused on the angle - integrated analysis of the occupied bands , @xcite a systematic investigation of the band dispersions and the fermi body by angle - resolved photoemission spectroscopy seems to be still missing .
in addition , existing magneto - transport measurements addressed only fractions of the fermi surface .
@xcite it is the purpose of the present work to make up for the apparent lack in experimental data and to present a comprehensive study of the electronic structure of @xmath0 .
an important difference between dhva and arpes is that the former requires very low temperatures whereas the latter does not .
this is crucial for the early transition - metal dioxides , some of which have a metal - insulator transition and thus do not show a fermi surface at low temperatures . in the present case
we have the opportunity to compare arpes and dhva because @xmath0 is metallic at low @xmath23 .
it turns out that the arpes and dhva results agree well , implying that arpes will be useful to study the fermi surfaces of the @xmath9 members @xmath1 and @xmath2 as well .
the paper is organized as follows : while sec .
[ theory ] gives an overview on the computational method and the theoretical results , sec .
[ experiment ] turns to a description of the arpes and dhva measurements and their results .
the theoretical and experimental findings are subject of a comparative discussion given in sec . [ discussion ] and sec .
[ summary ] summarizes our results .
the calculations are based on density - functional theory and the local density approximation .
@xcite they were performed using the scalar - relativistic implementation of the augmented spherical wave method ( see refs . and references therein ) . in the asw method ,
the wave function is expanded in atom - centered augmented spherical waves , which are hankel functions and numerical solutions of schrdinger s equation , respectively , outside and inside the so - called augmentation spheres . in order to optimize the basis set ,
additional augmented spherical waves were placed at carefully selected interstitial sites . the choice of these sites as well as the augmentation radii were automatically determined using the sphere - geometry optimization algorithm .
@xcite self consistency was achieved by a highly efficient algorithm for convergence acceleration .
@xcite the brillouin zone integrations were performed using the linear tetrahedron method with up to 518 * k*-points within the irreducible wedge .
@xcite in the present work , we used a new full - potential version of the asw method , which was implemented only very recently .
@xcite in this version , the electron density and related quantities are given by a spherical harmonics expansion inside non - overlapping muffin - tin spheres .
in the remaining interstitial region , a representation in terms of atom - centered hankel functions is used .
@xcite however , in contrast to previous related implementations , it is sufficient to work without having to employ a so - called multiple-@xmath24 basis set , permitting a very high computational speed of the resulting scheme .
the electronic properties of @xmath0 in the monoclinic as well as in the above - mentioned artificial rutile structure have already been discussed in much detail in our previous work .
@xcite for this reason , we will in the following concentrate on the changes coming with the consideration of the full potential rather than its muffin - tin approximation .
in particular , we will investigate the changes of the electronic states near the fermi energy .
furthermore , since a very detailed description of the monoclinic crystal structure and its relation to the rutile structure has been given in our previous work , we will not enter a renewed discussion here .
however , we refer the reader to our previous publication especially concerning the definition of the local coordinate systems to be used for the analysis of the electronic states .
@xcite while a first account of the monoclinic crystal structure was given by magnli and andersson , @xcite , our calculations are based on the refined data by brandt and skapski .
@xcite hence , in contrast to our previous work , @xcite we will exclusively deal with the measured monoclinic structure and not discuss our calculations for the related but artificial rutile structure .
in addition , since there are no indications for the formation of magnetic moments in @xmath0 , spin degeneracy was enforced in all calculations .
the resulting partial densities of states ( dos ) are shown in fig .
[ theofig1 ] .
they look very similar to those presented previously .
@xcite in particular , we recognize the 6.5ev wide o @xmath25 bands well below the fermi energy @xmath26 .
in contrast , the mo @xmath16 @xmath6 and @xmath27 manifolds are found around and well above @xmath26 .
worth mentioning is the pronounced dip of the @xmath6 partial dos at @xmath26 .
it results from the peierls - like instability of the @xmath28 states , which usually are designated as the @xmath7 states and which mediate the metal - metal overlap along the mo atom chains parallel to the rutile @xmath29-axis .
the splitting of these bands into bonding and antibonding branches as resulting from this peierls - like instability is clearly visible in fig .
[ theofig2 ] , @xmath6 dos .
selection of orbitals is relative to the local rotated reference frame , see ref . .
] which displays the three partial @xmath6 dos . in representing these partial dos
we have used the same local coordinate system as in our previous work with the local @xmath30-axis parallel to the rutile @xmath29-axis and the local @xmath31-axes pointing alternately along the @xmath32 $ ] and @xmath33 $ ] direction . @xcite in fig .
[ theofig2 ] , we observe the strong splitting of the @xmath28 states , which do no longer contribute to the metallic conductivity . furthermore , due to the splitting , the low lying , bonding branch is found at the lower edge of the @xmath6 group of bands and nearly separated from the higher lying @xmath34 and @xmath35 bands . the latter
are commonly designated as the @xmath21 bands .
the electronic bands along selected lines within the first brillouin zone , fig .
[ theofig3 ] , are displayed in fig .
[ theofig4 ] . with the interpretation of the partial dos at hand
, we easily recognize the o @xmath25 as well as the mo @xmath16 @xmath6 and @xmath27 states in the energy intervals from @xmath36 to @xmath37ev , @xmath38 to @xmath39ev , and @xmath39 to @xmath40ev , respectively . to be more detailed , we find a two bands between @xmath38 and @xmath41ev , which is separated from all other bands .
this split - off doublet is more easily identified in fig .
[ theofig5 ] , electronic bands on an expanded energy scale . ] where we show the near-@xmath26 bands on an expanded energy scale . according to the previous analysis ,
the split - off doublet is just the bonding @xmath7 band .
the one - dimensional behaviour of the @xmath7 bands is contrasted by the rather isotropic dispersion of the @xmath21 bands . as already pointed out in our previous work , hybridization between both types of bands
is rather small .
in contrast coupling of these states is only via charge conservation .
for this reason , we proposed to interpret the splitting of the former in the course of the structural distortions as an embedded peierls - like instability .
@xcite concentrating especially on the fermi body of monoclinic @xmath42 , we complement fig .
[ theofig5 ] by fig .
[ theofig6 ] , which displays cuts through the fermi body in the @xmath43-plane , the @xmath44-plane , and perpendicular to the @xmath45-axis as well as a three - dimensional view of the fermi surface sheets .
the fermi body is defined by the fermi level crossings of the @xmath34 and @xmath35 bands . in figs .
[ theofig5 ] and [ theofig6 ] , we identify three different fermi surface sheets : two hole - like surfaces near the y - point ( sheet 1 and 2 ) , which occupy 9 and 2% , respectively , of the volume of the 1st brillouin zone , as well as a peanut - shaped electron - like surface centered around the @xmath46-point and occupying 11% of the brillouin zone volume ( sheet 3 ) .
the latter sheet has two loopholes near the b - points , whereas it is closed along the direction @xmath46-a . as already mentioned in the introduction , use of the exact single - particle potential , i.e. of the so - called full potential , instead of its muffin - tin or atomic sphere approximation leads to important improvements .
these become obvious from comparing fig.[theofig5 ] of the present work to fig .
6 of ref .. apart from a broadening of the o @xmath25 bands we find distinct differences especially at the z- and the b - point . to be specific ,
those bands , which form the highest occupied states at these points in the present calculation , were found just above @xmath26 in the old calculation .
they generated small hole - like pockets of the fermi surface centered around these points , which were at variance with the experimental findings ( see below ) .
in contrast , in the new calculation these bands are found below @xmath26 and the hole - like pockets disappear .
another important difference can be observed at the @xmath46-point .
the maximum binding energy of the highest occupied band of the new calculation ( @xmath47ev ) agrees much better with experimental value of @xmath48ev ( see below ) than the binding energy of the old calculation ( @xmath49ev ) .
thus , by using the full - potential asw method , the agreement between theory and experiment is considerably improved .
@xmath0 single crystals were grown by chemical transport using @xmath50 as transport agent .
@xcite the crystals exhibit specular surfaces of a size up to @xmath51@xmath52 .
the surface orientation was determined by x - ray diffraction . for the arpes measurements
the as - grown surfaces were cleaned in situ by heating to @xmath53c for some minutes .
the quality and cleanness of the heated single - crystal surfaces were controlled by low energy electron diffraction ( leed ) and by scanning tunnelling microscopy ( stm ) . for the dhva measurements several rods of @xmath54 mm length and @xmath55 mm diameter were prepared .
the @xmath56-axis was the axis of rotation of the rods .
the residual resistance ratio was 97 for @xmath57 and 144 for @xmath58 .
experimentally , the electronic structure @xmath59 was investigated by arpes measurements using the geometry and emission angles defined in fig .
[ expfig1 ] .
is the wave vector in vacuum . ] due to energy conservation the binding energy @xmath60 of the photoelectron is determined by @xmath61 where @xmath62 is the photon energy and @xmath63 the work function of the crystal . for a detailed description of the photoemission process
we refer the reader to the literature .
@xcite on passing through the crystal surface , the component of the electron wave vector parallel to the surface @xmath64 remains unchanged with @xmath65 the perpendicular component @xmath66 is , strictly spoken , not conserved in the photoemission process but it is usually possible to reconstruct its approximate value by including information about the final states of the photoemission process .
if one assumes a free electron - like final state dispersion citenielsen03 , @xmath67 can be calculated as @xmath68 here , @xmath69 is the inner potential of the sample as referred to @xmath26 .
while the work function @xmath63 can be determined experimentally , the inner potential @xmath69 is usually estimated from an analysis of the measured band structure . as a consequence of eq .
( [ gl_ksenkr ] ) , the set of all @xmath64 vectors at a fixed photon energy describes a spherical surface in @xmath70-space .
a fermi vector @xmath71 is defined by the fermi level crossing of a valence band in the corresponding direction of @xmath70-space . as arpes permits a complete mapping of a solid s three - dimensional band structure
, the position of the fermi level crossings can be identified
. a three - dimensional arpes measurement can be performed by varying the emission angles and @xmath62 , causing a variation of @xmath64 and @xmath72 , respectively .
the arpes measurements were done at the beamline sgm3 at the synchrotron astrid in aarhus , denmark .
the beamline provides light in the energy range @xmath73ev @xmath74ev with an energy resolution of @xmath75 .
@xcite the uhv chamber is equipped with an arups10 electron analyzer ( vacuum generators ) with an angular resolution of @xmath76 and an energy resolution of @xmath77mev .
the analyzer is fixed on a computer controlled two - axis goniometer with an angular accuracy up to @xmath78 .
all arpes data were taken at a temperature of @xmath79k .
for the description of the measurement geometry , instead of the polar emission angle @xmath80 its projections on the @xmath81- and @xmath82-planes are used ( see fig .
[ expfig1 ] ) .
the resulting emission angles are called @xmath83 and @xmath84 .
the direction of normal emission was determined by a careful optical alignment of the highly reflective surface using a laser .
arpes spectra taken on the @xmath85 surface for variation of the emission angles @xmath83 and @xmath84 are shown in fig.[expfig2]a and b , surface under variation of the emission angle a ) @xmath83 and b ) @xmath84 ( @xmath86ev ) . ] respectively . at @xmath87ev ,
binding energies up to @xmath88ev have been measured . in both graphs a single peak
is observed , which displays a nearly parabolic dispersion . at normal emission ( @xmath89 and @xmath90 ) ,
it reaches a maximum binding energy of @xmath91ev . with increasing emission angles , both @xmath83 and @xmath84 , the peak approaches the fermi level and and crosses it .
as will be shown later by the symmetry of the fermi surface , the @xmath46-point is reached at @xmath92 ev .
hence , the maximum binding energy of @xmath91ev corresponds to the binding energy at the @xmath46-point .
we can thus compare the occupied bandwidth determined here directly with the binding energy of the highest occupied lda band at the @xmath46-point which is @xmath93ev .
analyzing the nearly parabolic dispersion of the arpes band , effective masses of @xmath94 and @xmath95 can be found for variation of the emissions angles @xmath83 and @xmath84 , respectively .
this is in excellent agreement with shubnikov - de
haas measurements of ref . , which provided for comparable geometry an effective mass of @xmath96 .
the fermi - level crossing of the highest occupied band , defining the corresponding fermi vector , can be identified by a strong increase of photoemission intensity at @xmath26 .
indeed , arpes can also provide a direct image of the fermi surface when the photoemission intensity in a narrow energy window around the fermi level is displayed as a function of emission angle or photon energy ( see ref . and references therein ) .
for a quantitative analysis of the arpes spectra a constant background as determined from the energy region of unoccupied states has been subtracted . to correct intensity variations caused by different emission angles ,
the spectra were normalized to this background .
the intensity dependence on @xmath62 is considered by normalization with the average total intensity of all spectra taken at one certain photon energy .
the work function has been determined by vleed measurements to @xmath97ev .
@xcite the final electron states are supposed to be free - electron like .
the two following methods are used to extract the fermi vectors from the arpes data : 1 .
the photoemission intensity is integrated over a @xmath98mev wide energy region near @xmath26 .
the maxima of this integrated intensity @xmath99 define the fermi vectors .
since the fermi surface is determined by strong variations of the intensity distribution @xmath99 , the gradient @xmath100 could be likewise used to define the fermi vectors .
@xcite however , this gradient method generally produces two fermi contour lines . whereas the contour line on the unoccupied side of the fermi surface coincides with the real fermi surface , the other line is an artefact of this technique .
@xcite both methods can be found in the literature . however , the gradient method often results in a better agreement with theoretical predictions .
@xcite in the following , arpes @xmath64 scans on the @xmath102 surface will be presented , which allow for a fermi surface mapping of planes perpendicular to the direction @xmath101 .
such fermi surface maps can be performed by variation of both emission angles at a fixed photon energy .
they provide spherical cuts through @xmath70-space . nevertheless
, for representation reasons , flat projections of the spherical paths are shown .
the theoretical data are produced in complete analogy : spherical paths in @xmath70-space are calculated , but flat projections are shown . in order to simplify the discussion , we define a cartesian coordinate system within the plane of the projection by @xmath103 and @xmath104 , which are the components of the vector @xmath64 parallel and perpendicular to the direction @xmath46-z , respectively . this situation is sketched in fig .
[ expfig4]a .
-space cuts corresponding to @xmath64 measurements in the a ) @xmath85 surface and b ) @xmath105 surface . the spherical curvature of the area in arpes measurements is neglected . ] in fig .
[ expfig5]a , mapping of the a ) photoemission intensity @xmath106 at @xmath26 , and b ) gradient of the photoemission intensity @xmath107 at @xmath26 , both resulting from an arpes measurement on the @xmath85 surface taken at a photon energy of @xmath87ev .
solid ( red ) lines point to structures of maximal intensity .
points 1 to 4 mark the fermi vectors in direction of @xmath103 and @xmath104 . ] an arpes @xmath64 scan through the fermi body of @xmath0 is displayed .
the measurement has been performed at a photon energy of @xmath92ev .
the photoemission intensity at @xmath26 is plotted as a function of @xmath103 and @xmath104 using the intensity method ( method 1 ) .
two different fermi surface structures , characterized by an enhanced photoemission intensity , can be identified , which are marked by solid ( red ) lines .
most prominent is the ellipsoidal structure , which is located around the origin and characterized by the points 1 to 3 with components @xmath108@xmath109 , @xmath110@xmath109 , and @xmath111@xmath109 , respectively .
in addition to this electron - like fermi surface a much less pronounced hole - like fermi surface contour is observed , albeit for negative values of @xmath112 only .
it comprises point 4 with component @xmath113@xmath109 and separates electron - like states with @xmath114 from hole - like states in the outer region @xmath115 .
figure [ expfig5]b shows the photoemission intensity gradient @xmath100 as a function of @xmath103 and @xmath104 ( method 2 ) , based on the data shown in fig . [ expfig5]a . in fig .
[ expfig5]b , the above - mentioned double structure of the gradient method is nicely visible for the electron - like fermi surface centered around the origin .
the real fermi surface contours , i.e. those not caused by the artefact of the gradient method , are marked by solid ( red ) lines .
the extent of this fermi surface amounts to @xmath116@xmath109 , @xmath117@xmath109 , and @xmath118@xmath109 , respectively .
the hole - like fermi surface contour for negative values of @xmath104 is also observed in fig .
. however , the photoemission gradient in this region is very small and , hence , the expected double fermi surface structure can not be detected .
only the real fermi surface contour is visible .
the extent of the hole - like fermi surface amounts to @xmath119@xmath109 . in general , the error of both methods of fermi surface determination is defined by the @xmath70-resolution of the arpes experiment and amounts to @xmath120@xmath109 . in fig .
[ expfig7 ] , : arpes and lda fermi surface contours . ]
the experimental arpes fermi surface contours as resulting from both the use of the intensity at @xmath26 and the gradient of the intensity are compared to the contours resulting from the lda calculations .
as mentioned above , the lda calculations are performed on spherical paths , so as to account for the spherical curvature of the experimental plane . however , in the figures , the flat projection on the plane perpendicular to @xmath121 is shown .
only the @xmath70-space region accessible by the arpes apparatus is included .
according to fig .
[ expfig7 ] , the lda calculations predict an electron - like , elliptic fermi surface around the @xmath46-point . note that , as will be shown later , @xmath122 reaches the @xmath46-point right at @xmath123ev . as a consequence , the @xmath64 mappings shown in figs .
[ expfig5]a and b as well as in fig .
[ expfig7 ] cover a plane perpendicular to @xmath121 , which includes the @xmath46-point ( see fig .
[ expfig4]a ) .
the theoretical findings are confirmed by the arpes measurements .
however , as compared to the lda results , the fermi surface contours determined by the intensity method are shifted towards the occupied states .
hence , the arpes fermi surface is smaller than the calculated one with the deviations of the measured fermi vectors ranging from @xmath124 to @xmath125% . here and in the following all deviations
are given with respect to the calculated values .
in contrast , the gradient method produces smaller shifts in both directions , away from as well as towards the occupied states . in this case
, the deviations of the measured from the calculated values range from @xmath126 to @xmath127% .
perpendicular to the @xmath46-z direction , cuts through two hole - like fermi surface sheets centered around the y - point are predicted by theory ( see also fig .
[ theofig6]a ) .
however , owing to the spherical curvature of the measurement plane , observation of both hole - like fermi surface contours is expected only for positive values of @xmath104 .
in contrast , for @xmath128 only one of these contours is cut by the measurement plane , which is indeed observed in the arpes experiments . in the region of negative @xmath104 its inner surface
can be seen in fig .
[ expfig7 ] , while both the outer part in @xmath128 and the two hole - like structures for @xmath129 are not detected .
compared to the lda calculations , also the hole - like arpes fermi surface contour shows a slight shift , which , for the results obtained by the intensity method ( method 1 ) , is directed towards the occupied states .
the corresponding arpes fermi surface is thus bigger than the calculated one , and the experimental deviations from the theoretical values amount to @xmath130 to @xmath131% . for the gradient method
the arpes contours generally agree well , but significant shifts with deviations between @xmath38% and @xmath130% are observed near some fractions of the fermi surface .
no systematic trend for the direction of these shifts can be discerned .
the experimental and theoretical values of the fermi vectors @xmath132 and @xmath133 in the plane perpendicular to @xmath121 through the @xmath46-point as well as the deviations of the theoretical values from the experimental ones as arising from the gradient method are summarized in tab .
[ exptab1 ] .
.magnitudes of fermi vectors in inverse .
the experimental fermi vectors are taken from @xmath64 scans on the @xmath85 surface ( @xmath87ev ) and analyzed using both the intensity maximum and its gradient .
the last lines gives the respective deviations of the experimental result as based on the gradient method from those arising from the lda calculations .
the plane perpendicular to @xmath121 through the @xmath46-point has been analyzed .
the points 1 to 4 are marked in fig .
[ expfig7 ] . [ cols="^,^,^,^,^ " , ] the dhva surface areas and the lda results parallel and perpendicular to @xmath134 are summarized .
all three theoretically predicted fermi surface sheets have been confirmed by the dhva measurements .
qualitatively , the shapes of the experimental and theoretical fermi surface contours agree well .
the quantitative comparison along selected high symmetry directions shows that there are slight discrepancies between the dhva and the lda results .
all dhva fermi surface areas are smaller than calculated by lda .
the discrepancies lie between @xmath135% and @xmath136% .
the peanut shaped , electron - like fermi surface sheet 3 shows the largest deviation to the lda calculations . due to the underlying ellipsoidal approximation ,
the shape of the dhva fermi surface sheets have to be treated with care .
for some directions the real fermi surface topology clearly differs from the assumed ellipsoidal shape .
especially , for sheet 2 ( @xmath137 and @xmath138 to @xmath134 ) and for sheet 3 ( @xmath138 to @xmath139 ) the lda results indicate non - negligible deviations .
a comparison of our dhva results to previously published magneto - transport data reveals a very good accordance .
@xcite in particular , the results for the fermi surface sheet 1 agree nearly perfectly , only differences of less then 1.5% were found .
in contrast , only little has been published for sheets 2 and 3 . yet
, our results for these fermi surface sheets , while being much more comprehensive , agree very well with the existing data , the differences being less than 6.5% . temperature dependent measurements of the magnetic oscillations allow the determination of effective masses of the charge carriers via eq .
( [ gl_redcoef ] ) .
an increase in temperature causes a reduction of the dhva signal . for a single orientation of the magnetic field
the effective masses of the three observed bands have been analyzed .
the angle between @xmath140 and @xmath141 amounts to @xmath142 .
the results are displayed in fig .
[ expfig18 ] . ) as a fit .
the angle between @xmath140 and @xmath141 amounts to @xmath142 . ]
the temperature has been varied between @xmath143k and @xmath54k .
the oscillation amplitudes ( normalized to one at @xmath144k ) as a function of the temperature are plotted .
a fit of the data using eq .
( [ gl_redcoef ] ) yields the effective masses @xmath145 .
the effective masses of the three fermi surface sheets differ clearly .
sheet 1 shows a slightly decreased effective mass of @xmath146 .
however , sheets 2 and 3 exhibit increased masses of @xmath147 and @xmath148 , respectively .
these values are in good agreement with existing shubnikov - de haas measurements.@xcite it has to be noted that the described geometry is not the one discussed in conjunction with the arpes measurements .
in general , we found excellent agreement between the experimental and theoretical fermi surface contours .
both the arpes and the de haas - van alphen results confirm the lda calculations . in particular , all three fermi surfaces sheets predicted by theory have been verified experimentally , with respect to both the shape and the dimensions . of course , while from the theoretical side all fermi surface sheets can be calculated in one go , the measurements are restricted by the respective experimental setups and limitations .
for instance , only certain cuts through the fermi body are usually accessible .
in addition , in the arpes measurements matrix element effects may also play a role , which might undermine the experimental detection of portions of the fermi surface .
this may explain the fact that the calculated fermi surface contours of all three sheets can only be partially observed by arpes . while the overall agreement of the results of all three methods is very good , there still remain small differences on a quantitative level .
these deviations have been included in tabs .
[ exptab1 ] and [ exptab2 ] . to be specific , the deviations of the arpes fermi vectors from the calculated ones amount to several percent but may go up even to @xmath79% for some experimental points .
yet , both negative and positive deviations occur for both the electron- and hole - like surfaces , precluding systematic errors .
this is different for the de haas - van alphen measurements .
all fermi surface contours resulting from these experiments , including those given in the literature , are systematically smaller than the calculated ones .
@xcite to be specific , the electron - like fermi surface sheet 3 is shifted up to @xmath136% in direction of the occupied states , whereas the hole - like fermi surface sheets 1 and 2 experience a shift of up to @xmath149% in the direction of the unoccupied states . however , we point out that in general the differences between the experimental and theoretical data are of a similar size as the differences between the different experimental techniques .
nevertheless , all these differences between the experimental and theoretical results , namely , both the unsystematic deviations of the arpes data and the systematically smaller dhva fermi surface contours , still conserve , within the above - mentioned limitations of the experimental setups , the volume of the fermi body , which , by construction , is correctly accounted for by the calculations .
@xcite for the same reason , non - stoichiometry of the samples can most likely be ruled out as a source of the observed quantitative differences . while the phase diagram for mo - o compounds excludes
a hole - doping of more than @xmath150%,@xcite a small defect concentration of @xmath151% has been detected by hall effect measurements.@xcite in general , hole - doping causes a lowering of the fermi level . according to the band structure shown in fig .
[ theofig5 ] , this is equivalent to a shift of the fermi surface contours towards the occupied states .
however , a shift of up to @xmath152% would necessitate a hole - doping of up to @xmath153% .
hence , non - stoichiometry will play only a minor role . of course , one might argue , that electronic correlations may affect the quality of the calculated results . indeed
, our previous calculations infer the existence of weak electron correlations in @xmath0.@xcite according to the present measurements , effective masses are only slightly enhanced . yet
, an enhancement may explain the reduction of the band width as it has been observed in the @xmath64 measurements on the @xmath154 surface .
however , magneto - transport measurements performed in this direction do not indicate an enhanced effective mass.@xcite furthermore , as has been demonstrated above part of the experimentally observed band width reduction could be accounted for by the inclusion of the full single - particle potential in the present calculations . going into more detail ,
the deviations between the experimental and theoretical data suggest that , as compared to the experiments , the calculated fermi body is slightly shortened parallel to the @xmath155-axis and dilated perpendicular to this axis . in the directions perpendicular to @xmath155 this presumption
is confirmed by the experimental data . in particular , the cuts through the electron - like arpes and dhva fermi body are smaller than the calculated results .
the assumed shortening parallel to @xmath155 is on somewhat weaker grounds .
only the waist of the `` peanut '' could be measured by arpes , but no experimental points exist at the top and bottom of the `` peanut '' .
in contrast , the dhva measurements in the direction parallel to @xmath155 might be affected by the underlying ellipsoidal approximation . yet , our assumption is well supported by the overall conservation of the volume of the enclosed fermi surface as discussed above .
nevertheless , in passing we mention that these deviation are of the order of a few percent only .
a detailed and comprehensive study of the electronic properties of @xmath0 has been presented .
this material has been analyzed using angle - resolved photoelectron spectroscopy ( arpes ) , de haas - van alphen ( dhva ) measurements , and electronic structure calculations based on a new full - potential implementation of the augmented spherical wave ( asw ) method .
the fermi surfaces determined by both kinds of experiments are in very good agreement with the theoretical predictions .
in particular , three different fermi surface sheets are correctly identified by all three approaches .
these sheets include an electron - like peanut - shaped fermi surface centered around the @xmath46-point , which has a volume of 11% of the brillouin zone , and two smaller hole - like fermi surfaces both centered around the y - point . as far as they can be detected within the limitations of the experimental geometries , slight differences between the epxerimental and theoretical results concern mainly the shape of the three fermi bodies .
to be specific , from the calculations , the peanut - shaped fermi body appears to be slightly shortened and elongated parallel and perpendicular to the @xmath141-axis . with the help of the new calculations , controversies concerning two additional fermi surface sheets centered around the z- and b - point
could eventually be resolved . in particular
, the occurence of these sheets could be attributed to the atomic - sphere approximation , which was used in our previous work .
in contrast , in the new full - potential asw calculations , these additional sheets do no longer appear .
furthermore , the full - potential treatment led to a much improved agreement between the arpes results and the calculations as concerns the occupied band width at the @xmath46-point .
in general , the very good agreement between the results obtained by the arpes and dhva measurements for metallic @xmath0 has important consequences for the investigation of those neighbouring transition - metal dioxides , which display a metal - insulator transition .
while the fermi surfaces of the metallic phases of these materials are not accessible to a low - temperature technique as dhva experiments , the good agreement found in the present study implies that arpes measurements will be useful to study the fermi surfaces of the high - temperature metallic phases .
furthermore , the good agreement between the different experimental techniques and the electronic structure calculations , with the deviations between measured and calculated data being of the same order as those between the different experimental techniques , support our original proposal that these calculations are well suited to study the electronic properties of the early transition - metal dioxides . in particular ,
our results are very encouraging concerning the validity of our calculations for the @xmath9 members @xmath1 and @xmath2 and support the interpretation of the metal - insulator transitions of the latter two compounds as embedded peierls - like instabilities .
many useful discussions with k .- h .
hck , r. horny , g. obermeier , and s. klimm are gratefully acknowledged .
thanks are due to s. g. ebbinghaus for performing the xrd - measurements .
this work was supported by the deutsche forschungsgemeinschaft through sfb 484 and the european community through access to research infrastructures ( ari ) . | a comprehensive study of the electronic properties of monoclinic @xmath0 from both an experimental and a theoretical point of view is presented .
we focus on the investigation of the fermi body and the band structure using angle resolved photoemission spectroscopy , de haas - van alphen measurements , and electronic structure calculations . for the latter , the new full - potential augmented spherical wave ( asw ) method has been applied .
very good agreement between the experimental and theoretical results is found . in particular , all fermi surface sheets are correctly identified by all three approaches .
previous controversies concerning additional hole - like surfaces centered around the z- and b - point could be resolved ; these surfaces were an artefact of the atomic - sphere approximation used in the old calculations .
our results underline the importance of electronic structure calculations for the understanding of @xmath0 and the neighbouring rutile - type early transition - metal dioxides .
this includes the low - temperature insulating phases of @xmath1 and @xmath2 , which have crystal structures very similar to that of molybdenum dioxide and display the well - known prominent metal - insulator transitions . |
non - superconducting prba@xmath0cu@xmath1o@xmath2 compound attracted considerable attention in the last two decades because of its intriguing magnetic and electronic properties within the reba@xmath6cu@xmath7o@xmath2 family of compounds ( re = rare earth atom).@xcite in spite of the great efforts from both experimental and theoretical sides , numerous problems remain unresolved such as the suppression of superconductivity and the nature of the long - range ordered state of pr sublattice with a unique , about an order of magnitude larger than for other rare earth atoms , ordering temperature .
although , there exists a discussion whether the ground state of prba@xmath0cu@xmath1o@xmath2 is really a non - superconducting , insulating material with a magnetically ordered cu and pr sublattices , herein we consider this modification of prba@xmath0cu@xmath1o@xmath2 as `` canonical '' and treat it solely herein .
we note that superconductivity in prba@xmath0cu@xmath1o@xmath2 was reported@xcite and a non - magnetic pr ground state was found by @xmath8pr nmr.@xcite it was suggested that these observations may result from a non - stoichiometric compound , i.e. when pr occupies only about a half of the rare - earth sites , and the other half is occupied by the non - magnetic ba.@xcite the theoretical model of fehrenbacher and rice , i.e. the hybridization of pr @xmath9 and the nearest neighbor o @xmath10 orbitals is the most accepted model to account for the absence of superconductivity in prba@xmath0cu@xmath1o@xmath2 .
it is suggested that the localization of holes in the hybridized @xmath9-@xmath10 orbitals renders the material non - superconducting.@xcite although the non - superconducting nature of prba@xmath0cu@xmath1o@xmath11 compound is a challenging and interesting problem , we concentrate herein exclusively on the low - temperature pr ordering .
prba@xmath0cu@xmath1o@xmath2 is an insulator for every @xmath12 and the cu spins in the cuo@xmath0 planes order antiferromagnetically at temperatures of @xmath13k and @xmath14k for @xmath15 and @xmath16 , respectively.@xcite as the temperature is further decreased , the pr sublattice also undergoes a phase transition@xcite in the temperature range @xmath17k depending on @xmath12 , which appears as an anomaly in the temperature dependence of thermodynamic quantities.@xcite mssbauer spectroscopy,@xcite neutron diffraction,@xcite and nmr@xcite showed that the pr magnetic moments order antiferromagnetically below @xmath18 . however , there is no consensus among these experiments with respect to the magnitude of the ordered pr moment , its direction , and the nature of this transition .
neutron diffraction studies found @xmath19 ( @xmath20 ) and @xmath21 ( @xmath22 ) ordered moment with direction tilted out from the @xmath4-plane.@xcite the rotation of cu moments within the cuo@xmath0 plane was found to accompany the pr ordering,@xcite which was suggested to result from the strong coupling between the pr and cu subsystems . however , the magnitude of the coupling , its character , and its influence on the magnitude of the ordered pr magnetic moment , and for the ordering temperature is yet unexplained .
we give herein a comprehensive description of the pr ordering in prba@xmath0cu@xmath1o@xmath23 on the basis of a model of localized @xmath24 electrons of the pr ion .
we construct the general form of the cu - pr interaction allowed by the symmetry and identify the pr order parameter from the rotation of cu spins below @xmath5 observed by neutron diffraction.@xcite although pseudo - dipole cu - pr interaction has been already proposed,@xcite we derive the cu - pr interaction in general including all pr multipole moments and arbitrary wave vectors .
we describe a crystalline electric field ( cef ) model which takes a ground state quasi - triplet composed by a doublet and a singlet states and also includes coupling between the pr and cu subsystems .
this low - energy scheme is consistent with the neutron diffraction results@xcite and also with the high - temperature susceptibility data.@xcite we also discuss the enhancement of the pr ordering temperature due to the cu - pr coupling and the magnitude of pr magnetic moment in the ordered state .
in prba@xmath0cu@xmath1o@xmath23 , the cu spins are antiferromagnetically ordered along the @xmath25 $ ] direction in the cuo@xmath0 planes with the ordering vector @xmath26 $ ] below the temperature @xmath27k ( phase afi).@xcite as the temperature is further decreased , the pr sublattice also undergoes an ordering at temperature @xmath28k .
this ordering is accompanied by the rotation of cu spins within the cuo@xmath0 planes in such a way that the new spin structure is non - collinear along the @xmath29-direction ( phase afiii).@xcite the cu spin rotation is characterized by the ordering vector @xmath30 $ ] . the cu spin structure in the cuo@xmath0 planes is shown in fig .
[ cu - pr - fig ] in the temperature range @xmath31 .
planes and pr dipole moments in the @xmath4-plane in the temperature range @xmath32 .
the central pr ion in the @xmath4-plane is located at position @xmath33 , and the numbers label the cu sites , i.e. cu[1 - 8 ] used in the text . ]
first , we construct the invariant form of the interaction between the multipole moments of a pr ion at position @xmath33 and the surrounding eight cu spins , cu[1 - 8 ] , shown in fig .
[ cu - pr - fig ] .
although the local symmetry at pr site is tetragonal , a cu - pr pair has lower symmetry .
for example , the cu[1]-pr pair has reflection symmetry with respect to the @xmath34 $ ] mirror plane .
the transformation of total angular momentum @xmath35 of pr ion under this operation is : @xmath36 , @xmath37 , and @xmath38 .
the same transformation is applied for the cu spin @xmath39 .
thus , spin operators @xmath40 ( @xmath41 ) and @xmath42 ( @xmath43 ) are odd , while @xmath44 ( @xmath45 ) is even under this reflection .
not only the bilinear products of the pr dipole moments and cu spins appear in the interaction but also the bilinear products of rank-3 octupole and rank-5 triakontadipole@xcite operators of pr ion and cu spins since they are allowed by the time - reversal symmetry .
the octupole operators @xmath46 and @xmath47 are even , while @xmath48 and @xmath49 are odd operators under the present transformation . in the case of tetragonal symmetry , there is one more independent magnetic operator , namely the rank-5 triakontadipole operator @xmath50 which is even under the present reflexion .
thus , the invariant form of the interaction for the cu[1]-pr pair is constructed as @xmath51 & = & ( s_{x}+s_{y})\left [ c_{11}(j_{x}+j_{y } ) + c_{12 } t^{\beta}_{z } + c_{13}a_{1u } \right ] \nonumber\\ & + & ( s_{x}-s_{y})\left [ c_{21}(j_{x}-j_{y } ) + c_{22 } j_{z } + c_{23 } t_{xyz } \right ] \nonumber\\ & + & s_{z}\left [ c_{31}(j_{x}-j_{y } ) + c_{32 } j_{z } + c_{33 } t_{xyz } \right ] , \label{int1}\end{aligned}\ ] ] where @xmath52 are constants which are not determined by the symmetry itself .
the pair cu[2]-pr is connected to the pair cu[1]-pr by a @xmath53 rotation around the @xmath29-axis . under this operation
the transformation of the angular momentum components is given by @xmath54 , @xmath37 , and @xmath55 .
the same transformation holds also for the cu spin components .
thus , the form of the interaction for the cu[2]-pr pair can be obtained from eq .
( [ int1 ] ) by applying the @xmath53 rotation . repeating further the appropriate rotations and reflections , the interaction can be obtained for all the eight cu - pr pairs .
the cu spin at position @xmath56 is expressed by fourier transformation as @xmath57 . using this expression
, we obtain the total interaction as @xmath58 = 8\sum_{\bf q } \left\ { ( c_{11 } + c_{21})(s_{x}^{\bf q}j_{x}+s_{y}^{\bf q}j_{y})c_{x}c_{y}c_{z } + ( c_{11 } - c_{21})(s_{x}^{\bf q}j_{y}+s_{y}^{\bf q}j_{x})s_{x}s_{y}c_{z } \right .
\nonumber\\ & + & \left .
c_{22}j_{z } \left [ s_{x}^{\bf q } s_{x } c_{y } s_{z } - s_{y}^{\bf q } c_{x } s_{y } s_{z } \right ] + c_{31 } \left [ j_{x } s_{z}^{\bf q } s_{x } s_{y } c_{z } - j_{y } s_{z}^{\bf q } c_{x } s_{y } s_{z } \right ] + c_{32 } j_{z } s_{z}^{\bf q } c_{x } c_{y } c_{z } \right .
\nonumber\\ & + & \left . c_{12 } t^{\beta}_{z } \left [ s_{x}^{\bf q } s_{x } c_{y } s_{z } + s_{y}^{\bf q } c_{x } s_{y } s_{z } \right ] + c_{13 } a_{1u } \left [ s_{x}^{\bf q } c_{x } s_{y } s_{z } + s_{y}^{\bf q } s_{x } c_{y } s_{z } \right ] + c_{23 } t_{xyz } \left [ s_{x}^{\bf q } c_{x } s_{y } s_{z } - s_{y}^{\bf q } s_{x } c_{y } s_{z } \right ] \right .
\nonumber\\ & + & \left .
c_{33 } t_{xyz }
s_{z}^{\bf q } s_{x } s_{y } c_{z } \right\ } , \label{int}\end{aligned}\ ] ] where @xmath59 denotes @xmath60 for @xmath61=@xmath12 , @xmath62 , and @xmath63 for @xmath61=@xmath64 .
now we discuss the multipole order of pr sublattice based on expression ( [ int ] ) . in phase afi ,
the cu spin component @xmath65 with @xmath66 is non - zero . in this phase
there is no coupling between the cu spins and pr magnetic moments because the wave vector @xmath67 leads to @xmath68 and @xmath69 , which gives that none of the terms survives in expression ( [ int ] ) . on the other hand ,
a coupling of cu spins and pr moments is present in phase afiii .
the non - collinear cu spin structure in phase afiii can be described by the appearance of the extra spin component @xmath70 in addition to the component @xmath71 .
the wave vector @xmath72 gives @xmath73 , @xmath74=1 , and @xmath75 , @xmath76 in expression ( [ int ] ) .
together with the condition @xmath77 , we find that the only surviving term in the form of cu - pr interaction is the coupling @xmath78 .
we note that the coupling @xmath79 is also allowed by symmetry since the @xmath25 $ ] and @xmath80 $ ] domains in phase afi are equivalent , which gives that the domains @xmath81 and @xmath82 are also equivalent .
thus , we conclude that the neutron diffraction result for the cu spin rotation below @xmath5 is compatible with pr dipole moments lying along the crystalline @xmath25 $ ] or @xmath80 $ ] direction , and no deviation from this direction is allowed by symmetry in contrast to the structure proposed in ref . .
since the @xmath83 cu spin components alternate along the @xmath84 and @xmath85 directions , the pr dipole moments also form an afm structure in the @xmath4-planes with the ordering vector @xmath72 .
the magnetic structure for the cu spins and pr dipole moments below @xmath5 has the structure shown in fig . [ cu - pr - fig ] .
we now turn to discuss a cef model of the pr ion @xmath9 electrons to describe the pr ordering in prba@xmath0cu@xmath1o@xmath23 .
the model consists of the coupled pr and cu subsystems and allows to calculate experimentally relevant quantities .
the pr ion has 4@xmath86 electronic configuration in prba@xmath0cu@xmath1o@xmath23 , which gives @xmath87 as the total angular momentum based on the hund s rule for the ground state . in the pr subsystem ,
we consider afm interaction between the in - plane @xmath88 and @xmath89 dipole moments described by the hamiltonian @xmath90 , \label{ham - pr}\end{aligned}\ ] ] where @xmath91 is the number of nearest - neighbor pr ions and @xmath92 is the coupling constant .
the hamiltonian ( [ ham - pr ] ) is taken in a quasi - triplet subspace , where the ground state is the doublet latexmath:[\ ] ] which ratio can be checked in the nmr or esr experiments .
acting on the local magnetic probe by changing the direction of the external magnetic field in the @xmath164 $ ] ( _ top _ ) and @xmath165 $ ] ( _ bottom _ ) planes with @xmath167 and @xmath168 , respectively . in the plot
we take the exemplifying values @xmath169 and @xmath170 .
the field angles are defined in the text .
, title="fig : " ] + acting on the local magnetic probe by changing the direction of the external magnetic field in the @xmath164 $ ] ( _ top _ ) and @xmath165 $ ] ( _ bottom _ ) planes with @xmath167 and @xmath168 , respectively . in the plot we take the exemplifying values @xmath169 and @xmath170 .
the field angles are defined in the text .
, title="fig : " ]
we studied the nature of the pr ordered phase in prba@xmath0cu@xmath1o@xmath23 in a quasi - triplet cef model of 4@xmath86 electrons , where we also included the symmetry allowed coupling between the pr and cu ions .
the reason to include cu - pr coupling is the cu spin rotation observed at the pr ordering temperature in neutron diffraction,@xcite which has not been observed for other rare earth reba@xmath0cu@xmath1o@xmath23 compounds . in the 4@xmath171 rare - earth series ,
the localized electrons of pr ion tend to hybridize rather strongly with conduction electrons as in e.g. prfe@xmath172p@xmath173
. this can be the reason why the cu - re coupling is much stronger in reba@xmath0cu@xmath1o@xmath23 for re = pr compared to the other rare - earth ions , re .
we showed by a general symmetry analysis that there is no coupling between the cu and pr subsystems in phase afi , but a coupling emerges in phase afiii .
we prove that only the afm ordering of @xmath174 dipole moments of pr ions is consistent with the cu spin rotation observed at @xmath100 by neutron diffraction.@xcite the interpretation of the neutron diffraction data in ref .
includes a tilting of the ordered pr magnetic moments out of the @xmath4-plane .
however , the presence of non - zero @xmath42 dipole component does not follow from the general symmetry analysis as it was shown in section [ cu - pr - interaction ] . upon identifying of the pr order parameter , we studied the pr ordering temperature in the cu - pr coupled quasi - triplet cef model by changing the interaction and cef parameters .
we found that the cu - pr coupling enhances the pr ordering temperature , @xmath5 , compared to the uncoupled ordering temperature , @xmath124 . by fixing the pr - pr and pr - cu interaction parameters to the values obtained by neutron diffraction,@xcite and the cef parameter to that obtained by high - temperature susceptibility measurements,@xcite we found that the pr ordering temperature is considerably enhanced due to the cu - pr coupling in the parameter range @xmath175k .
namely , the enhancement of the ordering temperature is as large as @xmath176 in this interval .
this observation explains the uniquely large ordering temperature for re = pr in the series reba@xmath0cu@xmath1o@xmath2 .
in addition , we predicted the magnetic field angle dependence of the local magnetic field acting at a pr site in the afm ordered phase of the pr ions . this can be directly compared to measured nmr or esr spectra which detect the local magnetic field by a local magnetic probe .
our model contains the minimal number of parameters which is required to account quantitatively for the experimental data .
given that the parameter values are assumed from independent measurements and no fit is performed , the agreement is reasonable .
first , the realistic estimate @xmath175k gives a @xmath5 of @xmath177k which is to be compared to the experimental @xmath5 of @xmath178k .
we note that this estimated range gives the pr transition temperature @xmath124 for the uncoupled system as @xmath179k , which is close to the typical values of the ordering temperatures for other rare - earth ions such as yb ( @xmath180k ) , nd ( @xmath181k ) or dy ( @xmath182k ) .
this fact may also indicate that the uniquely large ordering temperature for pr arises from the cu - pr coupling .
second , the interval @xmath175k is not very far from the estimate @xmath183k based on the analysis of the high - temperature susceptibility data.@xcite furthermore , the estimated interval for @xmath94 and also the splitting pattern of the quasi - triplet subspace is consistent with the results of neutron scattering,@xcite which found that the splitting of the ground state quasi - triplet does not exceed @xmath184mev above @xmath5 , and two inelastic lines are observed at energies @xmath185mev and @xmath186mev at @xmath187k in the ordered phase .
finally , for the interval @xmath175k our cef model gives the ordered pr magnetic moment at low temperatures as @xmath188 , which is to be compared to the observed value @xmath189.@xcite the quantitative discrepancies between the results of our model and the experimental data arise because of the simplicity of our model and minimal number of parameters . in summary
, the model we propose herein describes the main features of prba@xmath0cu@xmath1o@xmath23 system , and demonstrates the importance of the cu - pr coupling which may be the key point to understand the anomalous behavior of this compound such as the enhancement of the pr ordering temperature .
further studies which include also the itinerant character of @xmath24 electrons are necessary to give an extensive description of this rather complicated system .
this work was supported by the hungarian state otka grant k73455 and by the european research council grant nr .
ak acknowledges the postdoctoral programme of the magyary foundation and egt norway grants . | theoretical model is presented to describe the anomalous ordered phase of pr ions in prba@xmath0cu@xmath1o@xmath2 below @xmath3k .
the model considers the pr multipole degrees of freedom and coupling between the cu and pr subsystems .
we identify the symmetry allowed coupling of cu and pr ions and conclude that only an @xmath4-plane pr dipole ordering can explain the cu spin rotation observed at @xmath5 by neutron diffraction by boothroyd _
et al . _
[ a. t. boothroyd _
et al . _ ,
phys .
rev .
lett . * 78 * , 130 ( 1997 ) ] . a substantial enhancement of the pr ordering temperature is shown to arise from the cu - pr coupling which is the key for the anomalous magnetic behavior in prba@xmath0cu@xmath1o@xmath2 . |
rare earth titanates ( r@xmath0ti@xmath0o@xmath1 ) are an interesting family of materials showing rich ferroelectric @xcite and magnetic properties @xcite .
the common crystal structures @xcite for these materials are cubic pyrochlore for small rare earth ions ( sm@xmath3 lu@xmath3 ) and monoclinic for large rare earth ions ( la@xmath3 nd@xmath3 ) . the magnetic properties of the former have been explored extensively during the recent years @xcite . the subtle interplay among the various interactions including crystal field , exchange and dipole - dipole , coupled with geometrical frustration specific to the pyrochlore lattice @xcite , has led to exotic magnetic phases such as spin ice @xcite and spin liquid @xcite .
these phases are extremely sensitive to doping and external stimuli such as pressure @xcite and magnetic fields @xcite .
interesting dynamic processes have also been identified @xcite . on the other hand ,
the rare earth titanates with a monoclinic structure have received much less attention . due to the removal of geometric frustration ,
these materials are believed to possess magnetically ordered phases at low temperatures . however , the competition between the crystal field and exchange interactions ( plus dipole - dipole ) with close energy scales may lead to interesting and unusual dynamic freezing processes towards the ordered states . in this regard ,
it is essential to retrieve the magnitude of the crystal field , exchange and the dipole - dipole interactions , in order to study the magnetic freezing processes .
here we focus on the monoclinic nd@xmath0ti@xmath0o@xmath1 ( nto ) . some features of nto , such as the ferroelectricity @xcite , the optical magnetoelectric effect @xcite and the photo - catalysis effect @xcite , receive current interests in these materials . in this work , we study the dc susceptibility and specific heat of nto single crystals . the anisotropic paramagnetism is investigated in detail .
we found that crystal field interactions dominate the anisotropy , and exchange interactions @xmath4 are estimated by combining the results of the diluted samples .
the ground state of nd@xmath3 spins splits into five kramers doublets in the crystal field associated with the @xmath5 site in @xmath6 space group .
the magnetic field splits the ground state doublet into two singlets , as is evidenced by the specific heat at different fields , showing a two - level schottky scheme with the energy gap proportional to the field .
these studies lay solid ground for further investigation of this system under doping and external perturbations .
single crystals of nd@xmath0ti@xmath0o@xmath1 and nd@xmath7la@xmath8ti@xmath0o@xmath1 were grown by the floating zone method in flowing oxygen .
seed rods for the single crystal growth were polycrystalline nd@xmath0ti@xmath0o@xmath1 and nd@xmath7la@xmath8ti@xmath0o@xmath1 sintered by the standard solid state reaction in oxygen atmosphere , with starting materials of stoichiometric mixture of nd@xmath0o@xmath9 , la@xmath0o@xmath9 and tio@xmath0 powder ( alfa aesar ) .
oxygen atmosphere was used to avoid oxygen deficiency .
the lattice constants were obtained by fitting the peak positions of the x - ray diffraction pattern on polycrystals obtained by crushing nd@xmath0ti@xmath0o@xmath1 single crystals .
the single crystals were cleaved along @xmath10 direction , and the in - plane @xmath11 and @xmath12 directions were determined by the laue diffraction .
dc susceptibility were measured with a quantum design superconducting interference device magnetometer ( squid mpms-5 ) with fields of up to 5 t. dc susceptibility at higher fields were measured by the vibrating sample magnetometer option in a quantum design physical property measurement system ( ppms ) .
the in - plane angular dependence of the magnetization were measured using a commercial sample rotator in mpms , the diamagnetic background of the sample holder was carefully subtracted . for the susceptibility measurement along different crystalline directions ,
the alignment of the crystal along the direction of the magnetic field was maintained within the error of @xmath13 .
specific heat was measured in a ppms using the thermal relaxation technique .
the as - grown nto single crystals are of high quality and large size benefiting from the floating zone method .
the samples are monoclinic in space group @xmath6 , the diffraction peaks of the polycrystals obtained by crushing nto single crystals are well indexed as shown in fig . [
xrd](a ) , and resultant lattice parameters are @xmath14 , @xmath15 , @xmath16 , and @xmath17 , @xmath18 .
due to the layered structure , the crystal cleaves naturally along the ( 100 ) direction , as is shown in the xrd pattern in fig .
[ xrd ] ( b ) . in fig .
[ xrd ] ( c ) , the full width at half maximum ( fwhm ) of 0.09@xmath19 of the rocking curve further indicates good crystallinity of the sample .
the single crystals were cleaved and shaped into disks with typical dimensions of diameter @xmath20 1.5 mm and thickness @xmath20 0.2 mm .
the disk shape mitigated the effect of demagnetization factor in the in - plane anisotropy of magnetization .
however , we still corrected the demagnetization using the calculated value @xcite @xmath21 for @xmath22 and 0.12 for @xmath23 respectively .
in fig . [ pm ] ( a1)(c1 ) , we show @xmath24 at 1000 oe with @xmath25 and @xmath22 for a single crystal and a polycrystal sample , respectively . the zero - field - cooled and field - cooled @xmath24 overlap in the entire temperature range in study ( 2 300 k ) , showing typical paramagnetic behavior .
however , @xmath24 is clearly anisotropic when comparing the behavior of the single crystal with @xmath25 and @xmath22 , as well as that of a polycrystal . above 150 k ,
the system follows the curie - weiss law with the effective moment of 3.44 3.49 @xmath26 , close to the theoretical value of 3.62 @xmath26 ( an experimental value @xcite is 3.5 @xmath26 ) . the resultant weiss temperature @xmath27 varies from 10.4 k for @xmath25 to @xmath28 k for @xmath22 .
@xmath27 for the polycrystal , which can be taken as the average effect from all the crystalline directions , is @xmath29 k. the negative weiss temperature of the polycrystal indicates an effective antiferromagnetic correlation between nd@xmath3 spins . by comparing fig .
[ pm ] ( a1 ) with ( b1 ) , we see that the in - plane magnetization is more than three times larger than the @xmath22 case , this suggests that the nd@xmath3 spins prefer to align in the @xmath30 plane .
[ pm ] ( a2)(c2 ) show the isothermal magnetization curves for @xmath31 along different crystalline directions .
as is expected for a paramagnetic system , @xmath32 shows no hysteresis , is linear at high temperatures , and shows convex curvatures at low temperatures due to the spin polarization process .
the @xmath32 curves measured with the field applied along different directions again show strong anisotropy .
moreover , the in - plane @xmath32 curves themselves show anisotropy .
[ m - theta ] ( a)(d ) show the angular dependence of the in - plane magnetization .
a two - fold symmetry is found at all fields and temperatures .
the normalized in - plane anisotropy @xmath33 , where @xmath34 is the mean value of @xmath35 , i.e. , @xmath36 , is shown in fig .
[ m - theta ] ( e ) .
the magnetization along the @xmath12 axis is roughly two times larger than that along the @xmath11 axis . as we already noted
that the nd@xmath3 spins prefer to align in the @xmath30 plane , the @xmath12 axis is thus determined to be the easy axis of nd@xmath3 spins
. we also show @xmath37 with fields along different in - plane directions in fig .
[ m - theta ] ( f ) .
@xmath37 at different in - plane angles are parallel to each other at @xmath38 k , as expected for paramagnetism with the same effective moments but different weiss temperatures .
the angular dependence of weiss temperatures @xmath27 is shown in fig .
[ m - theta ] ( g ) .
magnetic anisotropy , in particular magnetization anisotropy in paramagnets were previously reported in rare earth compounds @xcite .
the primary origin of this anisotropy is the crystal electric field .
when an ion is placed within a crystal lattice , the degeneracy of the hund s rule ground state multiplet can be lifted by the electric field from the surrounding atoms . in a rare earth compound such as nto
, the cystal field can overcome the external field and pull the total moment resulting from the spin - orbit coupling into a certain direction .
the anisotropic exchange coupling also contributes to the magnetic anisotropy . to separate the contribution of crystal field from exchange to the magnetic anisotropy ,
a series of diluted samples nd@xmath7la@xmath8ti@xmath0o@xmath1 were studied .
nonmagnetic la@xmath3 has very close ionic radius to that of nd@xmath3 , it dilutes the nd@xmath3 spins and keeps the lattice almost unchanged .
thus , in a good approximation , the crystal field does not change with dilution while the exchange interaction decreases systematically with increasing la@xmath3 content .
because the weiss temperature of the system reaches as low as @xmath39 k for nd@xmath40la@xmath41ti@xmath0o@xmath1 , and even lower for samples with lower nd@xmath3 content , which makes it difficult to realize a reliable fitting using the curie - weiss law with the @xmath42 data in the temperature range of 2 300 k , only the data of nd@xmath40la@xmath41ti@xmath0o@xmath1 is presented here .
the nd@xmath3 content is determined indirectly by calculating the effective moment of the powder sample . by writing down the magnetic hamiltonian
, it is possible to extract both exchange and crystal field contributions to the angular dependent magnetization by assuming a constant crystal field contribution for the la@xmath3 diluted sample . following the approach in ref .
@xcite , the magnetic hamiltonian consists of two parts , the exchange energy @xmath43 and the crystal field energy @xmath44 ( the dipolar interaction can be assimilated in the exchange energy @xcite ) . the exchange interaction is anisotropic and can be expressed as @xmath45 where @xmath46 denote the three crystalline directions , and @xmath47 , @xmath48 and @xmath49 are the exchange constants for nd@xmath3 spins along @xmath50 and @xmath12 directions , respectively .
the crystal field part , for the low - symmetry @xmath5 site of the nd@xmath3 , is known to be @xmath51 where @xmath52 are crystal field parameters , and @xmath53 are stevens equivalent operators . it is known that the high temperature susceptibility of a system with only one spin species can be expressed as @xcite @xmath54 where @xmath55 is the curie constant , @xmath12 is the percentage of the magnetic ions on the spin lattice , in nd@xmath7la@xmath8ti@xmath0o@xmath1 case , @xmath56 .
@xmath57 and @xmath58 are the exchange tensor and the crystal field tensor , and can be expressed by the parameters in eq .
[ eq - exchange ] and eq .
[ eq - cf ] respectively
. when the three axes of the crystal are of at least second - order symmetry ( symmetry no lower than @xmath59 ) , eq .
[ eq - ki-1 ] can be expressed with the parameters in eq .
[ eq - exchange ] and [ eq - cf ] along each crystalline direction as @xmath60 here , @xmath61 denotes the crystalline direction @xmath10 , @xmath11 or @xmath12 , @xmath62 are the exchange constants as already appeared in eq .
[ eq - exchange ] , and @xmath63 is the @xmath61 direction component of tensor @xmath58 . in specific symmetries , @xmath63 can be expressed explicitly with the crystal field parameters @xmath52 .
the weiss temperature is then @xmath64 however , in the nto case , the symmetry of nd@xmath3 site is barely @xmath5 , an explicit expansion of @xmath58 is thus difficult .
therefore only the exchange parameter @xmath4 is extracted and the second term in eq .
[ eq - theta ] is left as a single parameter @xmath65 to be determined .
the results of the curie - weiss fitting for diluted samples suggest that the contributions to the weiss temperature from the exchange interactions @xmath66 and crystal field interactions @xmath67 are of opposite signs .
the fitted weiss temperature @xmath27 , and the contributions from exchange interaction and crystal field @xmath68 and @xmath65 in all three crystallographic orientations for nto and nd@xmath40la@xmath41ti@xmath0o@xmath1 are listed in table i. with these weiss temperatures , the exchange constants are determined to be @xmath69 k , @xmath70 k and @xmath71 k. it can be seen that the exchange constants are positive in all crystallographic orientations .
this illustrates the dominant role of the crystal field contribution to the magnetic anisotropy .
while the exchange interactions are always ferromagnetic , the apparent exchange deduced from the weiss temperature in different directions can be either positive or negative .
this is because the crystal field contributes a large negative part to the weiss temperature .
as seen from table i , @xmath65 is a couple of times larger in magnitude than @xmath68 along @xmath10 and @xmath11 axes and comparable in magnitude along the @xmath12 axis , making the weiss temperature large and negative along @xmath10 , @xmath11 directions and small but positive along the @xmath12 direction .
it is also worth to note that eq .
[ eq - ki ] is independent of the magnetic field .
indeed , we see that in fig .
[ m - theta ] ( e ) , the normalized in - plane anisotropy @xmath33 is nearly constant for different fields at the same temperature . only when spins are polarized significantly by the field , i. e. , m(h ) curve is not linear anymore , @xmath33 at different fields bifurcate , as is the case at 2 k. as we have discussed , the monoclinic nto does not accommodate geometrical frustration .
however , with the absence of a magnetic transition down to at least 2 k , and the average weiss temperature @xmath72 k , the frustration index as defined by @xmath73 ( ref .
@xcite ) of this system reaches about 21 , and is even higher along @xmath10 and @xmath11 directions . the large frustration index may indicate an unusual ground state and interesting spin freezing processes .
these investigations are currently under way .
the magnetic ground state of @xmath74 nd@xmath3 on the @xmath5 site in the @xmath6 space group should split into five kramers doublets due to the crystal field .
calculations on the nd@xmath3 in the same symmetry in other compounds indeed revealed the five doublets scheme @xcite , while a detailed calculation on nto is still lacking .
the specific heat measurement provides more information on the magnetic ground state .
[ specific - heat ] ( a)(c ) show the magnetic specific heat of nto for h @xmath75 a , h @xmath75 bc and that of nd@xmath40la@xmath41ti@xmath0o@xmath1 for h @xmath75 a at different fields .
the magnetic specific heat is obtained by subtracting the lattice contribution using the specific heat data of a nonmagnetic la@xmath0ti@xmath0o@xmath1 single crystal .
the zero - field specific heat in principle detects any magnetic phase transitions , and the magnetic entropy associates directly with the ground state degeneracy . in fig .
[ specific - heat ] ( a ) , the zero - field magnetic specific heat shows no anomaly for a magnetic transition , which is consistent with the paramagnetic behavior above 2 k. there is a progressive upturn as @xmath76 decreases , which signals the presence of a large portion of the magnetic entropy below 2 k. the magnetic entropy is therefore difficult to integrate due to the limited temperature range .
however , this problem is bypassed by using the specific heat in a magnetic field . as shown in fig .
[ specific - heat ] ( a ) , @xmath77 in a field shows a broad peak with the magnitude of around 3.6 j / mol@xmath78 k at low temperatures , and the maximum moves to higher temperatures with its magnitude almost unchanged as the field increases .
these indicate that the peaks are schottky anomalies associated with the ground state doublet splitting in the presence of a magnetic field .
the two - level schottky specific heat is , @xmath79 where @xmath80 is the gas constant , and @xmath81 is the energy gap between the two levels .
[ eq3 ] describes the data nicely as shown by the solid lines in fig .
[ specific - heat ] ( a ) .
it can be seen in fig .
[ specific - heat ] ( d ) that the fitting parameter @xmath81 scales linearly with magnetic field .
this further proves that the gap between the two schottky levels are caused by the zeeman splitting of the ground state doublet . using @xmath82 , the @xmath2 factor for nto with @xmath31 along @xmath10
is found to be 1.99 .
the magnetic specific heat is also studied in nd@xmath0ti@xmath0o@xmath1 with @xmath23 and in nd@xmath40la@xmath41ti@xmath0o@xmath1 with @xmath22 . for nto with @xmath23 ,
the @xmath2 factor is 3.10 .
the anisotropy in @xmath2 factor again points to the presence of strong crystal field , and these @xmath2 values should be corroborated with future crystal field calculations .
in conclusion , we have studied both single- and poly - crystalline nd@xmath0ti@xmath0o@xmath1 with angle - dependent dc susceptibility and specific heat . the anisotropic paramagnetism is observed and is mainly attributed to the crystal field effect .
the easy axis is determined to be the @xmath12-axis in the @xmath30 plane .
the contributions to weiss temperature from exchange interactions and crystal field interactions are separated .
the exchange interactions are ferromagnetic , while the crystal field contributes a large negative part to the weiss temperature , along all three crystallographic directions .
this leads to large negative @xmath27 values along @xmath10 and @xmath11 axes and small positive @xmath27 along the @xmath12 axis .
the @xmath74 ground state multiplet of nd@xmath3 splits into five kramers doublets .
the ground state doublet further splits into two singlets in a magnetic field , and this two - level ground state scheme is revealed by the magnetic specific heat .
@xmath2 factors show anisotropy too , again due to the crystal field .
these studies provide solid foundations for further investigations of this system , including the magnetic ground state and the spin freezing processes .
this work is supported by the nsfc ( grant no .
10634030 ) , us nsf dmr-0547036 , the fundamental research funds for the central universities of china ( grant no .
2010qna3026 ) , and national basic research program of china ( grant no .
2007cb925001 ) . shcherbakova
l g , mamsurova l g , and sukhanova g e , _ russ .
rev . _ * 48 * , 228 ( 1979 ) .
winfield g , azough f , and freer r , _ ferroelectrics _ * 133 * , 181 ( 1992 ) .
gardner j s , gingras m j p , and greedan j e , _ rev .
* 82 * , 53 ( 2010 ) .
bramwell s t , giblin s r , calder s , aldus r , prabhakaran d , and fennell t , _ nature _ * 461 * , 956 ( 2009 ) .
moessner r and ramirez a r , _ physics today _ * 59 * , 24 ( 2006 ) .
harris m j , bramwell s t , mcmorrow d f , zeiske t , and godfrey k w , _ _ * 79 * , 2554 ( 1997 ) . ramirez a p , hayashi a , cava r j , siddharthan r , and shastry b s , _ nature _ * 399 * , 333 ( 1999 ) .
bramwell s t and gingras m j p , _ science _ * 294 * , 1495 ( 2001 ) .
gardner j s , gaulin b d , berlinsky a j , waldron p , dunsiger s r , raju n p , and greedan j e , _ _ * 64 * , 224416 ( 2001 ) .
mirebeau i , goncharenko i n , cadavez - peres p , bramwell s t , gingras m j p , and gardner j s , _ nature _ * 420 * , 54 ( 2002 ) .
cao h , gukasov a , mirebeau i , bonville p , and dhalenne g , _ _ * 101 * , 196402 ( 2008 ) snyder j , ueland b g , slusky j s , karunadasa h , cava r j , mizel a , and schiffer p , _ _ * 91 * , 107201 ( 2003 ) xing h , he m , feng c , guo h , zeng h , and xu z a , _ _ * 81 * , 134426 ( 2010 ) .
cho b k , harmon b n , johnston d c , and canfield p c , _ _ * 53 * , 2217 ( 1996 ) .
skumryev v , kuzmin m d , gospodinov m , and fontcuberta j , _ _ * 79 * , 212414 ( 2009 ) .
boutron p , _ _ * 7 * , 3226 ( 1973 ) . ramirez a p , _ annual review of materials science _ * 24 * , 453 ( 1994 ) .
agnieszka m , gajek z , karbowiak m and rudowicz c , _ j. phys .
: condens matter _ * 20 * , 385205 ( 2008 ) .
.the weiss temperatures @xmath27 for nd@xmath0ti@xmath0o@xmath1 and nd@xmath40la@xmath41ti@xmath0o@xmath1 along @xmath10 , @xmath11 and @xmath12 directions .
the fitting errors are estimated to be within 3 k. the contributions from the exchange interactions and crystal field interactions are isolated as @xmath68 and @xmath65 using eq .
[ eq - theta ] . [ cols="^,^,^,^,^,^,^,^,^,^",options="header " , ] ( color online ) ( a ) the x - ray diffraction pattern of the polycrystals obtained by crushing nd@xmath0ti@xmath0o@xmath1 single crystals .
most peaks are indexed in the monoclinic structure in the space group @xmath6 , the index of some small peaks are not marked explicitly .
( b ) the x - ray diffraction pattern of a nd@xmath0ti@xmath0o@xmath1 single crystal .
crystals cleave naturally along the ( 100 ) direction .
( c ) the rocking curve of the ( 400 ) peak with fwhm = 0.09@xmath19 indicates good crystallinity . ]
( color online ) ( a1)-(c1 ) : zero - field - cooled and field - cooled susceptibility vs temperature at 1000 oe for @xmath25 , @xmath22 of a nd@xmath0ti@xmath0o@xmath1 single crystal and a polycrystal sample , respectively .
insets are fittings by the the curie - weiss law .
( a2)-(c2 ) : isothermal magnetization curves at 2 , 10 , 150 k for @xmath25 , @xmath22 and polycrystal respectively .
] ( color online ) ( a)-(d ) : the angular - dependent in - plane magnetization @xmath35 of nd@xmath0ti@xmath0o@xmath1 at 2 , 10 , 100 , 300k in the field of 0.5 , 2 , 5 t. ( e ) : the angular dependence of the normalized magnetization @xmath83 , where @xmath34 is the mean value of @xmath35 .
( f ) the temperature dependence of the inverse susceptibility of a nd@xmath0ti@xmath0o@xmath1 single crystal with the in - plane field along @xmath84 @xmath85 , @xmath86 , @xmath87 , @xmath88 , @xmath89 , @xmath90 , @xmath91 , @xmath92 and @xmath93 . (
g ) : the in - plane angular dependence of the weiss temperature by fitting @xmath37 above 200 k in ( f ) .
dashed lines labeled with @xmath11 and @xmath12 denote the in - plane crystalline @xmath11 and @xmath12 directions as determined by the laue diffraction . ]
( color online ) magnetic specific heat of ( a ) a nd@xmath0ti@xmath0o@xmath1 single crystal with @xmath22 ; ( b ) a nd@xmath0ti@xmath0o@xmath1 single crystal with @xmath23 ( the field is not applied intentionally along a specific in - plane direction ) ; ( c ) nd@xmath40la@xmath41ti@xmath0o@xmath1 with @xmath22 .
solid lines in ( a)-(c ) are two - level schottky fittings ; ( d ) the energy gap @xmath81 obtained by the two - level schottky fittings . ] | the anisotropic paramagnetism and specific heat in nd@xmath0ti@xmath0o@xmath1 single crystals are investigated .
angular dependence of the magnetization and weiss temperatures show the dominant role of the crystal field effect in the magnetization . by incorporating the results from the diluted samples , contributions to weiss temperature from exchange interactions and crystal field interactions
are isolated .
the exchange interactions are found to be ferromagnetic , while the crystal field contributes a large negative part to the weiss temperature , along all three crystallographic directions .
the specific heat under magnetic field reveals a two - level schottky ground state scheme , due to the zeeman splitting of the ground state doublet , and the @xmath2-factors are thus determined .
these observations provide solid foundations for further investigations of nd@xmath0ti@xmath0o@xmath1 . |
according to the hierarchical clustering scenario such as the cold dark matter cosmology , it is believed that clusters of galaxies ( cgs ) are formed through subcluster merger and/or absorption of smaller groups .
mergers of cgs are the most energetic events in the universe after the big bang , where the total kinetic energy of the two subclusters reaches @xmath2 ergs .
thus , it is most likely that they affect various properties of cgs . in intracluster medium ( icm ) , mergers make strong bulk - flow motion and shocks . these cause characteristic morphology of x - ray images ( elongation and/or substructures ) and complex electron temperature structures in icm ( e.g. , honda et al .
1996 ; churazov et al .
1998 ; donnelly et al .
1998 ; donnelly et al .
1999 ; markevitch , sarazin , & vikhlinin 1998 ; watanabe et al . 1999 ; watanabe , yamashita , & furuzawa 1999 ) .
in addition , rapid change of physical properties of icm probably induces nonequilibrium plasma ionization ( hanami et al .
1999 ) . for galaxies ,
mergers change drastically their environments through gravitational tidal field ( bekki 1999 ) and static and ram pressure of icm ( evrard 1991 ; fujita et al .
these may affect star formation activities of member galaxies ( e.g. , caldwell et al . 1993 ; tomita et al .
1996 ; wang , ulmer , & lavery 1997 ) . furthermore , high energy phenomena of cgs such as radio halos ( e.g. , giovannini et al . 1993 ; deiss et al .
1997 ; rttgering et al . 1997 ) and hard x - ray emission ( fusco - femiano et al .
1999 ) are well correlated to merger phenomena .
this suggests a part of huge energy of mergers is transported to acceleration of high energy relativistic particles ( sarazin 1999 ) and amplification of magnetic field , though their detailed physical processes are still unclear .
since cgs evolve in cosmological timescale and their initial conditions depend on the cosmology , cosmological n - body + hydrodynamical simulations are suitable to investigate the formation and evolution of cgs in general ( e.g. , evrard 1990 ; navarro , frenk , & white 1995 ; bryan & norman 1998 ; eke , navarro , & frenk 1998 ; suginohara & ostriker 1998 ; yoshikawa , itoh , & suto 1998 ) . however , the simulations starting from rather idealized initial conditions are more suitable to investigate mergers in detail ( roettiger , burns , & loken 1996 ; ishizaka & mineshige 1996 ) .
schindler & mller ( 1993 ) found that characteristic temperature structures occur in merging clusters through shock heating and adiabatic compression and expansion .
ishizaka ( 1997 ) found that the specific energy ratio of icm and galaxies , @xmath3 , is a good indicator of merging clusters and that it can be used to determine the phase of the mergers .
takizawa ( 1999 ) performed n - body and hydrodynamical simulations explicitly considering the relaxation process between ions and electrons and find that electron temperature distribution becomes significantly different than ion temperature one in later stages of mergers .
recently , roettiger , burns , & loken ( 1999 ) performed n - body + magnetohydrodynamical simulations to investigate evolution of magnetic field in icm during mergers .
some of the observational results ( e.g. , a754 in henriksen & markevitch 1996 ; a3395 in markevitch et al . 1998 ) imply off - center mergers .
furthermore , cosmological n - body simulations show collapsed dark halos have angular momentum which correspond to @xmath4 ( barnes & efstathiou 1987 ; ueda et al .
1994 ) , where @xmath5 is the ratio between the actual angular velocity and the angular velocity needed to provide rotational support ( see binney & tremaine 1987 ) .
it is probable that dark halos obtain a part of the angular momentum through off - center major mergers .
thus , it is an quite interesting problem to study evolution of off - center mergers and their observational consequences .
ricker ( 1998 ) studies gas dynamical evolution of off - center mergers in detail using the high resolution peacewise - parabolic method
. however their simulations do not contain the collisionless n - body component .
this is very problematic because dark matter and galaxies dominate gravity in typical cgs .
in addition , two - temperature nature of icm is not considered there .
thus , we perform n - body and hydrodynamical simulations explicitly considering the relaxation process between ions and electrons to investigate the dynamical evolution and observational consequences of off - center mergers of cgs . the rest of this paper is organized as follows . in
[ s : sim ] we describe the adopted numerical methods and initial conditions for our simulations . in [ s : res ] we present the results . in [ s : sd ] we summarize the results and discuss their implications .
in the present study , we consider cgs consisting of two components : collisionless particles corresponding to the galaxies and dark matter ( dm ) , and two - temperature gas corresponding to the icm .
when calculating gravity , both components are considered , although the former dominates over the latter .
radiative cooling and heat conduction are not included . the numerical method used here is fully described in 3.1 of takizawa ( 1999 ) .
thus , we briefly review the outline of the method in this subsection .
we used the smoothed - particle hydrodynamics ( sph ) method to solve the hydrodynamical equations for the gas component ( see monaghan 1992 ) .
as the standard sph codes for one - temperature fluid , we solve the continuity equation , the momentum equation , and the thermal energy equation for whole plasma with artificial viscosity to treat the shocks .
in addition to these equations , we solve one more equation for the normalized electron temperature , @xmath6 , where @xmath7 is the electron temperature and @xmath8 is the plasma mean temperature .
we assume that artificial viscous heating is effective only for ions and that only the coulomb coupling is considered in the relaxation process .
gravitational forces are calculated by the barnes - hut tree algorithm ( barnes & hut 1986 ) and softened using the plummer potential profile .
we set the softening length @xmath9 one - tenth of the initial core radius of the smaller subcluster in the simulation .
tree structure is also used to search for nearest neighbors in sph calculations ( hernquist & katz 1989 ) .
we consider mergers of two virialized subclusters of galaxies with masses @xmath10 and @xmath11 .
the initial configuration of each subcluster is the same as in takizawa ( 1999 ) .
the spatial distribution of dm in each subcluster is represented by the king distribution with core radii @xmath12 and @xmath13 .
we assume the velocity distribution of the dm particles to be an isotropic maxwellian .
we assume that the initial icm temperature is isothermal and equal to the virial temperature .
the icm is initially in hydrostatic equilibrium within the cluster potential of the dm and the icm , itself .
we set the initial conditions as follows .
two subclusters are initialized in the @xmath14plane , separated by a distance @xmath15 in the @xmath16direction and a distance @xmath17 ( the impact parameter ) in the @xmath18direction . the initial relative velocity , @xmath19 is directed along the @xmath16axis .
the coordinate system is taken in such a way that the center of masses is at rest in the origin . the total gas mass fraction @xmath20 is set to be 0.1 . in both run a1 and a2 , two subclusters have the same masses ; @xmath21 .
we set the core radii to be @xmath22 mpc .
each subcluster consists of 5000 collisionless particles and 5000 sph particles .
for the both runs , the binding energy between the two subclusters is the same as in run a of takizawa ( 1999 ) .
we set @xmath23 mpc .
then , the impact parameter @xmath17 and initial relative velocity @xmath24 are taken in such a way that @xmath25 for run a1 and @xmath26 for run a2 , respectively .
the parameter @xmath27 is the ratio between the actual angular velocity and the angular velocity needed to provide rotational support , where @xmath28 is the angular momentum of the two subclusters around the center of masses , @xmath29 is the binding energy between the two , @xmath30 is the gravitational constant , and @xmath31 is the total mass ( see binney & tremaine 1987 ) . in both run b1 and b2 , the mass ratio is @xmath32 .
the larger subcluster has the same mass as that in run a1 and a2 .
the particle numbers of the smaller one are one - fourth of the larger one . for the both runs ,
the binding energy between the two subclusters is the same as in run b of takizawa ( 1999 ) .
we set @xmath33 mpc .
again , the impact parameter @xmath17 and initial relative velocity @xmath24 are taken in such a way that @xmath25 for run b1 and @xmath26 for run b2 , respectively . the parameters in our calculations
are summarized in table [ tab : para ] .
first of all , let us see the evolution of electron temperature ( @xmath7 ) distribution and x - ray surface brightness distribution , which are important observational quantities in x - ray to investigate the dynamical properties of merging cgs . figure [ fig : xrsbtea1 ] shows the snapshots of x - ray surface brightness ( contours ) and emissivity - weighted @xmath7 ( colors ) distribution of run a1 viewed along the @xmath34-axis ( column 1 ) , @xmath35-axis ( column 2 ) , and @xmath36-axis ( column 3 ) at four epochs ( rows 1 - 4 ) during the merger .
the times relative to the most contracting epoch are listed above each panel : @xmath37 , @xmath38 , @xmath39 , and @xmath40 gyr .
x - ray surface brightness contours are equally spaced on a logarithmic scale and separated by a factor of 7.4 .
the blue , green , yellow , and red colors correspond to @xmath41 kev , 8 kev , 12 kev , and 16 kev , respectively .
each panel is 4 mpc on a side .
when we observe the cluster along the z - axis ( column 1 ) , which is nearly perpendicular to the collision axis , the morphology characteristic to the off - center merger is clarified .
when two subclusters approach each other , we see double peaks in the x - ray image of `` one cluster '' at @xmath37 gyr . just between the peaks electron temperature rises up to @xmath42 kev due to the shock .
the shock front is oblique to the collision axis due to the angular momentum between the subclusters . at the most contracting epoch ( @xmath43 gyr ) ,
two peaks merge to one peak in the x - ray image and the image elongates to the directions nearly perpendicular to the collision axis .
then the cluster expands and two shocks propagate in the opposite directions ( @xmath44 gyr ) .
since the system has angular momentum , the shape of the expanding shocks are similar to the arms of the spiral galaxies . however , due to the limitation of relaxation between ions and electrons , the post - shock hot regions have not so high electron temperature as in the contracting phase ( see [ s : tniep ] ) . on the other hand , when we observe the cluster along the y - axis ( column 2 ) , it looks like a head - on merger such as run a in takizawa ( 1999 ) .
the @xmath7 between the two peaks at @xmath37 gyr becomes lower than when viewed along the z - axis , because the both foreground and background cooler gas contaminates the hot gas at the shock .
this lead to the underestimation of the collision velocity though the morphology of this epoch is very simple and seems to be suitable for the estimation ( and we will discuss this in [
s : sd ] ) .
furthermore , when we observe the cluster along the x - axis ( column 3 ) , which is nearly along the collision axis , the morphology is rather different .
it looks like a spherical symmetric cluster both in the x - ray surface brightness and electron temperature distribution . except for the most contracting epoch ( @xmath43 gyr )
, there is no distinct temperature structure . at the most contracting epoch ,
temperature distribution is spherical symmetric but have a rather large gradient ( and we will discuss this in
[ s : sd ] ) .
since icm is optically thin , information along the line - of - sight direction is emissivity - weighted integrated and lost in large part . in this case , line - of - sight velocity distribution of galaxies is more suitable for analyzing the cluster dynamical properties ( see [ s : lvdg ] ) .
in addition , line - of - sight velocity of icm may also provide us with useful information , which will become observable with the x - ray spectrometer ( xrs ) after the launch of astro - e .
next , we describe the evolution of x - ray surface brightness and @xmath7 distribution of run b1 .
figure [ fig : xrsbteb1 ] shows the same conditions as figure [ fig : xrsbtea1 ] , but for run b1 .
the times are listed above each panel : @xmath37 , @xmath38 , @xmath39 , and @xmath45 gyr .
temperature color scale is adjusted for run b1 .
the blue , green , yellow , and red colors correspond to @xmath46 kev , 5 kev , 7.5 kev , and 10 kev , respectively .
as in run a1 , characteristic features of off - center mergers are most clearly seen when the cluster is viewed along the z - axis ( column 1 ) .
when two subclusters approach each other , the bow shock with an arc shape is formed just between the two as in head - on collisions . however , since the direction of the smaller subcluster s velocity is not coincident to the direction connecting the two , the shock is stronger in the upper region of the panel ( @xmath37 gyr )
. then the two peaks merge to one triangle image ( @xmath43 gyr ) .
the hot region associated to the bow shock is seen elongated slantingly backward with respect to the motion of the smaller subcluster . however
, this hot region is located not in the center of the cluster but in the upper half of the image due to an off - center collision .
then the gas expands and the two shocks propagate outward ( @xmath44 gyr ) .
as in run a1 , the expanding shocks form spiral - like structure and electron temperature there is significantly lower than the plasma mean temperature ( see [ s : tniep ] ) .
when we observe the cluster along the y - axis ( column 2 ) , its morphological feature in x - ray surface brightness is similar to the that of head - on merger as in run a1 .
however , this is not the case in the electron temperature distribution .
this is prominent especially in the expanding phase ( @xmath44 gyr ) .
the hot region located in the backward with respect to the motion of the smaller cluster is hardly seen .
since this hot component is not so spread , the foreground and background cooler gas contaminates it .
when we observe the cluster along the x - axis ( column 3 ) , non spherical structure can be seen on the contrary to the run a1 . at @xmath37
the geometry is very simple and seems to be very suitable for estimation of the collision velocity .
however , this lead to the underestimation of the collision velocity as in run a1 . as shown by takizawa ( 1999 )
, electron temperature distribution is significantly different than the plasma mean one in the expanding phase of mergers .
let us examine two - temperature nature of icm for each model .
figure [ fig : twtem ] shows the snapshots of electron temperature ( @xmath7 ) distribution ( column 1 ) , the plasma mean temperature ( @xmath8 ) one ( column 2 ) , and the normalized electron temperature ( @xmath6 ) one ( column 3 ) for each model at the expanding phase ( 0.25 gyr after the most contracting epoch ) viewed along the z - axis .
x - ray surface brightness distribution ( contours ) is overlaid for the panels in column 1 and 2 .
run a1 , a2 , b1 , and b2 are showed in row 1 , row 2 , row 3 , and row 4 , respectively . temperature color scale for @xmath7 and @xmath8 ( column 1 and 2 ) in row 1 and 2 ( for run a1 and a2 ) , and row 3 and 4 ( for run b1 and b2 )
are the same as in figure [ fig : xrsbtea1 ] and [ fig : xrsbteb1 ] , respectively . for @xmath47 distribution ( column 3 ) , the red , yellow .
green , and blue colors correspond to @xmath48 , and @xmath45 , respectively . for all models , spiral - shaped shocks are recognized . as in head - on mergers ( takizawa 1999 ) , electron temperature in the post shock regions is significantly lower than the plasma mean one in this phase
. however , even in the electron temperature distribution , spirally high temperature regions can be seen .
note that these regions are located not at the shocks , but @xmath49 mpc behind the shocks . in the higher angular momentum cases ( run a2 and b2 ) ,
double - peak structure can be seen even in the expanding phase whereas it can be seen only in the pre - merger contracting phase in head - on collisions .
thus , a cluster which has double peaks in x - ray surface brightness distribution and hot temperature regions not between the two peaks but around them , is most likely a candidate of a off - center collision .
this is the case of a3395 ( markevitch et al .
1998 ) . this will be discussed in [ s : sd ] .
as described in
[ s : xietmra1 ] and [ s : xietmrb1 ] , when we observe merging clusters nearly along the collision axis , line - of - sight ( los ) velocity distribution of galaxies has more useful information than the x - ray images .
let us discuss this issue . in the argument
below , we assume that dm particles distribution in velocity space well represents that of galaxies .
first , we describe the evolution of los velocity distribution of galaxies in run a1 .
figure [ fig : vhga1 ] shows the histograms of los velocities of n - body particles of run a1 viewed along the x - axis at four epochs .
for the pre - merger phase ( @xmath37 gyr ) , two components associated to the subclusters can be clearly recognized .
then at the most contracting epoch the distribution becomes boxier than the gaussian ( @xmath43 gyr ) .
this is confirmed from the evolution of the kurtosis ( fig .
[ fig : vmoma1 ] ) .
however , in the expanding phase , the distribution becomes closer to gaussian . to see the evolution of the distribution
quantitatively , we show the evolution of the skewness and kurtosis of the los velocity distribution viewed along the x - axis ( open triangles and solid lines ) , y - axis ( open squares and dotted lines ) , and z - axis ( filled square and short dashed lines ) , respectively . due to the symmetry of the equal mass collision
, it is quite natural that the skewness is almost zero in all directions . on the other hand ,
the kurtosis of the velocity distribution along the x - axis becomes significant negative values especially in the contracting phase although the kurtosis of the other components is almost nearly zero .
next , we investigate the behavior of run b1 .
figure [ fig : vhgb1 ] shows the same conditions as figure [ fig : vhga1 ] but for run b1 . in the contracting phase , a tail associated to the smaller subcluster is visible toward negative velocities .
this is more clarified in the skewness evolution in figure [ fig : vmomb1 ] , which is the same conditions as figure [ fig : vmoma1 ] but for run b1 . due to this tail component
the skewness becomes significantly large negative values .
note that the sign of the skewness depends on the relative position among two subclusters and the observer .
again , a kurtosis of the velocities along the x - axis becomes negative in the contracting phase although this is not so prominent as in run a1 . in the above cases , we use the all n - body particles ( @xmath50 for run a1 and @xmath51 for run b1 ) to calculating the moments of los velocity distribution . in real cases , however , it is difficult to measure los velocities for such numbers of galaxies . on the other hand ,
the standard deviations of a skewness and a kurtosis for a gaussian distribution are approximately @xmath52 and @xmath53 , respectively .
thus , for run a1 , we need @xmath54 galaxies los velocities to obtain a statistically significant negative kurtosis .
if the number of galaxies is only @xmath55 , it is difficult to distinguish a negative kurtosis due to mergers from that due to poor statistics . for run b1
, @xmath56 is required to detect a statistically significant negative kurtosis . on the other hand ,
we need @xmath54 galaxies to obtain a statistically significant skewness .
this will be discussed in [ s : sd ] .
we investigate evolution of cgs and their observational consequences during off - center mergers explicitly considering the relaxation process between the ions and electrons . in the contracting phase
, a bow shock occurs between the two subclusters .
x - ray images have two peaks and the high temperature region is located between the two . however , the observed temperature between the two peaks depends on the viewing angle . in the expanding phase spiral - shaped two shocks
propagate outward .
spatial distribution of icm electron temperature is significantly different in this phase as in head - on collisions .
morphological features in x - ray images and temperature distribution characteristic to off - center mergers are seriously affected by the viewing angle .
when we observe cgs during off - center merger in some directions , we can not distinguish them from head - on mergers only through x - ray images and temperature maps .
when we observe merging cgs nearly along the collision axis , los velocity distribution of galaxies has more useful information than x - ray images and temperature maps of icm . in the contracting phase ,
the distribution is boxier and have a negative kurtosis .
this is prominent when the two subclusters have nearly equal masses .
when the two subclusters have rather different masses , the skewness of the distribution is more suitable for a indicator of mergers in the contracting phase . to obtain statistically significant values of these moments ,
we need @xmath54 galaxies los velocities .
recently , markevitch et al . ( 1999 ) and kikuchi et al . ( 1999 ) estimated subcluster collision velocities in cygnus a and virgo using the electron temperature maps obtained by asca , respectively .
both of them have simple geometry ; hot temperature region is located just between the two peaks of x - ray images .
however , if we do not observe cgs just perpendicular to the collision axis , the observed temperature between the peaks is lower than the true temperature at the shock .
this is due to the contamination of cooler icm located foreground and background .
therefor , this method can lead to underestimation of the collision velocities .
we should regard their results as the minimum limit .
however , since estimated velocities through electron temperature maps depend on the true collision velocities and the viewing angles , they are related to the transverse components of the collision velocities . on the other hand , los components of collision velocities will be observable with xrs after the launch of astro - e .
therefor , we will be able to estimate true collision velocities combining the results of both electron temperature maps and los velocities .
when we observe merging clusters around the most contracting phase nearly along the collision axis through x - ray , they seem to be relaxed spherical ones in x - ray images .
in addition , the @xmath3 is nearly unity only around this epoch though a large @xmath3 value is a good indicator of mergers when observed along the collision axis ( ishizaka 1997 ; takizawa 1999 ) .
thus , it is difficult to resolve whether they are merging clusters or not .
one of their characteristic features is a large temperature gradient .
markevitch ( 1996 ) found that some of clusters have a larger radial gradient in electron temperature than that expected from the self - similar solution ( bertschinger 1985 ) and numerical simulations ( eg ; navarro et al .
1995 ; takizawa & mineshige 1998 ) in the plasma mean temperature .
one explanation to solve this discrepancy is energy nonequipartition of electrons and ions in the outer regions ( fox & loeb 1997 ; chize , alimi & teyssier 1998 ; ettori & fabian 1998 ; takizawa 1998 ) .
however , it is probable that some of them are hidden merging clusters as mentioned above .
again , measurement of icm los velocities with xrs of astro - e will clarify such hidden merging clusters . when the angular momentum of the systems are relatively high ( @xmath57 ) , double - peak structures in x - ray images survive after the most contracting epoch
. we can not find whether they are in a contracting or expanding phase only through x - ray surface brightness distribution .
however , electron temperature distribution enable us to determine the phase of mergers . if high temperature regions are located not between but around the peaks like a3395 ( markevitch et al . 1998 ) , they are in the expanding phase of off - center mergers .
if high temperature regions are between the peaks like cygnus a ( markevitch et al . 1999 ) , on the other hand , they are in contracting phase . if we assume the luminosity function of galaxies is the schechter form ( see binney & merrifield 1998 ) with the normalization through a typical galaxy number density in rich clusters , the absolute @xmath58 magnitude of the 1000-th luminous galaxy is @xmath59 .
this corresponds to @xmath60 in the apparent @xmath58 magnitude for an object at @xmath61 when the hubble constant is 75 km @xmath62 @xmath63 .
this is nearly equal to the survey limit of 2df ( @xmath64 ) .
thus , we will be able to discuss on merger dynamics for nearby clusters using such redshift survey data .
the author would like to thank k. ohta , t. t. takeuchi , and t. t. ishii for valuable discussion on the observational feasibility of galaxies los velocities in clusters .
the author also thanks k. masai and n. y. yamasaki for helpful comments and s. mineshige for continuous encouragement .
barnes , j. , & hut , p. 1986
, nature , 324 , 446 barnes , j. , & efstathiou , g. 1987 , apj , 319 , 575 bertschinger , e. 1985 , apjs , 58 , 39 bekki , k. 1999 , apj , 510 , l15 binney , j. , & merrifield , m. 1998 , galactic astronomy ( princeton : princeton university press ) binney , j. , & tremaine , s. 1987 , galactic dynamics ( princeton : princeton university press ) bryan , g. l. , & norman , m. l. , 1998 , apj , 495 , 80 caldwell , h. , rose , j. a. , sharples , r. m. , ellis , r. s. , & bower , r. g. 1993 , aj , 106 , 473 chize , j .-
p . , alimi , j .- m , & teyssier , r. 1998 , apj , 495 , 630 churazov , e. , gilfanov , m. , forman , w. , & jones , c. 1998 , apj , submitted ( astro - ph/9802166 ) deiss , b. m. , reich , w. , lesch , h. , wielebinski , r. 1997 , a&a , 321 , 55 donnelly , r. h. , markevitch , m. , forman , w. , jones , c. , david , l. p. , churazov , e. , & gilfanov , m. 1998 , apj , 500 , 138 donnelly , r. h. , markevitch , m. , forman , w. , jones , c. , churazov , e. , & gilfanov , m. 1999 , apj , 513 , 690 eke , v. r. , navarro , j. f. , & frenk , c. s. 1998 , apj , 503 , 569 ettori , s. & fabian , a. c. 1998 , mnras , 293 , l33 evrard , a. e. 1990 , apj , 363 , 349 evrard , a. e. 1991 , mnras , 248 , 8 fox , d. c. , & loeb , a. 1997 , apj , 491 , 459 fujita , y. , takizawa , m. , nagashima , m. , & enoki , m. 1999 , pasj , in press ( astro - ph/9904386 ) fusco - femiano , r. , fiume , d. d. , feretti , l. , giovannini , g. , grandi , p. , matt , g. , molendi , s. , and santangelo , a. 1999 , apj , 513 , l21 giovannini , g. , feretti , l. , venturi , t. , kim , k. t. , and kronberg , p. p. 1993 , apj , 406 , 399 henriksen , m. j. , & markevitch , m. l. 1996 , 466 , l79 hernquist , l. , & katz , n. 1989 , apjs , 70 , 419 honda , h. , hirayama , m. , watanabe , m. , kunieda , h. , tawara , y. , yamashita , k. , ohashi , t. , hughs , j. , p. , & henry , j. p. 1996
, apj , 473 , l71 ishizaka , c. , & mineshige , s. 1996 , pasj , 48 , l37 ishizaka , c. , 1997 ap & sps , 254 , 233 hanami , h. , tsuru , t. , shimasaku , k. , yamauchi , s. , ikebe , y. , & koyama , k. 1999 , apj , in press kikuchi , k. et al .
1999 , in preparation markevitch , m. 1996 , apj , 465 , l1 markevitch , m. , forman , w. r. , sarazin , c. l. , & vikhlinin , a. 1998 , apj , 503 , 77 markevitch , m. , sarazin , c. l. , & vikhlinin , a. 1999 , apj , in press ( astro - ph/9812005 ) monaghan , j. j. 1992 , ara&a , 30 , 543 navarro , j. f. , frenk , c. s. , & white , s. d. m. 1995 , mnras , 275 , 720 ricker , p. , m. 1998 , apj , 496 , 670 roettiger , k. , burns , j. o. , & loken , c. 1996 , apj , 473 , 651 roettiger , k. , stone , j. m. , & burns , j. o. 1999 , apj , in press ( astro - ph/9902015 ) rttgering , h. j. a. , wieringa , m. h. , hunstead , r. w. , and ekers , r. d. 1997 , mnras , 290 , 577 sarazin , c. l. 1999 , apj , in press ( astro - ph/9901601 ) schindler , s. , & mller , e. 1993 , a&a , 272 , 137 suginohara , t. , & ostriker , j. p. 1998 , apj , 507 , 16 takizawa , m. , & mineshige , s. 1998 , apj , 499 , 82 takizawa , m. 1998 , apj , 509 , 579 takizawa , m. 1999 , apj , in press ( astro - ph/9901314 ) tomita , a. , nakamura , f. e. , takata , t. , nakanishi , k. , takeuchi , t. , ohta , k. , & yamada , t. 1996 , aj , 111 , 42 ueda , h. , shimasaku , k. , suginohara , t. , & suto , y. 1994 , pasj , 46 , 319 wang , q. d. , ulmer , m. p. , lavery , r. j. 1997 , mnras , 288 , 702 watanabe , m. , yamashita , k. , furuzawa , a. , kunieda , h. , tawara , y. , & honda , h. 1999 , apj , submitted watanabe , m. , yamashita , k. , & furuzawa , a. 1999 , in preparation yoshikawa , k. , itoh , m. , & suto , y. 1998 , pasj , 50 , 203 | we investigate the dynamical evolution of clusters of galaxies and their observational consequences during off - center mergers , explicitly considering the relaxation process between ions and electrons in intracluster medium by n - body and hydrodynamical simulations . in the contracting phase
a bow shock is formed between the two subclusters .
the observed temperature between two peaks in this phase depends on the viewing angle even if the geometry of the system seems to be very simple like head - on collisions . around the most contracting epoch , when we observe merging clusters nearly along the collision axis , they look like spherical relaxed clusters with large temperature gradients . in the expanding phase , spiral bow shocks occur . as in head - on mergers ,
the electron temperature is significantly lower than the plasma mean one especially in the post - shock regions in the expanding phase .
when the systems have relatively large angular momentum , double - peak structures in the x - ray images can survive even after the most contracting epoch .
morphological features in both x - ray images and electron temperature distribution characteristic to off - center mergers are seriously affected by the viewing angle . when the clusters are observed nearly along the collision axis ,
the distribution of galaxies line - of - sight ( los ) velocities is a good indicator of mergers . in the contracting phase , an negative kurtosis and a large skewness
are expected for nearly equal mass collisions and rather different mass ones , respectively . to obtain statistically significant results , about @xmath0 galaxies
los velocities are required . for
nearby clusters ( @xmath1 ) , large redshift surveys such as 2df will enable us to study merger dynamics . |
we describe the calculation of the hadronic matrix elements that are required for the extraction of the @xmath1 and @xmath2 ckm matrix elements from experimental data @xcite . to reach the bottom quark mass our strategy is to interpolate between results from relativistic quarks with @xmath3 and results from lattice hqet @xcite .
here we discuss only the hqet simulations , as our clover form factor simulations have only just started .
all of our simulations use @xmath4 dynamical staggered configurations with a volume @xmath5 and @xmath6 .
the isgur - wise function is the qcd matrix element required in the extraction of @xmath1 from experimental data .
experimental measurements of the slope of the isgur - wise function vary from @xmath7 to @xmath8 , and the variations in theoretical predictions are nearly as large @xcite . initial attempts to calculate the isgur - wise function in lattice hqet had problems either with the signal to noise ratio @xcite or the renormalization factors @xcite .
the first complete calculation has been done recently by the kentucky group @xcite .
we use the same method as the kentucky group ( see also @xcite ) .
we ran at all permutations of the following velocities : @xmath9 , @xmath10 , @xmath11 and @xmath12 .
our sample size is @xmath13 configurations , and our wilson @xmath14 values are @xmath15 and @xmath16 .
a relative smearing function of @xmath17 was used between the quarks in the @xmath18 meson .
-9 mm -8 mm in fig .
[ fig : isgurwisefunc ] we plot the bare isgur - wise function for various time separations between the current and the @xmath18 source .
if the ground state has been isolated , then the isgur - wise function should be independent of this separation .
the data for @xmath19 are consistent within present errors .
it is traditional to report the slope of the isgur - wise function as a function of the dot product of the two meson velocities ( @xmath20 ) . in table
[ tb : rholinear ] we report fits to @xmath21 .preliminary fits to bare isgur - wise function data [ cols="^,^,^,^,^ " , ]
the observation of the decay @xmath0 allows a determination of @xmath2 , if the relevant qcd form factors can be calculated .
there have been a number of lattice qcd calculations of the required form factors ( see @xcite for reviews ) .
however , previous approaches suffer from the drawback that calculations are done at large @xmath23 , thus requiring a large extrapolation to @xmath24 , where measurements are currently made . to reach
a low @xmath23 requires large meson velocities not easily achieved for heavy mesons in the nrqcd- or propagating - quark - approaches . as we have shown @xcite ,
a good signal can be obtained for the hqet @xmath18 meson with a large velocity ( @xmath25 ) , so we propose the use of hqet light simulations to explore lower @xmath23 ( see @xcite for similar ideas ) .
it is not clear that hqet will be a good approximation to the dynamics of the @xmath18 meson , at these values of @xmath23 , nor that a sufficiently good signal will be obtained .
however , results should be very useful in helping to reduce the heavy quark extrapolation errors over simulations that only use clover quarks .
because @xmath26 light meson form factors have never been studied before using lattice hqet ( although the static limit was studied in @xcite ) , we have computed the matrix element for @xmath0 using hqet to check for a signal .
we use the setup described in @xcite with the heavy clover quark replaced by a hqet quark . in fig . [
fig : hqetlight ] we plot the ratio of three point functions to two point functions @xmath27 that is proportional to the @xmath28 matrix element , as a function of the operator time @xmath29 .
the @xmath18 source is fixed at @xmath30 , and the light meson source is fixed at @xmath31 .
this work is supported by the doe and the nsf .
the computations were carried out at ccs ( ornl ) .
this presentation was prepared with the kind additional support and assistance of the university of bielefeld and its zentrum fr interdisziplinre forschung . | we use simulations of heavy quark effective field theory to calculate the isgur - wise function , and we demonstrate the feasibility of calculating the matrix element for the @xmath0 decay in the lattice heavy quark effective theory ( hqet ) . |
since the direct experimental information about structure of susy model and values of susy particles masses are absent now we try to extract any possible data from the cosmology . from the recent wmap data @xcite the mass density in the universe is equal to its critical value @xmath1 ; the structured dark matter ( dm ) and
the isotropic and homogeneous dark energy ( de ) contribute to this density @xmath2 and @xmath3 , correspondingly . as it is known , baryons give @xmath4 of the total universe s mass . in our galaxy @xmath5 of total mass
is the mass of dm which forms the halo .
the lightest supersymmetric particles ( lsp ) neutralino are supposed to be the main contributors to the cold ( and hot ) dm .
the physics of lsp is discussed intensively in various aspects @xcite . to observe the neutralino appearance in space or land experiments one can use of the different ways .
for example , the high energy antiprotons , positrons or neutrino flux due to neutralino annihilation in the dark halo or in the center of the sun or the earth . in our work we have concentrated on the specific gamma - radiation @xcite which can be diffuse ( due to annihilation into fermions and bosons ) or monochromatic ( due to annihilation into @xmath6 or @xmath7 @xcite ) . to measure characteristics of these specific gamma - fluxes the special experimental programs have been elaborated with space telescopes ( egret , glast ) @xcite , @xcite .
now we have from the collider experiments only low restriction for the neutralino mass @xmath8 .
atlas and cms collaborations ( lhc ) will say something else only after 2007 .
but the absolute stability of neutralino ( or metastability ) can not be checked in collider experiments . moreover , some details of the standard cosmological scenario , which are based on the neutralino as the cdm , should be changed if neutralino s and other superparticles characteristics do not agree with the scenario .
thus we need in combined analysis of the lhc and astrophysical data to verify our ideas on cosmological and cosmogonic evolution @xcite .
neutralino in our galaxy or in other galaxies up to @xmath9 at the gravity separation stage can form massive dense clumps of different sizes @xcite , @xcite , @xcite .
due to the inner gravitational instability clumps can collapse and intensively annihilate .
our basic supposition is that the nature and origin of the rare high - energy grb can be understood as a result of the neutralino clump total annihilation .
we have shown that at the final stage of the process the powerful gamma - ray burst ( grb ) with photon energies over gev will occur . in the standard grb
the energies of photons are only fractions of mev ( see @xcite ) .
the specific energy and time distribution of this hard @xmath10-radiation can be related with the parameters of superparticles ( neutralino , in particular ) and give an important information on the susy model details and its compatibility with cosmology . as an interesting example we have considered the split higgsino scenario @xcite in the framework of famous split supersymmetry model @xcite and obtained the values of annihilation time of the neutralino clump , characteristic photon energies of the grb and the annihilation spectrum .
consideration of neutralino as a carrier of dm mass in the universe generates questions about their evolution and about astrophysical phenomenons which prove both the existence of neutralino and the reality of the evolution process with neutralino participation . according to high resolution cold dark matter simulations in the framework of the hierarchical principle of structure formation
the large virilized halos are formed through the constant merging of smaller halos produced at earlier times .
neutralino at the gravitational isolation stage @xmath9 formed dense massive clumps of different scales @xcite , @xcite .
it likes naturally that dm clumps are in the state of dynamical equilibrium as a result of compensation of own gravitational forces by forces of tidal interactions with each other . in consequence of inner instability
the clump can turn into irreversible collapse state .
the growth of density provides the growth of annihilation rate .
so neutralino clouds become more and more intense sources of relativistic particles . at final stage of this process
there is a gamma - ray burst .
specific energy and time distribution of the gamma radiation are connected with neutralino parameters and give an important information about structure of the mssm and about its correspondence with cosmology .
therefore our main statement consist in the nature of rare powerful cosmological gamma ray bursts is a result of the total annihilation of neutralino clumps .
we have analyzed the qualitative and quantitative descriptions of this process in the spherically symmetric collapse model with two assumptions .
first , annihilation products leave a clump during a time substantially smaller than the time of its macroscopic evolution . and
second , an annihilating clump is spatially homogeneous and isotropic since more dense regions annihilate faster than the rest and heterogeneities are vanishing .
the model , defined by equations of mass and momentum balance and the equation of non - relativistic gravitation theory , is : @xmath11 { \displaystyle}\frac{d}{dt}\int \rho_{\chi } v_i dv = -\int ( \rho_{\chi } v_i v_k + p_{ik})ds_k - \int \rho_{\chi } \nabla_i \phi dv - \int \nu_{dis}\rho_{\chi } v_i dv - \int \frac{2({\sigma}v)_{ann } \rho_{\chi}^2}{m_{\chi}}v_idv\ , , \\[5 mm ] { \displaystyle}\delta \phi = 4\pi g\rho_{\chi}\ , , \end{array } \label{e1}\end{aligned}\ ] ] where @xmath12 is the pressure tensor for neutralino gas .
it is absent in consequence of the second assumption .
last terms in the balance equations describe the momentary going away of annihilation products from the collapsing clump .
we have taken into account that annihilation and scattering are cross - channels . in the local form
we have the simple system : @xmath13 which integrates elementary . here
@xmath14 is the clump radius ; @xmath15 is the clump mass ; @xmath16 is the initial radius ; @xmath17 is the constant which fixed by initial velocity of the surface clump compression .
the results are in the parametric form @xmath18^{-2 } , \\[6 mm ] { \displaystyle}r(\zeta)=r_0(\zeta^2+\beta^2)^{-1 } , \\[6 mm ] { \displaystyle}\left(\frac{8\pi g\rho_{\chi(*)}}{3}\right)^{1/2}t(\zeta)=\frac{1}{\beta^2}\left(\frac{\zeta}{\zeta^2+\beta^2 } -(1-\beta^2)^{1/2}\right ) + \frac{1}{\beta^3}\left(\arctan\frac{\zeta}{\beta } -\arctan\frac{(1-\beta^2)^{1/2}}{\beta}\right)+ \\[6 mm ] { \displaystyle}+\frac{({\sigma}v)_{ann}}{3m_{\chi}}\left(\frac{3\rho_{\chi(*)}}{2\pi g}\right)^{1/2}\left(\ln(\zeta^2+\beta^2 ) + \frac{2\beta^2(\zeta^2+\beta^2 - 1)}{\zeta^2+\beta^2}\right ) , \end{array } \label{e6}\end{aligned}\ ] ] where @xmath19 are the initial clump mass and density . as we see @xmath20 for @xmath21 .
the annihilation rate is @xmath22^{4}}. \label{rad}\end{aligned}\ ] ] generally behavior of this function is that a slow growth at small times changes into quick evolution around narrow peak after that there is an exponential drop to zero .
this behavior explains by changing of the parameter hierarchy during the collapse .
we use the convenient time scale @xmath23 here @xmath24 is the time of the maximum . since we have @xmath25 then the annihilation rate does nt depend on initial conditions and expresses through the fundamental constants only .
the functions @xmath26 and @xmath27 of dimensionless time @xmath28 are shown at the fig .
[ fig : fig1 ] .
= = about 80 % of clump mass annihilates during the time @xmath29 effectively the annihilation takes place between @xmath30 ( before maximum ) and @xmath31 ( after maximum ) .
such event represents as a powerful gamma - ray burst with total energy about clump mass .
kinetic cross section can be parameterized in the form : @xmath32 , @xmath33 some parameter with the dimension of mass .
so the decay time of collapsing neutralino clump expresses through susy parameters @xmath34 and gravitational constant but it is absolutely independent from initial density and initial velocity of clump compressing . deeper analysis is possible only in the framework of concrete susy model .
so there exits a possibility to extract of the susy parameters from observable characteristics of the bursts and to find constraints on the susy model . in the framework of split higgsino scenario @xcite the neutralino mass @xmath35
is estimated from the statement that the dark matter in the universe has formed at the high symmetry phase of the cosmological plasma .
neutralino of such type in the halo annihilate into w- and z - bosons and fermions ( see fig . [
fig : fig2 ] ) .
diagrams of neutralino annihilation in the framework of split higgsino scenario.,scaledwidth=95.0% ] total kinetic cross section is @xmath36 the clump annihilation time in this model is @xmath37
the energy spectrum @xmath38 is determined by the distribution in multiplicity of the secondary hadrons resulted from the neutralino annihilation .
we use of the fact that the total secondary hadrons multiplicity is nearly twice larger than the charged hadrons multiplicity @xcite .
the average multiplicity of secondary charged hadrons @xmath39 was studied in @xmath40 processes @xcite .
it was established that this quantity is some universal function of energy @xmath41 with experimentally fixed parameters @xmath42 to choice a specific channel it is necessary to fix parameters @xmath43 . for neutralino annihilation @xmath44
further it is supposed that the part of charged hadrons @xmath45 does nt depend on energy and has the value @xmath46 , which is extracted from @xmath47-peak data @xcite . for characteristic photon energies
generation the following processes are most important @xmath48 . in the fermion annihilation channel ( fig .
[ fig : fig2 ] ) the total hadron multiplicity is described by following logarithmic function with good accuracy : @xmath49 an average energy of neutral pions in neutralino annihilation secondaries is @xmath50 . then in the @xmath51 decay maximal characteristic energy of photon is equal : @xmath52 analogously , for reactions @xmath53 and @xmath54 energies of photons are : @xmath55 in the boson annihilation channel ( fig .
[ fig : fig2 ] ) the total hadron multiplicity of @xmath56- and @xmath47-bosons decays is : @xmath57 here neutral pion average energy is @xmath58 and maximal characteristic photon energies are : @xmath59 in a wide energy region the multiplicity distribution can be described with a good accuracy by the negative binomial distribution ( nbd ) that depends on energy very weakly ( logarithmically ) @xcite : @xmath60^{n+k}}\ ; , \\
\displaystyle k^{-1}(\sqrt{s})= a+b\ln \sqrt s , \end{array } \label{nbd}\end{aligned}\ ] ] where @xmath61 . for various channels the coefficients in the function @xmath62
are different : @xmath63 @xmath64 using the nbd it is possible to find the approximate energy distribution of photons .
for example , in the hadronic channel with @xmath65 it is supposed that the branching ratios for various hadrons @xmath66 are nearly constants and equal to the corresponding branching ratios extracted from @xmath67-annihilation @xcite .
each annihilating neutralino pair gives a number of neutral pions @xmath68 with the probability @xmath69 .
these pions generate the following distribution of photons @xmath70 with energy @xmath71 in the multiplicity interval @xmath72 , which is connected with energy interval @xmath73 . for @xmath74-mesons and boson annihilation channel considerations
are analogous . then the number of photons with energy @xmath75 per one annihilation act is : @xmath76\times \right . \\ \vspace*{5 mm } \displaystyle \times \left[br(h)\langle n^{ff}_{ch}\rangle p\left(\frac{m_\chi}{2e_{\gamma } } ; \langle n^{ff}_{ch}\rangle , k_{ff}\right)+\frac12 br(wz)\langle n^{wz}_{ch}\rangle p\left(\frac{m_\chi}{4e_{\gamma } } ; \langle n^{wz}_{ch}\rangle , k_{wz}\right)\right]+ \\ \vspace*{5 mm } \displaystyle+ br(\eta^0/h)\left[br(\eta^0\to 3\pi^0)+\frac13 br(\eta^0\to \pi^+\pi^-\pi^0)\right]\times \\ \vspace*{5 mm } \times \displaystyle \left[br(h)\langle n^{ff}_{ch}\rangle p\left(\frac{m_\chi}{6e_{\gamma } } ; \langle n^{ff}_{ch}\rangle , k_{ff}\right)+\frac12 br(wz)\langle n^{wz}_{ch}\rangle p\left(\frac{m_\chi}{12e_{\gamma } } ; \langle n^{wz}_{ch}\rangle , k_{wz}\right)\right]+ \\ \vspace*{5 mm } \displaystyle + \frac13 br(\eta^0/h ) br(\eta^0\to \pi^+\pi^-{\gamma})\times
\\ \displaystyle \left.\times \left[br(h)\langle n^{ff}_{ch}\rangle p\left(\frac{m_\chi}{3e_{\gamma } } ; \langle n^{ff}_{ch}\rangle , k_{ff}\right)+\frac12 br(wz)\langle n^{wz}_{ch}\rangle p\left(\frac{m_\chi}{6e_{\gamma } } ; \langle n^{wz}_{ch}\rangle , k_{wz}\right)\right]\right\}. \label{dn / de } \end{array}\end{aligned}\ ] ] here @xmath77 is the total branching ratio for neutralino annihilation into @xmath56- and @xmath47-bosons ; charge hadron multiplicities @xmath78 and @xmath79 are defined in ( [ nff ] ) and ( [ nwz ] ) ; parameters @xmath80 and @xmath81 were determined with coefficients from ( [ abpp ] ) and ( [ abee ] ) , respectively . the spectrum is shown in fig . [ fig : fig3 ] .
gamma ray annihilation spectrum.,scaledwidth=50.0% ] the particular duration of the grb and the specific shape of its energy spectrum are the characteristic features of considering astrophysical event the collapse and total annihilation of the neutralino clump . calculated values of characteristic photon energies are close to the threshold of sensitivity for the satellite detector glast .
possibility of neutralino annihilation spectrum registration at this apparatus will be clarified when the glast will be put into operation .
we have proposed the new production mechanism of the cosmological gamma - ray bursts . in the framework of this mechanism
the gamma - ray burst represents as a result of collapse and total annihilation of the neutralino clump .
the grb signal when registered should contain an information on susy parameters and the details of the annihilation process .
we have calculated the characteristic photon energies and annihilation spectrum for @xmath82 in the framework of split higgsino scenario .
the particular duration of the grbs and specific shape of its energy spectrum are the characteristic features which can help to separate considering astrophysical events from the radiation background . from astrophysical data on grbs with high - energy photons
it is possible to extract an information about susy parameters in different models .
it can help to investigate compatibility of the susy model with cosmology , cosmogony and high - energy collider data .
a. cesarini , f. fucito , a. lionetto , a. morselli and p. ullio , astropart .
* 21 * , 267 ( 2004 ) [ arxiv : astro - ph/0305075 ] ; + v. berezinsky , a. bottino and g. mignola , phys .
b * 325 * , 136 ( 1994 ) [ arxiv : hep - ph/9402215 ] .
a. tasitsiomi and a. v. olinto , phys .
d * 66 * , 083006 ( 2002 ) [ arxiv : astro - ph/0206040 ] ; + g. jungman and m. kamionkowski , phys . rev .
d * 51 * , 3121 ( 1995 ) [ arxiv : hep - ph/9501365 ] ; + l. bergstrom , j. edsjo , p. gondolo and p. ullio , phys .
d * 59 * , 043506 ( 1999 ) [ arxiv : astro - ph/9806072 ] ; + v. berezinsky , v. dokuchaev and y. eroshenko , phys . rev .
d * 68 * , 103003 ( 2003 ) [ arxiv : astro - ph/0301551 ] ; + r. aloisio , p. blasi and a. v. olinto , astrophys .
j. * 601 * , 47 ( 2004 ) [ arxiv : astro - ph/0206036 ] ; + a. v. gurevich , k. p. zybin and v. a. sirota , sov .
usp . * 167 * , 913 ( 1997 ) [ arxiv : astro - ph/9801314 ] .
n. arkani - hamed , s. dimopoulos , jhep * 0506 * , 073 ( 2005 ) [ arxiv : hep - th/0405159 ] ; + g. f. giudice and a. romanino , nucl .
b * 699 * , 65 ( 2004 ) [ erratum - ibid .
b * 706 * , 65 ( 2005 ) ] [ arxiv : hep - ph/0406088 ] ; + n. arkani - hamed , s. dimopoulos , g. f. giudice and a. romanino , nucl .
b * 709 * , 3 ( 2005 ) [ arxiv : hep - ph/0409232 ] .
s. eidelman _ et al . _
[ particle data group ] , phys .
b * 592 * , 1 ( 2004 ) ; + i. m. dremin and j. w. gary , phys . rept . * 349 * , 301 ( 2001 ) [ arxiv : hep - ph/0004215 ] ; + p. achard _ et al . _
[ l3 collaboration ] , phys .
b * 577 * , 109 ( 2003 ) [ arxiv : hep - ex/0110072 ] . | rare astrophysical events cosmological gamma - ray bursts with energies over gev are considered as an origin of information about some susy parameters .
the model of generation of the powerful gamma - ray bursts is proposed . according to this model the gamma - ray burst represents as a result of the collapse and the total annihilation of the neutralino clump .
about 80 % of the clump mass radiates during @xmath0 second at the final stage of annihilation . the annihilation spectrum and its characteristic energies are calculated in the framework of split higgsino model . |
we propose a mechanical turbine - based engine to harness the contractions of soft , photo - responsive solids with a large stroke .
we thus take photo - active nematic liquid crystal elastomers ( lces ) as our working material .
related engines have been proposed before , for instance based on the bend response of strips of nematic photo - glasses connecting two wheels @xcite . in contrast , we proposed a two - wheel stretch engine @xcite using a nematic photo - lce .
the engine had analogies to that of steinberg _ et .
al _ who studied chemical to mechanical energy conversion @xcite . here ,
we extend the two - wheel approach of @xcite to mechanical turbine - based engines , see fig .
[ fig1 ] , in order to get a much higher conversion of light to mechanical work than in that paper .
there is a similarity to a turbine that converted chemical to mechanical energy @xcite .
here we analyse the mechanics and losses involved in such turbines . modelling the geometrical and material parameters of this turbine - based engine , along with the known photo - response of typical lces
, suggests that its efficiency can be as high as @xmath0 .
the basis of these two - wheel and turbine engines is that a nematic rubber strip passes respectively once around two wheels , or multiply around spindles , turning at the same rate .
the strip reduces its natural length on heating or illumination during its transition around one wheel ; since its length while on the wheel is fixed , then the tension rises . on leaving the wheel , it physically contracts to its new equilibrium length and , in thereby eliminating the tension that has arisen , does work on the wheels / spindles that are turning underneath it .
a net work is done because the wheels are of different diameters or the spindles are tapered .
this paper is about the mechanics , work cycle and efficiency if spindles are used rather than wheels . with many passes around the spindles ,
the work cycle comes to resemble that of a turbine .
classical elastomers are cross - linked polymer melts exhibiting liquid characteristics locally , but are solid - like on a macroscopic scale . incorporating molecular rods into the polymers of a simple elastomer leads to networks that combine orientational liquid crystal order with the extreme stretchiness of rubber , that is lces @xcite .
the shape of a monodomain lce is very sensitive to the change of the nematic order parameter @xmath1 ; network polymers are elongated by the directional order and mechanical shape change ensues .
the order decreases on increasing the temperature , which manifests as a uniaxial contraction , by a factor @xmath2 ( @xmath3 ) of the elastomer along the nematic director @xcite .
the contraction is especially rapid in the vicinity of the transition temperature to the isotropic state .
analogous shape changes occur in photoelastomers in which photoisomerizable dye molecules , rod - like in their _ trans _ ground state , are connected to the lce structure @xcite . here
, illumination causes the creation of bent - shaped _ cis _ isomers of the dye which act as impurities that reduce the nematic order , in turn leading to a contraction of the photoelastomer .
the presence of _ cis _ isomers raises the effective temperature of the photoelastomer from @xmath4 to a pseudo @xmath5 which depends on the _ cis _ concentration and mimics the disorder as if it were induced thermally @xcite .
it is important to note that mechanical deformations of elastomers are reversible , that is , on removal of heat or light , recovery elongations by a factor of @xmath6 occur .
these elongations can be huge , up to 400% @xcite .
the optothermal cycle of our engine is shown in fig .
[ fig2 ] . against stretch @xmath7 .
upper curve pseudo temperature @xmath8 ( illuminated , isotropic state ) , bottom lower @xmath4 ( nematic state ) .
the grayed area represents the work done by the turbine - based engine having a finite number of extension and contraction steps ( to be discussed in section [ friction ] ) . dashed lines two - wheel motor @xcite .
, width=328 ] for a nematic elastomer of shear modulus @xmath9 , the free energy per length of unstretched band , @xmath10 , and the tension @xmath11 depend on the stretch ratio ( deformation gradient ) @xmath7
@xcite : f(,t ) & = & a_0 ( ^2 + ) , + & = & = ( ) _ t = ( - ) , [ nrubber ] where @xmath12 is the cross sectional area of the unstretched elastomer , and @xmath13 is the tension reduced by the natural force scale in the problem .
the modulus @xmath9 is temperature dependent .
a simple , freely - jointed rod ( fjr ) model quite accurately describes a wide range of lces @xcite and , in particular , the development of photoforce @xcite . in this model ,
the coefficients @xmath14 and @xmath15 are the ratios of the effective step lengths of the network polymers at the reference temperature @xmath4 and the current temperature @xmath8 , respectively parallel and perpendicular to the director . before heating or illumination ,
the temperature is @xmath4 ; the extension factors @xmath7 are measured from the lengths in this initial state .
the step lengths depend directly on the aspect ratio of the units of the polymer , but these cancel in the ratios @xmath16 , thus thus depend on the nematic order parameters @xmath17 and @xmath18 in a simple way in the fjr model : = ( 1 + 2q)/(1 + 2 ) = ( 1-q)/(1-)[p - defs ] .
note that the aspect ratio of the polymer units also contributes to the order @xmath1 .
along a@xmath19b , the current temperature is actually @xmath20 ( nothing has changed ) and @xmath21 and @xmath22 trivially since @xmath23 .
the illuminated state c@xmath19d , with an elevated @xmath5 , is isotropic with @xmath24 and the parameters are @xmath25 and @xmath26 .
a free elastomer has @xmath27 , in either the force - free state d , with temperature @xmath8 , or a with temperature @xmath4 .
changing from the reference temperature @xmath4 to @xmath8 and setting @xmath27 in ( [ nrubber ] ) then gives a contraction @xmath28 of its natural length along its director @xcite .
thus @xmath29^{1/3}$ ] .
the large change between @xmath30 and @xmath31 is what makes lces promising working materials .
we henceforth describe optical response , thermal response being entirely analogous . to realize a continuously operating engine , we subject the elastomer to a cyclical process through states a b c d a by changing the force and illumination .
it is easy to see from fig .
[ fig2 ] that the net work delivered by the engine , per unit length of its lce working material in its initial but unstretched state at @xmath4 , as it is taken around the cycle , is : w & = & w - w + & = & a_0(_^ ( ; ) - _ 1^ ( ; t ) ) + & & a_0(i_1 - i_2 ) [ work1 ] + & = & f ( , ) - f(,t ) + a_0 ( 1 - ( ^2)^1/3 ) , where the @xmath32 are reduced force integrals .
@xmath33 is similar in form to the work done in the various gaseous @xmath34@xmath35 cycles .
we assume the moduli are comparable , @xmath36 .
notice that the work @xmath33 delivered is greater than the difference established in the free energies per length of band in its unstretched state at c and b by a ( positive ) term @xmath37 that arises during the cycle due to heat absorption .
one already sees from fig .
[ fig2 ] that the pre - stretch , @xmath38 , imposed before illumination , enhances work output considerably , just as a higher compression ratio improves conventional engines .
further , taking a turbine allows one to extend the cycle to zero force ( at a contraction equal to the natural illuminated length , @xmath2 ) which is impossible in two - wheel stretch engines @xcite ; see cycle a
c s a in fig .
however for a soft solid , formidable problems exist in realising the above @xmath33 .
these include sliding , and the consequential frictional losses , when finite tension differences in soft solids exist across the engine .
the remainder of this paper is concerned with overcoming these problems to get close to the above @xmath33 by using the turbine of fig .
. we will need to generalise the classical pulley result of euler @xcite to highly extensible belts .
the energy input per unit length of unstretched elastomer in the illumination process b
c is @xmath39 , with @xmath40 the appropriate photon energy and @xmath41 the number density of dye molecules . for our model of an isotropic elastomer , the internal energy is a function of temperature only and is unchanging along an isotherm @xcite .
thus the heat input per unit length of unstretched elastomer during the isothermal process c d equals the work done @xmath42 .
the efficiency is the ratio of the work done to the optical energy plus heat invested per cycle : = = ( 1- ) / ( 1 + ) [ eta1 ] where @xmath43 and @xmath44 are the reduced force integrals in eq . ( [ work1 ] ) and depend on the order along the isotherms @xmath4 and @xmath8 . from eq .
( [ nrubber ] ) , they are clearly just the free energy changes per length of band in its unstretched state ( reduced by @xmath45 ) between the relevant @xmath7s , and involve the @xmath1 and @xmath46 through the explicit expressions ( [ p - defs ] ) for the @xmath47s .
since @xmath48 , the efficiency is always @xmath49 .
the ratio of material constants has been estimated as @xmath50 @xcite .
thus the efficiency is determined by the value of the order parameter @xmath1 and the value of the pre - stretch @xmath38 . as an example we take a modest value for the dark - state order , @xmath51 , which would yield @xmath52 , or equivalently an elongation @xmath53 on recovering the nematic state from the isotropic , illuminated state .
in fact much greater elongations of @xmath54 are known . with @xmath51 and @xmath55 ,
one obtains @xmath56 .
we discuss dynamics and the influence on @xmath57 at the end .
we now follow the mechanical / thermal cycle in detail : a closed band of photo - lce is wound around the pair of spindles of fig .
[ fig1 ] , the tops of which are rigidly coupled by a loop of inextensible wire , thus ensuring they have equal angular velocities .
the spindles have slightly tilted axes which enables the elastomer to spontaneously follow a helical path down the spindles as they rotate .
the horizontal cross - section of the spindles can be approximated by circles , the radii of which grow from the top to attain a largest value ( @xmath58 ) at the middle section of the spindle , after which they then decrease .
initially , a free elastomer belt in the nematic , reference state a at temperature @xmath4 comes on to the upper part of the spindles with @xmath31 and hence @xmath59 .
isothermally , it passes helically from smaller to larger spindle radii , extending at each pass by a fraction equal to the ratio of the radii which material conservation requires as discussed later .
it suffers multiple - step extension during the process a b in fig .
state b , achieved when the belt enters the cylindrical section of the spindle with the largest radius @xmath58 ( fig . [ fig1 ] ) , has the highest ( pre-)stretch @xmath60 , and has force @xmath61 .
when in this region of radius @xmath58 , the elastomer is illuminated and assumed to reach @xmath8 before it leaves the region ; the extent of this region of the spindle is determined in part by dynamics since light absorption should proceed long enough to give sufficient dye conversion to attain the isotropic state before the band starts the contraction part of the cycle . along b c the process is isometric ( @xmath62 , an assumption of no slip along b
c ) . at the point c
, the force takes the highest value in the cycle @xmath63 , and the elastomer is in the isotropic state . after passing the region of maximum radius @xmath58
, the elastomer spirals down the lower part of the spindles , c@xmath19d .
it moves now repeatedly from larger to smaller radii , contracting at each pass by the ratio of the radii .
it releases its elastic energy as work done on the spindles ; in the final state d the stretch reaches the minimum value @xmath64 , the natural length of elastomer at @xmath8 . to remain with @xmath8 along c d , because of _ cis _
@xmath19 _ trans _ back reaction one must continue some level of illumination .
equally , to remain isothermal one needs external heat since entropy rises in a contracting elastomer .
( in practice much of this heat input can come directly or indirectly from the optical source : photons have energy in the range 2ev which is equivalent to @xmath65 for @xmath66k , that is , large energies per absorption are available from quantum efficiencies of less than 100% in the _ trans _ to _ cis _ transition .
equally , _
cis _ to _ trans _ back reactions yield energies per dye molecule that are large on the thermal energy scale and are spread over the network molecules . ) finally , on removal of illumination , at zero force the elastomer gradually recovers and elongates from @xmath2 and @xmath8 to @xmath67 and @xmath4 ( d a of the cycle ) .
the engine converts into mechanical work part of the optical energy elevating the elastomer to the effective temperature @xmath8 .
work is extracted by turning one spindle , the left in fig .
[ fig1 ] , against the external torque , @xmath68 .
the useful work done per cycle is less than that of eq .
( [ work1 ] ) because of sliding friction .
we analyze each pass of the elastomer around a spindle , considering differences in the tensions @xmath69 and @xmath70 of the incoming and outgoing sections ; see fig .
[ fig3](a ) . around its circumference .
( b ) an expanded view of a short section of belt at @xmath71 with a radial force and a frictional shear force acting on the belt .
, width=328 ] classically @xcite , there is an exponential increase of tension along a belt passing around a wheel ; see fig .
[ fig3](a ) where tension increases from @xmath69 to @xmath70 over parts of @xmath72 to @xmath73 : since the surface is curved , there is an inward force per unit length on the spindle from the belt of @xmath74 ; see fig .
[ fig3](b ) .
if the surface is rough , the normal force gives a frictional shear force ( per unit length on the circumference ) of @xmath75 .
the net shear force @xmath76 on an element @xmath77 balances the difference in tangential forces @xmath78 across the element .
clearly @xmath79 gives an exponential increase in @xmath80 .
there is an obvious limitation to the classical analysis that is important in soft solids where substantial stretch @xmath7 must accompany increasing @xmath11 : the belt has to slide on the spindle to extend so that @xmath11 increases with @xmath81 .
it can not translate with the circumferential speed @xmath82 for all its contact length with the spindle , but rather travel faster .
changing strain implies changing stored elastic energy .
sliding implies frictional losses during the transmission of power .
[ fig3](a ) assigns a @xmath83 corresponding to a @xmath80 .
in particular @xmath84 for the belt incoming on to the spindle , and @xmath85 when it emerges with @xmath70 .
take temperature constant along the belt in contact with the spindle ; increasing tension @xmath86 then implies @xmath87 .
soft solids are essentially incompressible under extension , hence material conservation dictates volume conservation .
the sectional area @xmath88 must diminish as the belt extends by @xmath7 .
the volume flux of belt on to the spindle is @xmath89 and must be matched by @xmath90 at a general point .
dividing through by @xmath12 gives v_1/_1 = v(x)/(x ) = v_2/_2 , [ speed ] which is the length of band per unit time passing , were it to be in its unstretched state .
this flux of length of band is equivalent to @xmath91 when the band is not sliding .
since the final stretch is greater than the initial , @xmath87 , it follows @xmath92 the belt starts going faster than the spindle .
there is _ sliding _ friction . neglecting inertia @xcite ,
we get : = _ k , [ fplength ] where @xmath93 is the coefficient of kinetic ( sliding ) friction giving @xmath94 .
since @xmath86 , friction inhibits sliding in the @xmath95 direction . solving ( [ fplength ] ) gives f(x ) = f_2 , [ tension ] which differs from the classical result in its use of kinetic friction coefficient , @xmath96 , and the passage of energy to friction and to elastic potential .
the tension varies from @xmath97 at @xmath73 down to @xmath98 at an @xmath99 given by x_1 = r .
[ x1 ] in the initial section @xmath100 , the tension retains its incoming value @xmath98 .
it is gradients @xmath101 that transfer force ( torque ) to the spindle in the region @xmath102 .
there is a region @xmath100 without torque if @xmath103 , that is if @xmath104 .
the power delivered to the spindle , in one step of a
b of fig .
[ fig2 ] with @xmath86 and the spindle turning clockwise ( see fig . [ fig3](a ) ) , is the speed of its surface , @xmath105 , times the force exerted on it : p = v_1 _ x_1^r x = v_1 _ x_1^r x = v_1 ( f_2 - f_1 ) .
[ pw ] the term @xmath106 is the power given by the spindle to the incoming band .
the power @xmath107 is the portion of the power @xmath108 delivered by the more tense band that actually finds its way to the spindle .
thus , all the useful power is delivered via the region that is slipping . when the band slips as soon as it makes contact with the wheel , slip is complete and @xmath109 in ( [ x1 ] ) ; then @xmath110 takes its maximal value , as does @xmath111 in eq .
( [ pw ] ) ; @xmath112 .
the power lost to friction in a step of a b is from the forces @xmath113 acting on elements @xmath77 moving at speeds @xmath114 relative to the elements : p = _ x_1^r x [ v(x ) - v_1 ] , [ pf ] with @xmath115 given by ( [ x1 ] ) .
the @xmath105 part is trivially @xmath116 , eq .
( [ pw ] ) .
the first term integrates by parts to @xmath117 .
the latter part has @xmath118 , and becomes : - _ x_1^r x & f(x ) & = - _ _ 1^_2 f ( ) + & = & - , [ elastic ] where we used eq .
( [ nrubber ] ) with @xmath119 , since the elastomer is at @xmath4 .
overall , the power lost to friction is : p = f_2(v_2 - v_1 ) - [ f(_2,t ) - f(_1,t ) ] , [ pff ] and does not involve the friction coefficient @xmath96 , unless slippage is complete ( @xmath120 ) . in the process a b of the engine , @xmath87 is realized by moving the elastomer belt in each pass from a spindle surface of a smaller radius @xmath121 to a surface of a larger radius @xmath122 .
the velocities are @xmath123 and @xmath124 . then using ( [ speed ] )
one gets @xmath125 , as required . in process c
d , in each turn around a spindle the isotropic elastomer with @xmath8 goes from a larger to a smaller radius , @xmath126 leading to @xmath127 and @xmath128 since @xmath129 .
hence the band goes from a higher to lower force , @xmath130 ; see fig .
[ fig3](a ) .
note that the spindle s direction of turn is fixed in the clockwise direction by the engine operation as a whole , independently of the relative size of the forces .
as before , there is an @xmath115 with no slip for @xmath131 , while slippage occurs for @xmath81 in the range @xmath102 . repeating the above ,
the power delivered and the frictional power lost are : p & = & v_1 ( f_2 - f_1 ) < 0 , + [ isopw ] p & = & f_2(v_2 - v_1 ) + [ f(_1 , ) - f(_2 , ) ] .
[ isopf ] it remains to analyze the elastomer belt going around surfaces of radius @xmath58 in the process b c .
now the elastomer at force @xmath132 and temperature @xmath4 comes on the spindle where the illumination begins ; see fig .
[ fig1 ] . in fig .
[ fig3](a ) we now have @xmath133 and @xmath134 .
we assume that the change @xmath4 to @xmath8 occurs without slippage and is complete before the elastomer reaches the point c in fig .
[ fig1 ] , possibly after several passes around the central section .
( force rising from @xmath132 to @xmath63 without slippage assumes @xmath135 where @xmath136 is the static friction coefficient .
elongation is fixed at @xmath60 and the variation of @xmath11 with @xmath81 is from the changing @xmath137 and @xmath138 factors in expression ( [ nrubber ] ) for @xmath11 . )
after @xmath115 , slippage then occurs and the force drops from @xmath63 to @xmath139 ( in fig .
[ fig3](a ) @xmath140 , @xmath141 ) .
the state s is marked in figs .
[ fig1 ] and [ fig2 ] .
the net power delivered to the spindles on transit of the cylindrical section with radius @xmath58 is p = r ( f - f ) .
[ pwcs ] for the frictional losses , expression ( [ isopf ] ) with the appropriate values of stretches , velocities and forces is still applicable since process c s is part of c d .
excluding the middle spindle section of radius @xmath58 , the power delivered to the spindle in each extension step of a b is given by expressions of the form ( [ pw ] ) and is clearly positive as one expects if @xmath142 .
however , the total power delivered in a b is actually negative ( see fig . [ fig2 ] ) since the band emerges stretched from a b .
this is due to the fact that one must add the highly negative contribution @xmath143 of ( [ pwcs ] ) to the set of positive contributions of the form ( [ pw ] ) .
similarly , in c d the highly positive term @xmath144 of ( [ pwcs ] ) is added to negative contributions ( [ isopw ] ) , making the total power delivered in c d positive since the elastomer contracts .
we now calculate the total power @xmath145 delivered to the spindles by the elastomer for a @xmath146stage contraction process c d and @xmath147stage extension process a b .
when the engine runs at a constant velocity @xmath148 , the net torque acting on each of the spindles is zero . for simplicity we shall neglect frictional forces at the bearings
. then one can express the balance of torques on the each spindle separately . beside the torques produced by the elastomer
, one should take into account the torques due to the inextensible wire which couples the spindles , and the external torque @xmath68 acting on the left spindle ( see fig . [ fig1 ] ) .
in such a way , one obtains a pair of equations which enable one to determine @xmath68 , and thus the useful power @xmath149 : p_nk = _
i=1^n f_i+1 ( r_i - r_i+1 ) - _
j=1^k f_j+1 ( r_j+1 - r_j ) .
[ powernk ] here , @xmath150 and @xmath151 are the forces in the @xmath152th step of c d and in the @xmath153th step of a
b process , respectively , while @xmath154 and @xmath155 are the corresponding spindle radii . in particular ,
@xmath156 and @xmath157 ( force @xmath69 would be @xmath158 ) .
clearly , since @xmath159 for all @xmath152 , the first sum on the right hand side of ( [ powernk ] ) , @xmath160 , is positive .
similarly , we have @xmath161 and @xmath162 ( force
@xmath163 would be zero ) .
on the other hand , @xmath164 is negative since @xmath165 .
using relations ( [ pff ] ) and ( [ isopf ] ) , the useful power is p_nk & = & _ i=1^n [ f(_i , ) - f(_i+1 , ) ] - _
i=1^n p_f , i + & - & _ j=1^k [ f(_j+1 , t ) - f(_j,t ) ] - _ j=1^k p_f , j + & = & w - _
i=1^n p_f , i - _
j=1^k p_f , j , [ areapowernk ] where @xmath166 and @xmath167 are the stretches , and @xmath168 and @xmath169 are the frictional power losses in the contraction and extension steps respectively .
we now sketch a geometrical interpretation of ( [ areapowernk ] ) .
the first term of the last line is the maximum useful mechanical power @xmath170 , with @xmath171 being the flux of unstretched length ( [ speed ] ) .
the maximum mechanical work @xmath33 done per unit length of unstretched elastomer in one cycle is given by the area in the solid line a b c d a in fig .
the useful work done per unit length of unstretched elastomer in one cycle @xmath172 is the gray shaded area enclosed in fig .
[ fig2 ] ( where for simplicity the change in radii in each contraction and extension step is taken to be the same ; see comments below ) .
thus , the work done in one cycle to overcome friction during the contraction process is equal to the area of the white regions between the solid curve c d and the gray shaded area ( corresponding frictional power losses are @xmath173 ) .
similarly , the area of regions between the gray shaded area and the solid curve a b corresponds to the term @xmath174 of ( [ areapowernk ] ) .
by contrast , the two - wheel cycle a
c s a dashed in fig .
[ fig2 ] has larger frictional losses : the areas between the isotherms a b and c s and the horizontal dashed lines respectively above and below them are much larger relative to the enclosed area a b c s a ( the maximum theoretical work ) .
the inaccessibility of an @xmath27 state means the enclosed area is smaller than the turbine s , even though the same optical energy input to go from b to c is required , further underscoring the superior efficiency of a turbine .
notice that the more steps @xmath146 and @xmath147 , the more useful work per cycle is done .
we shall compare the power @xmath16 corresponding to an infinite number of extension and contraction steps , with the power @xmath145 given by expression ( [ powernk ] ) , and determine how many steps @xmath146 and @xmath147 one needs to achieve , say , @xmath175 .
as an illustration , we take the change of radii in the contraction process c d to be the same for each step and equal to @xmath176 , where @xmath177 is the final radius in the contraction process . similarly , for the extension process a b we take @xmath178 , where @xmath179 is the initial radius in the extension process . from relations ( [ speed ] )
it follows : @xmath180 and @xmath181 .
here we used the facts that the stretch that corresponds to the radius @xmath177 is actually the natural stretch @xmath28 , and that the stretch corresponding to radius @xmath179 is @xmath67 .
the stretch steps in contraction and extension processes are then @xmath182 and @xmath183 , respectively .
the choice @xmath184 was made in fig .
[ fig2 ] , and cones taken in fig .
[ fig1 ] , but these restrictions are not necessary . the powers @xmath145 and @xmath16 can be expressed as p_nk & = & a_0 , + p & = & a_0 .
[ ppinf ] taking @xmath185 and @xmath55 , one finds @xmath186 for @xmath187 , @xmath188 . the maximum mechanical work done per unit length of unstretched elastomer in one cycle @xmath189 is given in eq .
( [ work1 ] ) .
the simple integrals above at constant temperatures return us to the free energies in eq .
( [ nrubber ] ) .
in summary , the optical contraction of photo - lces can be used to harness optical energy to generate mechanical energy .
our mechanical turbine - based engine utilizes more effectively the optothermal cycle than the two - wheel engine @xcite .
soft , extensible photo - solids deliver large amounts of work , but their extensions and contractions lead to sliding and hence concomitant frictional losses .
we analyzed such losses and calculated the fraction of work lost due to them .
that our proposed photo - conversion method has moving parts gives it a disadvantage over conventional photo - voltaics , though rubber is highly durable and tough for instance car tyres survive long use in harsh , abrasive conditions .
another difficulty , that could perhaps be solved by chemical design , is degradation from the effect of uv light .
the role of dynamics when attempting to attain the above efficiencies also remains a question . as we noted , the residence time in the isometric state b
c needs to be long enough that light penetrates deeply to convert the whole thickness of the band to the isotropic state . for intense beams
this requires a non - linear bleaching wave ( of light - induced dye conversion ) to traverse the thickness @xcite just as that part of the band leaves the section of the spindle with maximal radius @xmath58 .
longer residence simply gives time for back reaction without mechanical work being delivered .
light energy is wasted and the efficiency drops from the values emerging from the above analysis . thus light intensity , band thickness , residence in the isometric state , and rotational rate are all related aspects of the dynamics and need optimisation by future studies .
m. k. thanks the winton programme for the physics of sustainability and the cambridge international trust for financial support .
m. w. thanks professor peter palffy - muhoray for introducing him to these studies and for advice over many years .
16ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo
secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) @noop _ _ ( , , ) @noop * * , ( ) @noop * * , ( ) | liquid crystal elastomers are rubbers with liquid crystal order .
they contract along their nematic director when heated or illuminated .
the shape changes are large and occur in a relatively narrow temperature interval , or at low illumination , around the nematic isotropic transition .
we present a conceptual design of a mechanical , turbine - based engine using photo - active liquid crystal elastomers to extract mechanical work from light .
its efficiency is estimated to be 40% . |
the new generation of radioactive ion beam facilities will provide high intensity ( @xmath3 ions / s ) exotic beams ( for example , @xmath4kr , @xmath5sn , @xmath6xe or @xmath7cs ) .
one of the most interesting areas of research with radioactive beams will be the study of the complete fusion process @xcite where the fusion experiments with exotic beams can be performed to synthesize and study new isotopes of existing elements .
the central issue is whether the capture and fusion cross sections will be enhanced due to the large deformation of the neutron - rich or neutron - deficient projectile - nucleus .
however , one should bear in mind the smaller intensity of these beams in comparison with the intensity of stable beams .
our aim is to find the global trend in the production cross section of exotic nuclei as a function of the charge ( mass ) number of the projectile in complete fusion reactions .
based on this trend one can find a consensus between the cross sections resulting from a certain beam and the intensity of this beam . the goal of the present paper is to compare the fusion of stable @xmath8xe and radioactive @xmath9xe beams with the same target , @xmath1ca , in order to study the effects of the neutron excess and neutron transfer on the fusion process .
the target @xmath1ca is ideal for this purpose since this nucleus has the largest possible neutron excess and the systems @xmath9xe+@xmath1ca have positive neutron transfer @xmath10-values while all the corresponding reactions @xmath8xe+@xmath1ca display negative @xmath10-values . in the present paper
we demonstrate for the first time the possibilities for producing neutron - rich isotopes of @xmath2w in the complete fusion reactions @xmath11xe+@xmath1ca with rather large cross sections .
the nucleus @xmath12w was the heaviest isotope which has been synthesized in ( @xmath13,@xmath14 ) and ( @xmath15,@xmath16 ) reactions @xcite . in these experiments
the chemical extraction of @xmath12w was possible after long irradiation .
another method to produce the neutron - rich nuclei is fragmentation reactions @xcite .
cross - sections smaller than 0.4@xmath17b were measured for the isotopes @xmath18w in cold fragmentation of 950 mev / nucleon @xmath19au beams on be targets @xcite .
however , the production cross section decreases strongly with increasing neutron number .
the most neutron - rich w isotopes , up to @xmath19w , were observed in projectile fragmentation of @xmath20u at 1 gev / nucleon on be targets at the fragment separator ( frs ) at gsi @xcite . here , cross - sections smaller than
5 nb were measured for w isotopes with mass numbers a @xmath21 190 where the cross - section decreases by approximately one order of magnitude for every two neutrons more in the residual nucleus . in the present paper
we also compare the complete fusion reactions @xmath22xe+@xmath1ca with fragmentation reactions leading both to the production of neutron - rich w isotopes .
additionally , we will compare the complete fusion reactions @xmath23cs+@xmath24ga which lead to neutron - deficient rn isotopes with the respective yields from the fragmentation reactions .
because the capture cross section is equal to the fusion cross section for the reactions @xmath25xe+@xmath1ca treated in the present paper , the quantum diffusion approach @xcite for the capture is applied to study the complete fusion . with this approach many heavy - ion capture reactions at energies above and well below the coulomb barrier
have been successfully described @xcite . since the details of our theoretical treatment were already published in refs .
@xcite , the model will be only shortly described . in the quantum diffusion approach
@xcite the collisions of nuclei are described with a single relevant collective variable : the relative distance between the colliding nuclei .
this approach takes into consideration the fluctuation and dissipation effects in collisions of heavy ions which model the coupling with various channels ( for example , coupling of the relative motion with low - lying collective modes such as dynamical quadrupole and octupole modes of the target and projectile @xcite ) .
we have to mention that many quantum - mechanical and non - markovian effects accompanying the passage through the coulomb barrier are taken into consideration in our formalism @xcite .
the diffusion models , which include the quantum statistical effects , were also proposed in refs .
the nuclear deformation effects are taken into account through the dependence of the nucleus - nucleus potential on the deformations and mutual orientations of the colliding nuclei . to calculate the nucleus - nucleus interaction potential @xmath26 , we use the procedure presented in ref .
@xcite . for the nuclear part of the nucleus - nucleus potential , the double - folding formalism with a skyrme - type density - dependent effective nucleon - nucleon interaction
is used @xcite .
the nucleon densities of the projectile and target nuclei are specified in the form of the woods - saxon parameterization , where the nuclear radius parameter is @xmath27 fm and the diffuseness parameter takes the values @xmath28 fm for all nuclei .
the absolute values of the quadrupole deformation parameters @xmath29 of deformed nuclei were taken from refs . @xcite and @xcite for the known and unknown nuclei , respectively .
for the magic @xmath1ca and semimagic @xmath30xe nuclei in the ground state , we set @xmath31 and @xmath32 , respectively . the capture cross section is the sum of the partial capture cross sections @xcite @xmath33 where @xmath34 is the reduced de broglie wavelength , @xmath35 is the reduced mass ( @xmath36 is the nucleon mass ) , and the summation is over the possible values of the angular momentum @xmath37 at a given bombarding energy @xmath38 .
knowing the potential of the interacting nuclei for each orientation with the angles @xmath39 , one can obtain the partial capture probability @xmath40 which is defined by the probability to penetrate the potential barrier in the relative distance coordinate @xmath41 at a given @xmath37 .
the value of @xmath40 is obtained by integrating the propagator @xmath42 from the initial state @xmath43 at time @xmath44 to the final state @xmath45 at time @xmath46 ( @xmath47 is the momentum ) : @xmath48 . \label{1ab_eq}\end{aligned}\ ] ] here , @xmath49 is an internal turning point . the second line in ( [ 1ab_eq ] )
is obtained by using the propagator @xmath50 ( @xmath51 $ ] , @xmath52 , @xmath53 , @xmath54 , @xmath55 , @xmath56 , @xmath57 , @xmath58 ) calculated for an inverted oscillator which approximates the nucleus - nucleus potential @xmath59 in the variable @xmath41 as follows . at given @xmath38 and @xmath37 ,
the classical action is calculated for the realistic nucleus - nucleus potential . then the realistic nucleus - nucleus potential is replaced by an inverted oscillator which has the same barrier height and classical action .
so , the frequency @xmath60 of this oscillator is set to obtain an equality of the classical actions in the approximated and realistic potentials .
the action is calculated in the wkb approximation which is the accurate at the sub - barrier energies . usually in the literature the parabolic approximation with @xmath38-independent @xmath61 is employed which is not accurate at the deep sub - barrier energies .
our approximation is well justified for the reactions and energy range considered here @xcite .
we assume that the sub - barrier capture mainly depends on the two - neutron transfer with positive @xmath10-value .
our assumption is that , just before the projectile is captured by the target - nucleus ( just before the crossing of the coulomb barrier ) , the transfer occurs and leads to the population of the first excited collective state in the recipient nucleus @xcite .
so , the motion to the @xmath62 equilibrium starts in the system before the capture because it is energetically favorable in the dinuclear system in the vicinity of the coulomb barrier .
for the reactions under consideration , the average change of mass asymmetry is connected to the two - neutron transfer . since after the transfer the mass numbers , the isotopic composition and the deformation parameters of the interacting nuclei , and , correspondingly , the height @xmath63 and shape of the coulomb barrier are changed , one can expect an enhancement or suppression of the capture . if after the neutron transfer the deformations of the interacting nuclei increase ( decrease ) , the capture probability increases ( decreases ) .
when the isotopic dependence of the nucleus - nucleus potential is weak and after the transfer the deformations of the interacting nuclei do not change , there is no effect of the neutron transfer on the capture . in comparison with ref .
@xcite , we assume that the negative transfer @xmath64values do not play a visible role in the capture process .
our scenario was verified in the description of many reactions @xcite .
the primary neutron - rich products of the complete fusion reactions @xmath25xe+@xmath1ca of interest are excited and transformed into the secondary products with a smaller number of neutrons . since neutron emission is the dominant deexcitation channel in the neutron - rich isotopes of interest , the production cross sections of the secondary nuclei are the same as those of the corresponding primary nuclei .
this seems to be evident without special statistical treatment .
the calculations are performed by employing the predicted values of the mass excesses and the neutron separation energies @xmath65 for unknown nuclei from the finite range liquid drop model @xcite .
to analyze the isotopic trend of the fusion cross section , it is useful to use the so called universal fusion function ( uff ) representation @xcite .
the advantage of this representation appears clearly when one wants to compare fusion cross sections for systems with different coulomb barrier heights and positions . in the reactions where the capture and fusion cross sections coincide
, the elimination of the influence of the nucleus - nucleus potential on the fusion cross section with the uff representation allows us to conclude about the role of deformation of the colliding nuclei and the nucleon transfer between interacting nuclei in the capture and fusion .
on @xmath66 for the indicated reactions . ] on @xmath67 for the reactions @xmath25xe+@xmath1ca at fixed bombarding energies @xmath68 mev ( triangles ) , @xmath69 ( stars ) , @xmath70 mev ( circles ) . ] in ref .
@xcite the reduction procedure consists of the following transformations : @xmath71 the frequency @xmath72 is related with the second derivative @xmath73 of the nucleus - nucleus potential @xmath26 at the barrier radius @xmath74 and the reduced mass parameter @xmath17 . with these replacements one can compare the cross sections for different reactions . for the indicated neutron - rich isotopes
@xmath75w produced in the @xmath22xe+@xmath1ca reaction .
the vertical dashed lines show the range of energies for the production of given isotope . ] for the indicated neutron - rich isotopes @xmath76w produced in the @xmath77xe+@xmath1ca reaction .
the vertical dashed lines show the range of energies for the production of given isotope . ] in fig . 1 , one can see the comparison of the calculated functions @xmath78 for the reactions @xmath79xe+@xmath1ca with stable and radioactive beams .
as expected , at sub - barrier energies the enhancement of the complete fusion ( capture ) cross section is larger in the case of reactions with strongly quadrupole deformed projectile - nuclei and after neutron transfer .
the quadrupole deformation parameter @xmath29 of the projectile nucleus increases with changing mass number from @xmath67=136 to @xmath67=130 or to @xmath67=144 . for the reaction @xmath30xe+@xmath1ca with spherical target and projectile and without neutron transfer
the cross section is the smallest one at @xmath80 .
the sub - barrier cross sections for the reactions @xmath9xe+@xmath1ca with neutron transfer ( positive @xmath10-values ) are larger than those for the reactions @xmath8xe+@xmath1ca , where the neutron transfer is suppressed ( negative @xmath10-values ) .
since after two - neutron transfer the mass numbers and the deformation parameters of the interacting nuclei are changed and the height of the coulomb barrier decreases , one can expect an enhancement of the capture .
for example , after the neutron transfer in the reaction @xmath81xe(@xmath82)+@xmath1ca(@xmath31)@xmath83xe(@xmath84)+@xmath85ca(@xmath86 ) , the deformation of the target - nucleus increases and the mass asymmetry of the system decreases , and , thus , the value of the coulomb barrier decreases and the capture cross section becomes larger ( fig . 1 ) .
we observe the same behavior in the reactions with the projectiles @xmath87xe .
the complete fusion ( capture ) cross sections for the reactions @xmath0xe+@xmath1ca at different bombarding energies are presented in fig . 2 . the behaviour of the curves in fig
. 2 is determined by the quadrupole deformation and neutron transfer effects .
the isotopic dependency is rather weak at energies above the corresponding coulomb barriers . at sub - barrier energies
the fusion cross section decreases by about one order of magnitude with increasing mass number @xmath67 of the projectile from @xmath88 up to @xmath89 ( @xmath90 ) . for @xmath91
a steep increase can be observed for beam energies of 5 mev below the corresponding coulomb barriers . at energies near the coulomb barrier
the cross section changes in a similar way but the curve shows a much flatter slope . for the indicated neutron - deficient isotopes of rn produced in the @xmath92-channels ( @xmath93=2 - 4 ) of the @xmath94cs+@xmath24ga reaction .
] cs+@xmath24ga reaction . ] in figs . 3 and 4 we present the possibilities for future experiments to produce the neutron - rich isotopes @xmath2w in complete fusion reactions of @xmath11xe+@xmath1ca with radioactive beams .
the production cross sections of the neutron - rich @xmath95w isotopes , for example , are between the 10 @xmath17b and 100 mb levels meaning that they can be observed with rather low beam intensities and with the present experimental techniques .
the calculated cross sections are more than two orders of magnitude larger than in fragmentation reactions @xcite .
note also , that when the neutron number approaches the drip - line the production cross section in complete fusion decreases not so fast as in fragmentation reactions .
the availability of heavy radioactive beams at coulomb barrier energies at future facilities like fair , hie - isolde or spiral - ii will enable the experimental utilization of the above discussed effects for fusion reactions .
another competing method to produce heavy exotic isotopes is projectile fragmentation at relativistic energies which is for example used at the fragment separator ( frs ) at gsi . in the following ,
we give some comparative considerations on both methods since , depending on the region of the nuclear chart , fragmentation can lead to high yields of exotic nuclei . as an example , we consider the isotope @xmath96w .
cross - sections of up to about 2 mb are predicted for its production in the complete fusion reactions of @xmath22xe+@xmath1ca at e@xmath97 mev .
cross - sections on the same order are also measured in the fragmentation reactions leading to yields of @xmath98 ions / s . at the future super - frs facility
even yields of @xmath99 ions / s are predicted . in order to obtain at least the same yields of @xmath98 ions / s in fusion reactions , @xmath22xe beams with intensities of at least @xmath100 ions / s
are required .
the largest intensities for neutron - rich xe beams are expected at spiral - ii where @xmath101 of @xmath22xe projectiles per second are predicted which is , however , still eight orders of magnitude less than needed for an efficient application of fusion reactions to reach @xmath96w . as an other example , we discuss in the following the synthesis of neutron deficient rn ( @xmath102 ) isotopes in the complete fusion reactions .
figures 5 and 6 show the calculated excitation functions for fusion reactions of @xmath103cs beams with @xmath24ga target .
the survival probabilities of the excited compound nuclei in the neutron evaporation channels @xmath92 ( @xmath104 ) are calculated by employing the modified statistical code grogif @xcite with the same parameters as in ref .
the capture cross sections and fusion probabilities are calculated with the quantum diffusion approach @xcite and the dinuclear system fusion model @xcite , respectively .
radioactive cs beams are already now available with high intensities for a broad variety of isotopes and are therefore favourable projectiles . at rex - isolde for example , the isotopes @xmath105cs are provided with intensities around @xmath106 ions / s and for the future hie - isolde facility ten times higher intensities are expected at beam energies of @xmath21 5.5 mev / nucleon .
a comparison of the predicted yields for neutron deficient rn isotopes at the superfrs facility with the expected yields from fusion evaporation reactions with @xmath23cs beams at intensities of @xmath106 ions / s leads to the conclusion that the complete fusion is not superior to fragmentation for @xmath25rn isotopes with @xmath107 . for
these mass numbers at least 2 - 7 times lower yields can be obtained in the fusion reactions with the presently available beam intensities .
because of deformation and neutron transfer effects , a strong dependence of the sub - barrier complete fusion ( capture ) cross section on the isospin was found for the reactions @xmath0xe+@xmath1ca . at
fixed bombarding energy , the cross section increases with changing mass number of the projectile - nucleus from @xmath67=136 to @xmath67=130 or to @xmath67=150 .
the @xmath30xe+@xmath1ca reaction with magic and semimagic nuclei has the smallest cross section . the complete fusion ( capture )
cross sections for the reactions @xmath8xe+@xmath1ca without neutron transfer are smaller than those for the reactions @xmath9xe+@xmath1ca with neutron transfer .
we demonstrated the possibilities for producing neutron - rich isotopes of @xmath2w with relatively large cross sections for future experiments in the complete fusion reactions @xmath11xe+@xmath1ca with radioactive beams .
however , we found that for the production of neutron - rich w the fragmentation reactions are more preferable than the complete fusion reactions .
even if we consider here the formation of neutron - rich w isotopes as an example , our findings have general validity and are not restricted to specific isotopes .
exotic nuclei with large deformations which could be used as projectiles can equally be found in wide regions on the neutron - rich as well as on the neutron - deficient side of the nuclear chart .
we concluded also that the complete fusion @xmath23cs+@xmath24ga reaction with radioactive beam @xmath23cs is not superior to fragmentation for the production of neutron - deficient isotopes of @xmath108rn .
the fragmentation reactions result in slightly larger yields of these isotopes .
note that the choice of the method of production of the isotopes near the drip lines would be also affected by the purposes of the experiments and the available facilities .
99 w. loveland , phys .
c * 76 * , 014612 ( 2007 ) ; _ ibid _ * 75 * , 069801 ( 2007 ) .
haustein , e.m .
franz , s. katcoff , n.a .
morcos , h.a .
smith , jr . , and t.e .
ward , phys .
c * 14 * , 645 ( 1976 ) .
j. benlliure , k .- h .
schmidt , d. cortina - gil , t. enqvist , f. farget , a. heinz , a.r .
jurghans , j. pereira , and j. taieb , nucl .
* a660 * , 87 ( 1999 ) .
sargsyan , g.g .
adamian , n.v .
antonenko , and w. scheid , eur .
j. a * 45 * , 125 ( 2010 ) ; v.v .
sargsyan , g.g .
adamian , n.v .
antonenko , w. scheid , and h.q .
zhang , eur .
j. a * 47 * , 38 ( 2011 ) ; j. of phys .
: conf . ser .
* 282 * , 012001 ( 2011 ) ; epj web conf . *
17 * , 04003 ( 2011 ) ; v.v .
sargsyan , g.g .
adamian , n.v .
antonenko , w. scheid , c.j .
lin , and h.q .
zhang , phys .
c * 85 * , 017603 ( 2012 ) ; phys .
c * 85 * , 037602 ( 2012 ) .
sargsyan , g.g .
adamian , n.v .
antonenko , w. scheid , and h.q .
zhang , phys .
c * 84 * , 064614 ( 2011 ) ; phys .
c * 85 * , 024616 ( 2012 ) .
kuzyakin , v.v .
sargsyan , g.g .
adamian , n.v .
antonenko , e.e .
saperstein , and s.v .
tolokonnikov , phys .
c * 85 * , 034612 ( 2012 ) .
s. ayik , b. yilmaz , and d. lacroix , phys .
c * 81 * , 034605 ( 2010 ) .
h. hofmann , phys .
rep . * 284 * , 137 ( 1997 ) .
s. ayik , b. yilmaz , a. gokalp , o. yilmaz , and n. takigawa , phys .
c * 71 * , 054611 ( 2005 ) ; v.v .
sargsyan , z. kanokov , g.g .
adamian , and n.v .
antonenko , part . nucl . * 41 * , 175 ( 2010 ) ; g. hupin and d. lacroix , phys . rev .
c * 81 * , 014609 ( 2010 ) .
adamian _ et al .
e * 5 * , 191 ( 1996 ) .
s. raman , c.w .
nestor , jr , and p. tikkanen , at .
data nucl .
data tables * 78 * , 1 ( 2001 ) .
p. mller _ et al .
data nucl .
data tables * 59 * , 185 ( 1995 ) .
s. szilner _
et al . _ ,
c * 76 * , 024604 ( 2007 ) ; s. szilner _
c * 84 * , 014325 ( 2011 ) ; l. corradi _ et al .
c * 84 * , 034603 ( 2011 ) .
dasso , s. landowne , and a. winther , nucl . phys . *
a405 * , 381 ( 1983 ) .
adamian , a.k .
nasirov , n.v .
antonenko , and r.v .
jolos , phys .
nucl . * 25 * , 583 ( 1994 ) ; k. washiyama , d. lacroix , and s. ayik , phys . rev .
c * 79 * , 024609 ( 2009 ) ; s. ayik , k. washiyama , and d. lacroix , phys .
c * 79 * , 054606 ( 2009 ) .
canto , p.r.s .
gomes , j. lubian , l.c .
chamon , and e. crema , j. phys .
g * 36 * , 015109 ( 2009 ) ; nucl . phys . *
a821 * , 51 ( 2009 ) .
j. benlliure _ et al .
a * 660 * , 87 ( 1999 ) .
j. gilat , phys .
c * 1 * , 1432 ( 1970 ) ; o.v .
grusha _ et al .
_ , nucl . phys . *
a429 * , 313 ( 1984 ) ; o.v .
grusha , s.p .
ivanova , and yu.n .
shubin , vant , nuclear constants * 1 * , 36 ( 1987 ) ; a.s .
zubov , g.g .
adamian , n.v .
antonenko , s.p .
ivanova , and w. scheid , phys .
c * 68 * , 014616 ( 2003 ) . | the isotopic dependence of the complete fusion ( capture ) cross section is analyzed in the reactions @xmath0xe+@xmath1ca with stable and radioactive beams .
it is shown for the first time that the very neutron - rich nuclei @xmath2w can be reached with relatively large cross sections by complete fusion reactions with radioactive ion beams at incident energies near the coulomb barrier .
a comparison between the complete fusion and fragmentation reactions for the production of neutron - rich w and neutron - deficient rn isotopes is performed . |
in this letter we extend the analysis of @xcite , where a new method for the solution of oscillatory nonlinear problems was devised .
this method works by combining the lindstedt - poincar ( lp ) @xcite method to the linear delta expansion ( lde)@xcite , or optimized perturbation theory ( opt ) @xcite .
a detailed list of references on the lde can be found in @xcite .
the lp method is used to solve approximately nonlinear differential equations with periodic solutions , by introducing strained coordinates , which allow to eliminate the secular terms arising in the perturbative expansion ; in the lde method an insoluble theory is interpolated with a soluble one depending on some arbitrary parameter : an expansion is then carried out in terms of a suitably defined `` perturbation '' . since the `` perturbation '' is not a priori fixed and not necessarily expressed in terms of `` small '' parameters , the method is truly nonperturbative . as proved in @xcite the application of lde to the lp method , which is at the core of our method , allows to extend the analysis to the nonperturbative regime ( large nonlinearities ) and to obtain errors which are small even in regimes where the lp method fails completely . in this letter
we pursue two goals : on one hand to investigate the convergence of our method in the cases previously analyzed to higher orders ; on the other hand to test the method by applying it to more demanding problems , such as the van der pol equation .
the letter is organized as follows : in section [ sec : duff ] we study the duffing equation and discuss the convergence of the method ; in section [ sec : anharm ] we analyze more general anharmonic potentials ( sextic and octic ) ; finally , in section [ sec : vdp ] we consider the van der pol equation . finally in section [ sec : concl ] we draw our conclusions .
we start our analysis by considering the duffing equation , which describes the oscillations of a unit mass in a potential @xmath0 .
it reads : ( t ) + x(t ) = - x^3(t ) .
[ duf1 ] @xmath1 is a coupling which controls the strength of the nonlinear term .
the application of the lplde method requires to write eq.([duf1 ] ) as ^2 + ( 1 + ^2 ) x ( ) = , [ duf2 ] where @xmath2 is an arbitrary parameter and @xmath3 is the exact frequency of the system .
we also introduced a power - counting parameter @xmath4 , which allows to recover eq .
( [ duf1 ] ) when the value @xmath5 is taken .
the r.h.s . of eq.([duf2 ] ) is then treated as a perturbation , although its size is not fixed , being @xmath2 completely arbitrary .
when the perturbative expansion is carried out to finite order , a spurious dependence upon this parameter will show up in all the observables .
the principle of minimal sensitivity ( pms ) will then be used to choose a @xmath2 which minimizes this dependence .
this procedure is explained in detail in @xcite . to a given order @xmath6
we write the frequency and the solution as ^2 = _ n=0^n_max ^n _
n , x(t ) = _
n=0^n_max ^n x_n ( ) .
[ omegax ] and we take @xmath5 at the end . both the coefficients @xmath7 ( which are obtained by eliminating the resonant contributions at any given perturbative order ) and the solutions @xmath8 are @xmath2-dependent .
the dependence upon the arbitrary parameter is then eliminated by imposing the pms condition @xmath9 .
remarkably we have observed that the solution to this equation to any given order is approximated extremely well by the third order solution , i.e. @xmath10 , both for positive and negative values of @xmath1 .
this allows us to obtain expressions which are completely _
analytical_. we can also attach a more physical meaning to the pms condition by looking at fig .
[ fig000 ] : in this figure we plot the energy of the oscillator evaluated at @xmath11 , where all the energy is kinetic , and compare it with the exact expression for the energy , which is given by the potential energy at the inversion point .
the approximate solution found with eq .
( [ omegax ] ) violates the conservation of energy by an amount which depends upon @xmath2 .
however , when the optimal value of @xmath2 is used ( the vertical line in the figure ) such violation is nearly minimal .
.the vertical line is the optimal value to third order .
we assume @xmath12 and @xmath13.,width=340 ] we have noticed that the coefficients @xmath14 of eq .
( [ omegax ] ) can be written as : _ 0 = 1 + a^2 , _ 2 n = - , _ 2 n+1 = 0 [ alpha ] where the @xmath15 are purely numerical coefficients and @xmath16 .
such relations hold for all the @xmath14 that we have calculated . in fig .
[ fig0 ] we plot the logarithm of the coefficients , for @xmath17 ; the dashed line is a linear fit corresponding to : _
n = 0.0663 e^-1.46225 n [ kappa ] with @xmath18 even .
the first few coefficients are written in table [ table1 ] .
the expression in eq .
( [ omegax ] ) can be written equivalently as x_approx(t ) & = & _ n=0^n_max c_n^(approx ) .
[ xapprox ] where @xmath19 are the approximate fourier coefficients obtained with our method . as for the @xmath7 ,
also the @xmath19 are analytical and take the form c_n^(approx ) = _
m=0^n_max _ nm where @xmath20 are the corrections of order @xmath21 to the fourier coefficient corresponding to @xmath22 $ ] .
the latter can be written as : _ nm & = & , where @xmath23 is a numerical coefficient . in fig .
[ fig00 ] we display the logarithm of @xmath24 corresponding to different fourier modes and to different orders in our method : interestingly , we observe that also the @xmath24s decay exponentially with the order of the expansion . ) . the dotted line is a linear fit.,width=340 ]
.[table1 ] first few coefficients of the expansion of eq .
( [ alpha ] ) . [ cols="^,^,^,^",options="header " , ] for @xmath25 .
, width=340 ] for @xmath25 . left plot : coefficients of the @xmath26 ; right plot : coefficients of the @xmath27.,width=604 ]
in this work we have extended the analysis carried out in @xcite and we have applied the lplde method to much higher orders : we have observed that the results obtained converge quickly to the exact values , with errors which decay exponentially with the perturbative order .
this behaviour was observed for the duffing equation , for the sextic and octic oscillator .
remarkably the results obtained to _ any given order _ with our method are fully analytical .
the method was then tested on the van der pol equation where the calculations were performed up to order @xmath28 . in this case
, we were able to obtain good results for moderate values of @xmath1 ( @xmath29 ) , where the lp method already fails .
these results suggest that the extension of the method to higher orders could extend the region of application to larger values of @xmath1 .
moreover , although the optimal value of the variational parameter is not known analytically , as in the previous case , the use of a @xmath1 dependent fit allowed us to obtain also in this case a semi - analytical approximation for the frequency ( period ) which holds in the region of moderate @xmath1 . it remains to explore the application of the present method to a wider class of nonlinear problems with periodic solutions , as well as studying the van der pol equation to higher orders .
this will be done in future work .
the authors would like to thank a.aranda and r. senz for useful discussions .
they also acknowledge the support of the `` fondo alvarez - buylla '' of the university of colima and of conacyt grant no .
c01 - 40633/a-1 .
m. strasberg , _ recherche de solutions periodiques dequations differentielles non lineaires par de methodes de discretisation _ , p. jannsenns , j. mawhin and n. rouche , ed .
hermann , paris , 1973 , pp . | in this letter we apply a method recently devised in @xcite to find precise approximate solutions to a certain class of nonlinear differential equations .
the analysis carried out in @xcite is refined and results of much higher precision are obtained for the problems previously considered ( duffing equation , sextic oscillator ) .
fast convergence to the exact results is observed both for the frequency and for the fourier coefficients .
the method is also applied with success to more general polynomial potentials ( the octic oscillator ) and to the van der pol equation .
epsf .1 in |
the study of spin - orbit ( so ) effects in semiconductor nanostructures has been the object of many experimental and theoretical investigations in the last few years , see e.g. refs . and refs . therein .
it links the spin and the charge dynamics , hence opening the possibility of spin control by means of electric fields.@xcite it has been recently shown@xcite that the so interaction affects the optical properties of gaas quantum wells by inducing a coupling between charge density and spin density excitations in the long wavelength limit .
we extend here this study to the influence on the larmor resonance of the combined effect of both dresselhaus@xcite and bychkov - rashba@xcite so interactions , and use our results to discuss some features of the spin modes disclosed by inelastic light scattering@xcite and electron - spin resonance experiments.@xcite our approach is based on the solution of the equation of motion up to second order in the so intensity parameters.@xcite this method has been also used to derive the kohn theorem,@xcite and goes as follows .
we write the schrdinger equation for a @xmath3-particle system as @xmath4 , with @xmath5 and @xmath6 being the ground state ( gs ) and gs energy , respectively .
if one can find an operator @xmath7 such that @xmath8 , @xmath9 , it is possible to cast the schrdinger equation into an operator equation -the equation of motion- @xmath10 = \omega_n o^+_n$ ] , where @xmath11 is the excitation energy of the state @xmath12 .
the solutions of this equation are used to find the excitation energies of the system as well as its excited states in terms of their creation operators .
this work is organized as follows . in sec .
ii we apply the equation of motion method to the larmor mode in the presence of a so coupling .
the results are used in sec .
iii to discuss the spin modes in quantum wells , and are compared with the experimental results of refs . .
the operators describing the so rashba and dresselhaus interactions are respectively given by @xmath13_j\ ] ] and @xmath14_j~,\ ] ] where the @xmath15 s are the pauli matrices and @xmath16 represents the canonical momentum in terms of the vector potential @xmath17 which in the following we write in the landau gauge , * a*@xmath18 , with * b*=@xmath19 * a * = @xmath20 . in the effective mass , dielectric constant approximation , the quantum well hamiltonian @xmath21 can be quite generally written as @xmath22 , where @xmath23 is the kohn - sham ( ks ) one - body hamiltonian consisting of the kinetic , rashba , dresselhaus , exchange - correlation ks potential and zeeman terms , and @xmath24 is the residual coulomb interaction .
the @xmath25 hamiltonian reads where @xmath27 is the effective electron mass in units of the bare electron mass @xmath28 , @xmath29 , and @xmath30 .
although other approaches may be also considered , we have considered the exchange - correlation potential @xmath31 in the local - spin current density approximation ( lscda).@xcite it depends on the density @xmath32 , magnetization @xmath33 , and local vorticity @xmath34 , and is evaluated from the exchange - correlation energy per electron @xmath35 as @xmath36 . the last term in eq .
( [ eq3 ] ) is the zeeman energy , where @xmath37 is the bohr magneton , and @xmath38 is the effective gyromagnetic factor . for bulk
gaas , @xmath39 , @xmath40 , and the dielectric constant is @xmath41 . to simplify the expressions , in the following we shall use effective atomic units @xmath42 . in the following
, the residual coulomb interaction will be treated in the adiabatic time - dependent lscda ( tdlscda).@xcite we are going to see that , in the absence of so coupling , not only the exact hamiltonian , but also the one in which the residual interaction is treated in the tdlscda fulfill the equation @xmath43=\pm\omega_l s_{\mp}~,\ ] ] where @xmath44 and @xmath45 .
thus , if @xmath5 is the gs of the system , the states @xmath46 are eigenstates of @xmath21 with excitation energies @xmath47 .
note that a negative @xmath38 implies that the spin - up states are lower in energy than the spin - down ones , and that the actual physical solution of eq .
( [ eq4 ] ) is that corresponding to the @xmath48 operator .
this is the physical contents of the larmor theorem .
note also that in the absence of spin - orbit coupling , @xmath49=\omega_c \sum^n_j p^+_j$ ] , where @xmath50 is the cyclotron frequency .
this is the kohn theorem , which also holds in the adiabatic time - dependent local spin density approximation ( tdlsda ) and in the tdlscda , and can be generalized to the case of quantum wires and dots parabolically confined . since @xmath51= \omega_l s_- + 4\sum_{j=1}^n[\lambda_d p^+ \sigma_z+i\lambda_r p^-\sigma_z]_j~,\ ] ] the spin - orbit terms in eq .
( [ eq3 ] ) mix the transverse spin excitations induced by the operator @xmath48 with the spin - density excitations induced by @xmath52 , and thus larmor s theorem is not fulfilled . in the following , we use the equation of motion approach to find the eigenvalues and eigenstates of the ks hamiltonian @xmath23 eq .
( [ eq3 ] ) which arise from the so mixing , and will evaluate the spin wave dispersion relation @xmath53 by taking into account the effect of the residual interaction .
this is done by firstly solving the equation of motion @xmath54=\omega o^+~,\ ] ] and then calculate the transverse response @xmath55 per unit surface @xmath56 in the tdlscda : @xmath57 where @xmath58 , and @xmath59 is the ks transverse response per unit surface.@xcite the poles of @xmath55 yield @xmath53 . the transverse spin response without inclusion of spin - orbit coupling
has been studied in the past in the rpa@xcite and time - dependent hartree - fock@xcite approximations . up to second order in @xmath2 , eq .
( [ eq6 ] ) is straighforwardly solved by the operator @xmath60_j$ ] .
to do so , one has to use the commutators @xmath61=4\sigma_z$ ] and @xmath62=2\omega_c$ ] .
this yields a homogeneous system of linear equations for the coefficients @xmath63 , @xmath64 , @xmath65 and @xmath66 from which the energies @xmath67 are obtained by solving the secular equation valid up to @xmath68 order @xmath69 where @xmath70 .
this quartic equation can be exactly solved , yielding the excitation energies ( only positive solutions are physical ) . for each of them , the homogenous linear system , supplemented with the normalization condition
@xmath71\vert0\rangle=1 $ ] , determines the coefficients @xmath63 , @xmath64 , @xmath65 and @xmath66 .
we have found it more convenient to discuss the solutions of the above equation in the limits of small and large magnetic fields , which are more transparent and easier to compare with available experimental data . in the small @xmath0 limit @xmath72
we obtain a unique solution @xmath73 in the large @xmath0 limit we obtain @xmath74 which are mainly excited by the operators @xmath48 and @xmath75 , and @xmath76 which are mainly excited by the operators @xmath77_j$ ] and @xmath78_j$ ] . by mainly we mean that the coefficient of the corresponding operator entering the definition of is @xmath79 , whereas all the others are @xmath80 .
note that if @xmath81 , the two physical modes in the preceding equations are uncoupled .
equation ( [ eq9 ] ) shows that , at @xmath82 , to order @xmath68 there is no spin splitting due to the so coupling .
indeed , when @xmath83 not only @xmath84 and @xmath85 vary linearly with @xmath0 , but also @xmath86 does , implying that the solution of eq .
( [ eq8 ] ) goes to zero in this limit .
earlier electron - spin resonance measurements on gaas quantum wells@xcite seemed to indicate that a finite spin splitting was present in the @xmath87 limit .
however , subsequent experiments carried out by the same group@xcite covering a broader @xmath0 range point out that the spin splitting of a landau level is an exact quadratic function of @xmath0 , and that its extrapolation to @xmath87 leads to a vanishing spin splitting .
our result , which is not changed by the effect of the residual interaction , is thus in full agreement with the experimental findings of dobers et al.@xcite * we have checked that , at low @xmath0 fields , the dominant component of @xmath88 corresponding to the energy eq .
( [ eq9 ] ) is the spin - flip operator @xmath89_j$ ] . in this limit ,
the dresselhaus and rashba so interactions act `` in phase '' , whereas at high @xmath0 they partially compensate each other [ compare the energy given in eq .
( [ eq9 ] ) with these in eqs .
( [ eq11 ] ) and ( [ eq12 ] ) ] .
this arises from the structure of the secular eq .
( [ eq8 ] ) , where in the low @xmath0 limit , the second term dominates over the third one , whereas in the high @xmath0 limit both terms are equally important , yielding the solutions shown in eqs .
( [ eq11 ] ) and ( [ eq12 ] ) .
we have not been able to find a deeper explanation for this different behavior at low and high magnetic fields .
it is worth to mention that the independent particle hamiltonian can be exactly solved when only the rashba or dresselhaus so terms are included.@xcite the merit of eqs .
( [ eq9]-[eq11 ] ) is that they are exact to the relevant @xmath68 order when both so couplings are simultaneously taken into account . *
the excitation energy @xmath90 is the independent particle ( ks ) value for the spin splitting and violates larmor s theorem even if the so coupling is neglected . on the contrary , when the residual interaction is properly taken into account , the theorem eq .
( [ eq4 ] ) is recovered . in the following ,
we will concentrate on the large @xmath0 limit .
it is then possible to derive the spin wave dispersion relation , including spin - orbit effects , by solving the equation @xmath91 that gives the poles of the transverse response function eq .
( [ eq7 ] ) . to do so ,
we write the transverse spin response as@xcite @xmath92 where @xmath93\vert0\rangle$ ] , and the corresponding energies are given by eqs .
( [ eq11 ] ) and ( [ eq12 ] ) .
the calculation of the matrix elements in eq .
( [ eq14 ] ) must be done with care since @xmath5 and @xmath94 are not eigenstates of @xmath95 because of the spin - orbit coupling . neglecting terms in @xmath96 or smaller , one gets@xcite @xmath97 where @xmath98 is the gs elastic form factor .
note that @xmath99 and that @xmath100 goes to zero when @xmath101 . from eq .
( [ eq13 ] ) one finally obtains @xmath102 this is the main result of our work , together with the lack of so splitting we have found in the small @xmath0 limit . in the limit @xmath101 , eq .
( [ eq17 ] ) yields the independent particle spin splitting eq .
( [ eq11 ] ) , which crucially depends on the actual value of @xmath86 entering the definition of @xmath103 . in the @xmath104 limit ,
neglecting terms of order @xmath105 , eq .
( [ eq17 ] ) reduces to@xcite
@xmath106 this expression shows that , neglecting the so coupling , larmor s theorem is fulfilled in the adiabatic tdlscda ( it can be shown that the same holds in the adiabatic tdlsda ) , and that at high magnetic fields , the so interaction yields a @xmath0 independent contribution to the spin splitting . taking e.g. @xmath107 @xmath108ev , @xmath109 @xmath108ev , which have been recently used to reproduce the spin splitting in quantum dots@xcite and the splitting of the cyclotron resonance in quantum wells,@xcite we get @xmath110 @xmath108ev .
this is definitely a small amount , but it may have an influence on the fine analysis of some experimental results ( note the vertical scale in fig . [ fig1 ] ) .
using inelastic light scattering , davies et al.@xcite and kang et al.@xcite have measured charge and spin density excitations in 2d electrons systems confined in gaas quantum wells at high @xmath0 . in the following
we only discuss the results of ref . because the information presented in fig .
2 of this reference is especially well suited for the purpose of our work .
these results are represented in fig .
the data labeled @xmath111 correspond to wave - vector allowed scattering from the @xmath104 larmor mode ; indeed , the maximum in - plane @xmath112 allowed by the experimental geometry is small , @xmath113 @xmath114 , so that the @xmath111 mode of energy @xmath115 should correspond to the spin splitting energy eq .
( [ eq18 ] ) .
the data labeled @xmath116 is attributed to disorder - activated scattering , and would correspond to @xmath117 excitations of energy @xmath118.@xcite the difference between @xmath118 and @xmath119 is attributed to the exchange enhancement of @xmath118 above the zeeman energy.@xcite the dashed straight line represents the larmor energy taking for @xmath38 the bulk value , @xmath120 .
this overestimates @xmath115 , especially at high @xmath0 .
any sensible comparison with these results must take into account the @xmath0 dependence of @xmath38 .
this dependence has been clearly established in magnetoresistivity experiments,@xcite taking advantage of the fact that the electric - spin resonance affects the magnetoresistivity of the 2d electron gas , and this can be used to determine @xmath121 .
these experiments probe the one - electron energy levels and are not influenced by many - electron interactions , contrarily to magnetoquantum oscillations , which are strongly influenced by many - electron interactions .
the spin splitting obtained in ref .
is represented in fig .
[ fig1 ] as dots and crosses , which correspond to two different samples . in the lowest landau level , which is the common physical situation for all data represented in fig .
[ fig1 ] , these authors have fitted @xmath121 as @xmath122 .
the value of the parameters @xmath123 and @xmath124 turns out to sensibly depend on the experimental sample , and the possibility of a so shift at high @xmath0 values could not be considered there . moreover ,
the @xmath2 values are rather poorly known and dependent on , e.g. , the thickness of the experimental sample .
we have thus renounced to use the laws @xmath121 obtained in ref . in conjunction with eq .
( [ eq18 ] ) , to establish a clear evidence of spin - orbit effects on the @xmath115 energy obtained from resonant inelastic light scattering experiments , and have satisfied ourselves with the more limited scope of using eq .
( [ eq18 ] ) as a three - parameter law to fit @xmath115 as well as the spin splittings of ref .
, with the aim of seeing whether a reasonable value for these parameters can be extracted . the solid straight lines in fig .
[ fig1 ] represent the result of such linear fits , whose parameters are collected in table [ table1 ] . in the case of inelastic light scattering , the neglect of the so term in eq .
( [ eq18 ] ) yields an unrealistic @xmath125 , as this value should be smaller than that of bulk gaas due to the penetration of the electron wave functions into the al@xmath126ga@xmath127as barriers .
the dispersion of the electron - spin resonance datapoints@xcite seems to be smaller , and the analysis of the high @xmath0 data might be used to ascertain which so mechanism is dominating in a given sample .
this could be an alternative * or complementary * method to the recently proposed@xcite of using the anisotropy of the spin splitting in single - electron resonant tunneling spectroscopy in lateral quantum dots submitted to perpendicular or parallel magnetic fields .
the analysis of samples 1 and 2 would indicate that in the former , the dresselhaus so is the dominating mechanism , whereas in the later it is the bychkov - rashba one .
* we want to stress that we have extracted the experimental data from a careful digitalization of the original figures . due to the smallness of the effects we are discussing
, we can not discard that this procedure may have had some effect on the value of the parameters determined from the fit , and our analysis should be considered as qualitative to some extent . however , we find it encouraging that the parameters obtained from the fits are meaningful , and within the range of values found in other works.@xcite * we finally discuss briefly the @xmath117 @xmath116 mode . from eqs .
( [ eq17 ] ) and ( [ eq18 ] ) , we have that @xmath128 . at high @xmath112
, this difference is sensibly determined by @xmath86 . in our calculation , as well as in time - dependent hartree - fock@xcite and exact diagonalization@xcite calculations , we have found values of @xmath86 of the order of @xmath129 mev . hence , @xmath130 is about a factor 40 larger than the measured @xmath131 , which has been obtained in the @xmath132 limit where short range correlations are very important in determining the actual value of @xmath133 .
this can be seen by assuming for @xmath100 the independent particle value . using a slater determinant made of fock - darwin single particle wave functions to describe the gs of the system at @xmath134
, one finds @xmath135 , where @xmath136 is the magnetic length @xmath137 . in the small @xmath112 limit , @xmath138 . for @xmath139 t , @xmath140 and @xmath141 mev , which is about one tenth of the experimental result as shown in fig .
light scattering experiments at small @xmath112 and high @xmath0 are thus very sensitive to correlation effects in the elastic form factor , which is the key quantity to reproduce the experimental findings .
this work has been performed under grants fis2005 - 01414 from dgi ( spain ) and 2005sgr00343 from generalitat de catalunya .
e. l. has been suported by dgu ( spain ) , grant sab2004 - 0091 . .
squares and diamonds , from ref . ; dots ( sample 1 ) and crosses ( sample 2 ) , from ref . .
the dashed line corresponds to @xmath142 with @xmath143 , and the solid lines are fits using the law eq .
( [ eq18 ] ) .
, width=604 ] | we have studied the relevance of spin - orbit coupling to the dispersion relation of the larmor resonance observed in inelastic light scattering and electron - spin resonance experiments on gaas quantum wells .
we show that the spin - orbit interaction , here described by a sum of dresselhaus and bychkov - rashba terms , couples zeeman and spin - density excitations . we have evaluated its contribution to the spin splitting as a function of the magnetic field @xmath0 , and
have found that in the small @xmath0 limit , the spin - orbit interaction does not contribute to the spin splitting , whereas at high magnetic fields it yields a @xmath0 independent contribution to the spin splitting given by @xmath1 , with @xmath2 being the intensity of the bychkov - rashba and dresselhaus spin - orbit terms . |
we say that a hyperbolic riemann surface is of finite type if it has the finite area with respect to the underlining hyperbolic metric .
such surfaces are either closed or are obtained from closed surfaces after removing at most finitely punctures .
all riemann surfaces in this paper are hyperbolic and of finite type ( except the unit disc / upper half space which is the universal cover of such surfaces ) .
let @xmath0 and @xmath1 be two finite type riemann surfaces that are both either closed or both have at least one puncture .
then there is always a common holomorphic ( possibly branched ) cover of @xmath0 and @xmath1 . however a generic pair of such surfaces will not have a common holomorphic , unbranched finite degree cover . except the universal cover , from now on all covers in this paper
will be assumed to be holomorphic , unbranched and finite degree , so every time we use the term cover this will be understood . since we assume that both @xmath0 and @xmath1 are either closed or have at least one puncture , one can find covers @xmath8 and @xmath9 , of @xmath0 and @xmath1 respectively , that are quasiconformally equivalent ( there is a quasiconformal map between them ) .
this is equivalent to saying that @xmath8 and @xmath9 have the same genus @xmath10 and the same number of punctures @xmath11 ( both @xmath8 and @xmath9 are of the type @xmath12 ) .
.1 cm the well - known ehrenpreis conjecture asserts that for a given @xmath2 one can find covers @xmath8 and @xmath9 , of @xmath0 and @xmath1 respectively , so that @xmath8 and @xmath9 are quasiconformally equivalent and the distance between them is less than @xmath7 . since @xmath8 and @xmath9 are quasiconformally equivalent they belong to the same moduli space @xmath13 , where @xmath10 is the genus and @xmath11 is the number of punctures in @xmath8 .
recall that @xmath13 is the space of all hyperbolic metrics one can put on a surface that has genus @xmath10 and @xmath14 punctures .
the distance between them is measured in terms of a metric that exists on @xmath13 .
originally the problem was posed in terms of any natural metric on the corresponding @xmath13 ( see @xcite ) .
note that there are two cases of this conjecture , the first is when @xmath0 and @xmath1 have punctures , and the second when they are both closed .
in fact there is no easy way showing that one can find the corresponding covers @xmath8 and @xmath9 so that the distance between them in their moduli space is strictly less than the distance between @xmath0 and @xmath1 in their moduli space ( providing that @xmath0 and @xmath1 are homeomorphic ) .
recall that the teichmller space is the universal holomorphic cover of @xmath13 .
the two standard ( and most studied ) metrics on @xmath15 that are well defined on @xmath13 are the teichmller metric and the weil - petersson metric . in this paper
we prove this conjecture in the case of punctured surfaces and where the distance is measured with respect to the normalised weil - petersson metric .
we stress that our results do not imply the case when the distance is measured with respect to the teichmller metric .
let @xmath16 be a finite type surface and @xmath17 be a cover ( the genera of @xmath16 and @xmath18 is @xmath10 and @xmath19 respectively and the number of punctures is @xmath11 and @xmath20 respectively ) .
then the covering map @xmath21 induces an embedding @xmath22 .
this embedding is an isometry of @xmath15 onto its image with respect to the teichmller metric .
however , if one takes the traditional definition of the weil - petersson metric this embedding increases the distance by multiplying it by the square root of the degree of the cover .
therefore it is important that we normalise the weil - petersson metric on @xmath15 ( and therefore on @xmath13 ) by dividing it by the square root of the hyperbolic area of a surface of the type @xmath12 ( every such surface has the area equal to @xmath23 ) .
after this normalisation the embedding @xmath24 becomes an isometry with respect to the weil - petersson metric as well .
.1 cm let @xmath0 and @xmath1 be two finite type riemann surfaces that both have at least one puncture .
then given @xmath2 one can find covers @xmath5 and @xmath6 of @xmath0 and @xmath1 respectively , so that @xmath5 and @xmath6 are quasiconformally equivalent and the weil - petersson distance between them is less than @xmath7 .
.1 cm we construct the covers @xmath5 and @xmath6 explicitly .
we believe that it can be recovered from our construction that the degree of the covers @xmath3 and @xmath4 is of the order @xmath25 , where @xmath26 is a polynomial that depends on @xmath0 and @xmath1 .
we believe that the degree of this polynomial is independent of @xmath0 and @xmath1 .
.1 cm it has been shown in @xcite that for every @xmath2 there are covers of the modular torus ( this is the punctured torus that is isomorphic to @xmath27 where @xmath28 is a finite index subgroup of @xmath29 ) that are not conformally the same but are @xmath7 close in the corresponding moduli space and with respect to the teichmller metric .
see @xcite for equivalent formulations of the ehrenpreis conjecture ( see also @xcite , @xcite for related results ) .
.1 cm we say that a riemann surface @xmath30 is modular if @xmath30 is isomorphic to @xmath27 where @xmath28 is a finite index subgroup of @xmath29 .
such surfaces are characterised by having an ideal triangulation where all the shears are equal to zero ( a shear coordinate that corresponds to an edge @xmath31 of an ideal triangulation is the signed hyperbolic distance between the normal projections to @xmath31 of the centres of the two ideal triangles that contain @xmath31 as their edge ) .
give a riemann surface @xmath0 , for every @xmath32 we construct a finite degree cover @xmath33 such that @xmath34 has an ideal triangulation where the shear coordinates are small " .
then there exists a modular surface @xmath35 such that all the shear coordinates of the corresponding ideal triangulation are equal to zero .
one has to make this precise and in particular be able to estimate the weil - petersson distance between @xmath34 and @xmath35 .
we show that the weil - petersson distance between @xmath34 and @xmath35 less than @xmath36 .
it can be recovered from the construction that the degree of the cover @xmath33 is less than @xmath37 , where @xmath38 is a polynomial in @xmath39 .
.1 cm in section 2 .
we discuss the weil - petersson distance and obtain the needed estimates of this distance in terms of the shear coordinates . in section 3 .
we develop the method of construction finite degree covers of @xmath0 by gluing ideal immersed ideal triangles in @xmath0 . in section 4 .
we discuss measures on triangles and the notion of transport of measure .
we state theorem 4.1 which claims existence of certain measures on the space of immersed ideal triangles in @xmath0 .
we prove theorem 1.1 using theorem 4.1 . in section 5 .
we construct the measures from theorem 4.1 and prove this theorem .
heavier computations needed in the proof of theorem 4.1 are done in section 6 and section 7 . in the appendix
we prove an ergodic type theorem about the geodesic flow on a finite area hyperbolic surface with cusps .
this theorem is most likely known but in the absence of an appropriate reference we offer a proof .
let @xmath0 be a riemann surface of the type @xmath12 .
let @xmath40 be the line element for the hyperbolic metric on @xmath0 ( here @xmath41 is the local parameter ) .
denote by @xmath42 the banach space of holomorphic quadratic differentials on @xmath0 with the norm given by @xmath43 here @xmath44 represents the hyperbolic area of @xmath0 .
note that if the @xmath45 than elements of @xmath42 have at most first order poles at the punctures .
.1 cm by @xmath46 we denote the banach space of beltrami differentials on @xmath0 .
these are measurable @xmath47 forms with the finite supremum norm @xmath48 for @xmath49 .
we introduce the equivalence relation on @xmath46 by saying that @xmath50 if @xmath51 for every @xmath52 .
the equivalence class of @xmath53 is denoted by @xmath54 $ ] .
the space @xmath55 is a finite dimensional vector space .
the induced supremum norm on @xmath55 is given by @xmath56||_{\infty}=\inf_{\nu \in [ \mu]}||\nu||_{\infty}.\ ] ] .1 cm the teichmller space @xmath15 is a complex manifold and its tangent space @xmath57 at @xmath0 is identified with the vector space @xmath58 .
the corresponding cotangent space @xmath59 is identified with @xmath42 .
the weil - petersson pairing on @xmath42 is given by @xmath60 the induced scalar product on @xmath61 is called the weil - petersson product .
the corresponding norm of a vector @xmath54 \in l^{\infty}(s)/ \sim$ ] is given by @xmath56||_{wp}=\sup_{\phi \in q(s ) } { { 1}\over{2\pi(2{{\bf { g}}}-2+{{\bf { n}}})||\phi||_2 } } \left|\int\limits_{s}\mu\phi \ , dxdy \right|\ ] ] below we show how to define the norm @xmath62||_{wp}$ ] in terms of harmonic beltrami differentials .
the induced riemannian metric on @xmath15 is called the weil - petersson metric , and by @xmath63 we denote the corresponding distance on @xmath15 . .1
cm this definition of the weil - petersson distance is a modification of the usual one .
if @xmath64 than the usual weil - petersson distance between @xmath26 and @xmath65 is given by @xmath66 .
clearly we have just rescaled the distance by the factor @xmath67 . if @xmath8 is a riemann surface of the type @xmath68 that covers @xmath0 and if @xmath69 is the induced embedding , than @xmath24 is an isometric embedding with our definition of the weil - petersson distance .
also , according to @xcite we have that @xmath70 is strictly less than the teichmller distance between @xmath26 and @xmath65 .
.1 cm we have let @xmath49 .
then @xmath56||^{2}_{wp}\le 9||[\mu]||_{\infty } \left ( \inf_{\nu \in [ \mu]}\int\limits_{s}|\nu|\rho^{2}(z)\ , dxdy \right).\ ] ] let @xmath28 be the covering group of mbius transformations acting on the upper half plane @xmath71 so that the riemann surface @xmath0 is isomorphic to @xmath27
. let @xmath30 be a fundamental domain for the action of @xmath28 .
the lift of @xmath49 to @xmath71 we also denote by @xmath72 . recall that the density of the hyperbolic metric on @xmath71 is given by @xmath73 where @xmath74 .
let @xmath75 denote another complex parameter .
set @xmath76 where @xmath77 .
the function @xmath78 is the bergman kernel for @xmath71 .
the following are the well known properties of @xmath78 .
we have * @xmath79 . * for any mbius transformation @xmath80 we have latexmath:[\[\label{kernel-1 } * for every @xmath77 we have @xmath82 .1 cm let @xmath83(w)={{-12}\over{\pi } } \int\limits_{{{\bf h } } } \overline{\mu(z ) } k(z , w)\ , dxdy.\ ] ] the differential @xmath84(w)$ ] is called the harmonic beltrami differential .
the weil - petersson norm @xmath62||_{wp}$ ] can be expressed as ( see @xcite ) @xmath56||^{2}_{wp}=\int\limits_{s_0 } v^{2}|\phi[\mu](w)|^{2}\ , dudv.\ ] ] let @xmath85 $ ] be any beltrami dilatations from @xmath54 $ ] and use the same notation for the lifts of @xmath86 and @xmath87 to @xmath71
. then @xmath88(w)=\phi[\nu](w)=\phi[\nu_1](w)$ ] .
we have @xmath56||^{2}_{wp } \le \sup_{w \in { { \bf h}}}|v^{2 } \phi[\nu_1](w)| \left(\int\limits_{s_0 } |\phi[\nu](w)|\ , dudv \right).\ ] ] from ( [ kernel-2 ] ) we have @xmath89(w)|\ , dudv.\ ] ] .1 cm we have @xmath90(w)|\ , dudv= { { 12}\over{\pi } } \int\limits_{s_0 } \left|\int\limits_{{{\bf h}}}\overline{\nu(z ) } k(z , w)\ , dxdy \right|\ , dudv \le\ ] ] @xmath91 we partition @xmath71 into the sets @xmath92 , @xmath93 .
this gives @xmath90(w)|\ , dudv \le { { 12}\over{\pi } } \sum\limits_{g \in g } \int\limits_{s_0 } \left(\int\limits_{g(s_0)}|\nu(z)||k(z , w)|\ , dxdy \right)\ , dudv.\ ] ] we have @xmath94 @xmath95 @xmath96 since @xmath97 and from ( [ kernel-1 ] ) we get @xmath90(w)|\ , dudv \le { { 12}\over{\pi } } \sum\limits_{g \in g } \int\limits_{g^{-1}(s_0 ) } \left(\int\limits_{s_0}|\nu(z)||k(z , w)|\ , dxdy \right)\ , dudv=\ ] ] @xmath98 we exchange the integrals in the above inequality to get @xmath90(w)|\ , dudv \le { { 12}\over{\pi } } \int\limits_{s_0 } \left(\int\limits_{{{\bf h}}}|\nu(z)||k(z , w)|\ , dudv \right)\ , dxdy.\ ] ] from ( [ kernel-2 ] ) we have that @xmath99 which shows that @xmath90(w)|\ , dudv \le 3\int\limits_{s_0 } y^{-2}|\nu(z)|\ , dxdy=3\int\limits_{s } \rho^{2}(z)|\nu(z)|\ , dxdy,\ ] ] where @xmath100 is the density of the hyperbolic metric on @xmath0 . together with ( [ wp-1 ] ) this proves the lemma .
the notation we introduce here remains valid throughout this section .
fix a surface @xmath0 of genus @xmath10 and with @xmath45 punctures ( here @xmath0 is not assumed to be a riemann surface , in fact we will equip @xmath0 with various complex structures ) .
by @xmath101 we denote the set of punctures ( recall that @xmath102 is a closed surface and @xmath0 is obtained by removing the set @xmath103 from @xmath102 ) .
let @xmath104 be an ideal triangulation of @xmath0 .
this means that @xmath104 is a triangulation of @xmath102 where the vertex set is exactly equal to @xmath103 ( from now on all triangulations will be ideal triangulations and we will not use the term ideal anymore ) . by @xmath105
we denote the ordered set of edges of the triangles from @xmath104 . here
@xmath106 denotes the total number of edges . for topological reasons we have @xmath107
let @xmath108 denote the total number of triangles in @xmath104 .
we have @xmath109 and therefore the equality @xmath110 holds .
we say that two triangulations @xmath104 and @xmath111 of @xmath0 are isotopic if there is a homeomorphism @xmath112 that pointwise preserves the set @xmath103 and that is homotopic to the identity map on @xmath0 modulo the set @xmath103 .
.1 cm fix a triangulation @xmath104 of @xmath0 .
we define the set @xmath113 as follows .
let @xmath114 and let @xmath115 be the subset of edges from the set @xmath116 that have @xmath117 as an end point .
also , let @xmath118 be defined so that for @xmath119 we have @xmath120 if the endpoints of @xmath31 represent different punctures on @xmath0 and @xmath121 if the two endpoints represent the same puncture on @xmath0 .
let @xmath122 .
then @xmath123 if for every @xmath124 we have @xmath125 clearly @xmath126 is a linear subset of @xmath127 . in particular , we have @xmath128 .
.1 cm now we define the map @xmath129 . recall that @xmath15 is the space of complete hyperbolic metrics on @xmath0 with marking . to a point @xmath130
we associate an element @xmath131 that is represented by the marked hyperbolic metric with the following properties .
there exists a triangulation @xmath132 ( with the corresponding set of edges @xmath133 ) of @xmath0 that is isotopic to @xmath104 and such that all the edges in @xmath133 are geodesics with respect to this metric and so that the following holds ( a triangulation will be called a geodesic triangulation if all the edges are geodesics in the corresponding hyperbolic metric ) .
let @xmath134 and let @xmath135 and @xmath136 be the two triangles from the triangulation @xmath132 that have @xmath137 as an edge .
we lift @xmath137 , @xmath135 and @xmath136 to the universal cover @xmath71 ( the upper half space ) .
the edge @xmath137 lifts to the geodesic that connects @xmath138 and @xmath139 .
one of the triangles @xmath135 or @xmath136 ( depending on how we lift @xmath137 ) lifts to the ideal triangle with vertices at @xmath140 and the other triangle lifts to the triangle with vertices @xmath141 .
this requirement defines the map @xmath142 .
this definition does not depend on how we lift the geodesic @xmath137 ( there are exactly two ways we can lift it ) .
the number @xmath143 can also be defined as the signed hyperbolic distance between the points on the geodesic @xmath137 that are the orthogonal projections of the centres of the triangles of @xmath135 and @xmath136 respectively ( equivalently one can see these two points as the feet of the perpendiculars from the two vertices of @xmath135 and @xmath136 respectively that are opposite to @xmath137 ) .
.1 cm this maps was originally defined by thurston ( this concept has been developed by bonahon , penner , fock , chekhov ... ) .
the corresponding coordinates on @xmath15 are called the shear coordinates .
the map @xmath129 is well defined real analytic homeomorphism .
see @xcite for the proof of this theorem .
this parametrisation of @xmath15 depends on the choice of the triangulation @xmath104 and different triangulations produce different parametrisation .
.1 cm recall that the farey tessellation @xmath144 is the ideal triangulation of @xmath71 with the property that for any two adjacent triangles the feet of the perpendiculars that are dropped from the vertices of these two triangles , that are opposite to their common edge , coincide . to make @xmath144 be a unique tessellation satisfying this property we require that one its triangles is the one with the vertices at @xmath145 .
it is well known that this triangulation is preserved by the action of the group @xmath29 .
let @xmath104 be a triangulation of a finite type surface @xmath0 ( @xmath0 has at least one puncture ) .
then the riemann surface that corresponds to @xmath146 is obtained as the quotient of @xmath71 by a finite index subgroup of @xmath29 .
let @xmath1 be the riemann surface that corresponds to @xmath146 .
we lift the corresponding geodesic triangulation @xmath132 of @xmath1 to @xmath71 and assume that the triangle with vertices at @xmath145 belongs to this lift . let @xmath28 be the corresponding covering group of mbius transformation acting on @xmath71 . since @xmath1 corresponds to @xmath146
we conclude that the corresponding tessellation of @xmath71 is the farey tessellation @xmath144 and @xmath28 preserves @xmath144 . on the other hand , every mbius transformation that preserves @xmath144 must be in @xmath29 .
this shows that @xmath28 is a subgroup of @xmath29 .
the fact that @xmath28 is of finite index follows from the assumption that @xmath0 is of finite type so @xmath1 has finite hyperbolic area .
this simple proposition is important for us .
.1 cm for @xmath123 we define the supremum norm @xmath147 as usual ( this norm does not depend on @xmath104 of course ) .
next , we define a norm on @xmath126 that depends on @xmath104 .
we define the oscillation norm @xmath148 for @xmath123 as follows .
let @xmath28 be the covering group of mbius transformations acting on @xmath71 so that @xmath0 is isomorphic to @xmath27 ( here @xmath0 has the complex structure that corresponds to @xmath149 ) .
let @xmath150 be a k - tuple of edges from @xmath116 .
we say that this k - tuple is a k - tuple of consecutive edges if we can find the lifts @xmath151 to @xmath71 so that each curve @xmath152 has @xmath139 as its endpoint , and so that each curve @xmath152 is to the left of the curve @xmath153 , where @xmath154 .
in this case we also say that the k - tuple of consecutive edges @xmath150 is left oriented ( one similarly defines a right oriented k - tuple of consecutive edges ) .
.1 cm set @xmath155 where the supremum is taken among all consecutive k - tuples @xmath156 ( and for any @xmath157 ) .
note that if @xmath158 is equal to the number of all edges that enter the puncture that corresponds to @xmath139 ( an edge is counted twice if the puncture on @xmath0 that corresponds to @xmath139 is equal to both of its endpoints ) then by definition of @xmath126 we have @xmath159 .
this shows that the supremum in the above definition is achieved and this shows that @xmath148 is a well defined non - negative real number .
note that @xmath160 .
.1 cm our aim here is to estimate from above the weil - petersson distance @xmath161 for a given @xmath123 .
we make this estimate in terms of the vector @xmath162 ( and under certain assumptions on @xmath162 ) . until the end of this section
@xmath123 is a fixed vector .
.1 cm let @xmath163 \to { { \bf { t } } _ { g , n } } $ ] be given by @xmath164 .
the map @xmath165 is differentiable ( since @xmath142 is differentiable ) , and we compute its first derivative in order to estimate the distance . for @xmath166 $ ]
let @xmath167 be the riemann surface that corresponds to the point @xmath168 .
fix @xmath169 $ ] and identify @xmath15 with @xmath170 in the standard way . in order to estimate the distance
@xmath161 we estimate the weil - petersson norm of the vector @xmath171 in the tangent space of @xmath15 at the point @xmath172 .
first we construct an explicit quasiconformal map @xmath173 so that the pair @xmath174 represents the point @xmath175 in @xmath176 ( the requirement is that @xmath177 is homotopic to the identity as a map of @xmath0 onto itself ) .
the homotopy class of map @xmath177 has been studied by penner - saric in @xcite . passing to the universal cover of @xmath167 , they explicitly construct the quasisymmetric map of the unit circle that determines the homotopy class of @xmath177 in terms of the corresponding ideal triangulation of the unit disc .
this has various important applications .
let @xmath178 be the geodesic triangulation of @xmath167 that is homotopic to @xmath104 .
by @xmath179 we denote the corresponding set of edges ( these edges are now geodesics ) . for a triangle @xmath180
we denote by @xmath181 the corresponding triangle in @xmath178 ( for @xmath119 we denote by @xmath182 the corresponding element of @xmath179 ) .
let @xmath183 be the geometric centre of @xmath181 .
for adjacent triangles @xmath184 let @xmath185 denote the geodesic segment between the centres @xmath186 and @xmath187 .
we define the map @xmath177 at the centres @xmath188 by setting @xmath189 .
we define @xmath177 on each @xmath190 so that @xmath191 and by the requirement that @xmath177 stretches the hyperbolic distance by the factor @xmath192 where @xmath193 and @xmath194 are the corresponding hyperbolic lengths .
if @xmath195 is the common edge for @xmath135 and @xmath136 then we have @xmath196 because the point @xmath197 is always the middle point ( in terms of the hyperbolic distance ) of the geodesic segment @xmath185 .
.1 cm fix @xmath180 and let @xmath198 be triangles adjacent to @xmath199 .
the triangle @xmath181 is partitioned into sets @xmath200 where these three sets are separated by the segments @xmath201 , @xmath202 . in particular , the boundary of set @xmath203 is the union of the curve @xmath204 and the two geodesic rays lying on the edges that separate the pairs @xmath205 and @xmath206 respectively .
denote by @xmath207 and @xmath208 the corresponding edges that separate the pairs @xmath205 and @xmath206 respectively .
the map @xmath177 is already defined on the part of the boundary of @xmath209 and we define it on the rest of @xmath209 as follows .
.1 cm we use the same notation for the lifts of triangles from @xmath178 and the edges from @xmath179 to the universal cover @xmath71 .
we assume that for every @xmath166 $ ] the triangle @xmath210 lifts to the triangle in @xmath71 that has the vertices at @xmath145 ( our definition of @xmath177 does not depend on this normalisation ) .
we may assume that @xmath211 is to the left of the triangle @xmath181 in the universal cover .
then the edges @xmath212 and @xmath213 lift to the geodesics in @xmath71 with the endpoints @xmath214 and @xmath215 respectively . we have already seen that the two vertices of the triangle @xmath211 are @xmath138 and @xmath139 .
the third vertex of @xmath211 is at the point @xmath216 ( see figure [ wp ] ) .
similarly , the vertices of @xmath217 are at @xmath218 .
we need to keep track of the vertices for @xmath211 and @xmath217 in order to be able to verify later that the map @xmath177 is well defined on the geodesics @xmath219 and @xmath220 .
.1 cm set @xmath221 let @xmath222 \to ( 0 , \infty)$ ] so that @xmath223 is the graph of the function @xmath224 .
the function @xmath224 has all derivatives everywhere except at the point @xmath225 ( but it has both left and the right derivatives at this point ) . at the point @xmath226
we have that the function @xmath227 depends only on the value of the shear coordinates @xmath228 and @xmath229 .
if we fix an upper bound on the sum @xmath230 we have that the set of such functions @xmath222 \to ( 0 , \infty)$ ] is compact in the @xmath231 topology on both intervals @xmath232 $ ] and @xmath233 $ ] .
this shows that there is a constant @xmath234 that depends only on @xmath230 so that @xmath235 .1 cm the set @xmath203 is given by @xmath236 the map @xmath177 is already defined on @xmath237 .
let @xmath238 \to [ 0,1]$ ] and @xmath239 \to ( 0,\infty)$ ] be the functions so that @xmath240 the functions @xmath241 and @xmath242 depend only on the values @xmath228 and @xmath229 and both of them have all derivatives at every point in @xmath243 except @xmath225 ( but it has the left and the right derivative at this point ) .
note that @xmath241 fixes the points @xmath244 . also , @xmath245 and @xmath246 .
again , by fixing an upper bound on the sum @xmath230 we have that the set of all such functions @xmath238 \to [ 0,1]$ ] and @xmath239 \to ( 0,\infty)$ ] is compact in the @xmath231 topology on the intervals @xmath232 $ ] and @xmath233 $ ] .
this shows that there are functions @xmath247 \setminus{{{1}\over{2 } } } \to { { \bf r}}$ ] that depend only on @xmath230 so that @xmath248 for every @xmath249 , @xmath250 .
moreover , the functions @xmath251 and @xmath252 depend on the real variables @xmath228 and @xmath229 and have all derivatives with respect to these variables too .
if we fix an upper bound on the sum @xmath230 then the set of functions @xmath253 and @xmath254 is compact in the variables @xmath255 , @xmath228 and @xmath229 in the @xmath231 topology . in particular , the derivatives of @xmath256 and @xmath257 with respect to @xmath228 and @xmath229
are bounded from above , and this bound depends only on the sum @xmath230 .
this shows that there is a constant @xmath258 that depends only on the upper bound on the sum @xmath230 , so that @xmath259 ) will be used in the case when the sum @xmath230 is small .
.1 cm we define the map @xmath177 so that @xmath260 .
we have @xmath261 where the function @xmath262 is defined as follows . by @xmath263
we denote a k - tuple of consecutive edges ( see the definition above ) , starting with the edge @xmath212 .
let @xmath264 . then @xmath265 . if @xmath266 then let @xmath267 be the smallest integer so that @xmath268 in this case let @xmath269 $ ] be determined by the identity @xmath270 if @xmath271 then @xmath269 $ ] is given by @xmath272 .
.1 cm the function @xmath273 is defined for @xmath274 and @xmath275 and it is continuous there , while it is differentiable outside a discrete set of smooth curves . from ( [ y - condition ] ) for @xmath266 we have @xmath276 where defined .
for @xmath266 set @xmath277 for @xmath278 set @xmath279 .1 cm for @xmath266 it follows from ( [ p - derivative ] ) that @xmath280 for @xmath278 we have @xmath281 .1 cm as we stated above we define @xmath282 as @xmath283 clearly @xmath284 is a homeomorphism .
it is differentiable outside a discrete set of smooth curves . by repeating the same process
we define @xmath177 on every triangle .
every triangles can mapped by a mbius transformation to the triangle with vertices @xmath145 and this how we defined the map @xmath177 on every triangle .
it is directly seen that the definition of @xmath177 on an edge @xmath285 does not depend on the choice of the triangle that has @xmath285 in its boundary .
for example consider the triangle @xmath286 that is adjacent to @xmath287 ( recall that @xmath181 has the vertices at @xmath288 and @xmath211 has the vertices @xmath289 ) .
let @xmath290 . then @xmath291 .
we define @xmath292 by @xmath293 where @xmath294 .
it is elementary to see that @xmath295 on @xmath219 .
we define @xmath177 on @xmath286 by setting @xmath295 .
.1 cm the beltrami dilatation of @xmath177 is given by @xmath296 set @xmath297 if @xmath298 from ( [ alpha - beta-1 ] ) and ( [ gamma - derivative ] ) we compute @xmath299 @xmath300 @xmath301 if @xmath278 then @xmath302 going back to our path @xmath163 \to { { \bf { t } } _ { g , n } } $ ] we have that @xmath303 represents the tangent vector @xmath171 in the tangent space of @xmath15 at the point @xmath172 .
.1 cm next , we derive two estimates needed in the proofs of theorem 2.1 and theorem 2.2 below . from ( [ u - estimate ] ) , ( [ alpha - beta-2 ] ) , ( [ mu ] ) , ( [ mu-1 ] ) we have @xmath304 which shows that @xmath305 set @xmath306 . since @xmath160 we get @xmath307 ) that the teichmller norm of the tangent vector @xmath171 is less or equal to @xmath308 for every @xmath309 .
let @xmath310 , @xmath311 , be a sequence of surfaces equipped with triangulations @xmath312 and let @xmath313 be a sequence of vectors so that @xmath314 , @xmath315 . since @xmath316 the constant @xmath317 in ( [ mu - supremum - estimate ] ) does not depend on @xmath318 .
this implies that the teichmller distance between the points @xmath319 and @xmath320 tends to @xmath138 when @xmath315 .
.1 cm let @xmath321 .
let @xmath157 such that @xmath322 and such that @xmath323 .
assume that for every @xmath324 we have @xmath325 .
for @xmath264 which satisfies that @xmath326 from ( [ y - condition ] ) we get that @xmath327 .
then from ( [ mu ] ) we have @xmath328 where @xmath329 is a constant that depends only on the upper bound for @xmath330 that is @xmath329 is a universal constant .
.1 cm furthermore @xmath331 @xmath332 in the first integral above we estimate @xmath333 by ( [ mu - estimate ] ) and in the second integral we estimate @xmath333 by ( [ mu - supremum - estimate ] ) .
note that @xmath334 and since @xmath322 we have @xmath335 . from the assumption @xmath323 we have @xmath336 , therefore @xmath337 we conclude @xmath338 where @xmath339 depends only on the upper bound for @xmath330 , that is
@xmath339 is a universal constant .
.1 cm let @xmath0 be a surface of type @xmath12 ( where @xmath340 ) and let @xmath104 be a triangulation on @xmath0 ( where @xmath116 denotes the corresponding set of edges ) .
fix @xmath123 .
let @xmath341 and denote by @xmath342 the number of edges for which the absolute value of the corresponding shear coordinate is greater than @xmath343 .
then there is a constant @xmath344 that depends only on @xmath345 so that for any @xmath157 such that @xmath323 and @xmath322 we have .1 cm @xmath346 important point in this theorem is that the constant @xmath344 does not depend on choice of the triangulation @xmath104 or on the type @xmath12 .
it only depends on the upper bound on @xmath345 .
let @xmath347 . then @xmath348 .
we enlarge the set @xmath349 to the set @xmath350 as follows .
an edge from @xmath116 belongs to @xmath350 if it belongs to a consecutive k - tuple of edges that starts at an edge from @xmath349 ( here we take both the left and the right oriented consecutive k - tuples ) . note that @xmath351 by @xmath352 we denote the set of triangles that have at least one of its edges in @xmath350 .
note that @xmath353 we need to estimate @xmath354 for every @xmath169 $ ] . here
\to { { \bf { t } } _ { g , n } } $ ] is the path defined above and we have @xmath355 .1 cm assume that a triangle @xmath199 does not belong to @xmath352 and let @xmath212 be its edge .
as above , by @xmath356 , we denote the consecutive k - tuple of edges . for each @xmath357
, we have @xmath358 . applying
( [ mu - int-1 ] ) on @xmath209 ( and repeating the same on @xmath359 and @xmath360 ) we get @xmath361 on the other hand , for a triangle @xmath362 from ( [ mu - supremum - estimate ] ) we get the estimate @xmath363 this implies that ( we express the hyperbolic area of @xmath364 as @xmath365 ) @xmath366 where @xmath367 is a constant that only depends on @xmath345 . applying lemma 2.1 and the last estimate we conclude that there is a constant @xmath368 that depends only on @xmath345 such that @xmath369 together with ( [ wp - integral ] ) this proves the estimate ( [ wp - main ] ) and this theorem .
the following theorem is the only result regarding weil - petersson distances that will be used later in this paper .
we use the notation from the previous theorem .
let @xmath370 and assume that for every @xmath371 we are given a finite type surface @xmath34 ( with at least one puncture ) with an ideal triangulation @xmath372 ( the corresponding set of edges is @xmath373 ) that determines a point @xmath374 with the following properties * there exists a universal constant @xmath375 such that @xmath376 for every @xmath371 .
* let @xmath377 .
then there exists a polynomial @xmath38 such that @xmath378 * we have @xmath379 .
then for some polynomial @xmath380 we have @xmath381 in particular the weil - petersson distance between the points @xmath382 and @xmath383 tends to zero when @xmath384 .
let @xmath158 be an integer that is given by @xmath385 then @xmath322 and @xmath386 for @xmath39 large enough .
the proof follows directly from ( [ wp - main ] ) for this choice of @xmath158 .
the aim of this section is to describe how to construct finite degree covers of a given punctured riemann surface of finite type .
we also need to keep the track of the geometry of these covers which is the motivation for the construction below . from now until the end of the paper @xmath0 denotes a fixed riemann surface of type @xmath12 ( we assume that @xmath45 ) .
the corresponding set of punctures in the boundary of @xmath0 ( or simply cusps ) is denoted by @xmath103 .
we number the cusps in @xmath103 as @xmath387 .
we also fix an ideal geodesic triangulation @xmath388 on @xmath0 where @xmath389 denotes the corresponding set of the edges .
most estimates we obtain will depend on the choice of @xmath0 and @xmath388 .
.1 cm let @xmath390 denote the set of all immersed ideal geodesics on @xmath0 ( all geodesics are taken with respect to the underlining hyperbolic metric ) .
a geodesic @xmath391 on @xmath0 is said to be ideal if both of its endpoints are in the set @xmath103 ( note that these geodesics can have self intersections ) .
we will say that a triangle @xmath199 on @xmath0 is an immersed ideal triangle if it lifts to an ideal triangle in the universal cover .
the set of all immersed ideal triangles is denoted by @xmath392 . if @xmath393 then its edges belong to @xmath390 .
.1 cm a geodesic @xmath394 does not carry a natural orientation .
that is , we have a choice of two orientations on each such @xmath391 . by @xmath395
we denote the corresponding set of oriented geodesics .
in particular , the set of oriented geodesics from @xmath389 will be called @xmath396 . for @xmath397 the corresponding unoriented geodesic
is typically denoted by @xmath394 .
.1 cm let @xmath398 denote the space of all formal sums of triangles from @xmath392 over non - negative integers .
similarly , by @xmath399 we denote the space of all formal sums of oriented geodesics over all integers subject to the following rule .
if @xmath400 and @xmath401 represent the same geodesic with opposite orientations then we have the identity @xmath402 .
.1 cm we define the boundary operator @xmath403 as follows .
let @xmath393 and let @xmath404 , @xmath202 , denote its edges .
then @xmath405 where @xmath406 are the corresponding oriented geodesics so that the triangle @xmath199 is to the left of each @xmath406 .
in other words , the orientation of @xmath406 agrees with the orientation that @xmath404 inherits as the part of the boundary of @xmath199 ( such @xmath407 is called an oriented edge of the triangle @xmath199 ) .
the operator @xmath408 is extended to @xmath409 by linearity . .1 cm note that the space @xmath398 naturally embeds into the corresponding group of 2-chains on @xmath0 .
also , @xmath399 naturally embeds into the corresponding group of 1-chains on @xmath0 . the operator @xmath408 we defined agrees with the restriction of the standard boundary operator on @xmath398 .
let @xmath410 denote the first homology of the surface @xmath102 relative to the boundary of @xmath0 which is the set of cusps @xmath103 .
the class of each element of @xmath399 represents an element in @xmath410 .
in particular , for every @xmath411 we have that @xmath412 is equal to zero in @xmath410 ( although @xmath412 is not necessarily equal to zero in @xmath395 ) .
.1 cm by @xmath28 we always denote a fuchsian group so that @xmath0 is isomorphic to @xmath27 .
the lift of @xmath388 , @xmath389 , @xmath396 , @xmath392 , @xmath390 and @xmath395 under this uniformisation will be denoted by @xmath413 , @xmath414 , @xmath415 , @xmath416 , @xmath417 and @xmath418 .
also , by @xmath419 we denote the set of cusps in the boundary of @xmath71 ( we identify @xmath420 with @xmath421 ) , that is , @xmath419 is the set of fixed points of all parabolic elements in @xmath28 .
we can identify the quotient @xmath422 with the set of punctures in the boundary of @xmath0 which we denoted by @xmath103 . if @xmath423 by @xmath424_g$ ] we denote the orbit of @xmath199 under @xmath28 .
we identify @xmath424_g$ ] with the corresponding element of @xmath392 .
.1 cm the set @xmath425 is defined as the set of pairs @xmath426 where @xmath423 and @xmath427 is an oriented edge of @xmath199 . if @xmath93 then @xmath428 as well .
set @xmath429 ( this definition does not depend on the choice of the group @xmath28 ) .
the equivalence class of @xmath426 is denoted by @xmath430_g \in { { \mathcal { t}}}^{*}(s)$ ] . set @xmath431_g)=[t]_g \in { { \mathcal { t}}}(s)$ ] .
then @xmath432 is well defined .
let @xmath393 and let @xmath433 denote its oriented edges .
it can happen that @xmath434 .
this is why we can not define @xmath435 as the set of pairs @xmath426 where @xmath393 and @xmath397 is an oriented edge of @xmath199 . if @xmath436 then @xmath437_g \ne [ ( t_2,-\gamma^{*})]$ ] . to prove this ,
assume that @xmath437_g = [ ( t_2,-\gamma^{*})]$ ] .
then there is @xmath93 so that @xmath438 . also , such @xmath439 is not the identity map .
but then @xmath440 . this implies that @xmath439 fixes the endpoints @xmath441 of @xmath442 .
since @xmath443 we have that @xmath439 is a parabolic element of @xmath28 and therefore @xmath444 is the unique fixed point of @xmath445 .
this contradicts the fact that @xmath446 .
.1 cm fix @xmath423 and let @xmath447 be the standard rotation around the centre of @xmath199 for the angle @xmath448 ( this rotation is of the order three )
. for @xmath449 set @xmath450 .
the map @xmath451 is a bijection and @xmath452 is the identity map .
it is obvious that @xmath453 and @xmath454 have no fixed points . if @xmath93 then @xmath455 .
this shows that the map @xmath456 given by @xmath457_g)=[(t,{{\operatorname { r}}}_t(\gamma^{*}))]_g$ ] is well defined .
assume that @xmath457_g)=[(t,\gamma^{*})]_g$ ] .
then there is @xmath93 so that @xmath458 and so that @xmath459 .
this implies that @xmath439 and @xmath460 agree at the vertices of @xmath199 which shows that @xmath461 .
this is not possible since @xmath28 is torsion free .
the conclusion is that the map @xmath462 has no fixed points . since @xmath463 is the identity map ,
we conclude that @xmath464 has no fixed points either .
note that @xmath465 .
.1 cm let @xmath466 be a finite set of labels , and let @xmath467 be a labelling map .
we say that the pair @xmath468 is a labelled collection of triangles if the following holds : * there exists a bijection @xmath469 so that @xmath470 is the identity map . * we have @xmath471 .
since @xmath462 and @xmath464 have no fixed points , we see that @xmath472 and @xmath473 have no fixed points either .
.1 cm choose @xmath397 and let @xmath474 be its lift to @xmath71 .
we say that @xmath475 if @xmath476_g$ ] for some @xmath423 so that @xmath477
. then @xmath478 .
it follows from the above discussion that the sets @xmath479 and @xmath480 are disjoint ( if the were not there would be an order two element in @xmath28 as we showed above ) .
let @xmath481 , @xmath482 , be the set of different edges of triangles from @xmath483 .
choose an orientation for @xmath407 for each @xmath404 .
then @xmath466 is the disjoint union @xmath484 .
.1 cm set @xmath485 .
let @xmath486 , where @xmath487 is the number of elements in the preimage @xmath488 . since @xmath472 has no fixed points , and since @xmath465 we have that @xmath489 , for some integer @xmath490 .
then each @xmath491 induces an element in @xmath398 given by @xmath492 .
if there is no confusion we will denote this element of @xmath398 by @xmath491 as well .
we define @xmath493 as the boundary of the corresponding element of @xmath398 .
every @xmath411 where @xmath494 and @xmath490 are positive integers , induces a labelled collection of triangles as follows ( this will be used in the proof of theorem 1.1 in the next section ) .
the corresponding set of labels is @xmath495 ( note that the set @xmath466 has @xmath496 elements ) .
let @xmath497 be a lift of @xmath498 .
let @xmath499 , @xmath500 , be its oriented edges , so that @xmath501 .
the corresponding labelling map @xmath502 is given by @xmath503_g$ ] , for @xmath504 .
the required bijection @xmath469 is given by @xmath505 .
let @xmath491 be a labelled collection of triangles .
let @xmath506 be a bijection .
we say that the pair @xmath507 is an @xmath508 @xmath509 if the following holds : * the map @xmath510 is an involution , that is , @xmath511 is the identity map , and @xmath510 has no fixed points . * if @xmath512 and @xmath513 then there are triangles @xmath514 and @xmath427 such that @xmath515_g$ ] and @xmath516_g$ ] ( recall that @xmath517 denotes the opposite orientation of @xmath442 ) .
if @xmath507 is an admissible pair , we have two bijections @xmath518 . the group generated by these two bijections is denoted by @xmath519 .
.1 cm one can construct examples of admissible pairs as follows .
let @xmath520 be a cover , where @xmath8 is a finite type surface ( recall that in this paper all the covers , except the universal cover , are assumed to be finite degree , regular and holomorphic ) . then there is a fuchsian group @xmath521 so that @xmath8 is isomorphic to @xmath522 .
let @xmath523 be a geodesic triangulation of @xmath8 and let @xmath524 denote its lift to @xmath71 .
by @xmath525 we denote the set of pairs @xmath426 where @xmath526 and @xmath442 is an oriented edge of @xmath199 .
same as above , the group @xmath527 acts on @xmath525 . by @xmath430_{g_{1}}$ ] we denote the corresponding orbit , and by @xmath528 we denote the corresponding set of orbits .
same as above , we define the map @xmath529 .
then @xmath530 is a bijection of order three .
.1 cm set @xmath531 and let @xmath532 . for @xmath430_{g_{1 } } \in \tau^{*}_1(s_1)$ ] set @xmath533_{g_{1}})=[(t,\gamma^{*})]_g \in { { \mathcal { t}}}^{*}(s)$ ] .
one can verify that @xmath471 .
then @xmath468 is a labelled collection of triangles .
we now define the involution @xmath510 .
.1 cm let @xmath534 .
we want to define @xmath535 .
let @xmath526 and @xmath427 be its oriented edge , so that @xmath536_{g_{1}}$ ] .
let @xmath135 be the unique triangle in @xmath524 that is adjacent to @xmath199 along @xmath442 .
set @xmath537_{g_{1}}$ ] .
clearly @xmath538 so @xmath539 is well defined .
moreover , @xmath510 is of order two .
if @xmath510 had a fixed point , then there would exist an element of order two in @xmath28 which is not possible .
we have that @xmath507 is an admissible pair .
the above constructed pair @xmath507 is called a @xmath540 @xmath541 @xmath509 .
if @xmath542 , @xmath543 , are virtual triangulations pairs , then we can construct a new admissible pair @xmath507 as follows . set @xmath544 ( here we assume that the sets of labels @xmath545 are mutually disjoint )
. on each @xmath546
the map @xmath472 agrees with @xmath547 .
also , on each @xmath545 the map @xmath502 agrees with the map @xmath548 .
then @xmath491 is a labelled collection of triangles .
we define @xmath510 to agree with @xmath549 on each @xmath545 .
we have that @xmath507 is an admissible pair .
let @xmath542 , @xmath543 , be admissible pairs .
we say that an admissible pair @xmath507 is the union of @xmath542 , @xmath543 , if the following holds .
* there exist injections @xmath550 so that the set @xmath466 is the disjoint union of the sets @xmath551 . * for each @xmath552 we have @xmath553 .
* we have @xmath554 and @xmath555 .
we have the following lemma
. for every admissible triangulation pair @xmath507 there exist virtual triangulation pairs @xmath542 , @xmath556 , so that @xmath507 is the union of @xmath542 . we divide @xmath466 into the orbits of the group @xmath519 .
denote these orbits by @xmath545 where @xmath557 goes through some finite set of labels . let @xmath548 , @xmath549 , and @xmath547 denote the restrictions of the the maps @xmath502 , @xmath510 , and @xmath472 on the set @xmath545 .
then each pair @xmath558 is an admissible pair , and @xmath507 is the union of @xmath542 .
moreover , the group @xmath559 acts transitively on @xmath545 .
therefore , in order to prove the lemma , it suffices to prove that every admissible pair @xmath507 for which the group @xmath519 acts transitively on @xmath466 is in fact a virtual triangulation pair .
.1 cm assume that @xmath507 is an admissible pair , so that the group @xmath519 acts transitively on @xmath466 .
let @xmath560 be the space of triples @xmath561 where @xmath449 and where @xmath562 $ ] . there is a unique involution @xmath563 such that @xmath564 if and only if @xmath565 and @xmath566 .
define @xmath567 by @xmath568 .
it is easy to verify that the group @xmath569 is isomorphic to @xmath570 and that it acts freely on @xmath560 .
moreover , the collection of triangles from @xmath416 that appears in a given orbit under this action , is an ideal triangulation of @xmath71 .
.1 cm the group @xmath28 naturally acts on @xmath560 by @xmath571 ( note that @xmath430=[(g(t),g(\gamma^{*}))]$ ] ) . if @xmath572 then by definition we have @xmath573=[(t_2,\gamma^{*}_2)]$ ] .
that is , there is @xmath93 so that @xmath574 .
in particular , this shows that the group generated by @xmath28 and @xmath569 acts transitively on @xmath560 .
also , every element from @xmath28 commutes with every element from @xmath569 .
.1 cm fix @xmath575 and consider the orbit @xmath576 .
let @xmath527 be the subgroup of @xmath28 so that @xmath577 if @xmath439 preserves the orbit @xmath578 .
if @xmath579 then there is @xmath93 so that @xmath574 .
moreover , since @xmath439 commutes with elements from @xmath569 we conclude that @xmath439 preserves the entire orbit @xmath578 .
this shows that the map @xmath580 given by @xmath581 is an injection .
since the group @xmath519 acts transitively on @xmath466 we find that @xmath165 is a bijection .
.1 cm let @xmath8 be the riemann surface isomorphic to @xmath522 .
set @xmath582 and denote by @xmath583 the union of triangles that appear in @xmath584 . then @xmath583 is an ideal triangulation of the surface @xmath8 .
since @xmath583 contains finitely many such triangles ( the map @xmath585 is a bijection onto a finite set @xmath466 ) , we see that @xmath8 is a finite type surface , and the group @xmath527 has finite index in @xmath28 .
now it follows that the admissible pair @xmath507 is a virtual triangulation pair , that via the map @xmath165 corresponds to the triangulation of @xmath8 .
.1 cm consider a labelled collection of triangles @xmath491 as an element of @xmath398 .
assume that @xmath507 is an admissible pair .
we saw above that @xmath466 is the disjoint union @xmath484 where @xmath586 are the different edges of triangles from @xmath587 .
then @xmath588 .
we conclude that the sets @xmath589 and @xmath590 have the same number of elements , which implies that @xmath591 does not figure in @xmath592 .
we have that @xmath592 is equal to zero in the space @xmath399 .
.1 cm on the other hand , if @xmath491 is a labelled collection of triangles so that @xmath592 is zero in @xmath399 then the corresponding sets @xmath479 and @xmath480 have the same number of elements .
again , from the fact that @xmath466 is the disjoint union @xmath484 we can construct an appropriate involution @xmath510 .
however , there are many ways in which we can do this ( unless each @xmath589 has one element ) .
this is where the geometric considerations start .
our aim will be to construct these involutions so that the pair of triangles that are related by the involution satisfies that the hyperbolic distance between the orthogonal projections of the centres of these triangles to their common edge , is as small as possible .
as we said , @xmath28 denotes a fuchsian group so that @xmath0 is isomorphic to @xmath27 . for every @xmath593 by @xmath594
we denote the orbit of @xmath595 under the group @xmath28 .
each such orbit corresponds to a cusp @xmath114 .
fix @xmath117 that is we fix the orbit of some @xmath593 .
let @xmath596 be a mbius transformation , so that @xmath597 and so that the parabolic element in in @xmath598 that generates the corresponding cyclic subgroup of @xmath598 , is the translation @xmath599 ( there are many such mbius transformations @xmath600 but we choose one for each orbit ) .
the group @xmath601 is called the normalised group with respect to the cusp @xmath595 . if @xmath602 then @xmath603 .
there are exactly @xmath11 normalised groups @xmath604 ( where @xmath11 is the number of punctures in the boundary of @xmath0 ) . from now on @xmath28 is one of the normalised groups ( if we do not specify which one , then it can be any one ) .
the constants we introduce below may depend on the choice of the group @xmath28 .
however , since we consider only finitely many such groups , these constants depend only on @xmath0 .
let @xmath605 @xmath593 .
let @xmath600 be the corresponding mbius transformation that conjugates @xmath28 onto @xmath604 .
we have : * set @xmath606 where @xmath607 is the imaginary part of @xmath608 .
we say that @xmath609 is the height of the point @xmath610 with respect to the cusp @xmath593 .
* let @xmath423 and denote by @xmath611 its vertices .
set @xmath612 .
note that @xmath613 . *
set @xmath614 .
* let @xmath593 and let @xmath615 be such so that @xmath595 is not one of its endpoints
. denote by @xmath616 the point on @xmath391 so that @xmath617 for any other point @xmath618 .
* let @xmath619 .
set @xmath620 .
we say that @xmath621 is the @xmath622-horoball at @xmath595 .
* let @xmath623 .
we set @xmath624 and @xmath625 .
we establish the following related definitions . if @xmath626 are two triangles so that @xmath627 for some @xmath628 then @xmath629 .
this shows that @xmath630 is well defined for every @xmath393 .
if @xmath631 are equivalent under @xmath28 then the projection of the horoballs @xmath632 and @xmath633 to @xmath0 agree . therefore we can define the @xmath622-horoball @xmath634 for every @xmath635 .
it is well known that any two @xmath138-horoballs on @xmath0 either coincide or are disjoint ( note that the closures of two zero horoballs for @xmath29 may touch , but if a fuchsian group @xmath28 is a covering group of a riemann surface , then this can not happen ) .
therefore , for any @xmath605 there can exist at most one cusp @xmath593 so that @xmath636 .
moreover , given @xmath605 and any interval @xmath637 there are at most finitely many cusps @xmath593 so that @xmath638 $ ] .
this shows that @xmath639 is well defined .
if @xmath93 then @xmath640 .
this shows that for @xmath641 the value @xmath642 is well defined . also , by @xmath643 we denote the projection of the set @xmath624 to @xmath0 .
we have @xmath644 .
.1 cm in the following proposition we introduce certain constants that will be used in the proof of the correction lemma below .
the constant @xmath645 will be used throughout the paper . since @xmath0 is of finite type
the following numbers are well defined .
there exist a large enough number @xmath646 a sufficiently negative number @xmath647 and an integer @xmath648 with the following properties : * let @xmath593 and let @xmath649 denote the number of non - equivalent geodesics from @xmath414 that end at @xmath595
. then the integer @xmath650 is well defined .
* let @xmath593 and let @xmath651 .
then @xmath652 is large enough so that @xmath610 does not belong to any @xmath652-horoball , except possibly the one at @xmath595 .
* we have @xmath653 and @xmath654 .
* let @xmath655 and let @xmath441 be its endpoints .
let @xmath656 be one of the two triangles that have @xmath391 in their boundary .
then the number @xmath657 is sufficiently negative , so that for every point @xmath658 and @xmath618 we have @xmath659 and @xmath660 . if @xmath661 then for every @xmath593 we have @xmath662
there are only @xmath663 non - equivalent geodesics in @xmath414 . also , the set @xmath664 is compact in @xmath71 .
the propositions follows from the basic compactness argument .
h if @xmath665 is an arbitrary geodesic segment ( or finite or infinite length ) , by @xmath666 we denote the number of ( transverse ) intersections between @xmath391 and edges from @xmath414 . in particular ,
if @xmath615 the intersection number @xmath666 is finite . for a geodesic segment @xmath667 by @xmath668
we denote the number of ( transverse ) intersections between @xmath391 and edges from @xmath389 .
let @xmath615 and let @xmath618 .
let @xmath441 be the endpoints of @xmath391 .
we define the @xmath669 @xmath670 @xmath671 as @xmath672 where @xmath673 if @xmath674 and @xmath675,[e^{{{\bf h}}_{c_{2}(z ) } } ] \},\ ] ] if @xmath676 . here
$ ] and @xmath678 $ ] denote the integer parts of @xmath679 and @xmath680 respectively . note that if @xmath658 then @xmath681 . since the 0-horoballs are disjoint , only one of the numbers @xmath682 can be non - negative for any @xmath605 . if @xmath427 represents @xmath391 together with a choice of orientation on @xmath391 then we set @xmath683 .
.1 cm the following lemma will be used in the proof of theorem 1.1 ( in the next section ) .
let @xmath1 be a finite type riemann surface that covers @xmath0 and let @xmath684 be a geodesic triangulation on @xmath1 ( by @xmath685 we denote the corresponding set of edges ) .
let @xmath686 be the vector so that @xmath1 is the underlying riemann surface for the point @xmath687 ( see section 2 . ) .
let @xmath688 .
then @xmath689 .
let @xmath527 be a finite index subgroup of @xmath28 so that @xmath1 is isomorphic to @xmath522 .
by @xmath690 and @xmath691 we denote the corresponding lifts of @xmath684 and @xmath685 .
we have @xmath692 so we can define the height @xmath630 for every @xmath693 ( and by projecting @xmath199 to @xmath1 we define @xmath630 for @xmath694 ) .
.1 cm the sets of cusps for @xmath0 and @xmath1 agree .
choose @xmath593 and let @xmath28 be normalised , and @xmath695 .
let @xmath696 be a k - tuple of consecutive edges that all end at @xmath139 .
let @xmath697 denote the corresponding shear coordinates on @xmath31 @xmath482 .
let @xmath698 be the triangle that has @xmath31 as its edge , and so that @xmath699 is to the left of @xmath700 where the orientation @xmath700 is chosen so that @xmath701 is to the right of @xmath700
. then @xmath702 and @xmath703 @xmath704 since @xmath705 is the maximum of all the sums of the type @xmath706 the lemma follows .
next , we determine a set of independent generators for the homology group @xmath410 .
let @xmath707 be an ideal fundamental polygon for @xmath28 so that the edges of @xmath707 are geodesics from @xmath414 ( every such @xmath707 has @xmath708 edges ) .
choose an orientation on every such edge , so that the identification ( given by @xmath28 ) respects the orientation . denote by @xmath709 the set of oriented geodesics , whose lifts to the universal cover @xmath71 are the oriented geodesics from the boundary of @xmath707 ( if we forget about the orientation , the corresponding set of geodesic is called @xmath710 ) . the set @xmath711 contains @xmath712 different elements .
note that the complement of the set @xmath711 in @xmath0 is simply connected .
this shows that the geodesics from @xmath711 generate the group @xmath410 . on the other hand ,
the dimension of the group is @xmath410 is @xmath712 .
this shows that the set @xmath711 is a set of independent generators .
the following proposition is elementary .
let @xmath411 .
assume that @xmath713 where @xmath714 and @xmath715 .
then @xmath716 for every @xmath557 that is @xmath717 in @xmath399 .
we have already observed that the homology class of @xmath412 is equal to zero in @xmath410 . since @xmath711 are independent generators the proposition follows .
given a labelled collection of triangles @xmath491 our aim is to equip @xmath491 with the corresponding involution to produce an admissible pair .
this of course not always possible . as we saw above , only if @xmath718 in @xmath395 one can do this .
moreover , we want to be able to construct this involution so that the centres of the triangles that are paired are as close as possible . given an arbitrary @xmath491 we can add more labels to the set @xmath466 and
expand the domain and the range of the map @xmath502 so that we are able to construct the corresponding involution .
the purpose of this subsection is to prove the lemma which tells us how many labels we need to add .
.1 cm we first prove a few propositions .
all the distances mentioned below are considered to in the hyperbolic metric ( unless stated otherwise ) .
there exists a constant @xmath719 which depends only on @xmath0 so that the following holds .
let @xmath720 and let @xmath721 @xmath695 be normalised .
let @xmath722 with the endpoints @xmath441 , @xmath723 , @xmath724 , and let @xmath725 .
assume that for @xmath726 the geodesic @xmath727 intersects at least @xmath728 geodesics from @xmath389 that end at @xmath139 .
also , assume that @xmath729 .
let @xmath730 be the geodesic that is orthogonal to @xmath727 and that contains @xmath731 .
let @xmath732 and @xmath733 be the endpoints of @xmath730 .
we have : * there exists @xmath734 so that the geodesic that connects @xmath735 and @xmath139 belongs to @xmath414 and so that @xmath735 is the closest point to @xmath255 subject to the condition @xmath736 .
let @xmath404 , @xmath737 , be the geodesic that connects @xmath738 with @xmath735 and let @xmath199 be the triangle bounded by @xmath727 , @xmath739 and @xmath740 .
then the distance between the centre @xmath741 and the point @xmath731 is less than @xmath742 . *
assume @xmath743 ( the analogous statement holds for @xmath744 ) .
there exists @xmath734 so that the geodesic that connects @xmath735 and @xmath139 belongs to @xmath414 and so that @xmath735 is the closest point to @xmath745 subject to the condition @xmath746 .
let @xmath404 , @xmath737 , be the geodesic that connects @xmath738 with @xmath735 and let @xmath199 be the triangle bounded by @xmath727 , @xmath739 and @xmath740 .
then the distance between the centre @xmath741 , and the point @xmath731 is less than @xmath742 .
the existence of such @xmath735 in both cases follows from the fact that @xmath747 .
we prove the first statement .
assume that @xmath748 ( the other case is done in the same way ) .
clearly @xmath749 it is elementary to see that @xmath750 let @xmath751 . if @xmath752 then @xmath753 .
this implies @xmath754 .
then one shows that @xmath755 .
we prove the existence of the constant @xmath742 by contradiction .
assume that there is no such @xmath742 .
then there exists the corresponding sequence of geodesic @xmath756 and points @xmath757 , @xmath311 , so that @xmath758 when @xmath315 ( here @xmath759 is the corresponding sequence of triangles ) .
let @xmath760 where @xmath761 , @xmath762 , @xmath763 and @xmath764 , are the corresponding sequences of points .
since @xmath765 , we have that the sequence of triangles @xmath766 after passing onto a subsequence if necessary , converges to a non - degenerate ideal triangle @xmath767 in @xmath71 . from the choice of @xmath768
we have that @xmath769 for every @xmath311 .
.1 cm if @xmath770 then from ( [ dis-0 ] ) we conclude that the sequence @xmath771 converges to @xmath772 .
if @xmath773 remains bounded then from ( [ dis-1 ] ) , and from the fact that @xmath774 ( which means that this sequence is bounded from below as well ) , we conclude that the sequence @xmath771 converges to @xmath772 . also , @xmath775 and @xmath776 .
this shows that the sequence of geodesic @xmath777 converges to a proper geodesic in @xmath71 .
moreover , @xmath778 converges to a point @xmath779 . since @xmath780 , @xmath315
, we obtain a contradiction .
the second statement is proved in a similar way .
there exist constants @xmath781 so that the following holds . let @xmath722 and suppose @xmath782 .
let @xmath783 .
let @xmath784 be the geodesic , so that letting @xmath785 the distance @xmath786 is the smallest among all the distances between @xmath731 and the intersection points between @xmath727 and geodesics from @xmath414 ( if there are two such closest points @xmath787 we choose either one of them ) .
assume in addition that if for some @xmath593 the geodesic @xmath727 intersects at least @xmath728 geodesics from @xmath414 that all have @xmath595 as their endpoint , then @xmath731 does not belong to the horoball @xmath788 .
then , 1
. we have @xmath789 and the smaller angle between @xmath727 and @xmath739 is greater than @xmath790 .
2 . let @xmath791 be an endpoint of @xmath739 .
let @xmath792 be the geodesics that connect @xmath444 with the two endpoints of @xmath727 respectively .
let @xmath423 be the triangle bounded by @xmath727 @xmath793 and @xmath794
. then @xmath795 . moreover , there are points @xmath796 and @xmath797 so that the distance between the centre of the triangle @xmath199 and any of the points @xmath798 is less than @xmath799 .
we first prove ( 1 ) .
let @xmath756 , @xmath311 , be a sequence of geodesics , and @xmath800 a sequence of points , so that @xmath756 and @xmath757 share the above stated properties of @xmath727 and @xmath731 .
let @xmath801 and @xmath802 be the corresponding geodesics and points .
assume that either @xmath803 or that the angle between @xmath756 and @xmath804 tends to zero , when @xmath805 .
since @xmath806 we can choose @xmath807 so that after passing to a subsequence if necessary , we have @xmath808 .
therefore , we may assume that @xmath809 .
let @xmath810 be a geodesic in @xmath71 where @xmath811 .
if @xmath803 then the geodesic @xmath810 does not intersect any geodesics from @xmath414 . we conclude that @xmath812 .
if the sequence @xmath813 is bounded , then the sequence of geodesics @xmath804 tends to a geodesic @xmath814 .
if the angle between @xmath756 and @xmath804 tends to zero , when @xmath805 then @xmath815 and we again conclude that @xmath812 .
.1 cm let @xmath816 denote the endpoints of @xmath812 .
since @xmath817 we conclude from proposition 3.1 that @xmath818 belongs to both horoballs @xmath819 and @xmath820 .
therefore , for @xmath821 large enough , and for every @xmath822 we have @xmath823 @xmath824 @xmath825 note that @xmath826 is contained in the horoball @xmath827 ( the same is true for @xmath828 ) . on the other hand , since @xmath811 and @xmath829 ( since by the assumption of the lemma we have that @xmath756 does not belong to @xmath414 ) , we conclude that the number of geodesics from @xmath414 that have either @xmath830 or @xmath828 as their endpoints , and that are intersected by @xmath756 tends to @xmath139 .
let @xmath831 be large enough , so that @xmath832 intersects at least @xmath833 geodesics from @xmath414 that all have @xmath830 as their endpoint ( the case when these geodesics end at @xmath828 is similar ) . from this and from ( [ horo-1 ] )
we obtain a contradiction with the assumption that @xmath834 does not belong to a @xmath835-horoball of any cusp @xmath836 when @xmath832 intersects at least @xmath728 geodesics from @xmath414 that end at @xmath595 .
.1 cm next we prove ( 2 ) .
every geodesic from @xmath414 that intersects @xmath727 either intersects exactly one of the geodesics @xmath793 or .
@xmath794 or this geodesic ends at @xmath444 . since @xmath784 has @xmath444 as its endpoint , we see that @xmath837 .
.1 cm we prove the last part of ( 2 ) by contradiction .
the argument is very similar as in the proof of ( 1 ) .
let @xmath838 , @xmath311 , be a sequence of geodesics , and @xmath839 , @xmath311 , a sequence of points .
let @xmath840 and @xmath841 be the corresponding sequences of geodesics , and let @xmath759 be the corresponding sequence of triangles , where @xmath842 denotes the centres of @xmath759 .
assume that at least one of the distances @xmath843 @xmath844 or @xmath845 tends to @xmath139 when @xmath315 .
same as above , we can assume that @xmath846 and @xmath811 .
from the conclusion ( 1 ) of this lemma that was proved above , we conclude that @xmath759 converges to an ideal triangle @xmath767 . in particular , @xmath847 and @xmath848 .
also , @xmath849 where @xmath850 is the centre of @xmath767 .
since @xmath851 we see that @xmath843 does not tend to @xmath139 .
this means that one of the distances @xmath844 or @xmath845 tends to @xmath139 when @xmath315 .
this implies that at least one of the geodesics @xmath852 or @xmath853 is contained in @xmath854 .
since @xmath646 we have that the set @xmath854 is a union of disjoint horoballs ( the base points of these horoballs are in @xmath419 ) in @xmath71 .
since every geodesic in @xmath71 is a connected set , we conclude that at least one of the geodesics @xmath852 or @xmath853 is contained in a horoball in @xmath71 .
this is a contradiction .
we will use the following definition : let @xmath655 be a geodesic .
let @xmath855 be the two triangles that are adjacent along @xmath391 .
by @xmath856 we denote the intersection point between @xmath391 and the geodesic segment that connects the centres of the triangles @xmath135 and @xmath136 .
we now prove the correction lemma .
let @xmath722 be a geodesic , and let @xmath857 be the geodesic @xmath727 with a chosen orientation .
let @xmath858 be a point .
then there exists an ideal polygon @xmath859 with a triangulation @xmath860 that has the following properties : 1 .
the geodesic @xmath727 is an edge of @xmath859 and all other edges of @xmath859 are geodesics from @xmath861 where @xmath861 is the lift of @xmath710 to @xmath71 .
the polygon @xmath859 is to the right of @xmath862 .
3 . there exists a constant @xmath863 which depends only on @xmath0 so that the centres of any two adjacent triangles from @xmath859 are within the @xmath864 hyperbolic distance .
if @xmath865 is the triangle adjacent to @xmath727 then the distance between the centre of @xmath199 and the point @xmath731 is less than @xmath866 .
if @xmath865 is a triangle adjacent to another edge @xmath391 of @xmath859 then the hyperbolic distance between the centre of @xmath199 and the point @xmath867 is less than @xmath866 .
4 . there exist constants @xmath868 which depend only on @xmath0
so that @xmath869 . here @xmath870 stands for the total number of triangles in @xmath859 .
first consider the case when @xmath871 .
we allow any orientation on @xmath872 .
then we have @xmath873 .
moreover , let @xmath441 be the two endpoint of @xmath727 .
in this case we have that @xmath874 .
we build the corresponding polygon as follows .
let @xmath199 be the triangle from @xmath413 that is to the right of @xmath872 .
if an edge of @xmath199 belongs to the set @xmath861 we do not add any more triangles along this edge .
if an edge of @xmath199 does not belong to @xmath861 we glue the corresponding triangle from @xmath413 to the right of this edge .
we repeat this process until all the edges of the polygon @xmath859 we obtained are all from @xmath861 ( except the edge @xmath727 ) .
this process has to end because the geodesics from @xmath861 are the edges of an ideal fundamental polygon for @xmath28 .
this way we have also constructed the corresponding triangulation @xmath860 of the polygon @xmath859 . .1 cm we repeat this construction for every @xmath873 . if @xmath628 and @xmath875 then the corresponding polygon for @xmath793 is the image of the corresponding polygon for @xmath727 under the map @xmath600 .
so , we need to consider only @xmath663 different geodesics from @xmath414 .
since @xmath876 is a compact set in @xmath71 by the compactness argument we find that there exist constants @xmath877 and @xmath878 so that for every such @xmath879 the corresponding polygon @xmath859 satisfies that @xmath880 . also , the distances between the centres of adjacent triangles in @xmath859 are bounded by @xmath881
. the distances between the centres of boundary triangles , and the corresponding points on the edges of @xmath859 are bounded by @xmath882 .
.1 cm we now prove the general case by induction .
we first determine the constants @xmath864 , @xmath344 from the statement of the lemma ( the constant @xmath883 has already been defined ) .
set @xmath884 let @xmath885}\over{2 } } -2-n(s)}}.\ ] ] recall that we chose @xmath652 large enough so that the denominator is a positive constant .
set @xmath886 the numbers that appear in the definition of @xmath344 and @xmath864 , will appear in the arguments below .
.1 cm the proof is by induction on @xmath887 .
the case @xmath888 has been done above .
fix @xmath311 and assume that the statement of the lemma is true for every non - negative integer that is less than @xmath318 .
we now prove that the statement for @xmath318 .
there are four cases to consider .
.1 cm @xmath889 @xmath890 in this case we assume that @xmath891 . moreover , we assume that if for some @xmath593 @xmath727 intersects at least @xmath728 geodesics from @xmath414 that all end at @xmath595 then the point @xmath731 does not belong to @xmath892 ( then @xmath731 does not belong to @xmath788 either , so we may apply proposition 3.4 ) .
we allow any orientation @xmath872 .
we have @xmath893 .
.1 cm we apply proposition 3.4 .
that is , there are geodesics @xmath792 so that there exists a triangle @xmath423 that is bounded by @xmath727 , @xmath793 and @xmath794
. moreover , there are points @xmath796 and @xmath797 so that the distance between the centre of the triangle @xmath199 and any of the points @xmath798 is less than @xmath799 . also , we have that @xmath894 .
we choose the orientations of @xmath895 and @xmath896 so that @xmath727 is to the left of both of them .
.1 cm since @xmath897 we have @xmath898 and @xmath899 . by the induction hypothesis , the statement is true for both pairs @xmath900 and @xmath901 .
we glue the corresponding triangulated polygons to the right of @xmath895 and @xmath896 respectively .
together with the triangle @xmath199 this gives the needed polygon @xmath859 .
we have ( using @xmath902 ) @xmath903 by the induction hypothesis , the distances between the centres of adjacent triangles in the two polygons we have glued along @xmath793 and @xmath794 are less than @xmath864 . also , by the induction hypothesis , the distances between the points @xmath904 and @xmath905 and the centres of the corresponding triangles in those glued polygons , that contain the points @xmath904 and @xmath905 in the their boundaries respectively , are less than @xmath866 .
since the distance between the centre of the triangle @xmath199 and any of the points @xmath798 is less than @xmath906 we have that the distance between the centre of @xmath199 and either one of the two adjacent triangles to @xmath199 in @xmath859 is less than @xmath864 .
this shows that all the distances between the corresponding points in @xmath859 are within @xmath864 or @xmath866 as required .
this settles the first case .
.1 cm @xmath889 @xmath907 let @xmath441 be the endpoints of @xmath727 .
in this case we assume that @xmath908 .
for the sake of the argument assume @xmath909 .
we allow any orientation on @xmath727 .
normalise @xmath28 that is @xmath910 and @xmath911 .
also , assume that the orientation of @xmath872 is such that the pair @xmath912 has the positive orientation on @xmath872 ( the other case is done analogously ) .
let @xmath913 . then @xmath914=m - k$ ] .
let @xmath734 be the point on @xmath915 defined as follows .
we require that the geodesic that connects @xmath735 and @xmath139 belongs to @xmath414 .
then , @xmath735 is the smallest such point subject to the condition @xmath916 let @xmath917 be the geodesic that connects @xmath918 and @xmath735 ( we fix the orientation of @xmath474 so that @xmath139 is to the left of @xmath474 ) .
since @xmath28 is normalised , we have @xmath919 and this yields @xmath920}\over{4}}+1+n(s).\ ] ] let @xmath921 .
since @xmath922 we have that @xmath923 and therefore by the choice of @xmath652 in proposition 3.1 the point @xmath924 does not belong to any @xmath652-horoball ( except possibly the one at @xmath139 which is not an endpoint of @xmath739 ) .
this implies that @xmath925 .1 cm let @xmath740 be the geodesic that connects @xmath735 and @xmath139 ( we fix the orientation of @xmath926 so that @xmath918 is to the left of @xmath926 ) . by the choice of @xmath735 we have that @xmath927 .
let @xmath928 be the point whose imaginary part is @xmath929 . since @xmath930
we have that @xmath931 does not belong to any @xmath652-horoball except possibly the one at @xmath139 .
we have @xmath932 \le { { e^{{{\bf h}}_{\infty}(z_0)}}\over{4 } } \le { { [ e^{{{\bf h}}_{\infty}(z_0)}]}\over{4}}+1.\ ] ] by the induction hypothesis , we have that the statement of the lemma is true for both pairs @xmath933 and @xmath934 .
we construct the polygon @xmath859 by gluing the corresponding polygon for @xmath739 to the right of @xmath474 and the corresponding polygon for @xmath740 to the right of @xmath926 .
we also add the triangle @xmath423 bounded by @xmath727 @xmath739 and @xmath740 to @xmath859 . combining this with the above estimates ( and the definition of @xmath344 ) we have @xmath935}\over{4}}+1+n(s))+c({{[e^{{{\bf h}}_{\infty}(z_0)}]}\over{4}}+1)+2k \le\ ] ]
@xmath936}\over{2}}+2+n(s))+2k \le\ ] ] @xmath937)+k + \big(k+1- c({{[e^{t_{1}(s)}]}\over{2}}-2-n(s))\big ) \le c{{\mathcal { k}}}(\gamma_0,z_0)+k\ ] ] note that the distance between the centre of @xmath199 and any of the points @xmath938 is less than @xmath939 .
one now shows that the distances between the corresponding points in @xmath859 are within @xmath864 or @xmath866 in much the same way as in the case 1 .
.1 cm @xmath889 @xmath940 the only case left to consider is when @xmath891 but when there exists @xmath593 so that @xmath731 belongs to @xmath892 and where @xmath727 intersects at least @xmath728 geodesics from @xmath414 that end at @xmath595 . in this case
@xmath727 can not have @xmath595 as its endpoint .
moreover , @xmath893 .
assume that @xmath28 is normalised , @xmath721 .
let @xmath441 , @xmath724 , be the endpoints of @xmath727 .
we first consider the case when the orientation of @xmath872 is such that the cusp @xmath595 is to the left of @xmath727 .
.1 cm let @xmath793 be the geodesic that is orthogonal to @xmath727 and that contains @xmath731 .
let @xmath732 be the corresponding endpoint of @xmath793 .
since @xmath727 intersects at least @xmath728 geodesics from @xmath414 that end at @xmath139 we can choose a point @xmath734 so that the geodesic that connects @xmath735 and @xmath139 belongs to @xmath414 and so that @xmath735 is the closest such point to the point @xmath255 subject to the condition @xmath941 ( see proposition 3.3 ) .
let @xmath942 @xmath737 , be the geodesic that connects @xmath738 with @xmath735 .
choose the orientation of @xmath407 so that @xmath139 is to the left of @xmath407 .
let @xmath423 be the triangle bounded by @xmath727 , @xmath739 and @xmath740 .
let @xmath943 , @xmath737 , be the orthogonal projections of the centre @xmath741 to the geodesic @xmath404 .
since @xmath944 , @xmath737 , it is easily seen that @xmath945 .
we conclude that the point @xmath943 does not belong to the @xmath652-horoballs at @xmath738 or @xmath735
. therefore , @xmath946 .1 cm from proposition 3.3 we have that @xmath947 . by the induction hypothesis , the statement of the lemma is true for the pair @xmath948 .
we glue the corresponding polygons to the right of @xmath407 .
let @xmath423 be the triangle bounded by @xmath727 , @xmath739 and @xmath740 .
we add this triangle to obtain the required polygon @xmath859 .
we have @xmath949 by proposition 3.3 the distance between the centre of @xmath199 and the point @xmath731 is less than @xmath742 . by the definition of @xmath924 and
@xmath931 the distance between the centre of @xmath199 and the point @xmath943 is less than @xmath950 ( the distance between the centre of an ideal triangle , and its orthogonal projections to one of its sides is less than @xmath950 ) . in much the same way as before one shows that the distances between the corresponding points in @xmath859 are within the required bounds .
.1 cm [ three ] @xmath889 @xmath951 the assumptions are the same as in case 3 . except that the orientation of @xmath872 is such that @xmath139 is to the right of @xmath872 .
let @xmath745 denote the second endpoint of @xmath793 and assume that @xmath743 ( the case @xmath744 is treated similarly ) .
recall @xmath893 .
let @xmath952 denote the number of geodesics from @xmath414 that all end at @xmath139 and that are intersected by @xmath727 .
let @xmath953 denote the number number of geodesics from @xmath414 that are intersected by @xmath727 and that are to the left of those geodesics we counted for @xmath952 .
let @xmath954 denote the number of geodesics from @xmath414 that are intersected by @xmath727 and that are to the right of those counted for @xmath952 . then @xmath955 .
.1 cm choose a point @xmath734 ( see figure [ three ] ) so that the geodesic that connects @xmath735 and @xmath139 belongs to @xmath414 and so that @xmath735 is the closest such point to the point @xmath745 subject to the condition @xmath956 ( see proposition 3.3 ) .
let @xmath158 be the integer so that @xmath957-\log_{2}[c_3-c_2 ] < k \le \log_{2}[c_2-c_1]-\log_{2}[c_3-c_2]+1,\ ] ] where @xmath958 $ ] and @xmath959 $ ] denote the corresponding integer parts . note that @xmath960n_{\infty}(s)-1 $ ] .
this gives @xmath961n_{\infty}(s).\ ] ] .1 cm first consider the case @xmath962 .
let @xmath963 , @xmath964 , be given by @xmath965.\ ] ] let @xmath966 , @xmath967 , denote the geodesic that connects @xmath968 and @xmath969 .
by @xmath970 denote the geodesic that connects @xmath971 and @xmath969 , @xmath967 ( with this definition , the geodesics @xmath727 and @xmath972 agree ) .
let @xmath973 be the geodesic that connects @xmath444 and @xmath139 , and @xmath974 the one that connects @xmath975 and @xmath139 .
let @xmath976 be the triangle bounded by @xmath404 , @xmath977 and @xmath978 .
by @xmath979 denote the triangle bounded by @xmath973 , @xmath974 and @xmath980 .
.1 cm note that by the construction , the distance between the centres of adjacent triangles @xmath981 and @xmath976 is less than @xmath982 .
it is directly seen that the distance between @xmath983 and @xmath984 is less than @xmath985 . by proposition 3.3
the distance between @xmath731 and the centre of the triangle @xmath135 is less than @xmath742 .
.1 cm let @xmath986 , @xmath987 , be the points that are the orthogonal projections of the centres of triangles @xmath699 to @xmath988 .
since @xmath989 we have that the centre of @xmath699 is high enough with respect to @xmath139 so that @xmath990
. therefore , @xmath991 .
since for both endpoints of @xmath404 , @xmath992 , we have that there are geodesics from @xmath414 that connect those points to @xmath139 we have that @xmath993n_{\infty}(s)-1+m_1+m_2,\ ] ] for @xmath992 . for the pair @xmath994 we have @xmath995n_{\infty}(s).\ ] ] .1 cm on the other hand , let @xmath996 and @xmath997 be the points whose imaginary parts are equal to @xmath998}\over{8}}$ ]
. then @xmath999}\over{8}},\ ] ] and @xmath1000}\over{8}}.\ ] ] also , it is easily seen that the distances between the centre of the triangle @xmath979 and the points @xmath1001 and @xmath1002 are less than @xmath1003 .
.1 cm we now apply the induction hypothesis on the pairs @xmath1004 , @xmath987 , @xmath1005 , @xmath1006 .
we put the corresponding orientations on these geodesics , so that @xmath727 is to the left of them .
after gluing the corresponding polygons to the right of the corresponding geodesics , and adding the triangles @xmath699 and @xmath979 we obtain the polygon @xmath859 .
we have @xmath1007n_{\infty}(s)+cm_2+c(m_1+{{[c_2-c_1]}\over{8}})+c{{[c_2-c_1]}\over{8}}+kk\le\ ] ] @xmath1008}\over{4}}+2^{k-1}[c_3-c_2]n_{\infty}(s)\big ) \le\ ] ] @xmath1009}\over{4}}+ { { [ c_2-c_1]}\over{2}}n_{\infty}(s)+{{1}\over{2}}\big ) \le\ ] ] @xmath1010n_{\infty}(s)-1)+k+\left({{3}\over{2}}c - c{{[c_2-c_1]}\over{4}}+kk \right ) \le c{{\mathcal { k}}}(\gamma_0,z_0)+k.\ ] ] the last part of the above estimate follows from ( [ k - correction ] ) , the fact that @xmath1011 and the property @xmath1012 .
.1 cm let @xmath1013 in the above construction .
let @xmath1014 and @xmath1015 be the geodesics that connect @xmath444 and @xmath918 with @xmath139 , respectively .
let @xmath979 be the triangle bounded by @xmath727 , @xmath1014 and @xmath1015 .
let @xmath996 and @xmath1016 be the corresponding points ( chosen in the same way as above ) .
we now glue the corresponding polygons ( whose existence follows from the induction hypothesis ) , to the appropriate sides of @xmath1014 and @xmath1015 .
we also add the triangle @xmath979 to obtain the polygon @xmath859 .
the rest is proved in the same way as in the case @xmath962 .
let @xmath1018 and @xmath1019 denote the unit normal bundles of @xmath390 and @xmath417 respectively . here
@xmath1018 is the union of the normal bundles @xmath1020 , where @xmath394 .
every point in @xmath1020 is a pair @xmath1021 where @xmath618 and @xmath1022 is the unit vector in the tangent space @xmath1023 at @xmath610 that is orthogonal to @xmath391 .
every point @xmath1024 corresponds to a unique pair @xmath1025 , where @xmath1026 and @xmath618 , such that the vector @xmath1022 points to the left of @xmath442 .
.1 cm connected components of @xmath1018 are the normal bundles @xmath1027 , where @xmath397 , and @xmath1018 is the disjoint union @xmath1028 .
the normal bundle @xmath1027 is the subset of the normal bundle @xmath1020 that contains the unit vectors that are pointing to the left of @xmath442 .
note that for a fixed geodesic @xmath391 and a chosen orientation @xmath442 the space @xmath1027 is a connected 1-manifold .
this manifold is identified with @xmath391 by the obvious projection map @xmath1029 .
therefore , the set @xmath1018 is a 1-manifold with countably many components .
note that the group @xmath28 acts on @xmath1019 and @xmath1030 . since every point in @xmath1019 corresponds to a unique pair @xmath1025 .
therefore , we define the combinatorial length mapping @xmath1031 as @xmath1032 , where @xmath1022 points to the left of @xmath442 . the induced map @xmath1033 is also denoted by @xmath1034 .
let @xmath391 be a geodesic in @xmath71 .
we define the @xmath1035 @xmath1036 @xmath1037 as follows .
let @xmath1038 and let @xmath618 denote the orthogonal projection of @xmath787 to @xmath391 .
then @xmath1039 is the point @xmath1021 such that the geodesic ray that starts at @xmath1021 contains @xmath787 . unless otherwise stated , all measures in this paper are positive , borel measures .
if @xmath1040 is a topological space , then @xmath1041 denotes the space of positive , borel measures on @xmath1040 ( necessarily finite ) .
if @xmath1040 is a countable set , we equip @xmath1040 with the discrete topology .
in particular , by @xmath1042 and @xmath1043 we denote the spaces of positive , borel measures on the set of triangles @xmath392 and the manifold @xmath1018 .
the corresponding spaces of measures on @xmath416 and @xmath1019 are denoted by @xmath1044 and @xmath1045 .
note that @xmath398 can be seen as the subset of @xmath1042 .
each @xmath411 induces a measure in @xmath1042 in the obvious way ( if @xmath1046 then the corresponding measure @xmath1047 satisfies that @xmath1048 ) .
we define the @xmath1049 operator as follows .
the set @xmath392 is a countable set , so every measure from @xmath1042 is determined by its value on every triangle in @xmath392 .
let @xmath393 and let @xmath481 , @xmath202 , denote its edges .
set @xmath1050 where @xmath741 is the centre of @xmath199 .
choose @xmath1047 .
let @xmath1051 be the atomic measure , supported on @xmath1052 , @xmath1053 and @xmath1054 such that @xmath1055 .
we define @xmath1056 as @xmath1057 if @xmath1047 is a finite measure , then @xmath1058 is a finite measure as well .
moreover , the total measure of @xmath1058 is three times the total measure of @xmath72 . in this subsection
we define the notion of equivalent measures .
the following is the standard result in measure theory .
let @xmath1059 , @xmath737 , be two finite measures on the real line @xmath915 .
let @xmath878 .
the following conditions are equivalent : 1 . for every @xmath1060 the following inequalities hold @xmath1061 \le \mu_2(-\infty , a+k ] , \ , \mu_2(-\infty , a ] \le \mu_1(-\infty , a+k].\ ] ] 2
. the total @xmath1062 and @xmath1063 measures of @xmath915 coincide , that is @xmath1064 .
there exist mappings @xmath1065 \to { { \bf r}}$ ] so that @xmath1066 and @xmath1067 . here
@xmath86 denotes the lebesgue measure on the interval @xmath1068 $ ] and @xmath1069 denotes the push - forward of the measure @xmath86 under @xmath1070 .
moreover , the mapping @xmath1070 is non - decreasing and left continuous .
3 . there exists a topological space @xmath1040 with a borel measure @xmath1071 so that the following holds .
there exist mappings @xmath1072 so that @xmath1073 and @xmath1067 .
if either of these three conditions is satisfied we say that @xmath1062 and @xmath1063 are @xmath883-equivalent measures
. if @xmath1074 is a translation , then the measures @xmath1062 and @xmath1063 are @xmath883-equivalent , if and only if the measures @xmath1075 and @xmath1076 are . here
@xmath1077 denotes the push - forward of the measure @xmath1078 by @xmath439 .
since @xmath439 is a homeomorphism the similar statement is true for the pull - backs of @xmath1062 and @xmath1063 .
also , note that the relation of being @xmath883-equivalent is symmetric .
moreover , if a pair of measures @xmath1079 is @xmath1080 equivalent , and a pair @xmath1081 is @xmath1082-equivalent , then the measures @xmath1083 and @xmath1084 are @xmath1085-equivalent . the implications @xmath1086 and @xmath1087 are elementary .
we prove @xmath1088 .
this also proves @xmath1089 since one can take @xmath1090 $ ] .
.1 cm note that ( [ transport - condition ] ) implies that the total @xmath1062 and @xmath1063 measures of @xmath915 coincide , that is @xmath1064 . set @xmath1091
define @xmath1070 as follows .
set @xmath1092 \ge x\},\ ] ] for @xmath737 .
let @xmath1093 .
the fact the infimum of the set @xmath1094 belongs to this set follows from the countable additivity property of measures . in particular , @xmath1095 note that @xmath1070 is non - decreasing .
it is elementary to check that @xmath1066 .
we now show that @xmath1096 .
let @xmath1097 .
it follows from ( [ transport - condition ] ) that @xmath1098 this shows that @xmath1099 .
we conclude @xmath1100 .
similarly one shows @xmath1101 .
this proves the proposition .
for @xmath394 let @xmath1102 be an isometric parametrisation ( from the euclidean metric on @xmath915 to the hyperbolic on @xmath391 ) .
every other isometric parametrisation is obtained by pre - composing @xmath1103 by a translation on @xmath915 .
let @xmath1104 and let @xmath397 .
the restriction of the measure @xmath1105 to @xmath1027 is denoted by @xmath1106 .
same as before , we identify ( in the obvious way ) , the manifold @xmath1027 and @xmath391 .
therefore , the pull - back measure @xmath1107 is well defined .
let @xmath1108 and let @xmath878
. for @xmath394 let @xmath1109 denote the two orientations on @xmath391 .
let @xmath1102 be any isometric parametrisation .
we have * if for every @xmath394 the measures @xmath1110 and @xmath1111 , @xmath737 , are @xmath883-equivalent , then we say that the measures @xmath1105 and @xmath1112 are @xmath883-equivalent . *
if for some @xmath394 the measures @xmath1113 and @xmath1114 are @xmath883-equivalent , we say that the measure @xmath1105 is @xmath883-symmetric on @xmath391 .
if @xmath1105 is @xmath883-symmetric on every @xmath394 then we say that @xmath1105 is @xmath883-symmetric .
the above definition does not depend on the choice of the parametrisation @xmath1103 ( see the above remark ) .
.1 cm the following propositions are needed in the proof of theorem 1.1 below .
let @xmath1115 .
let @xmath1116 be discrete measures with finitely many non - trivial atoms , and suppose that @xmath1117 and @xmath1118 are @xmath883-equivalent .
then there are measures @xmath1119 , @xmath737 , so that @xmath1120 and @xmath1121 and @xmath1122 are @xmath883 equivalent . also , @xmath1123 has atoms of rational weights , and the weight of any atom of @xmath1124 is at most @xmath7 .
let @xmath1125 and @xmath1126 denote respectively the points where @xmath1117 and @xmath1118 have non - trivial atoms .
set @xmath1127 and @xmath1128 .
let @xmath953 be the total number of atoms @xmath1129 and let @xmath954 be the total number of atoms @xmath1130 .
set @xmath1131 . also , let @xmath1132 be the minimum of all non - zero weights of atoms of both measures , and @xmath1133 the maximum of all non - zero weights .
.1 cm since @xmath1117 and @xmath1118 are @xmath883-equivalent , we have that ( [ transport - condition ] ) holds .
since @xmath1117 and @xmath1118 have finitely many atoms , the condition ( [ transport - condition ] ) becomes a finite systems of linear inequalities with integer coefficients , in @xmath1134 and @xmath1135 .
each such inequality has the form @xmath1136 or @xmath1137 where every @xmath1138 , @xmath1139 is either @xmath950 or @xmath138 . if we treat @xmath1134 and @xmath1135 as real variables , then we conclude that this system of linear inequalities has a non - trivial solution .
in fact , the set of solutions of each above inequality is a half - space in @xmath1140 ( each half space contains the origin in @xmath1140 ) .
the set of the solutions of the entire system is the intersection of all these half - spaces in @xmath1140 .
we denote this set by @xmath1141 .
let @xmath1142 and let @xmath1143 .
we have that @xmath1144 is a convex polyhedron ( possibly degenerate ) in @xmath1140 . by the krein - milman theorem
, @xmath1144 is the closure of the convex combinations of the extreme points on @xmath1144 ( there are finitely many extreme points ) .
each extreme point is the unique solution of a certain system of equations with integer coefficients .
therefore , every extreme point is a rational point in @xmath1140 .
we conclude that the rational points in @xmath1140 are dense in every @xmath1144 and therefore the rational points are dense in @xmath1145 .
.1 cm if @xmath1134 and @xmath1135 is the only solution to this system , then these numbers have to be rational .
if this is not the only solution , then we can choose rational solutions @xmath1146 and @xmath1147 to be as close to @xmath1134 and @xmath1135 as we want .
fix any @xmath1148 and let @xmath1146 and @xmath1147 be rational numbers that satisfy all the above inequalities , and such that @xmath1149 and @xmath1150 .
let @xmath622 be a rational number so that @xmath1151 then @xmath1152 and @xmath1153 and also @xmath1154 and @xmath1155 satisfy all the above inequalities .
moreover , the following inequalities hold @xmath1156 and @xmath1157 choose @xmath1158 small enough so that @xmath1159 .
.1 cm let @xmath1121 be the measure with the same non - trivial atoms as @xmath1117 and @xmath1160 .
similarly define @xmath1161 .
set @xmath1162 and @xmath1163 . the measures
@xmath1164 and @xmath1165 are non - negative , and the weight of any atom under either of these two measures is less than @xmath7 . since @xmath1154 and @xmath1155 satisfy the same inequalities as @xmath1134 and @xmath1135 we conclude that @xmath1121 and @xmath1122 are @xmath883-equivalent .
let l be a finite set .
by @xmath1166 we denote the counting measure , that is if @xmath1167 then @xmath1168 is equal to the number of elements in @xmath1169 . on the other hand
, we say that a borel measure is integral if the measure of every set is an integer .
the following proposition is left to the reader .
let @xmath1170 be a finite set , and let @xmath1171 be a function .
suppose that @xmath1172 such that @xmath1173 are integral measures
. then we can write @xmath1174 where @xmath1169 and @xmath1175 are disjoint , such that @xmath1176 and @xmath1177 .
let @xmath1178 and @xmath1179 be finite sets of labels , and let @xmath1180 and @xmath1181 be labelling maps .
suppose that the measures @xmath1182 and @xmath1183 are @xmath883-equivalent .
then we can find a bijection @xmath1184 such that @xmath1185 . since @xmath1105 and @xmath1186 are @xmath883-equivalent
, we conclude that the total mass of @xmath1105 is equal to the total mass of @xmath1186 .
set @xmath1187 . by proposition 4.1
, we can find non - decreasing and left continuous functions @xmath1188 \to { { \bf r}}$ ] such that @xmath1189 where @xmath1190 for @xmath1191 ( here @xmath86 is the lebesgue measure on @xmath1192 $ ] ) . since @xmath1193 is an integral measure , we have that @xmath1194 where @xmath1195 is the least integer greater or equal to @xmath622 .
therefore , we can find bijections @xmath1196 , @xmath1191 , such that @xmath1197 for every @xmath1198 .
.1 cm for @xmath1199 we let @xmath1200 . then @xmath1201 which proves the proposition .
let @xmath1132 and @xmath1133 be finite sets of real numbers .
let @xmath1178 and @xmath1179 be finite sets of labels , and @xmath1202 and @xmath1203 labelling maps .
let @xmath1204 be the atomic measures , that are supported on @xmath1132 and @xmath1133 respectively , and @xmath1205 for every @xmath1206 and @xmath1207 for every @xmath1208 ( here @xmath1209 , @xmath1210 denote the numbers of the corresponding preimages in @xmath1178 and @xmath1179 respectively ) .
suppose that @xmath1211 and @xmath1212 where @xmath1213 are integral measures , such that @xmath1117 , @xmath1214 are @xmath1080-equivalent , and @xmath1118 , @xmath1215 are @xmath1082-equivalent .
denote by @xmath953 the total mass of @xmath1117 and by @xmath954 the total mass of @xmath1118 .
then , there is a bijection @xmath1184 so that for @xmath953 elements @xmath1199 we have @xmath1216 and for the remaining @xmath954 elements @xmath1217 we have @xmath1218 .
note that @xmath1182 and @xmath1183 .
the proof follows from proposition 4.3 and proposition 4.4 .
we end this subsection with the following two elementary propositions that will be used in section 6 .
let @xmath1219 such that @xmath1105 and @xmath1186 are @xmath883-equivalent , and such that @xmath1186 and @xmath1220 are @xmath1170-equivalent
. then @xmath1105 and @xmath1220 are @xmath1221-equivalent .
one directly verifies that the condition @xmath1222 from proposition 4.1 holds for the measures @xmath1105 and @xmath1220 . let @xmath1223 be a measure space and let @xmath1224 , @xmath737 , be such that the mapping @xmath1225 $ ] is measurable for every @xmath619 .
suppose that @xmath1226 and @xmath1227 are @xmath883-equivalent for every @xmath1228 .
then the measures @xmath1229 and @xmath1230 are @xmath883-equivalent . if @xmath1231 is a measurable mapping , and if @xmath1232 is @xmath883-symmetric for every @xmath255 , then the measure @xmath1233 is also @xmath883-symmetric .
one verifies directly that the condition @xmath1222 from proposition 4.1 holds for the measures @xmath1234 and @xmath1235 .
the remainder of the paper after this subsection is devoted to proving the next theorem . from now on @xmath38 denotes a polynomial that only depends on @xmath0 .
in particular we have @xmath1236 and @xmath1237 .
there exist a constant @xmath1238 that depends only on @xmath0 , so that for every @xmath371 there exists a finite measure @xmath1239 so that the total measure @xmath1240 satisfies the inequality @xmath1241 and with the following properties . there exist measures @xmath1242 so that the measure @xmath1243 can be written as @xmath1244 and the following holds 1 .
let @xmath1245 denote the restriction of the measure @xmath1246 to the set of triangles @xmath393 for which @xmath1247
. then @xmath1248 2 .
the measure @xmath1249 is @xmath1250-symmetric , and the measure @xmath1251 is @xmath65-symmetric , for any @xmath1252 .
we have @xmath1253 4 .
we have @xmath1254 we will explicitly construct the required measure @xmath1246 . in the remainder of this section we prove theorem 1.1 assuming theorem 4.1 and its notation .
first , we show that we may assume that the measure @xmath1246 from theorem 4.1 , has finite support .
assume that theorem 4.1 holds .
then there exist a constant @xmath1238 that depends only on @xmath0 so that for every @xmath371 there exists a finite measure @xmath1255 with finite support , so that the total measure @xmath1256 satisfies the inequality @xmath1257 and with the following properties .
there exist measures @xmath1258 so that the measure @xmath1259 can be written as @xmath1260 and the following holds 1 .
the measure @xmath1261 is @xmath1250-symmetric , and the measure @xmath1262 is @xmath65-symmetric , for any @xmath1263 .
we have @xmath1264 3 .
we have @xmath1265 let @xmath1255 be such that @xmath1266 where @xmath1245 was defined in the statement of theorem 4.1 .
then @xmath1267 has finite support since it is supported only on triangles from @xmath392 which satisfy @xmath1268 ( and there are only finitely many such triangles ) . by the assumption in theorem 4.1 we have @xmath1269 .
this shows that @xmath1257 for @xmath39 large enough .
note that the measure @xmath1270 consists of countably many atoms , that are obtained as the feet of the centres of the triangles from @xmath392 .
the measure @xmath1267 consists of finitely many atoms .
it remains to construct the decomposition @xmath1260 .1 cm for a discrete positive measure @xmath1220 we let @xmath1271 be the set of points @xmath619 for which @xmath1272 .
we observe that @xmath1273 and @xmath1274 have disjoint support because if @xmath626 are triangles that share a geodesic @xmath391 , and such that @xmath1275 then @xmath1276 ( recall that @xmath1277 is an element of @xmath1020 so if @xmath1275 then @xmath135 and @xmath136 are on the same side of @xmath391 ) .
we let @xmath1278 be the restriction of @xmath1105 to @xmath1279 and @xmath1280 the restriction of @xmath1105 to @xmath1281 .
one defines @xmath1282 , @xmath1283 , @xmath1284 and @xmath1285 likewise
. then @xmath1286 and @xmath1287 , @xmath1288 and @xmath1289 .
.1 cm we aim to symmetrise " @xmath1278 so that it is @xmath1250-symmetric . to this end for each @xmath394 we choose an orientation @xmath397 for @xmath391 .
let @xmath1290 be the restriction of @xmath1105 to @xmath1027 and @xmath1291 be the restriction of @xmath1105 to @xmath1292 , and think of @xmath1290 and @xmath1291 as measures on @xmath391 .
we define @xmath1293 , @xmath1294 , @xmath1295 and @xmath1296 likewise .
since @xmath1105 is @xmath1250-symmetric by proposition 4.1 we can write @xmath1297 such that @xmath1298 satisfy that @xmath1299 for all @xmath1300
. then we let @xmath1301 be defined by @xmath1302 .
set @xmath1303 .
also , let @xmath1304 . by construction @xmath1305 and @xmath1306 are @xmath1250-equivalent . now consider @xmath1305 and @xmath1306 as measures on @xmath1027 and @xmath1292 respectively .
let @xmath1307 .
then @xmath1308 is @xmath1250-symmetric .
.1 cm we let @xmath1309 .
the reader can verify that there exists a measure @xmath1310 such that @xmath1311 is @xmath1250-symmetric ( the measure @xmath1312 is constructed in the obvious way using the maps @xmath1313 ) .
note that if @xmath1314 satisfy @xmath1315 then @xmath1316 .
it follows that @xmath1317 @xmath1318 for @xmath39 large enough .
.1 cm we likewise find @xmath1319 such that @xmath1320 is @xmath65-symmetric and @xmath1321 satisfy @xmath1322 also @xmath1323 this shows that @xmath1308 and @xmath1320 satisfy the conditions @xmath1222 and @xmath1324 of this proposition .
.1 cm we then let @xmath1325 . then @xmath1326 .
moreover @xmath1327 @xmath1328 since @xmath1329 .
this proves the proposition .
the idea is to use the measures @xmath1267 to construct certain admissible pairs @xmath1330 . in turn
, this will enable us to construct the corresponding covers of @xmath0 that are required by theorem 1.1 .
.1 cm from now on we suppress the dependence on @xmath39 that is we set @xmath1331 , @xmath1332 , @xmath1333 and @xmath1334 . since @xmath72 is finitely supported , the measure @xmath72 is atomic and it has finitely many atoms .
the measure @xmath1058 is atomic and it has finitely many atoms , as well .
let @xmath318 denote the total number of non - trivial atoms for @xmath1058 .
let @xmath1170 be the maximum of the combinatorial length function @xmath1034 over the @xmath318 points where @xmath1058 is supported .
let @xmath1335 .1 cm for @xmath393 choose a number @xmath1336 so that @xmath1337 is a rational number , and set @xmath1338 . if @xmath1339 then we set @xmath1340 as well .
this is how we define two new measures @xmath1341 .
both these measures are atomic ( finitely many atoms ) , and @xmath1342 has the same set of atoms as @xmath72 ( the weights of atoms of @xmath1342 are rational numbers ) . in particular , @xmath1343 has at most @xmath318 atoms , and each of them has the weight at most @xmath7 . note that the total measure of @xmath1342 satisfies @xmath1344 .1 cm the measure @xmath1345 is also atomic , with finitely many atoms as well .
let @xmath394 and let @xmath1346 denote the two orientations on @xmath391
. since @xmath1105 is @xmath1250-symmetric , the two measures on @xmath915 that arise as the restrictions of @xmath1105 on @xmath1347 and @xmath1348 respectively , are atomic ( finitely many atoms ) , @xmath1250-equivalent measures . applying proposition 4.2 on these two measures , and repeating the same for each @xmath394
we construct the measures @xmath1349 with the properties : * @xmath1350 .
* @xmath1351 is @xmath1250-symmetric .
* for any atom @xmath1352 of @xmath1112 we have @xmath1353 .
we repeat the same for @xmath1117 .
we construct the measures @xmath1354 with * @xmath1355 .
* @xmath1351 is @xmath65-symmetric .
* for any atom @xmath1352 of @xmath1164 we have @xmath1356 . we have @xmath1357 note that from the fourth inequality in proposition 4.6 , we have @xmath1358 .1 cm set @xmath1359 . since @xmath1342 is atomic ( finitely many atoms ) with rational weights ,
so is the measure @xmath1360 . since @xmath1351 and @xmath1361 are also atomic ( finitely many atoms ) with rational weights , and
since all the measure in question are positive , we conclude that @xmath1214 is also atomic ( finitely many atoms ) and with rational weights .
moreover , we have @xmath1362 @xmath1363 that is , for @xmath39 large enough we have @xmath1364 .1 cm let @xmath14 be a large enough integer so that the weights of atoms of the measures @xmath1342 , @xmath1351 , @xmath1361 and @xmath1214 are all integers .
we multiply the measures in question by @xmath14 .
set @xmath1365 , @xmath1366 , @xmath1367 and @xmath1368 .
we have that @xmath1369 is still @xmath1250-symmetric , and @xmath1370 is @xmath65-symmetric .
moreover , we have the following estimate for the total measure of @xmath1370 @xmath1371 also , the total measure of @xmath1372 satisfies that @xmath1373 .
.1 cm now we apply the correction lemma ( lemma 3.3 ) .
let @xmath1374 be an atom of @xmath1375 ( note that the same point can be an atom of the measure @xmath1369 or @xmath1370 ) .
choose a lift of @xmath1025 to @xmath1019 and denote it also by @xmath1025 .
we apply lemma 3.3 to this pair , to obtain the corresponding polygon @xmath859 in @xmath71 that is to the right of @xmath442 .
project the triangles from the triangulation @xmath860 to @xmath0 .
let @xmath1376 and let @xmath1377_g = t \in { { \mathcal { t}}}(s)$ ] be its projection .
let @xmath1378 be the measure that is supported on @xmath199 and so that @xmath1379 .
set @xmath1380 where we sum over all such triangles @xmath199 . from lemma 3.3
we have the bound on the number of triangles in @xmath860 and thus we obtain the following estimate of the total measure @xmath1381 here @xmath344 and @xmath883 are the constants from lemma 3.3 .
note that each atom of @xmath1382 has an integer weight .
.1 cm let @xmath864 be the constant from lemma 3.3 , and let @xmath1383 be a geodesic that lifts to an edge of a triangle from @xmath860 .
denote by @xmath1384 the restriction of @xmath1375 to the point @xmath1385 .
then the two measures on @xmath727 that are the restrictions of the measure @xmath1386 on @xmath1387 are @xmath864-equivalent .
this follows from lemma 3.3 , that is , the two atoms of @xmath1386 in @xmath1387 ( one on each side of @xmath727 ) , are within the hyperbolic distance @xmath864 ( these two atoms have the same weight by the definition of @xmath1382 ) .
.1 cm repeat this process for every non - trivial atom of @xmath1375 and set @xmath1388 where we sum over all non - trivial atoms for @xmath1375 .
we have @xmath1389 and @xmath1390 .1 cm let @xmath1383 .
then the two measures on @xmath727 that are the restrictions of the measure @xmath1391 to @xmath1387 are @xmath864-equivalent .
let @xmath1392 .
then by lemma 3.3 all the atoms of @xmath1393 on @xmath727 are within the @xmath1394 hyperbolic distance from the point @xmath1395 . if we can show that the total measures are equal , that is of @xmath1396 that would show that the measure @xmath1391 is @xmath864-symmetric .
.1 cm set @xmath1397 .
again , @xmath1398 has finitely many atoms , and all the weights are integers .
the above estimate for @xmath1399 implies that , for @xmath39 large enough , the total measure of @xmath1398 satisfies the following inequalities @xmath1400 .1 cm set @xmath1401 .
then the following equality holds @xmath1402 .1 cm we have the following estimates on the total measures of @xmath1403 @xmath1404 .1 cm since @xmath1398 has finitely many atoms with integer weights , we can consider @xmath1398 as an element of @xmath398 .
we construct the labelled collection of triangles @xmath491 that corresponds to @xmath1405 ( see remark before definition 3.1 ) .
let @xmath16 denote the total number of elements of @xmath466 ( clearly @xmath16 is equal to the total measure of @xmath1406 ) .
from ( [ number - triangles ] ) we have @xmath1407 .
.1 cm fix @xmath1408 and choose an orientation @xmath442 on @xmath391 .
we have already seen that the restriction of the measure @xmath1398 on @xmath1020 can be written as the sum of three measures from @xmath1409 where each of these three measures produces a pair of measures on @xmath391 that are equivalent ( for some constant ) .
this implies that the sets @xmath1410 have the same number of elements .
this shows that such @xmath442 does not figure in the formal sum @xmath1411 .
so @xmath1412 where @xmath714 and @xmath715 . by proposition 3.2
we have that every @xmath716 that is @xmath718 in @xmath399 .
thus for every @xmath394 the total measures of @xmath1413 on @xmath1027 and on @xmath1292 are the same .
as indicated above , this proves that the measure @xmath1401 is @xmath864-symmetric .
.1 cm set @xmath1414 .
then @xmath1415 is @xmath1416-symmetric , and from ( [ total - measure-2 ] ) , ( [ total - measure-1 ] ) , for @xmath39 large enough , we have @xmath1417 .1 cm let @xmath394 and consider the restriction of the measure @xmath1413 to @xmath1020 .
this produces a pair of measures in @xmath1418 .
apply proposition 4.5 to this pair of measures .
choose an orientation @xmath442 on @xmath391 .
recall the notation from proposition 4.5 .
we say that @xmath1419 if @xmath1025 is a non - trivial atom of the measure @xmath1413 .
we say that @xmath1420 if @xmath1421 is a non - trivial atom of the measure @xmath1413 .
we identify @xmath1178 with @xmath1422 and @xmath1179 with @xmath1423 .
we define the involution @xmath1424 by @xmath1425 where @xmath1426 is the bijection from proposition 4.5 .
this is how we construct the admissible pair @xmath507 .
.1 cm fix @xmath534 .
let @xmath1427 such that @xmath1428 $ ] and @xmath1429 $ ] .
let @xmath1430_g={\operatorname{pr}}_s({\operatorname{lab}}_{{{\mathcal { c}}}}(a))=t \in { { \mathcal { t}}}(s)$ ] and @xmath1431_g={\operatorname{pr}}_s({\operatorname{lab}}_{{{\mathcal { c}}}}(\sigma_{{{\mathcal { c}}}}(a)))=t ' \in { { \mathcal { t}}}(s)$ ] . also , let @xmath397 be the projection of @xmath474 to @xmath0 and finally let @xmath394 be the corresponding unoriented geodesic .
we have @xmath1432 where @xmath1433 .
let @xmath1434 denote the number of elements @xmath534 so that for the corresponding edge @xmath391 of @xmath1435 we have @xmath1436 . from proposition 4.5 and from ( [ total - measure-3 ] ) ,
we conclude that @xmath1437 .
.1 cm by lemma 3.1 , there exists finitely many virtual triangulation pairs @xmath1438 so that @xmath1439 is their union . for each @xmath557 there
exists a finite cover @xmath1440 of @xmath0 so that @xmath1441 is a triangulation of @xmath1440 ( by @xmath1442 we denote the corresponding set of edges ) . also , by @xmath1443 we denote the corresponding shear coordinates .
note that the riemann surface @xmath1440 corresponds to the point @xmath1444 in the corresponding teichmller space .
for each @xmath1440 we have the following : * since @xmath1268 for every @xmath1445 from lemma 3.2 we have @xmath1446 .
* we have @xmath1447 .
one can verify the following elementary proposition .
let @xmath1448 , @xmath482 , and set @xmath1449 , @xmath1450 .
then @xmath1451 let @xmath1452 denote the number of edges from @xmath1442 for which the corresponding shear coordinate is greater than @xmath1250 .
we have @xmath1453 also , @xmath1454 from the above proposition we have that for at least one surface @xmath1440 , say for @xmath8 , we have that @xmath1455 where @xmath1456 is the total number of edges in @xmath1457 . .1 cm let @xmath384 . from theorem 2.2
we have that the weil - petersson distance between @xmath1444 and @xmath1458 , tends to @xmath138 when @xmath384 .
note that by proposition 2.2 , the riemann surface that corresponds to @xmath1458 is isomorphic to the quotient of @xmath71 by a finite index subgroup of @xmath29 .
.1 cm theorem 1.1 states that for any two punctured riemann surfaces @xmath0 and @xmath1 of finite type , and for every @xmath2 we can find finite covers @xmath5 and @xmath6 of @xmath0 and @xmath1 respectively , so that the weil - petersson distance between them is less than @xmath7 .
we first find @xmath1459 and @xmath1460 finite covers of @xmath0 and @xmath1 respectively , so that @xmath1459 and @xmath1460 are @xmath1461-close ( in the weil - petersson sense ) to two riemann surfaces @xmath1462 and @xmath1463 where @xmath1462 and @xmath1463 are isomorphic to @xmath522 and @xmath1464 respectively , where @xmath1465 are finite index subgroups of @xmath29 .
set @xmath1466 and let @xmath1467 be the corresponding riemann surface .
then there are covers @xmath5 and @xmath6 of @xmath0 and @xmath1 respectively , so that @xmath5 and @xmath6 are @xmath1461-close ( in the weil - petersson sense ) to @xmath16 .
moreover , the weil - petersson distance between @xmath5 and @xmath6 is at most @xmath7 .
this proves the theorem .
let @xmath1468 denote the group of orientation preserving mbius transformations of @xmath71 .
by @xmath1469 we denote the subgroup of @xmath1468 whose elements preserve @xmath139 . by @xmath1470
we denote the set of all geodesics in @xmath71 .
also , let @xmath1471 denote the set of all ideal triangles in @xmath71 .
.1 cm denote by @xmath1472 the unit tangent bundle of @xmath71 .
elements of @xmath1472 are pairs @xmath1473 where @xmath605 and @xmath1474 is a unit vector at @xmath610 . if @xmath1475 then @xmath1476 is a well defined element .
the quotient @xmath1477 is isomorphic to the unit tangent bundle @xmath1023 . also , if @xmath1478 or @xmath1479 then @xmath1480 denotes the restriction of the corresponding unit tangent bundle over @xmath1481 .
.1 cm we parametrise @xmath1472 by @xmath1482 . here
@xmath41 , and @xmath1483 is the positively oriented angle that @xmath1474 makes with the positive part of the @xmath255-axis .
the liouville volume form on @xmath1472 is given by @xmath1484 .
the corresponding measure on @xmath1472 is called the liouville measure . the total measure
@xmath1485 is finite . .1
cm let @xmath1486 .
by @xmath1487 we also denote the map @xmath1488 .
clearly the map @xmath1489 is a diffeomorphism of @xmath1472 of order three ( this is the first operation on @xmath1472 we define in this subsection ) . also ,
the map @xmath1489 commutes with every element from @xmath1468 ( and in particular @xmath1489 commutes with every element of the group @xmath28 ) .
we have that @xmath1490 is a well defined diffeomorphism .
moreover , @xmath1489 is measure preserving ( it preserves the liouville measure on @xmath1472 ) , that is @xmath1491 where @xmath1492 denotes the jacobian . .1
cm fix @xmath1493 .
let @xmath1494 be the natural parametrisation of the geodesic ray that starts at @xmath610 and that is tangent to the vector @xmath1474 at @xmath610 that is @xmath1495 .
we use the same notation for the induced map @xmath1496 .
by @xmath1497 we also denote the corresponding geodesic ray @xmath1498 .
.1 cm as usual , for @xmath1499 the geodesic flow @xmath1500 is given by @xmath1501 .
moreover , for @xmath1475 we have @xmath1502 .
therefore , the flow is well defined on @xmath1477 .
we use the same notation for the induced flow @xmath1503 .
the geodesic flow @xmath1504 is measure preserving .
let @xmath1505 denote the map such that @xmath1506 is the end point of the geodesic ray @xmath1497 .
let @xmath1507 denote the map such that @xmath1508 is the geodesic that connects the points @xmath1509 and @xmath1510 .
let @xmath1511 denote the map such that @xmath1512 is the triangle with the vertices @xmath1509 @xmath1513 and @xmath1514 .
since the angle between @xmath1474 and @xmath1515 is @xmath448 we find that the point @xmath610 is the centre of the triangle @xmath1516 .
the triangle @xmath1517 is bounded by the geodesics @xmath1518 , @xmath1519 and @xmath1520 .
moreover , we endow the geodesic @xmath1521 with the induced orientation , so that the endpoints @xmath1522 and @xmath1510 correspond to @xmath1523 and @xmath701 , respectively .
.1 cm we now define the other three operations on @xmath1472 .
let @xmath1493 .
let @xmath1524 be the rotation of order two about the point @xmath1525 .
set @xmath1526 .
clearly the map @xmath1527 is a diffeomorphism of order two .
also the map @xmath1528 commutes with every element from @xmath1468 and the induced diffeomorphism @xmath1529 is well defined .
note that @xmath1530 and @xmath1531 .
the group @xmath1532 generated by @xmath1533 is isomorphic to @xmath570 .
it is easy to see that there exists a finite orbit under the action of this group if and only if the surface @xmath0 is modular , that is @xmath0 is isomorphic to @xmath27 where @xmath28 is a finite index subgroup of @xmath29 .
this indicates the relevance of this group action to the ehrenpreis conjecture .
let @xmath1493 , @xmath1534 , and @xmath619 .
let @xmath1535 be the hyperbolic transformation which fixes the points @xmath1509 and @xmath1510 , and so that the signed hyperbolic distance between the points @xmath787 and @xmath1536 , is equal to @xmath622 . here
the signed distance between @xmath787 and @xmath1536 is positive if and only if the points @xmath1509 , @xmath787 , @xmath1536 , and @xmath1510 , sit on the geodesic @xmath1518 , in this order , with respect to the induced orientation on @xmath1518 .
the definition of the map @xmath1537 does not depend on the choice of @xmath1534 .
clearly , the collection of transformations @xmath1537 , @xmath619 , is an one - parameter abelian group .
set @xmath1538 .
the collections of maps @xmath1539 , @xmath619 , is an one - parameter abelian group of diffeomorphisms , and @xmath1540 commutes with every element from @xmath1468 .
note that @xmath1541 , and @xmath1542 .
in fact , @xmath1539 , is the equidistant " flow .
this means that for @xmath1543 , the points @xmath1544 move along the line that is equidistant from the geodesic @xmath1518 ( the distance between @xmath1544 and @xmath1518 is @xmath1545 ) .
.1 cm let @xmath1493 so that @xmath1546 .
let @xmath1547 , be the geodesic that connects @xmath1509 and @xmath139 , and let @xmath1548 , be the reflection through @xmath391 .
set @xmath1549 , and @xmath1550 .
the maps @xmath1551 are defined almost everywhere on @xmath1472 , because the subset of @xmath1472 on which @xmath1552 has zero measure in @xmath1472 .
the maps @xmath1553 and @xmath1554 commute with every element of @xmath1469 .
note that @xmath1554 is of order two .
also , @xmath1555 .
we set @xmath1556 . the map @xmath1557 does not commute with the entire group @xmath1468 and the map @xmath1553 does not give rise to a map on @xmath1023 .
let @xmath593 , and let @xmath721 be normalised . then @xmath1553 commutes with the translation for @xmath950 , so the maps @xmath1553 and @xmath1554 , are well defined almost everywhere on @xmath1558 .
we have the induced maps @xmath1559 , and @xmath1560 , that are defined on each @xmath138-horoball in the normalised setting . also , note that @xmath1561 .
we have @xmath1562 everywhere on @xmath1472 and @xmath1563 almost everywhere on @xmath1472 .
moreover we have that the relations @xmath1564 and @xmath1565 hold for every @xmath619 .
also @xmath1566 is the identity mapping on @xmath1472 .
we have already observed that @xmath1528 is of order two , that is @xmath1566 is the identity mapping on @xmath1472 .
if @xmath1567 , and @xmath1568 , for some @xmath619 , then @xmath1569 , and @xmath1570 .
this yields the relation @xmath1565 .
the relation @xmath1564 , is proved similarly .
.1 cm since all four maps @xmath1528 , @xmath1540 , @xmath1554 , and @xmath1553 commute with @xmath1469 , and @xmath1469 acts transitively on @xmath71 , we have that the functions @xmath1571 , @xmath1572 , @xmath1573 , and @xmath1574 , are constant functions almost everywhere on @xmath1472 . since @xmath1575 we have that @xmath1576 everywhere on @xmath1472 .
it follows from @xmath1564 , that @xmath1577 . since @xmath1578
, we conclude that @xmath1579 everywhere on @xmath1472 . from @xmath1580 , we find that @xmath1581 . since @xmath1491 it follows from the definition of @xmath1553 that @xmath1563 .
let @xmath1493 , and let @xmath1582 .
* denote by @xmath1583 the unique number in @xmath1584 that is equal to the positively oriented angle between the vector @xmath1474 and the geodesic ray @xmath1585 , where @xmath1585 denotes the geodesic ray that starts at @xmath610 and ends at @xmath1586 .
* by @xmath1587 we denote the unique horoball that contains @xmath610 and meets @xmath420 at @xmath1586 . *
let @xmath1588 .
denote by @xmath1589 $ ] the signed hyperbolic distance between the horocircles @xmath1590 and @xmath1591 .
that is , @xmath1589 $ ] is non - negative if and only if @xmath1592 is contained in @xmath1593 ( in this case we say that @xmath610 is closer to @xmath1586 than @xmath1594 is ) . *
suppose that @xmath1595 .
then by @xmath1596 we denote the point on @xmath1497 that is the closest to @xmath1586 . also , if @xmath1595 , then there exists a unique point on @xmath1497 that is the closest to @xmath1586 .
this shows that @xmath1596 is well defined .
if @xmath1597 , then such a point does not exist .
.1 cm note that for @xmath77 , we have @xmath1598=-\delta_p[w , z]$ ] , and @xmath1599| \le { { \bf d}}(z , w)$ ] . for points
@xmath1600 , we have @xmath1601-\delta_p[z , w_1]=\delta_p[w_1,w_2]$ ] .
let @xmath593 , and @xmath1588 .
then @xmath1602 $ ] measures the difference in heights , that is @xmath1603.\ ] ] it follows from the definition that @xmath1604 , if and only if @xmath1605 . if @xmath1606 , then @xmath1607 . if @xmath1605 , then there exists @xmath1608 , so that @xmath1609 .
most calculations in the remainder of this paper are based on the following two elementary identities @xmath1610=\log(\csc(|{{\theta}}_p(z , v)|)),\ ] ] and @xmath1611 it follows that @xmath1612<\log 2,\ ] ] and when @xmath1613 is small , we have @xmath1614=\log 2- o(|{{\theta}}_p(z , v)|^{2}).\ ] ] let @xmath1493 , and @xmath1582 , such that @xmath1605 .
let @xmath1608 be the number so that @xmath1615
. then for every @xmath1616 , we have @xmath1617 < t.\ ] ] also , @xmath1618 the first inequality follows from ( [ ang - estimate-2 ] ) .
the second inequality follows from ( [ ang - estimate-1 ] ) , and the fact that for @xmath1619 , we have @xmath1620 .
let @xmath619 , @xmath1582 , and @xmath1493 , such that @xmath1621 .
let @xmath1543 .
then @xmath1622| \ge |t|-\log 6.\ ] ] let @xmath1623 , and @xmath1624 .
since @xmath1621 , we have @xmath1625 .
it follows from the definition of @xmath1626 , that the hyperbolic distance between the points @xmath787 and @xmath1627 , is equal to @xmath1628 .
from the first inequality in the previous proposition we have @xmath1629| \ge |t|-\log 2.\ ] ] since @xmath1630 , we have @xmath1622| \ge |\delta_p[w , w_t]|- \log 3 \ge |t|-\log 6.\ ] ] if @xmath1586 is an endpoint of @xmath1518 , then @xmath1631|=|t|>|t|-\log 6 $ ] .
let @xmath1493 , and let @xmath1632 , @xmath1633 , so that @xmath1634 , for some @xmath1635 .
let @xmath1636 be the triangle with the vertices @xmath1637 .
there exists a universal constant @xmath375 , so that for @xmath343 small enough , we have @xmath1638 .
we consider the unit disc model @xmath1639 for @xmath71 .
we may assume that @xmath1640 , and that @xmath1641 .
then @xmath1642 , @xmath1633 , where @xmath1643 is the unit circle .
since @xmath1644 , we have that @xmath1645 , where @xmath1646 is the euclidean distance .
recall that @xmath610 is the centre of the triangle @xmath1517 .
the centre of an ideal triangle , as the function of triples of points on @xmath1643 , is smooth in some neighbourhood the triple @xmath1647 .
therefore , there exists a universal constant @xmath375 , so that @xmath1648 , for @xmath343 small .
let @xmath1493 , and set @xmath1649 .
let @xmath1632 , @xmath1650 , so that @xmath1634 , for some @xmath1635 . by @xmath1547
, we denote the geodesic with the endpoints @xmath1651 and @xmath1652 .
there exists a universal constant @xmath375 , so that for @xmath343 small enough we have @xmath1653 , @xmath1654 , and @xmath1655 .
we consider the unit disc model @xmath1639 for @xmath71 .
we may assume that the geodesic @xmath1518 connects the points @xmath1656 and @xmath950 , on @xmath1643 , and that @xmath1657 . also , we may assume that the @xmath255-coordinate of @xmath610 is equal to zero .
then @xmath1658 , and @xmath1659 .
since @xmath1634 , @xmath1650 , we have that for some universal constant @xmath863 , and for @xmath343 small enough , the inequalities @xmath1660 , and @xmath1661 , hold . here
@xmath1662 , and @xmath1663 , are the euclidean distances .
the function @xmath1664 , as a function of @xmath1651 , and @xmath1652 , is smooth in some neighbourhood of the pair @xmath1665 .
this shows that there is a universal constant @xmath375 , so that for @xmath343 small enough , we have @xmath1666 .
first we give a short overview of the construction that follows . in order to construct the measures from the statement of theorem 4.1 , for every @xmath1667 we define a map ( almost everywhere on @xmath1472 with respect to the liouville measure @xmath1668 on @xmath1472 ) @xmath1669 such that for almost every @xmath1493 we have * @xmath1670 , for every @xmath93 .
* @xmath1671 .
moreover we have the induced map @xmath1672 where @xmath1673 is the geodesic with the endpoints @xmath1674 and @xmath1675 .
also we have the map @xmath1676 where @xmath1677 is the triangle with vertices @xmath1678 , @xmath1679 .
the maps @xmath1680 , @xmath1681 and @xmath1682 are @xmath28 equivariant so we have the induced maps from @xmath1023 to @xmath419 , @xmath390 and @xmath392 , respectively .
note that if @xmath1683 is a non - negative integrable function on @xmath0 , then @xmath1684 .
the measure @xmath1047 from theorem 4.1 will be constructed in this way for a suitable choice of @xmath1685 .
.1 cm our goal is to let @xmath39 be large and show that for most `` points @xmath1493 we have @xmath1686 .
this would imply that @xmath1687 , where @xmath1105 is @xmath1688 symmetric and @xmath1186 is small .
this seems to be a good candidate for the choice of measure @xmath72 in theorem 4.1 .
however with this choice we have @xmath1689 because of the thin part of @xmath0 . in order to overcome this problem we set @xmath1690 , and consider the measure @xmath1691 ( note that @xmath1692 for @xmath1693 ) . but
this introduces an imbalance , that is the measure @xmath1694 is no longer almost symmetric outside @xmath1695 because @xmath1696 for @xmath610 that is is outside @xmath1695 .
that is why we introduce the map @xmath1697 where @xmath1698 is the
correctable '' part .
then we let our measure @xmath1047 be given by @xmath1699 where @xmath1700 , for @xmath1701 .
then we show that the measure @xmath1058 has the desired decomposition .
.3 cm we now return to the construction .
recall that @xmath0 is a fixed surface of type @xmath12 , and @xmath1702 .
as always , @xmath28 is one of the @xmath14 normalised fuchsian groups @xmath1703 , @xmath1704 , such that @xmath27 is isomorphic to @xmath0 . recall that the interiors of the different @xmath138-horoballs on @xmath0 are disjoint .
this implies that different @xmath950-horoball are disjoint on @xmath0 .1 cm let @xmath1493 . for any @xmath1705 ,
the point @xmath1706 is either in the interior of @xmath1707 , or in one of the @xmath950-horoballs .
let @xmath1708 , be the ordered set of cusps from @xmath419 , so that the ray @xmath1497 intersects each horoball @xmath1709 , and which are ordered so that the ray @xmath1497 visits the horoball @xmath1709 before it visits the horoball @xmath1710 if and only if @xmath1711 .
in other words , let @xmath1712 , so that @xmath1713 . then @xmath1714 , if and only if @xmath1715 .
since the @xmath950-horoballs are disjoint , and each horoball is a convex subset of @xmath71 , we have that @xmath1716 , for @xmath1717 . for a point @xmath1493 ,
the set @xmath1718 is either : empty , finite but non - empty , or infinite . it can be show that the set of points in @xmath1472 , for which the corresponding set @xmath1718 is infinite , has the full measure in @xmath1472 ( we do not use this result ) .
we now define the maps @xmath1680 , and @xmath1719 , @xmath737 , described above . given @xmath1720 , to every point @xmath1721 , we associate a triangle from @xmath392 .
fix @xmath1667 .
let @xmath1493 , and let @xmath1708 , be the corresponding ordered set of cusps that @xmath1497 intersects .
* define @xmath1722 , if + @xmath1723 + and if this inequality does not hold for any cusp @xmath1724 , where @xmath1725 .
* let @xmath1726 , be such that @xmath1727 is the first point of entry of the ray @xmath1497 in the horoball @xmath1728 .
if @xmath1473 does not get absorbed , then @xmath1729 . *
if both @xmath1674 and @xmath1675 are well defined , then by @xmath1730 , we denote the geodesic with the endpoints @xmath1674 and @xmath1675 .
if all three cusps @xmath1674 , @xmath1675 , and @xmath1731 , are well defined , then by @xmath1732 , we denote the triangle with the vertices @xmath1674 , @xmath1675 , and @xmath1731 . since @xmath1667 , we have that @xmath1733 .
this implies that @xmath1734 , so @xmath1735 .
this shows that @xmath1673 and @xmath1677 are well defined .
it is useful to pause here and observe that if @xmath593 , is such that @xmath1736 , then @xmath1737 .
this follows from proposition 5.2 .
we will often use this observation in the arguments that follow .
we will see below that the function @xmath1738 is defined almost everywhere on @xmath1472 .
let @xmath93 .
we have @xmath1739 .
this shows that the set where @xmath1680 is defined is invariant under @xmath28 .
the induced map @xmath1740 , is also denoted by @xmath1680 .
the absorption time @xmath1741 is defined in the same way .
the maps @xmath1742 , and @xmath1743 , are defined accordingly .
.1 cm fix @xmath1667 , and @xmath1493 .
then for every @xmath1744 , we have @xmath1745 . moreover ,
suppose that @xmath1746 , for some @xmath791 , and that @xmath1747 . if @xmath1748 is the the second cusp ( after @xmath444 ) so that @xmath1497 visits the horoball @xmath1749 , then @xmath1750 . for each cusp
@xmath1751 , such that @xmath1497 enters this cusp , we let @xmath1752 , so that @xmath1753 is the point of entry of the ray @xmath1497 in @xmath1709 . by @xmath1754 ,
we denote the exit time , that is @xmath1755 is the exit point of the ray @xmath1497 from @xmath1709 . if @xmath1756 , then @xmath1757 . if @xmath1497 never leaves some @xmath1709 , then @xmath1758 .
.1 cm it follows from proposition 5.2 that if @xmath1759 , then @xmath1760 \ge { { t_{{{\operatorname { ex}}}}(i)+t_{{{\operatorname { ent}}}}(i)}\over{2}}-\log 2.\ ] ] if @xmath1756 , then @xmath1761 \ge { { t_{{{\operatorname { ex}}}}(i)+{{\bf h}}_{c_{i}}(z)}\over{2}}-\log 2.\ ] ] if @xmath1762 , and @xmath1763 , then either @xmath1759 , or @xmath1764 and @xmath1756 . in either case
, we have @xmath1765 , since @xmath1766 ( the identity @xmath1767 would contradict the assumption @xmath1762 ) . in the first case , from ( [ diff-1 ] ) we obtain @xmath1768 . in the second case , from ( [ diff-2 ] ) we have @xmath1769 .
this proves the first part of the proposition .
.1 cm assume that @xmath1770 .
let @xmath1771 , be the cusp so that the horoball @xmath1749 is the second @xmath950-horoball that the ray @xmath1497 enters .
by ( [ diff-1 ] ) we have @xmath1772 since @xmath1770 , and since @xmath1747 , we conclude @xmath1750 .
we want to show that the absorption map is well defined almost everywhere on @xmath1472 ( and on @xmath1023 ) .
we saw in the previous proposition that the ray @xmath1497 , will get absorbed if it leaves @xmath1773 after the time @xmath1774 . in order to show
that @xmath1680 is defined almost everywhere , we need to estimate the liouville measure of the geodesics segments of a given length that stay in @xmath1773 .
the following lemma is probably known , but we prove it in the appendix .
let @xmath1775 , and let @xmath1776 be such that @xmath1777 if the segment @xmath1778 $ ] is contained in @xmath1773 . then there are constants @xmath1779 , that depend only on @xmath0 , such that @xmath1780 .
we have fix @xmath1667 .
the map @xmath1740 , is defined almost everywhere .
the proof below uses lemma 5.1 .
however one does not need lemma 5.1 to prove that the absorptions map is defined almost everywhere on @xmath1472 .
the fact that the geodesic flow @xmath1504 is ergodic implies that the geodesic ray @xmath1497 leaves the thick part @xmath1773 , after the time @xmath1774 , for almost every @xmath1721 .
then it follows from proposition 5.6 that almost every @xmath1473 will be absorbed .
however lemma 5.1 will be used later in a similar manner , so we decide to state it here and present its first application .
let @xmath1781 , and let @xmath1782 be such that @xmath1783 if @xmath1729 , and @xmath1784 .
by proposition 5.6 we have that @xmath1785 , for every @xmath1775 ( here @xmath1786 is the set defined in the statement of lemma 5.1 ) .
since @xmath1787 when @xmath1788 , we find that @xmath1789 , so the map @xmath1680 is defined almost everywhere on @xmath1790 , for every @xmath1781 . for every @xmath1667 , we define the map @xmath1791 , as @xmath1792 . that is , @xmath1793 is the foot of the centre of the triangle @xmath1732 . the map @xmath1794 commutes with the action of the group @xmath28 , and we have that the induced map @xmath1795 , is well defined .
there exists a universal constant @xmath375 , and @xmath370 , so that for @xmath371 , we have that @xmath1796 .
it follows from proposition 5.2 that @xmath1797 . by proposition 5.4 , for @xmath39
large enough we have @xmath1798 , for some universal constant @xmath375 .
this proves the proposition .
denote by @xmath1799 , the space of measures from @xmath1800 , that are absolutely continuous with respect to the liouville measure @xmath1668 .
let @xmath1801 .
then @xmath1802 , and @xmath1803 , are well defined since @xmath1681 and @xmath1682 are defined almost everywhere on @xmath1023 .
.1 cm let @xmath1683 , be an integrable , non - negative function .
we have the induced map @xmath1804 , given by @xmath1805
. then @xmath1806 . since @xmath1807 , and since @xmath1808 , we have @xmath1809 where @xmath1810 , is the operator defined at the beginning of section 4 .
.1 cm let @xmath1493 , and set @xmath1811 .
we say that @xmath1812 ( or that @xmath1473 is above the geodesic @xmath1518 ) if @xmath1813 and @xmath1814 ( or that @xmath1473 is below the geodesic @xmath1815 ) if @xmath1816 . since for almost every @xmath1493 we have that either @xmath1817 or @xmath1816 we see that @xmath1818 has full measure in @xmath1472 . the sets @xmath1819 and @xmath1820 are disjoint and @xmath1821 .
the sets @xmath1819 and @xmath1820 are invariant under @xmath28 and the sets @xmath1822 and @xmath1823 are defined accordingly ( recall that @xmath639 is well defined for @xmath1824 ) .
fix @xmath1667 , and let @xmath1493 .
we say that @xmath1825 if @xmath1826 , and if @xmath1686 ( here we assume that both @xmath1673 and @xmath1827 are well defined ) .
the set @xmath1828 is invariant under @xmath28 and by @xmath1829 we denote the corresponding subset of @xmath1023 .
fix @xmath1667 and let @xmath593 .
let @xmath1830 set @xmath1831 .
let @xmath1832 where the union is taken over all @xmath593 .
the set @xmath1833 is invariant under @xmath28 and by @xmath1834 we denote the corresponding subset of @xmath1023 .
note that @xmath1835 so @xmath1836 .
let @xmath1667 and let @xmath593 .
let @xmath1837 set @xmath1838 let @xmath1839 .
we say that @xmath1840 , if @xmath1841 , and @xmath1842 , if @xmath1843 .
the sets @xmath1844 and @xmath1845 are invariant under @xmath28 and the corresponding quotients are denoted by @xmath1846 and @xmath1847 . in the definition of @xmath1845
, we consider the maps @xmath1848 , as the maps of @xmath1849 . since @xmath1812 , we have that @xmath1850 , and @xmath1851 , are well defined .
assume that @xmath1852 .
then @xmath1853 for some @xmath593 and @xmath1854 ( by definition of @xmath1833 ) .
moreover , neither @xmath1497 or @xmath1855 gets absorbed by the cusp @xmath595 .
this shows that every such @xmath1473 belongs to @xmath1844 , that is @xmath1856 we have @xmath1857 , and @xmath1858 .
let @xmath1839 , and let @xmath1859 .
there exists a cusp @xmath593 , so that @xmath1860 .
set @xmath721 , and @xmath695 .
it follows from the definition of @xmath1553 , that @xmath1861 .
also @xmath1862 , and @xmath1863 .
this implies that @xmath1864 , and @xmath1865 .
.1 cm since @xmath1839 , we have @xmath1866 . in order to show @xmath1867
, we need to show that @xmath1868 .
assume that @xmath1869 .
then @xmath1870 .
but then @xmath1871 , so @xmath1872 , which is a contradiction .
similarly we show @xmath1873 .
putting this together proves @xmath1874 . in the same way we show @xmath1875 .
.1 cm the equality @xmath1858 , follows directly from the definition .
we are yet to see that @xmath1829 , @xmath1834 , and @xmath1847 , are non - empty sets ( see lemma 5.3 ) .
the following proposition is elementary and the proof is left to the reader .
let @xmath1667 . if @xmath1876 then @xmath1877 .
let @xmath1667 , and let @xmath1839 .
let @xmath593 , so that @xmath1860 .
set @xmath1878 . by @xmath1879 , we denote the geodesic with the endpoints @xmath1674 and @xmath1880 ( providing that @xmath1674 exists ) .
by @xmath1881 , we denote the triangle with the vertices @xmath1674 , @xmath1675 , and @xmath1880 ( providing that @xmath1882 , @xmath1883 , exist ) . from the definition of the set @xmath1844
, we have that @xmath1884 ( if @xmath1885 , @xmath1650 , then @xmath1886 ) .
this shows that the maps @xmath1880 , @xmath1879 , and @xmath1887 are well defined almost everywhere on @xmath1844 .
note that these three maps commute with the action of @xmath28 , so we have the induced maps @xmath1888 , @xmath1889 , and @xmath1890 . the edges of the triangle @xmath1887 , are @xmath1673 , @xmath1879 , and @xmath1891 .
.1 cm for @xmath1839 , let @xmath1892 , that is @xmath1893 is the foot of the centre of the triangle @xmath1881 , with respect to the geodesic @xmath1730
. set @xmath1894 , and @xmath1895 .
the notation @xmath1896 indicates that the point @xmath1896 belongs to the vertical edge of @xmath1887 that is on the left - hand " side of @xmath1887 , with the normalisation @xmath1897 .
similarly , the point @xmath1898 belongs to the vertical edge of @xmath1887 that is to the right - hand " side of @xmath1887 .
there exists @xmath370 , so that for @xmath371 , the following holds .
let @xmath1840 , and set @xmath1899 .
assume that @xmath1887 and @xmath1900 , exist . then @xmath1901 .
similarly , if @xmath1842 , for @xmath1902 , we have @xmath1903 , providing that @xmath1887 and @xmath1900 , exist .
suppose @xmath1840 .
let @xmath1904 , and set @xmath721 .
let @xmath1905 , @xmath1906 , and @xmath1907 ( the identity @xmath1908 follows from the definition of @xmath1553 ) .
since @xmath1909 , we have by proposition 5.6 that if @xmath1910 , then @xmath1911 , is the first @xmath950-horoball that the geodesic ray @xmath1912 enters after leaving @xmath1913 . with this normalisation
, we have that the euclidean diameter of the horoball @xmath1911 , is at most @xmath950 . since the ray @xmath1912 ends at the point @xmath1652
, we conclude that @xmath1914 , where @xmath1915 is the euclidean distance between the real numbers @xmath444 and @xmath1652 .
let @xmath1916 .
similarly , we see that @xmath1917 .
.3 cm since @xmath1840 , we have @xmath1918 . similarly we see that @xmath1919 . from the fact that @xmath1920
, we have that @xmath1921 .
this implies that @xmath1922 for @xmath39 large enough . since @xmath1923 the proposition follows .
let @xmath1924 .
then @xmath1925 , and @xmath1926 , are well defined since @xmath1681 and @xmath1682 are defined almost everywhere on @xmath1846 . by definition , we have @xmath1927 if @xmath1674 is defined , then the @xmath39-combinatorial length @xmath1928 is defined as follows .
let @xmath391 be the geodesic ray that connects @xmath610 and @xmath1674 .
let @xmath666 , be the number of ( transverse ) intersections between the ray @xmath391 and the edges from @xmath414 .
if @xmath610 does not belong to @xmath1728 , then @xmath1929 . if @xmath610 belongs to @xmath1728 , then @xmath1930 .
again , the function @xmath1931 , is defined almost everywhere .
there exists @xmath370 , so that for @xmath371 , the following holds . for @xmath1493 ,
we have @xmath1932 let @xmath1839 .
we have @xmath1933 @xmath1934 and @xmath1935 let @xmath391 denote the geodesic ray that connects @xmath610 and @xmath1674 , and let @xmath730 denote the geodesic ray that connects @xmath610 and @xmath1675 .
then every edge from @xmath414 , that intersects ( transversely ) the geodesic @xmath1673 , has to intersect one of the rays @xmath391 or @xmath730 .
this shows @xmath1936 assume that @xmath1937 belongs to the @xmath950-horoball of one of the cusps @xmath1674 or @xmath1675 , say @xmath1937 belongs to the @xmath950-horoball at the cusp @xmath1674 .
it follows from proposition 5.8 that for @xmath39 large enough , we have @xmath1938 therefore , we have that @xmath1939 .
this proves the first inequality in this proposition .
.1 cm assume now that @xmath1839 . since @xmath1940 , we have @xmath1941 .
let @xmath391 and @xmath730 be as above .
then every edge from @xmath414 , that intersects ( transversely ) the geodesic @xmath1673 , has to intersect one of the rays @xmath391 or @xmath730 .
this proves the second inequality @xmath1942 .
.1 cm if we prove @xmath1943 then the third inequality would follow from the second .
note that @xmath1944 .
we have @xmath1945 the geodesic @xmath1673 intersects every edge from @xmath414 , that is intersected by @xmath1879 .
in addition , the geodesic @xmath1673 intersects at least @xmath1946 edges from @xmath414 , that all have @xmath1880 as their endpoints ( since @xmath1879 has @xmath1880 as an endpoint , we see that @xmath1879 can not intersect ( transversely ) any edge from @xmath414 , that ends at @xmath1880 ) .
we have @xmath1947 which proves the third inequality .
the fourth inequality is proved in the same way .
we have let @xmath1667 .
define @xmath1948 , by @xmath1692 , if @xmath1693 , and by @xmath1949 note that @xmath1950 is continuous on @xmath0 , and @xmath1951 .
the induced function @xmath1952 , is also denoted by @xmath1950 .
the reason that the factor @xmath225 appears in the definition of @xmath1950 , is lemma 5.5 below .
in fact , we could replace @xmath225 , with any number between @xmath138 and @xmath950 . also , note that if @xmath1953 , then @xmath1954 . for @xmath1667 , the set @xmath1955 ,
is defined so that @xmath1956 , if @xmath1957 , and if the inequalities @xmath1958 , hold .
note that if @xmath1956 , then for @xmath1959 , we have @xmath1960 , where @xmath1961 ( recall definition 3.4 for the definition of @xmath630 ) .
the following lemma will be proved in section 7 .
there exists @xmath370 , so that for @xmath371 , we have @xmath1962 the following lemma will also be proved in section 7 .
there exists @xmath370 , so that for @xmath371 , we have @xmath1963 where @xmath1964 , and @xmath1965 .1 cm we now give a better description of the set @xmath1966 .
the results that follow in this subsection will not be used in the proof of theorem 4.1 below , so the reader may skip the rest of this subsection and go to the next subsection and the proof of theorem 4.1 .
.1 cm let @xmath1967 , so that @xmath1968 .
let @xmath391 be the geodesic that contains @xmath1969 and @xmath924 , and assume that @xmath1970 , where @xmath731 is the midpoint of the geodesic segment between @xmath1969 and @xmath924 .
let @xmath1971 , so that @xmath1972 , for some @xmath1635 , and let @xmath730 be the geodesic that contains @xmath1973 and @xmath1974 .
there exists a universal constant @xmath375 , so that for @xmath343 small enough , we have @xmath1975 .
we work in the unit disc @xmath1639 .
the point @xmath1976 , corresponds to the point @xmath1977 .
we may assume that @xmath1978 .
then @xmath391 is the geodesic that connects @xmath1656 and @xmath950 .
moreover , @xmath1979 , and @xmath1980 , for some @xmath1981 .
let @xmath1982 , be the endpoints of @xmath730 . from the assumption @xmath1972
, and since @xmath1983 , we see that there exists a universal constant @xmath863 , so that @xmath1984 .
the mbius transformation @xmath600 is uniquely determined by the conditions @xmath1985 , @xmath1986 , and @xmath1987 .
moreover , @xmath600 depends smoothly on @xmath1988 near @xmath1656 , and @xmath1989 near @xmath950 .
therefore , there exists a universal constant @xmath375 , so that @xmath1990 . since @xmath1985 , we have @xmath1991 , and this proves the proposition .
let @xmath1992 , such that @xmath1993 .
let @xmath1994 , @xmath1995 , be the point on the geodesic segment between @xmath1996 and @xmath1997 , so that @xmath1998
. then @xmath1999 , for some @xmath2000 .
we have the following elementary inequality in hyperbolic geometry @xmath2001 for some universal constant @xmath863 .
there exists @xmath370 , so that for @xmath371 , the following holds .
let @xmath2002 , such that @xmath2003 , and @xmath2004 .
suppose that @xmath2005=0 $ ] .
let @xmath1582 , and suppose @xmath2006 \ge r$ ] .
then @xmath2007 moreover , we have latexmath:[\[\label{geom - estimate-2 } the rays @xmath1497 and @xmath2009 end at the same point at @xmath420 .
since @xmath2005=0 $ ] , and since @xmath2004 , for any @xmath1775 , applying ( [ long - range - estimate ] ) we obtain @xmath2010 let @xmath2011 , so that @xmath1609 . by proposition 5.2
we have @xmath2012 \ge r$ ] .
let @xmath2013 , and @xmath2014 .
also , let @xmath2015 , and @xmath2016 .
by ( [ distance-1 ] ) we have @xmath2017 , so it follows from the previous proposition that for @xmath39 large enough , we have @xmath2018 this shows that ( [ geom - estimate-1 ] ) holds .
.1 cm we may assume that @xmath2019 .
let @xmath2020 , denote the naturally parametrised horocircle ( with respect to @xmath139 ) , so that the oriented angle between the vectors @xmath2021 and @xmath1474 , is @xmath2022
. then for every @xmath2023 , we have @xmath2024=0.\ ] ] moreover , there exists @xmath2025 , so that @xmath2026 .
assume @xmath2027 ( the other case is handled in the same way ) .
the number @xmath2028 depends only on the upper bound of the hyperbolic distance between @xmath610 and @xmath1594 which is bounded above by @xmath2029 .
.1 cm by @xmath2030 denote the naturally parametrised horocircle ( with respect to @xmath1586 ) , so that the oriented angle between the vectors @xmath2031 and @xmath1474 is positive .
it follows from proposition 5.2 that @xmath2032 .
therefore , the angle between the vectors @xmath2021 and @xmath2031 is at most @xmath2033 .
we find that there exists a constant @xmath878 that depend only on @xmath2028 with the following properties .
for @xmath2034 we have @xmath2035 denote the naturally parametrised geodesic that contains the geodesic ray @xmath2009 such that @xmath2036 .
for @xmath39 large enough , we have that @xmath2037 intersects the geodesic @xmath391 .
this is true because when @xmath384 we have that the horoball at @xmath1586 that contains @xmath610 converges on compact sets in @xmath71 to the horoball at @xmath139 that contains @xmath610 and on the other hand we have that @xmath1105 intersect @xmath391 orthogonally at @xmath1594 . let @xmath2038 .
let @xmath2039 , be such that @xmath2040 . then for @xmath39 large enough
, we have @xmath2041 .
we have @xmath2042|=|\delta_{\infty}[z , z'']|=|\delta_{\infty}[z,\alpha_r(s')]|$ ] so it follows from ( [ hyp - estimate-1 ] ) that @xmath2043 .1 cm note @xmath2044=0 $ ] .
it follows from ( [ geom - estimate-1 ] ) that @xmath2045-\delta_p[{{\bf z}}_{\max}(\gamma , p),z'']| \le { { \bf d}}({{\bf z}}_{\max}(\gamma_{(z , v)},p),{{\bf z}}_{\max}(\gamma , p ) ) \le re^{-r}.\ ] ] since @xmath2046 , we have for @xmath39 large , that @xmath2045-\delta_p[{{\bf z}}_{\max}(\gamma_{(z',v')},p),z']|\le re^{-r}+ke^{-r}=2re^{-r},\ ] ] which proves ( [ geom - estimate-2 ] ) .
the following proposition states that if two unit vectors in @xmath1472 are nearby and related by the horocyclic flow then either they are absorbed by the same cusp or the first vector just misses " being absorbed by some cusp or it just barely gets absorbed by some cusp .
let @xmath1959 and let @xmath2047 such that @xmath2048 and such that * @xmath2003 .
* @xmath2049 and @xmath2005=0 $ ] .
assume that @xmath2050 .
then at least one of the following two conditions holds 1 .
there exists a cusp @xmath593 , such that @xmath2051 and @xmath2052 .
2 . there exists a cusp @xmath593 , such that @xmath2053 and @xmath2054 , for some @xmath2055 .
set @xmath2056 .
we have @xmath2057 .
first consider the case when the geodesic ray @xmath2009 does not intersect @xmath2058 . in this case
we show that the condition ( @xmath2059 ) holds .
it follows from ( [ geom - estimate-1 ] ) that @xmath2060 since @xmath2052 and since in this case @xmath2061 does not belong to @xmath2058 , we conclude that @xmath2062 .
let @xmath1608 be such that @xmath2063 .
in order to show that ( @xmath2059 ) holds it remains to prove that @xmath2064 .
let @xmath2065 so that @xmath2066 is the first point of entry of the geodesic ray @xmath1497 into @xmath2058 .
then @xmath2067 and by the assumption @xmath2048 we have @xmath2068 . since @xmath2069 and since by proposition 5.2 we have @xmath2070 , we have @xmath2071 . on the other hand ,
since @xmath1736 we have @xmath2072 .
then from proposition 5.2 we have @xmath2073 .
this shows that the condition ( @xmath2059 ) holds .
.1 cm from now on we assume that @xmath2009 enters the horoball @xmath2058 but that @xmath2057 .
then there are two possible reasons why @xmath2057 .
the first one is that @xmath2074={{\bf h}}_c({{\bf z}}_{\max}(\gamma_{(z',v')},c))-{{\bf h}}_c(z')<r$ ] .
since @xmath2075={{\bf h}}_c({{\bf z}}_{\max}(\gamma_{(z , v)},c))-{{\bf h}}_c(z ) \ge r$ ] , from ( [ geom - estimate-2 ] ) we find that @xmath2076 .
so in this case the condition ( @xmath950 ) holds because we already know that @xmath2052 .
.1 cm the second reason is that the geodesic ray @xmath2009 gets absorbed before entering the horoball @xmath2058 .
set @xmath2077 . if @xmath2078 is such that @xmath2079 then @xmath2080 .
assume that @xmath2081 does not belong to @xmath2082 .
then we show that the condition ( @xmath2059 ) holds ( with respect to the cusp @xmath2083 ) .
since @xmath2084 by ( [ geom - estimate-1 ] ) we have that @xmath2085 . since @xmath2086 , from ( [ geom - estimate-2 ] ) we get that @xmath2087 this yields that @xmath2088 .
we have already seen that @xmath2080 .
this proves the statement .
.1 cm it remains to examine the case @xmath2089 .
we show that in this case ( @xmath2059 ) holds .
again let @xmath2078 such that @xmath2079 .
it follows from ( [ geom - estimate-2 ] ) that @xmath2087 on the other hand , we have that @xmath2090 because otherwise we would have that @xmath2091 . the last two estimates
put together give us that @xmath2092 .
this proves the proposition .
the following proposition replaces the hypotheses @xmath2049 and @xmath2005=0 $ ] of proposition 5.15 with the condition that @xmath610 and @xmath1594 are in @xmath2093 .
let @xmath1959 and @xmath593 .
let @xmath2094 where @xmath2048 such that @xmath2003 and such that @xmath2095 .
if @xmath2050 then the following holds * there exists a cusp @xmath593 , such that @xmath2053 and @xmath2054 for some @xmath2055 .
since @xmath2095 we have that @xmath2096 .
since @xmath2097 ( because @xmath1959 ) , it follows from proposition 5.6 that @xmath1497 gets absorbed by the first @xmath950-horoball it hits after leaving @xmath2058 .
the same is true for the ray @xmath2009 .
.1 cm let @xmath1220 be the horocircle at @xmath1509 that is tangent to @xmath2098 .
let @xmath2099 , @xmath1995 , be the two geodesics that start at @xmath1509 and that are tangent to @xmath2098 .
let @xmath2100 be the subsegment of @xmath1220 that is bounded by the points @xmath2101 ( since @xmath2099 starts at the same point on @xmath915 where @xmath1220 touches @xmath915 , there exists a unique intersection point @xmath2101 in @xmath71 ) .
we have that the hyperbolic length of @xmath2100 is equal to @xmath950 and that @xmath2100 is contained in @xmath2102 .
let @xmath924 and @xmath2103 be the points of intersection between @xmath2100 and the geodesic rays @xmath1497 and @xmath2009 respectively .
observe that @xmath2104=0 $ ] and @xmath2105 .
also @xmath2106 .
.1 cm let @xmath2107 be such that the ray @xmath1912 is contained in the ray @xmath1497 .
similarly let @xmath2108 be such that the ray @xmath2109 is contained in the ray @xmath2110 .
since @xmath2106 from proposition 5.6 we have that @xmath2111 and @xmath2112 . also @xmath2113 .
we now apply the previous proposition to @xmath2114 and @xmath2115 . since @xmath2116 we see that the condition ( @xmath950 ) from the previous proposition can not hold so we have that the condition ( @xmath2059 ) holds for @xmath2114 and hence for @xmath1473 .
this proves the proposition .
we can now define three
error sets `` and show that any point of the bad set '' @xmath1966 belongs to one of these three sets .
let @xmath1667 .
define the sets @xmath2117 , @xmath202 , as follows .
we say that @xmath2118 if * we have @xmath2048 .
* we have @xmath2119 .
* let @xmath593 such that @xmath2120 .
then @xmath595 is an endpoint of @xmath1673 .
we say that @xmath2121 if * there exists a cusp @xmath593 , such that @xmath2122 and @xmath2052 .
* we have @xmath2048 and @xmath1473 does not belong to @xmath2123 .
we say that @xmath2124 if * there exists a cusp @xmath593 , so that @xmath2125 and so that @xmath2054 for some @xmath2055 .
* we have @xmath2048 and @xmath1473 does not belong to @xmath2126 .
note that the set @xmath2127 , @xmath202 , is invariant under the action of @xmath28 and the corresponding quotient is denoted by @xmath2128 .
suppose that @xmath2129 .
then @xmath2130 , for some @xmath202 , or @xmath2131 for some @xmath2132 .
we refer the reader to the flow chart for the logic of the proof below .
assume that @xmath2133 and that @xmath2048 .
.1 cm if @xmath2134 then by the definition of @xmath1828 we find that either @xmath2135 or @xmath2136 .
suppose that @xmath2135 ( the other case is handled in the same way ) , and set @xmath2137 . then @xmath2003 and @xmath2005=0 $ ] .
also , @xmath2138 .
then by proposition 5.15 we have that @xmath1473 belongs to one of the sets @xmath2139 , @xmath2140 .
.1 cm if @xmath2141 then by proposition 5.10 we have @xmath2142 for some @xmath593 .
suppose that @xmath1473 does not belong to @xmath1833 .
there are two reasons why this can happen .
the first one is that @xmath1473 does not belong to the set @xmath1831 defined in definition 5.6 .
set @xmath2143 .
then @xmath1673 or @xmath2144 has @xmath595 as its endpoint .
if @xmath1673 has @xmath595 as its endpoint then @xmath2145 .
if @xmath2146 then by proposition 5.15 we have that @xmath1473 belongs to one of the sets @xmath2139 , @xmath2140 . if @xmath2147 then @xmath2148 belongs to one of the sets @xmath2139 , @xmath2140
.1 cm assume @xmath2149 .
if @xmath1473 does not belong to @xmath1833 then there exists @xmath2150 such that * @xmath2151 .
* @xmath2144 does not have @xmath595 as its endpoint .
. then @xmath2153 .
the condition @xmath2152 means that @xmath2050 or @xmath2154 .
applying the previous proposition to @xmath1473 and @xmath2155 ( or @xmath2148 and @xmath2156 ) we conclude that @xmath1473 or @xmath2148 belongs to the set @xmath2157 .
.1 cm assume that @xmath2158 and that @xmath2048 .
then @xmath1860 for some cusp @xmath593 and @xmath595 is not an endpoint of @xmath1673 ( by the definition of @xmath1844 ) .
. set @xmath2160 ( note that @xmath1850 is well defined since @xmath2159 ) .
then @xmath2144 since by proposition 5.9 @xmath2161 .
suppose that @xmath2162 . then by the previous proposition we have that @xmath2163 .
the case @xmath2164 is treated in the same way .
recall the statement of theorem 4.1 .
there exist a constant @xmath1238 that depends only on @xmath0 , so that for every @xmath371 there exists a finite measure @xmath1239 so that the total measure @xmath1240 satisfies the inequality @xmath1241 and with the following properties . there exist measures @xmath1242 so that the measure @xmath1243 can be written as @xmath1244 and the following holds 1 .
let @xmath1245 denote the restriction of the measure @xmath1246 to the set of triangles @xmath393 for which @xmath1247 . then @xmath1248 2 .
the measure @xmath1249 is @xmath1250-symmetric , and the measure @xmath1251 is @xmath65-symmetric , for any @xmath1252 .
we have @xmath1253 4 .
we have @xmath1254 let @xmath2165 be defined as follows .
if @xmath1473 does not belong to @xmath1846 then @xmath2166 .
for @xmath1701 we have @xmath2167 if @xmath1839
we have that @xmath2168 .
therefore , @xmath2169 .
it follows from the definition of @xmath1950 that @xmath2170 depends only on that @xmath2171 and is decreasing in @xmath2171 .
since @xmath2172 we have that @xmath2173 is a non - negative function .
.1 cm we define the measure @xmath2174 since @xmath2175 on @xmath1023 and since @xmath2176 for every @xmath2177 we conclude that the total measure of @xmath2178 is approaching @xmath1485 when @xmath384 . since @xmath2179 on @xmath1023 and since @xmath2180 for every @xmath2181 , we have that the total measure of @xmath2182 tends to zero when @xmath384 .
so for @xmath39 large enough we have that the total measure @xmath1240 satisfies the inequality @xmath2183 let @xmath1245 be the restriction of the measure @xmath1246 to the set of triangles @xmath393 for which @xmath1247 .
let @xmath1721 be such that the triangle @xmath1677 belongs to the support of @xmath1245 . then at least one of the inequalities @xmath2184 , @xmath1679 , is satisfied , so @xmath2185 . from proposition 5.12 and
lemma 5.2 , we have that for @xmath39 large enough , the following holds , @xmath2186 .1 cm next , we define the measures @xmath1249 , @xmath1251 , and @xmath2187
. let @xmath2188 if @xmath1953 , and @xmath138 otherwise .
let @xmath2189 if @xmath2190 , and @xmath138 otherwise .
let @xmath2191 if @xmath2190 , and @xmath138 otherwise .
let @xmath2192 if @xmath2193 , and @xmath138 otherwise , and let @xmath2194 if @xmath2195 , and @xmath138 otherwise .
.1 cm set @xmath2196 and @xmath2197 set @xmath2198 .
in addition , let @xmath2199 @xmath2200 we have @xmath2201 so we conclude from ( [ boundary - measure-0 ] ) and ( [ boundary - measure-1 ] ) that @xmath1244 . in order to prove theorem 4.1
, it remains to prove that the measures @xmath1249 , @xmath1251 , and @xmath2187 satisfy the corresponding properties .
.1 cm we first show that @xmath2202 is @xmath1250-symmetric .
let @xmath615 and let @xmath474 and @xmath926 denote the two orientations on @xmath391 .
let @xmath2203 , @xmath737 , so that @xmath2204 if @xmath2205 . by the definition of the set
@xmath1828 we have that @xmath2206 and @xmath2207 are disjoint , and @xmath2208
. for @xmath2209 set @xmath1649 . combining proposition 5.5 and proposition 5.8 , we have that @xmath2210 for @xmath39 large enough . from the property @xmath1324 of proposition 4.1 , and since @xmath1576
we see that the measures @xmath2211 and @xmath2212 are @xmath1250-equivalent on @xmath391 , where @xmath2213 is the restriction of the measure @xmath2214 on @xmath2215 .
this shows that @xmath2202 is @xmath1250-symmetric .
.1 cm next , we show that @xmath2216 is @xmath2217-symmetric ( and therefore this measure is @xmath1250-symmetric )
. let @xmath1839 .
by proposition 5.9 we have that @xmath1840 if and only if @xmath2218 .
also , for almost every such @xmath1473 we have that @xmath1673 , @xmath2219 , and @xmath2220 are well defined .
we only need to consider such points .
then , by proposition 5.11 we have that @xmath2221 for @xmath39 large enough .
.1 cm let @xmath442 be the orientation on @xmath2222 so that the endpoint @xmath1674 comes before the endpoint @xmath1880 on @xmath442 .
let @xmath2223 so that @xmath2224 if @xmath2225 .
set @xmath2226 .
we have @xmath2227 and for @xmath2228 we have @xmath2229 . since @xmath1563 and from the property @xmath1324 of proposition 4.1 , we see that the measures @xmath2230 and @xmath2231 are @xmath2217-equivalent on @xmath391 . here
@xmath87 is the restriction of the measure @xmath2232 on @xmath2206 , and @xmath2233 is the restriction of the measure @xmath2232 on @xmath2207 .
.1 cm since both measures @xmath2202 and @xmath2216 are @xmath1250-symmetric , we see that @xmath1249 is @xmath1250-symmetric .
.1 cm from proposition 5.10 we have that if @xmath2190 then @xmath2119 . for @xmath39 large enough
, we have @xmath2234 the following lemma will be proved in the next section .
the measure @xmath1251 is @xmath65-symmetric , for any @xmath1263 .
it remains to analyse @xmath2187 . from the definition of @xmath2187 ( and from ( [ b - subset - c ] ) ) we see that if either of the points @xmath1937 , @xmath1893 , @xmath1896 , or @xmath1898 belongs to the support of @xmath2187 we have that @xmath2129 ( recall the definition of @xmath1966 from lemma 5.3 ) .
note that @xmath2235 for every @xmath1721 . from lemma 5.2 ,
lemma 5.3 , and proposition 5.12 , we have @xmath2236 @xmath2237 for @xmath39 large enough .
this completes the proof of theorem 4.1 .
we have the following preliminary propositions .
let @xmath1493 and set @xmath1811 .
let @xmath1582 .
set @xmath2238 and denote by @xmath2239 a unit vector at @xmath787 that is tangent to @xmath391 .
then @xmath2240| < 2|{{\theta}}_{p}(w , u)| \sinh(\log \sqrt{3}).\ ] ] recall that @xmath2241 .
let @xmath1475 be the rotation centred at @xmath787 , and so that @xmath2242 .
then @xmath600 is a rotation for the angle @xmath2243 .
let @xmath2244 and @xmath2245 ( here we use that @xmath1528 commutes with @xmath600 ) .
then @xmath2246=0 $ ] .
.1 cm recall @xmath2247 . using the formula that says that the hyperbolic circumference of the circle of radius @xmath2023 is @xmath2248
we conclude that @xmath2249 since @xmath2246=0 $ ] and from the previous inequality we get @xmath2240| < |\delta_{p}[z',z'_1]|+ { { \bf d}}(z , z')+{{\bf d}}(z_1,z'_1 ) < 2|{{\theta}}_{p}(w , u)| \sinh(\log \sqrt{3}).\ ] ] let @xmath1220 be a geodesic in @xmath71 and let @xmath2250 be the natural parametrisation of @xmath1220 so that @xmath2251 , and so that @xmath139 is to the left of @xmath2252 , where @xmath2252 is the push - forward of the orientation on @xmath915 by the map @xmath1103 .
let @xmath2253 be the unique point so that @xmath2254 , and so that @xmath2255 . set @xmath2256 .
let @xmath2257 , and set @xmath2258 . then @xmath2259 .
let @xmath2260 , and @xmath2261 .
we have @xmath2262 , and @xmath2263 .
note that @xmath2264 $ ] .
let @xmath2265 , and let @xmath2266 be the unit vector so that @xmath2267 .
such vector exists because @xmath595 is not an endpoint of @xmath1220 ( when @xmath2268 there are two such vectors and we choose either one ) .
also , @xmath2269 . from proposition 6.1
we have @xmath2270 the identity ( [ ang - estimate-2 ] ) yields @xmath2271 and therefore , we have @xmath2272 since for @xmath2273 , we have @xmath2274 , we find @xmath2275 .
this shows @xmath2276 which proves the first inequality .
.1 cm it follows from proposition 5.2 that @xmath2277 . since @xmath2278 , we have @xmath2279 , which proves @xmath2263 .
let @xmath2280 and let @xmath2281 then @xmath2282 we compute @xmath2283 we apply proposition 6.2 to the right hand side of the above inequality and since @xmath2284 for @xmath2285 $ ] , we get @xmath2286 the identity @xmath2287 completes the proof of the proposition . for any @xmath2288 let @xmath2289 and @xmath2290 be the measures on @xmath915 that are supported on @xmath2291 $ ] and given by @xmath2292 and @xmath2293 let @xmath2294 be the measure on @xmath915 that is supported at the point @xmath138 and such that @xmath2295 then the measures @xmath2296 and @xmath2290 are @xmath65-equivalent for @xmath2297 .
we write @xmath2298 , @xmath2299 and @xmath343 for @xmath2289 , @xmath2290 and @xmath2294 respectively . to prove that the corresponding measures are @xmath65-equivalent we use proposition 4.1 .
observe that for any @xmath619 , @xmath2300 \le \mu^{-}(-\infty , t],\ ] ] and @xmath2301 \le qe^ { { { 2r - q(0 ) } \over{2}}}.\ ] ] moreover when @xmath2302 we have @xmath2303 finally by the definition of @xmath343 we have @xmath2304 .1 cm we now show that for any @xmath619 we have @xmath2305\le \mu^{-}(-\infty , t+q).\ ] ] for @xmath2306 the left hand side of the above inequality is @xmath138 .
for @xmath2307 the inequality follows from ( [ 1 ] ) , ( [ 2 ] ) and ( [ 3 ] ) .
for @xmath2308 the inequality follows from ( [ 4 ] ) .
this shows ( [ 5 ] ) .
.1 cm by @xmath2309 symmetry we have @xmath2310 and replacing @xmath2311 with @xmath2312 we obtain @xmath2313 thus by ( [ 4 ] ) we have @xmath2314 \le ( \mu^{+}+\delta)(-\infty , t+q).\ ] ] we have verified the hypothesis from proposition 4.1 and this proves the proposition .
let @xmath1667 and let @xmath615 .
we need to show that the restriction of the measure @xmath1251 to @xmath1020 is @xmath65-symmetric .
for @xmath39 large , we have that the support of the measure @xmath1251 is deep into the thin part of @xmath1020 .
in fact , it follows from proposition 5.10 that @xmath2315 .
note that for @xmath2316 we have that @xmath2317 . combining this with proposition 5.5
, we have that for @xmath39 large enough , the restriction of the measure @xmath1251 to @xmath1020 is supported in @xmath2318 .
therefore , for each cusp @xmath593 we consider the restriction of the measure @xmath1251 to @xmath2319 which we denote by @xmath2320 . to prove lemma 5.4 , it is enough to show that each @xmath2320 is @xmath65-symmetric .
there are only finitely many cusps @xmath593 , so that @xmath2320 is a non - zero measure , and the restriction of @xmath1251 to @xmath1020 is the finite sum of these measures @xmath2320 .
note that if @xmath595 is an endpoint of @xmath391 then @xmath2321 is the zero measure .
.1 cm let @xmath721 ( then @xmath695 ) .
we say that @xmath2322 , where @xmath2323 , if the following holds * we have @xmath2316 .
* we have @xmath2324 . * @xmath2325 .
we say that @xmath2326 , if @xmath2327 .
set @xmath2257 , and let @xmath2328 . by definition of @xmath1833 , we have @xmath2329 . for @xmath2330 and @xmath619 set @xmath2331 .
also @xmath2332 .
we have fix @xmath2333 .
then there exists @xmath2334 such that the set @xmath2335 is a symmetric interval @xmath2336 or @xmath2291 $ ] .
observe that @xmath2337 , for any @xmath619 .
the condition @xmath2338 it equivalent with the following two conditions .
the first one is that @xmath595 is not an endpoint of @xmath2339 or of @xmath2340 .
the second one is that @xmath2341 .
one directly verifies that each of these two conditions is satisfied on a symmetric interval .
there exists @xmath370 , so that for @xmath371 , the following holds .
let @xmath2333 .
for every @xmath2342 , we have @xmath2343 where @xmath2344 is the natural parametrisation of @xmath391 such that @xmath2345 .
assume that @xmath721 ( then @xmath695 ) .
since @xmath2346 , and since @xmath2347 is above @xmath2348 , we have that @xmath2349 . by proposition 5.6
we have that @xmath2350 , is the first @xmath950-horoball that the geodesic ray @xmath2351 enters after leaving @xmath1913 . since with this normalisation
the euclidean diameter of any @xmath950-horoball ( except the one based at @xmath139 ) is at most @xmath950 , and since the ray @xmath2351 ends at @xmath2352 , we have @xmath2353 .
similarly , @xmath2354 . from @xmath2355 , we have that @xmath2356 let @xmath2357 , be the unique mbius transformation so that @xmath2358 ( note that @xmath2359 ) .
then for @xmath2360 , we have @xmath2361 , where @xmath2362 , and @xmath2363 .
the coefficients @xmath2364 , are determined by the conditions @xmath2365 , and @xmath2366 .
we have @xmath2367 and @xmath2368 .
since @xmath2369 , for @xmath2370 $ ] we have that @xmath2371 since @xmath2372 and since @xmath1277 does not increase the hyperbolic distance we obtain @xmath2373 together with proposition 5.8 this completes the proof .
let @xmath2333 .
define the mappings @xmath2374 , by @xmath2375 and let @xmath2376 where the measures @xmath2377 were defined in proposition 6.4 .
we have that @xmath2378 .
note that @xmath2379 is supported at the point @xmath2380 , where @xmath2381 is the unit normal vector at @xmath610 that points towards the cusp @xmath595 ( so @xmath2382 ) .
the measure @xmath2383 is supported at the point @xmath2380 , where @xmath2384 is the unit normal vector at @xmath610 that points away from the cusp @xmath595 ( so @xmath2385 ) .
.1 cm let @xmath2386 be the atomic measure that on @xmath1020 that is supported at the point @xmath2387 , where @xmath2388 , and the mass of @xmath2386 is @xmath2389 ( recall the definition of @xmath2390 from proposition 6.3 ) . for every @xmath2333 the measure
@xmath2391 is @xmath2392-symmetric .
we can by abuse of notation write @xmath2393 and @xmath2386 as measures on @xmath391 rather than on @xmath1020
. then we must show that @xmath2394 and @xmath2395 are @xmath2392-equivalent .
.1 cm we let @xmath2396 and @xmath2397 .
by the previous proposition for @xmath2398 we have @xmath2399 , for @xmath39 large enough .
therefore by proposition 4.1 the measures @xmath2393 and @xmath2400 are @xmath225-equivalent .
also @xmath2401 .
the proof now follows from proposition 6.4 and proposition 4.6 .
let @xmath1481 be a borel subset of @xmath2402 .
for @xmath2403 , @xmath2404 , let @xmath2405 since @xmath1540 is a measure preserving flow ( one - parameter abelian group ) and since @xmath2406 , there exists a non - negative constant @xmath2407 that depends only on @xmath1481 , so that @xmath2408 .
the constant @xmath2407 is called the cross sectional area of the set @xmath1481 . this way
we construct a positive borel measure @xmath2409 on @xmath2402 .
similarly , we get the measure @xmath2410 on @xmath2411 . if @xmath2412 , then @xmath2413 , and @xmath2414 ( because @xmath1576 , and @xmath1564 ) .
this shows that the restriction of the map @xmath2415 , satisfies that @xmath2416 .
if @xmath2417 , then the measure @xmath2321 is the zero measure .
.1 cm it follows from the definition of @xmath1251 and from ( [ b - subset - c ] ) that @xmath2418 then lemma 5.4 follows from the previous proposition and proposition 4.7 .
one sees from proposition 6.4 that @xmath2419 .
the following definition and propositions will be used throughout this section .
recall that for @xmath1493 the map @xmath1494 is the natural parametrisation of the geodesic ray that starts at @xmath610 and that is tangent to the vector @xmath1474 at @xmath610 that is @xmath1495 . by @xmath1778
$ ] we denote the corresponding geodesic segment .
let @xmath1493 and @xmath619 .
the geometric intersection number @xmath2420 is defined as the number of ( transverse ) intersections between the geodesic segment @xmath1778 $ ] and the edges from @xmath414 . for @xmath1721
the intersection number @xmath2421 is defined in the same way . next we establish the connection between the @xmath39-combinatorial length @xmath1928 ( definition 5.7 ) and the intersection number @xmath2422 .
let @xmath1721 such that @xmath2423 .
then @xmath2424 let @xmath1473 also denotes a lift of @xmath1473 in @xmath1472 .
let @xmath391 be the geodesic ray that connects @xmath610 and @xmath2425 .
the assumption @xmath2426 implies that @xmath610 does not belong to the horoball @xmath2427 and that is why @xmath1929 .
let @xmath924 be the intersection point between @xmath391 and the horocircle @xmath2428 ( since @xmath391 ends at @xmath1674 there is exactly only one such intersection point @xmath924 ) .
let @xmath739 be the geodesic sub - segment of @xmath391 bounded by the points @xmath610 and @xmath924
. by definition the point @xmath1727 also belongs to the horocircle @xmath2429 .
let @xmath2430 .
then the euclidean distance between the points @xmath924 and @xmath2431 is at most @xmath950 and the corresponding @xmath745-coordinate of both these points is equal to @xmath2432 .
.1 cm let @xmath1105 be an edge from @xmath414 such that @xmath391 intersects @xmath1105 .
then @xmath1105 does not end at @xmath139 ( since @xmath391 ends at @xmath139 we see that @xmath391 can not ( transversely ) intersect any other geodesic that ends at @xmath139 ) .
moreover we have that the segment @xmath739 intersects @xmath1105 . but then @xmath1105 either intersects intersects the segment @xmath2433 $ ] or the horocyclic segment between the points @xmath924 and @xmath1727 . every point on this horocyclic segment
has the @xmath745-coordinate equal to @xmath2432 so we conclude that if @xmath1105 intersects this horocyclic segment it has to end at @xmath139 .
therefore we find that @xmath1105 intersects the segment @xmath2433 $ ] .
this proves the proposition .
let @xmath1721 .
there exists a constant @xmath2434 that depends only on @xmath0 such that is @xmath2435 then @xmath2436 . if @xmath2437 then @xmath2438 .
in particular for any @xmath1721 we have @xmath2439 . if @xmath2435 then the geodesic segment remains in @xmath2440 . since the space of geodesic segments of the hyperbolic length @xmath950 that remain in @xmath2440 is compact , we find that there exists @xmath2434 so that every such segment intersects at most @xmath2441 edges from @xmath389 .
.1 cm assume that @xmath610 does not belong to @xmath1773 ( this equivalent to saying that @xmath2442 )
. then @xmath2443 for exactly one cusp @xmath114 .
let @xmath2444 .
let @xmath1493 also denote a lift of @xmath1473 so that @xmath2445 . then the geodesic segment @xmath2446 $ ] can intersect only edges from @xmath414 that end at @xmath139 .
this is why @xmath2446 $ ] can intersect at most as many edges from @xmath414 as the horocyclic segment ( horocyclic with respect to @xmath139 ) of the hyperbolic length @xmath950 that begins at @xmath610 .
therefore @xmath2446 $ ] intersects at most @xmath2447 edges from @xmath413 .
this is easily seen by observing that @xmath2448 agrees with the @xmath745-coordinate of the point @xmath605 . there exists a constant @xmath2449 ( that depends only on @xmath0 ) and @xmath370 so that for @xmath371 the following holds .
let @xmath1493 . if @xmath2450 then @xmath2451 if @xmath2452 for some @xmath593 then @xmath2453 here @xmath2454
if @xmath2455 and @xmath2456 . otherwise set @xmath2457 .
note that if @xmath2458 gets smaller the upper bound in ( [ cl - estimate - thin ] ) increases .
however once @xmath2459 is small enough so that @xmath2460 then the upper bound in ( [ cl - estimate - thin ] ) decreases sharply .
in fact @xmath2461 takes values in the interval @xmath2462 $ ] .
assume first that @xmath2463 . then by definition @xmath1853 for some @xmath593 .
moreover we have @xmath2455 and @xmath2464 ( see ( [ ang - estimate-1 ] ) ) . in this case
we have @xmath2465 because @xmath610 belongs to the @xmath950-horoball at the cusp which absorbs @xmath1497 .
if @xmath2450 we have that @xmath2466 so the inequality ( [ cl - estimate - thick ] ) holds for such @xmath1473 .
if @xmath2467 then ( [ cl - estimate - thin ] ) holds as well . from now on we assume that @xmath2426
. then @xmath2468 by proposition 7.1 .
.1 cm let @xmath1493 and let @xmath2469 .
by proposition 5.6 we have that the segment @xmath2470 $ ] is contained in @xmath1707 and therefore by proposition 7.2 we have @xmath2471,\tau(g ) ) \le k(s)({{\bf t}}_r(z , v)+1)\ ] ] it remains to estimate @xmath2472,\tau(g))$ ] .
fix @xmath2473 .
let @xmath2474 , @xmath2475 , and let @xmath2476 be the ordered set of points where the segment @xmath1778 $ ] intersects the @xmath950-horocircles .
note that there can be at most @xmath2477 such segments , that is @xmath2478 .
each segment @xmath2479 $ ] is either contained in @xmath1707 or in some @xmath2058 . if @xmath2479 $ ] is contained in @xmath1707 then by proposition 7.2 we have @xmath2480,\tau(g ) ) \le k(s)(|t_{i+1}-t_i|+1).\ ] ] .1 cm assume that @xmath2479 $ ] is contained in some @xmath2058 ( note that @xmath2481 ) .
there are two cases .
the first one is when @xmath1853 ( then clearly @xmath2482 ) .
then the entire segment @xmath2483 $ ] is contained in @xmath2058 .
let @xmath2484 .
then the segment @xmath2483 $ ] can intersect at most @xmath2485 edges from @xmath414 . on the other hand from ( [ ang - estimate-1 ] ) we have that @xmath2486 if @xmath2487 and @xmath2488 if @xmath2489 .
this shows that @xmath2490,\tau(g ) ) \le 2n(s)\chi_r(z , v)e^{{{\bf h}}(z)}.\ ] ] for any other segment @xmath2479 $ ] that is contained in some @xmath2058 we have that @xmath2491 where @xmath2484 .
this follows from the fact that @xmath2481 , and from @xmath2492 .
then the segment @xmath2479 $ ] can intersect at most @xmath2485 edges from @xmath414 , that is @xmath2493,\tau(g ) ) \le 2n(s)e^{r+1}.\ ] ] .1 cm since @xmath2494 and from ( [ ocena - nulta ] ) , ( [ ocena - prva ] ) and ( [ ocena - druga ] ) we obtain @xmath2495,\tau(g ) ) \le 2n(s)\chi_r(z , v)e^{{{\bf h}}(z)}+(2(3r+{{\bf h}}(z))+1)2n(s)e^{r+1}+(2(3r+{{\bf h}}(z))+1)k(s),\ ] ] ( here we also used that there are at most @xmath2477 segments @xmath2479 $ ] ) .
.1 cm let @xmath2450 . if @xmath1853 then from @xmath2496 we obtain @xmath2495,\tau(g ) ) \le 2n(s)e^{11r}+(26r+1)2n(s)e^{r+1}+(26r+1)k(s).\ ] ] together with ( [ thick - cl ] ) this proves ( [ cl - estimate - thick ] ) .
.1 cm suppose that @xmath2452 for some @xmath593 .
then similarly from ( [ ocena - nulta ] ) , ( [ ocena - prva ] ) and ( [ ocena - druga ] ) we get @xmath2495,\tau(g ) ) \le 2n(s)\chi_r(z , v)e^{{{\bf h}}(z)}+(2(3r+{{\bf h}}(z))+1)2n(s)e^{r+1}+(2(3r+{{\bf h}}(z ) ) + 1)k(s).\ ] ] since @xmath2497 for @xmath39 large enough we have @xmath2498 together with ( [ thick - cl ] ) this proves ( [ cl - estimate - thin ] ) .
let @xmath28 be a normalised group and let @xmath2499 , that is @xmath2500 .
let @xmath2501 , where @xmath2502 , @xmath1995 , and let @xmath1220 be the geodesic segment between @xmath924 and @xmath931 .
moreover let @xmath2503 be the vertical geodesic segment connecting @xmath1996 with the point @xmath2504 ( that is @xmath2503 is orthogonal to the @xmath138-horocircle at @xmath139 ) .
then @xmath2505 let @xmath655 .
if @xmath391 intersects @xmath1220 then @xmath391 either intersects @xmath2100 or @xmath2506 or the horocyclic segment between the points @xmath2507 and @xmath2508 . since @xmath2500 and
since @xmath28 is normalised we have that if @xmath391 intersects this horocyclic segment then @xmath391 ends at @xmath139 .
on the other hand there are at most @xmath2509 such vertical geodesics in @xmath414 .
this proves the proposition .
let @xmath1705 and set @xmath2510 we have there exists @xmath370 so that for @xmath371 and every @xmath2511 we have @xmath2512 where @xmath2513 is the constant from lemma 5.1 .
fix @xmath2511 .
we have @xmath2514 one finds that @xmath2515 where @xmath11 is the number of cusps in @xmath103 . since @xmath2516 for @xmath39
large enough we have @xmath2517 .1 cm it follows from proposition 5.6 that for every @xmath1721 and every @xmath2518 we have that @xmath2519 , where @xmath1786 is the set defined in lemma 5.1 .
if @xmath2520 then @xmath2521 it follows from lemma 5.1 that @xmath2522 since @xmath2516 for @xmath39 large enough we have @xmath2523 together with ( [ prva ] ) this proves the proposition .
we are ready to prove lemma 5.2 .
let @xmath2524 we have @xmath2525 .
it follows from proposition 7.3 and from the definition of @xmath2526 that @xmath2527
@xmath2528 @xmath2529 we estimate each of the four integrals on the right hand side .
cm from proposition 7.5 we have @xmath2530 for @xmath39 large enough .
.1 cm next we estimate the second integral @xmath2531 we have that @xmath2532 .
let @xmath1751 such that @xmath2533_g = c_i(s)$ ] and set @xmath2534 .
let @xmath2535 .
then @xmath2536 in the polar coordinates on @xmath1472 .
we have @xmath2537 where @xmath344 is some universal constant .
this shows that for @xmath39 large enough we have @xmath2538 @xmath2539 .1 cm next we estimate the third integral .
recall that if @xmath2540 is an integrable function on a measure space @xmath2541 then @xmath2542 set @xmath2543 and @xmath2544 .
we find @xmath2545 this together with proposition 7.5 gives @xmath2546 @xmath2547 so for @xmath39 large enough we have @xmath2548 .1 cm if @xmath2549 then @xmath2550 .
this implies that @xmath2551 we have estimated all four integrals and we find that @xmath2552 this proves lemma 5.2 . from proposition 5.16
we have @xmath2553 @xmath2554 @xmath2555 so in order to prove lemma 5.3 we need to estimate from above the integrals on the right hand side .
.1 cm we first estimate the integral @xmath2556 let @xmath2557 .
then @xmath1956 . let @xmath1473 also denote a lift of @xmath1473 to @xmath71 .
then @xmath2120 , for some @xmath593 and @xmath595 is an endpoint of @xmath1673 . here
= c_i(s)$ ] for some @xmath114 .
let @xmath2559 be the set of those @xmath2145 so that @xmath2142 and @xmath2560 .
let @xmath2561 be the set of those @xmath2145 so that @xmath2142 and @xmath2562
. then @xmath2563 .
.1 cm let @xmath2564 .
since @xmath1956 we have @xmath2565 .
it follows from proposition 5.2 that @xmath2566 .
this implies that @xmath2567 , @xmath1995 .
since @xmath1736 we have @xmath2465 . combining this with ( [ cl - estimate - thin ] ) ( recall the definition of @xmath2461 ) we get @xmath2568 a single lift of the set @xmath2559 to @xmath71 ( with respect to the normalised group @xmath721 ) is contained in the set @xmath2569 passing with the integration to the universal cover we get for @xmath39 large that @xmath2570 @xmath2571 @xmath2572 we repeat this for every @xmath114 . after repeating the same argument for @xmath2573 , @xmath1995 , we obtain @xmath2574 for @xmath39 large enough .
.3 cm it is somewhat more delicate to estimate the integrals over the sets @xmath2575 and @xmath2576 .
we need to prove several propositions first .
let @xmath2577 . then for @xmath622 large enough , we have @xmath2578 denote the above integral by @xmath2579 .
consider @xmath2580 .
it follows from proposition 7.2 that @xmath2581
so we conclude that for any @xmath1721 we have @xmath2582 it follows that @xmath2583 since @xmath2584 and since @xmath1504 is measure preserving , we have @xmath2585 @xmath2586 @xmath2587 where @xmath2588 .
we have @xmath2589 where @xmath375 depends only on @xmath0 . on the other hand , by passing to the group @xmath2444 and using the expression for the volume element @xmath2590 we have @xmath2591 @xmath2592 combining these estimates we have @xmath2593 and therefore @xmath2594 . for
@xmath622 large enough we have @xmath2595 which proves the lemma .
we have already discussed ( in section 6 ) the cross sectional area for two dimensional subsets of @xmath1472 with respect to the equidistant flow @xmath1540 .
let @xmath1478 be a domain , and let @xmath2596 be a smooth unit vector field on @xmath1481 .
then @xmath2597 is a surface sitting inside the three dimensional manifold @xmath1472 .
let @xmath2598 .
define @xmath2599 . for simplicity , set @xmath2600 .
we use the same notation for @xmath1479 .
we say that a unit vector field @xmath2239 defined on a domain @xmath1478 is transverse to @xmath1481 if @xmath2601 for every @xmath1781 .
assume that @xmath2239 is transverse to @xmath1481 .
since @xmath1504 preserves the volume , we have that the quotient @xmath2602 does not depend on @xmath1705 .
we call @xmath2602 the cross sectional area of @xmath1481 with respect to @xmath2596 .
in fact , the flow @xmath1504 induces an area form @xmath2603 on @xmath1481 so that the cross sectional area of @xmath1481 agree with the area of @xmath1481 with respect to @xmath2603 .
the two form @xmath2603 is obtained by contracting @xmath2604 by the vector field @xmath2596 .
one can verify that since each vector @xmath2596 has the unit length , and since @xmath2605 we have that @xmath2606 where @xmath2607 .
that is the density of the two form @xmath2603 is always bounded above by the density of the hyperbolic metric .
.1 cm let @xmath2608 be the projection map , that is on each slice @xmath2609 , @xmath2610 , the map @xmath165 agrees with @xmath2611 . since @xmath2239 is transverse to @xmath1481 the map @xmath165 is well defined .
.1 cm let @xmath2612 , @xmath2613 , and @xmath2614 be integrable functions .
if for every such @xmath622 we have @xmath2615 for every @xmath2616 then @xmath2617 and if @xmath2618 for every @xmath2616 then @xmath2619 .1 cm we are particularly interested in the vector fields @xmath2620 and @xmath2621 . the vector @xmath2620 is the unique vector such that the point @xmath2622 corresponds to the coordinates @xmath2623 .
observe that @xmath2624 is transverse to @xmath71 .
the vector @xmath2621 is the unique vector such that the point @xmath2625 corresponds to the coordinates @xmath2626 .
one can verify that @xmath2627 and that @xmath2628 .
it is not surprising that @xmath2628 since the corresponding set @xmath2629 is two dimensional for any set @xmath1478 and therefore the cross sectional area of @xmath1481 with respect to @xmath2621 is equal to zero .
.1 cm fix @xmath114 . for @xmath2500 set @xmath2630 where @xmath1995 .
set @xmath2444 .
note that if @xmath2631 and @xmath655 then @xmath391 has @xmath139 as its endpoint .
this implies that for @xmath2632 we have @xmath2633 and @xmath2634 for @xmath2635 .
.1 cm fix @xmath2500 . for every @xmath1705 the set of points
@xmath2636 is a horocircle in @xmath71 that bounds a horoball at @xmath139 ( recall @xmath41 ) .
moreover , there exist numbers @xmath2637 and @xmath2638 ( that depend only on @xmath622 ) , such that for @xmath2639 we have @xmath2640 . the functions @xmath2641 and @xmath2642 are increasing in @xmath622 .
it is elementary to verify that @xmath2643 and @xmath2644 . .1
cm we apply proposition 7.4 to the segment @xmath2645 $ ] .
we have @xmath2646 this shows that for every @xmath2647 we have @xmath2648 @xmath2649 that is @xmath2650 .1 cm since the vector @xmath2651 is obtained by translating the vector @xmath2621 for @xmath2652 and @xmath2641 does not depend on @xmath255 we get @xmath2653 for @xmath2500 and any @xmath117 we have @xmath2654 and @xmath2655 for @xmath622 large enough . fix @xmath114 and let @xmath2444 .
let @xmath1478 be the set given by @xmath2656 .
let @xmath2657 be the corresponding set ( see the above definition ) .
note that @xmath2658 .
moreover , the set @xmath2659 injects into @xmath2660 under the standard covering map .
therefore , from proposition 7.6 for @xmath622 large enough we have @xmath2661 @xmath2662 @xmath2663
.1 cm let @xmath2664 be the restriction of the projection map @xmath2665 introduced above .
it holds that @xmath2666 for every @xmath1775 where @xmath2667 is a constant that depends only on @xmath0 .
the constant @xmath2667 bounds above the number of intersections between the geodesic segment @xmath2668 $ ] and @xmath413 when @xmath2669 . from ( [ psi-2 ] ) we have @xmath2670 combining this with ( [ a - estimate-3 ] ) yields @xmath2671 for @xmath622 large enough . since @xmath2627
this gives @xmath2672 @xmath2673 @xmath2674 from ( [ a - estimate-2 ] ) and ( [ a - estimate-4 ] ) we obtain @xmath2675 observe that for @xmath2676 $ ] we have @xmath2677 .
therefore @xmath2678 combining this with ( [ a - estimate-1 ] ) we conclude that for every @xmath2500 we have @xmath2679 .
the second estimate in the proposition follows from the first one , and from ( [ a - estimate-2 ] ) .
if instead of @xmath2620 we use @xmath2680 to define @xmath2681 that is @xmath2682 we obtain the same estimate @xmath2683 .
.1 cm again fix @xmath114 .
let @xmath2684 and @xmath2685 .
for @xmath2500 set @xmath2686 set @xmath2687 .
we want to estimate @xmath2688 . for every @xmath1705 and @xmath2500 the set of points
@xmath2689 is a horocircle in @xmath71 that bounds a horoball at @xmath139 ( recall @xmath41 ) . since @xmath2690 we see that @xmath2691 for any value of @xmath622
this implies that the point @xmath2689 lies below the the horocircle @xmath2692 for @xmath39 large enough .
.1 cm fix @xmath2500 .
recall the functions @xmath2641 and @xmath2642 defined above ( and the corresponding point @xmath2639 ) .
similarly , we define the functions @xmath2693 and @xmath2694 as follows .
let @xmath2695 so that @xmath2696 lies on the same horocircle @xmath2692 as the points @xmath610 and @xmath1594 , and so that @xmath2697
. then @xmath2698 .
the functions @xmath2693 and @xmath2694 are increasing in @xmath622 and do not depend on @xmath610 .
it is elementary to verify that @xmath2699 and @xmath2700 .
.1 cm as we explained ( and stated ) above from proposition 7.4 we have @xmath2701 similarly we apply proposition 7.4 to the segment @xmath2702 and get @xmath2703 ( here we used that the euclidean length of the horocyclic segment between the points @xmath1594 and @xmath2696 is at most @xmath2059 ) .
this yields that @xmath2704 so from proposition 7.7 , for @xmath39 large enough we get @xmath2705 that is @xmath2706 if instead of @xmath2620 we use @xmath2680 to define @xmath2688 , that is @xmath2707 we get that the same estimate ( [ i - estimate ] ) holds . .1 cm we are ready to estimate the integral @xmath2708 note that for @xmath114 the vector fields @xmath2709 are well defined on @xmath2710 .
.1 cm let @xmath2711 .
let @xmath2712 .
we also use @xmath1473 to denote a lift of @xmath1473 to @xmath71
. then @xmath2713 and @xmath2714 for some cusp @xmath593 .
set @xmath721 .
it follows from ( [ ang - estimate-3 ] ) that for @xmath39 large enough , we have @xmath2715 for some @xmath2716 . combining this with the fact that @xmath2450 ( recall that @xmath2717 ) , this implies that @xmath2718 where @xmath2719 is either equal to @xmath2624 or to @xmath2720 .
.1 cm recall that if @xmath2721 then @xmath1473 does not belong to @xmath2722 so for @xmath1679 , by proposition 7.1 we have @xmath2723 in particular for @xmath2724 we have the estimate @xmath2725 where @xmath2726 is the projection map .
define @xmath2727 by @xmath2728 and @xmath2729 by @xmath2730 @xmath2731 we apply ( [ psi-1 ] ) separately for @xmath2732 and @xmath2733 . adding the two inequalities , we obtain @xmath2734 @xmath2735 passing to the universal cover for each cusp @xmath117 and applying the fubini theorem to the last integral , yields that ( recall @xmath2736 from the remark after the proof of proposition 7.7 ) @xmath2737 @xmath2738 where @xmath11 is the number of cusps in @xmath103 . by using ( [ i - estimate ] ) and proposition 7.7 and
since @xmath2739 we have @xmath2740 @xmath2741 @xmath2742 for @xmath39 large enough . .3
cm let @xmath2743 . recall that if @xmath2163 then for some @xmath2055 and some @xmath593 we have that @xmath2054 and @xmath2744 this implies that if @xmath2745 then @xmath2746 where @xmath2719 is either equal to @xmath2624 or to @xmath2720 .
.1 cm recall that if @xmath2745 then @xmath1473 does not belong to @xmath2722 , for @xmath1679 . by proposition 7.1
we have @xmath2747 in particular , for @xmath2748 we have the estimate @xmath2749 where @xmath2750 is the projection map .
define the maps @xmath2751 and @xmath2729 as above .
we apply ( [ psi-2 ] ) separately for @xmath2732 and @xmath2733 . adding the two inequalities we obtain @xmath2752 @xmath2753 passing to the universal cover for each cusp @xmath117 and applying the fubini theorem to the last integral , yields that @xmath2754 where @xmath11 is the number of cusps in @xmath103 . by using ( [ i - estimate ] ) and proposition 7.7
we have @xmath2755 @xmath2756 @xmath2757 since @xmath2758 .
.1 cm putting together ( [ e-1-estimate ] ) , ( [ e-2-estimate ] ) , and ( [ e-3-estimate ] ) , and replacing these in ( [ h - estimate ] ) , we get @xmath2759 which proves lemma 5.3 .
to prove lemma 5.1 it is enough to prove the following somewhat more general theorem .
let @xmath0 be a hyperbolic non - compact finite - area riemann surface and let @xmath2760 be compact .
then we can find @xmath2761 such that @xmath2762 \subset k \}.\ ] ] satisfies @xmath2763 for any @xmath1705 .
we may assume that @xmath2764 for some @xmath2765 .
we fix a proper ideal triangulation @xmath104 for @xmath0 , and let @xmath2766 denote the set of geodesic edges of @xmath104 . for any @xmath2767 , we can record the sequence of left and right turns taken by the geodesic ray @xmath2768 to obtain a sequence @xmath2769 , where @xmath2770 .
this sequence is finite if and only if @xmath2771 .
we let @xmath2772 be the times such that @xmath2773 , and we observe that we can find @xmath2774 ( @xmath2775 ) such that @xmath2776 as long as @xmath2777 \subset { { \bf { th}}}_s(h_0)$ ] . moreover , if @xmath2778 \subset { { \bf { th}}}_s(h_0)$ ] ( where @xmath2779 ) , then @xmath2780 for @xmath2781 , where @xmath2782 , depends only on @xmath2783 .
.1 cm we let @xmath2784 we will show that there are @xmath2785 such that @xmath2786 ; this will complete the proof of the theorem .
.1 cm to this end , we fix a triangle @xmath2787 ( @xmath2788 is the lift of @xmath104 to @xmath71 ) .
we observe that if @xmath2789 , then the sequence @xmath2790 depends only on @xmath2791 .
we let @xmath2792 . for any @xmath2793 , where @xmath2794 are positive integers
, the set @xmath2795 is an open interval in @xmath2796 whose endpoints are joined by an element of @xmath2797 . therefore @xmath2798 is a disjoint union of @xmath2799 such intervals . for any such interval @xmath2800 , we let @xmath2801 be the least interval such that @xmath2802 .
the following two facts are central to our argument : here @xmath2806 where @xmath2807 and @xmath2808 , is a mbius invariant of @xmath2809 .
it follows that @xmath2810 where we fix a unit disk model @xmath1639 for @xmath71 , and let @xmath2811 be the arc length for @xmath2812 .
therefore @xmath2813 so @xmath2814 where @xmath2815 .
since for any interval @xmath2816 we have @xmath2817 we find that @xmath2818 where @xmath344 depends only on @xmath2783 and @xmath104 .
l. ahlfors , _ lectures on quasiconformal mappings .
_ second edition . with supplemental chapters by c. j. earle , i. kra , m. shishikura and j. h. hubbard .
university lecture series , 38 .
american mathematical society , providence , ri , ( 2006 ) . | let @xmath0 and @xmath1 be two hyperbolic finite area surfaces with cusps .
we show that for every @xmath2 there are finite degree unbranched covers @xmath3 and @xmath4 , such that the weil - petersson distance between @xmath5 and @xmath6 is less than @xmath7 in the corresponding moduli space . |
in 1952 , lee and yang in their famous papers @xcite first proposed a new and effective method for investigation of phase transitions .
they studied the distribution of zeros of the partition function considered as a function of a complex magnetic field ( fugacity ) , and proved the circle theorem , which states that the zeros of the partition function of ising ferromagnet lie on the unit circle in the complex fugacity plane ( _ yang - lee zeros _ ) . after these pioneer works of lee and yang , fisher @xcite , in 1964 , initiated the study of partition function zeros in the complex temperature plane ( _ fisher zeros _ ) .
these methods are then extended to other type of interactions and were widely applied @xcite .
recently , binek @xcite showed that the density of yang - lee zeros can be determined experimentally from the field dependence of the isothermal magnetization data .
the fractal structure of fisher zeros in q - state potts model on the diamond lattice was obtained by derrida , seze and itzykson @xcite .
they showed that the fisher zeros in hierarchical lattice models are just the julia set corresponding to the renormalization transformation .
bosco and goulard rosa @xcite investigated yang - lee zeros of the ferromagnetic ising model on the cayley tree and associated these with the julia set of the cayley tree recursion map .
they studied the case when the complex magnetic field was applied only to spins on lattice surface and found that yang - lee zeros are distributed along the unit circle .
the lattice models with multisite interaction ising and heisenberg models have been used for modelling the physical systems such as the binary alloys @xcite , classical fluids @xcite , liquid bilayers @xcite , solid @xmath0 @xcite , rare gases @xcite and anisotropic magnets ( @xmath1 , @xmath2 , etc . ) .
these systems have rather complicated phase diagrams and unusual properties . indeed , the monte carlo results give evidence of the presence of phase transitions at nonzero values of the magnetic field for a number of ferromagnetic multisite interaction models @xcite .
there are few analytical results for these systems due to their greater complexity .
here we proposed a numerical algorithm for investigation of complex zeros of partition functions of models on the cayley - type lattices .
this letter is organized as follows . in section 2
we give the model and derive the generalized recurrence equation . in section 3 it is shown that the zeros of partition function can be associated with the set of external parameters ( _ kt , magnetic field _ ) at which the recurrence function has neutral periodic cycles . in section 4
we present some definitions and theorems from complex analytic dynamics and give the numerical algorithm for obtaining yang - lee and fisher zeros of the models . in the last section we give the explanation of the results and make conclusions .
we consider a multisite interaction ising model on the cayley - type lattice .
the cayley tree and husimi lattice are the well known representatives of this class of recurrence lattices . in general
, the cayley - type lattice is constructed of @xmath3-polygons .
it is characterized by @xmath3 , the number of edges ( the number of sites ) of the polygon ( @xmath4 - usual cayley tree , @xmath5 - husimi lattice ) and by @xmath6 , the number of @xmath3-polygons that go out from each site ( fig.1 ) .
( 2.5,1 ) ( 2.5,1 ) ( 0,1)(1.5,0)2(1,-1)1 ( 0,1)(1,-1)2(1,0)1.5 ( 1,0 ) ( 2.5,0 ) ( 1.5,1 ) ( 0,1 ) ( 2.5,1 ) ( 2.5,1 ) ( 0,0)(1,1)2(1,0)1.5 ( 0,0)(1.5,0)2(1,1)1 ( 0,0 ) ( 1.5,0 ) ( 1,1 ) ( 2.5,1 ) ( 1,2.5 ) ( 1,2.5 ) ( 0,1)(1,-1)2(0,1)1.5 ( 0,1)(0,1.5)2(1,-1)1 ( 0,1 ) ( 1,0 ) ( 1,1.5 ) ( 0,2.5 ) ( 1,2.5 ) ( 1,2.5 ) ( 0,0)(1,1)2(0,1)1.5 ( 0,0)(0,1.5)2(1,1)1 ( 0,0 ) ( 0,1.5 ) ( 1,2.5 ) ( 1,1 ) ( 20,18 ) ( 10,3)(-1,4)2(1,4)1 ( 10,3)(1,4)2(-1,4)1 ( 10,11)(3,3)2(4,1)4 ( 10,11)(4,1)2(1,1)3 ( 10,11)(-3,3)2(-4,1)4 ( 10,11)(-4,1)2(-1,1)3 ( 7.5,2 ) ( 10,2 ) ( 6.5,6 ) ( 8,7 ) ( 11,6 ) ( 11,7 ) ( 13,9.5 ) ( 14,11 ) ( 17,14 ) ( 17,15 ) ( 13,14 ) ( 10.5,14 ) ( 7,14 ) ( 6,14 ) ( 2,15 ) ( 0.5,14 ) ( 3.5,11 ) ( 6,9.5 ) ( 10,11.7)(0,0)0 ( 13.7,12.3)(0,0)1 ( 13,13.5)(0,0)1 ( 17,14.5)(0,0)1 ( 6.3,12.3)(0,0)1 ( 7,13.5)(0,0)1 ( 3,14.5)(0,0)1 ( 10.7,7)(0,0)1 ( 9.3,7)(0,0)1 ( 10,2.4)(0,0)1 ( 12.7,11)(0,0)2 ( 12.7,9.5)(0,0)2 ( 14,10.1)(0,0)2 ( 15,10.7)(0,0)2 ( 16.5,10.7)(0,0)2 ( 15.5,12.3)(0,0)2 ( 18,13.7)(0,0)2 ( 19.5,13.7)(0,0)2 ( 18.7,15.3)(0,0)2 ( 18.3,16)(0,0)2 ( 18.3,17.5)(0,0)2 ( 16.7,16.5)(0,0)2 ( 14.3,15)(0,0)2 ( 14.3,16.5)(0,0)2 ( 12.7,15.5)(0,0)2 ( 12,15.3)(0,0)2 ( 10.5,15.3)(0,0)2 ( 11.5,13.7)(0,0)2 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ fig .
the recursive structure of @xmath7-polygon cayley - type lattice with @xmath8 .
the numbers @xmath9 stand for shells .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ one of the pecular properties of the cayley - type lattice is that in the thermodynamical limit ( the number of shells are tended to @xmath10 ) the number of surface sites becomes proportional to the number of inner sites of the lattice ( for more detailes see @xcite ) . therefore , the models defined on the cayley - type lattice exhibit quite unusual behavior @xcite-@xcite . in the infinite dimensional euclidean space
the cayley - type lattices are usually viewed as ramified trees with constant vertex connectivity . however
, these can be embedded in the two - dimensional space of constant negative curvature ( the hyperbolic plane ) with fixed bond angles and length @xcite .
the hamiltonian of the multisite interaction ising model on a @xmath3-polygon cayley - type lattice has the form @xmath11 where @xmath12 takes on values @xmath13 , the first sum goes over all the @xmath3-polygons of the lattice and @xmath14 is the product of all spins on a @xmath3-polygon ; the second sum goes over all the sites on the lattice .
the cayley - type lattice has the advantage that for models defined on it one can derive the exact recurrence relation and use the theory of dynamical systems for investigation thermodynamical properties of the models @xcite .
let us denote the statistical weight of the @xmath15-th generation branch with the basic spin at the state @xmath16 as @xmath17 .
by cutting of the @xmath18-th generation branch at the basic @xmath3-polygon one obtains @xmath19 interacting @xmath20-th generation branches @xmath21 where @xmath22 are the spins on the basic @xmath3-polygon , @xmath23 and @xmath24 is the statistical weight of the basic @xmath3-polygon @xmath25 \label{stweight}\ ] ] where @xmath26 , @xmath27 .
cutting apart the lattice at the central site one can obtain for the partition function @xmath28 the following expression @xmath29 using eq.([brec ] ) and introducing an auxiliary quantity @xmath30 , we can get the recurrence relation for @xmath31 , and express the thermodynamic quantities such as the magnetization , the specific heat , the free energy etc .
, in terms of @xmath31 @xcite .
@xmath32 where @xmath33 , \ ] ] and @xmath34 is the simple combinations of @xmath3 elements taken @xmath35 at a time . using the binomial theorem ,
one can obtain after some calculations the following general formula for the recurrence relation of the multisite interaction ising model on @xmath3-polygon cayley - type lattice in an external magnetic field @xmath36 , where @xmath37 here we used the substitution @xmath38 . in the following
we shall use some definitions from the theory of dynamical systems .
for every @xmath39 on the riemann sphere ( @xmath40 ) @xmath41 , @xmath42 generates the forward orbit of @xmath39 , which is denoted by @xmath43 .
if @xmath44 for some @xmath15 in @xmath45 we say that @xmath46 is a periodic point , and @xmath47 is called a periodic orbit or periodic cycle . if @xmath15 is the smallest integer with that property , then @xmath15 is the period of the cycle .
usually , if @xmath48 , @xmath49 is called a fixed point . obviously , @xmath46 is a fixed point of @xmath50 , if @xmath46 is a periodic point of period @xmath15 .
( one should not confuse the iterations of @xmath51 with the powers of @xmath51 , i.e. @xmath52 is different from @xmath53^n$ ] ) . to characterize the stability of a periodic point @xmath46 of period @xmath15 we have to compute the derivatives .
the complex number @xmath54 @xmath55 is called the eigenvalue of @xmath46 . using the chain rule of differentiation
we see that this number is the same for each point in a cycle .
a periodic point @xmath46 is called ( 1 ) attracting ( stable ) if @xmath56 and @xmath57 for some integer @xmath15 , ( 4 ) irrationally neutral if @xmath58 , but @xmath59 is never @xmath60 .
let us now consider the partition function of the model @xmath61 ^q+1 ) , \label{partit2}\ ] ] where @xmath62 is an analytic function and @xmath63 for any @xmath64 , @xmath65 and @xmath15 ; @xmath39 is an initial point of iterations ( @xmath66 corresponds to the free boundary condition ) . from eq.([partit2 ] ) one can see that @xmath67 is a rational function and the free energy of the system has the form @xmath68 , \label{free}\ ] ] where @xmath69 and @xmath70 are polynomials in @xmath71 and @xmath72 , where all coefficients are positive ( see eq.([xrec ] ) ) . since @xmath62 is an analytic function , the free energy @xmath73 has a singularity if and only if @xmath74 or @xmath75 .
it is obvious from eq.([partit2 ] ) and eq.([free ] ) that @xmath70 is the denominator of @xmath76 function .
since @xmath77 can not be a periodical point for @xmath78 function , the condition @xmath75 will not produce singularity of the free energy @xmath73 in the thermodynamical limit .
hence , the singularities of @xmath73 ( phase transition points ) correspond to zeros of the partition function @xmath79 ( @xmath74 ) , i.e. yang - lee or fisher zeros @xcite .
let us now consider the problem of phase transitions on the cayley - type lattices in terms of recurrence relations and dynamical systems theory . at high temperatures @xmath80 ( @xmath81 is the paramagnetic phase transition point )
the recurrence relation has only one stable fixed point @xmath46 , corresponding to the paramagnetic phase .
when the temperature is lowered , two different scenarios occur depending on the type of interactions . for ferromagnetic interactions ( @xmath82 ) at @xmath83
the paramagnetic fixed point either becomes unstable and there arise two different attracting fixed points , each corresponding to one of the two ferromagnetic phases with opposite directions of magnetization @xcite , or the
fixed point remains stable and there arise an additional attracting fixed point ( see fig .
2 ( d ) ) . for antiferromagnetic interactions ( @xmath84 )
the paramagnetic fixed point at @xmath83 becomes unstable and there occur a period doubling bifurcation cascade and even chaos @xcite .
one says that for given external parameters ( @xmath85 , @xmath65 ) the system is in the stable equilibrium state when the iterations of @xmath78 approach a stable ( attracting ) @xmath86-cycle , and the system undergoes phase transition when , by changing the external parameters , the iterations of @xmath78 pass from one ( nonstable ) @xmath86-cycle to the other ( stable ) @xmath87-cycle ( phase transitions between the modulated phases @xcite ) .
the values of external parameters at which the phase transition occurs may be obtained from the following conditions @xmath88 we see that the phase transition points could be associated with neutral periodic cycles of the mapping .
we call the set of parameter values , at which the rational function @xmath51 from eq.([xgenrec ] ) has neutral periodic cycles , the mandelbrot - like set of @xmath51 . thus , to investigate the yang - lee and fisher zeros one can study the mandelbrot - like sets of @xmath51 .
in this section we give the required definitions and theorems from complex analytic dynamics on the riemann sphere @xcite and develop a numerical algorithm for obtaining the yang - lee and fisher zeros of the model .
one can consider any rational function @xmath89 , where @xmath90 and @xmath91 are polynomials in @xmath92 as an holomorphic map on a riemann sphere .
the poles of the rational function are simply the points of riemann sphere @xmath93 , that are mapped to @xmath77 ( the upper pole of the riemann sphere ) .
the degree of @xmath78 @xmath94 is defined as @xmath95 also , the degree of @xmath51 is the number ( counted with multiplicity ) of inverse images of any point of @xmath96 .
@xmath97 is a critical value of @xmath51 if the equation @xmath98 has a solution the multiplicity of which is greater than one .
such a solution @xmath99 is called a critical point . in local
coordinates this is equivalent to the condition @xmath100 ( at least when @xmath101 ) .
to discuss the behavior of @xmath78 near @xmath77 one usually invokes another transformation , the reflection @xmath102 at the unit circle , which exchanges @xmath103 with @xmath77 ; i.e. to treat the behavior of @xmath104 near @xmath77 , one is to consider @xmath105 in the vicinity of @xmath103 .
below we give some theorems from complex analytic dynamics without proofs .
the reader should consult carleson and gamelin @xcite for proofs .
_ theorem _ 1 .
the number of critical points of @xmath51 is at most @xmath106 .
if @xmath46 is an attracting fixed point for @xmath51 , we define the basin of attraction of @xmath46 , denoted by @xmath107 , to consist of all @xmath92 such that @xmath108 is defined for all @xmath109 and @xmath110 .
the connected component of @xmath107 containing @xmath46 is called the immediate basin of attraction of @xmath46 and is denoted by @xmath111 . if @xmath112 is an attracting cycle of period @xmath15 , then each of the fixed points of @xmath108 has its basin and @xmath113 is simply the union of these basins
. _ theorem _ 2 .
if @xmath46 is an attracting periodic point , then the immediate basin of attraction @xmath111 contains at least one critical point . _
theorem _ 3 .
if @xmath46 is a rationally neutral periodic point , then each immediate basin of attraction associated with the cycle of @xmath46 contains a critical point .
_ theorem _ 4 .
the total number of attracting and neutral cycles is at most @xmath106 .
the numerical algorithm is based on theorems 1 - 4 and the well known fact that the convergence of iterations to neutral periodic cycle is very weak ; i.e. one has to make a number of iterations ( infinite number of iterations ) in order to approach the neutral periodic cycle .
of course , the initial point does not belong to the periodic cycle or the julia set of mapping ( for the definition of the julia set see ref .
the algorithm is as follows : one finds all the critical points of the mapping and investigates the convergence of all the orbits of critical points ( critical orbits ) to any attracting periodic cycle .
if all critical orbits converge , for example , after @xmath15 iterations , one says that the system is in stable equilibrium state , otherwise the system undergoes phase transition . of course , the last statement is not rigorous since a weak convergence to an attracting periodic cycle is possible .
depending on the choice of @xmath15 and @xmath114 - the accuracy of approaching the attracting cycle , the phase transition pictures may change .
our complutergraphical experiments showed that @xmath115 and @xmath116 are optimal values for @xmath15 and @xmath114 , and the pictures generated by this algorithm will not qualitatively change when @xmath15 is increased and/or @xmath114 decreased from their optimal values . in our algorithm
we supposed that the cases where neutral periodic cycles are irrational are very rare and negligible .
one can easily find all the critical points of mapping @xmath51 from eq.([xgenrec ] ) as is described in the previous section .
the critical points are as follows : @xmath117 with multiplicity @xmath118 , @xmath119 with multiplicity @xmath120 and @xmath121 solutions of equations @xmath122 and @xmath123 .
the degree of our mapping @xmath51 is @xmath124 and , according to theorem 1 of section 4 , these all are critical points of @xmath125 .
one can easily find that the most of the critical orbits after the first iteration are intersected and can consider only the orbits starting at points @xmath103 , @xmath126 , @xmath60 , @xmath127 for @xmath128 and @xmath129 , @xmath130 for @xmath4 . also , the orbits of @xmath103 and @xmath77 intersect when @xmath3 is odd and the orbits of 1 and -1 intersect when @xmath6 is odd . bellow we give some pictures generated by our algorithm .
2 shows computergraphical study for fisher zeros of the ising model on the cayley tree ( @xmath4 , @xmath8 ) for @xmath131 and different values of magnetic field @xmath132 .
we have experimental evidence that all critical orbits converge in white regions .
physical phase transitions take place at the points where black regions intersect the real axis .
we see that for @xmath133 ( fig . 2 ( a ) ) the second order phase transition occurs for @xmath134 in accordance with baxter s result @xcite and for @xmath135 ( fig . 2 ( c ) ) there is no phase transition for @xmath136 . for @xmath137 ( fig .
2 ( b ) ) the first order phase transition occurs in _ cat face_-like region .
figure 2 ( d ) shows the neutral fixed point of @xmath78 function corresponding to phase transition in the
_ cat face_-like region of fig .
a care should be taken in calculations for small @xmath138 because one of the attracting fixed points tends to infinity when the temperature is lowered . since the free energy of the ising model on the cayley tree is an even function of @xmath132 , figs .
2 ( a)-(c ) are invariant under the transformation @xmath139 and show phase transitions for both ferromagnetic and antiferromagnetic interactions ( @xmath140 is equivalent to @xmath141 and @xmath139 ) . in fig .
3 we present yang - lee zeros of the ferromagnetic ising model on the cayley tree . figs .
3 ( a)-(b ) show that the phase transition exists at positive temperatures for @xmath142 and @xmath143 in accordance with fig .
the set of yang - lee zeros at the critical point @xmath134 is shown on fig .
one can see that in this case a real phase transition occurs at zero magnetic field only as in usual ferromagnetic systems .
it is interesting to note that the set of yang - lee zeros of ferromagnetic ising model on the cayley tree resembles the boundary of the mandelbrot set of the quadratic mapping @xmath144 . in this case
the yang - lee zeros are not located on the unit circle because in the thermodynamical limit the contribution of surface and inner spins to the partition function of the model on cayley tree are of the same order . in fig .
3 ( d ) we give the improved set of fisher zeros of ising model on the husimi lattice ( @xmath5 , @xmath145 ) previously obtained in ref . @xcite . in conclusion , we should like to note that the numerical algorithm presented in this letter may be applied for obtaining yang - lee and fisher zeros of any model on the cayley - type lattice , for which one - dimensional recurrence relation can be derived .
it is remarkable that there appear the well - known mandelbrot set figures .
this phenomenon is known as the universality of the mandelbrot set @xcite .
the authors would like to thank s. dallakyan , n. izmailian and s. karabegov for fruitful discussions .
this work was partly supported by the grants intas-97 - 347 and istc a-102 .
99 t. d. lee and c. n. yang , _ phys .
rev . _ * 87 * : 404 , 410 ( 1952 ) . m. e. fisher , in _ lectures in theoretical physics _ , edited by w. e. brittin , ( university of colorado press , boulder , 1965 ) , vol . *
7c * , p. 1 .
p. f. arndt , _ phys .
lett . _ * 84 * , 814 ( 2000 ) ; j. l. monroe , _ j. stat .
phys . _ * 65 * , 445 ( 1991 ) ; c. itzykson , r. b. pearson , j. b. zuber , _ nucl . phys . _ * b220 * [ fs8 ] , 415 ( 1983 ) ; d. ruelle , _ phys .
lett . _ * 26 * , 303 ( 1971 ) ; w. t. lu , f. y. wu , _ physica _ * a 258 * , 157(1998 ) , cond - mat/0001409 ( 2000 ) ; y. park , m. fisher , cond - mat/9907429 ( 1999 ) , p. f. arndt , t. heinzel , v. rittenberg , _ j. stat .
phys . _ * 97 * , 1 ( 1999 ) .
binek , _ phys .
* 81 * , 5644 ( 1998 ) .
b. derrida , l. de seze , c. itzykson , _
phys . _ * 33 * , 559 ( 1983 ) .
f. a. bosco , s. goulart rosa , _ europhys .
_ , * 4 ( 10 ) * , 1103 ( 1987 ) ; the fractal dimension of the julia set was calculated incorrectly in this paper , for details see j. l. monroe , _ europhys .
_ , * 29 ( 2 ) * , 187 ( 1995 ) .
r. mosseri and j. f. sados , _ j. phys
. france lett . _ * 43 * , 1249 ( 1982 ) ; j. a. miranda - neto and fernando moraes , _ ibid . _
* 2 * , 1657 ( 1992 ) ; _ ibid . _
* 3 * , 29 ( 1993 ) .
d. f. styer , m. k. phani , j. l. lebowitz , _ phys.rev . _ * b34 * , 3361 ( 1986 ) .
m. grimsditch , p. louberyre , a. polian , _ phys .
rev . _ * b33 * , 7192 ( 1986 ) .
m. roger , j. h. hetherington , j. m. delrieu , _ rev .
phys . _ * 55 * , 1 ( 1983 ) .
h. l. scott , _ phys .
rev . _ * a37 * , 263 ( 1988 ) .
j. a. barker , _ phys .
lett . _ * 57 * , 230 ( 1986 ) .
j. r. heringa , h. w. j. blote , a. hoogland , _ phys .
lett . _ * 63 * , 1546 ( 1989 ) .
ghulghazaryan , _ int .
* b14 * , 589 ( 2000 ) . r. j. baxter , _ exactly solved models in statistical mechanics _
( academic press , london , 1981 ) .
t. p. eggarter , _ phys .
rev . _ * b9 * , 2989 ( 1974 ) .
e. muller - hartmann , j. zittartz , _ phys rev .
lett . _ * 33 * , 893 ( 1974 ) .
thompson , _ j. stat .
_ * 27 * , 441 ( 1982 ) .
n. s. ananikian , n. sh .
izmailian , k. a. oganessyan , _ physica _ * a254 * , 207 ( 1998 ) ; n. s. ananikian , _ et al _ , fractals * 5 * , 175 ( 1997 ) ; n. s. ananikian , s. k. dallakian , _ physica _ * d107 * , 75 ( 1997 ) . c. s. o. yokoi , m. j. de oliveira , s. r. salinas , _ phys . rev .
_ * 54 * , 163 ( 1985 ) ; h. r. tragtenberg , c. s. o. yokoi , _ phys
_ * e52 * , 2187 ( 1995 ) .
l. carleson , t. w. gamelin , _ complex dynamics _
( springer - verlag , new - york , inc . ,
1993 ) .
peitgen , p. h. ritcher , _ the beaity of fractals _
( springer - verlag , berlin , heidelberg , 1993 ) .
a. e. alahverdian , n. s. ananikian , s. k. dallakian , _ phys .
rev . _ * e57 * , 2452 ( 1998 ) .
a. douady , j.h .
hubbard , _ ann .
ecole normale sup .
_ , * 18 * , 287 ( 1985 ) .
figure 1 . the recursive structure of @xmath7-polygon cayley - type lattice with @xmath8 . the numbers @xmath9 stand for shells .
figure 2 . the fisher zeros of ising model on the cayley tree .
( a ) @xmath146 , @xmath133 ; ( b ) @xmath147 , @xmath137 ; ( c ) @xmath131 , @xmath135 ; ( d ) the plot of @xmath78 function from eq .
( [ xgenrec ] ) corresponds to the phase transition point @xmath148 , @xmath147 , @xmath137 of ( b ) .
( a)-(c ) the yang - lee zeros of ising models on the cayley tree ; ( a ) @xmath4 , @xmath8 , @xmath149 ; ( b ) a close up from left ; it is noteworthy that the set in question may be obtained by mapping @xmath150 from the mandelbrot set boundary of ( a ) ; ( c ) @xmath4 , @xmath8 , @xmath134 ; ( d ) the fisher zeros of multisite interaction ising model on husimi lattice @xmath151 , @xmath152 , @xmath5 , @xmath145 . | a general analytical formula for recurrence relations of multisite interaction ising models in an external magnetic field on the cayley - type lattices is derived . using the theory of complex analytical dynamics on the riemann sphere , a numerical algorithm to obtain yang - lee and fisher zeros of the models is developed .
it is shown that the sets of yang - lee and fisher zeros are almost always fractals , that could be associated with mandelbrot - like sets on the complex magnetic field and temperature planes respectively . |
gene trees are fundamental to molecular systematics .
traditionally , a gene tree is reconstructed from dna sequence variation at individual genetic loci in a group of species and is taken as the phylogenetic tree of the species due to sequencing technology limitations . however , when gene copies are sampled from various species , the resulting gene tree might disagree with the species tree .
as such , the relationship between gene trees and species trees has been the focus of many studies ( see for example @xcite ) .
it has long been recognized that gene trees can be used to estimate species divergence time , ancestral population sizes and even the containing species tree although they may not accurately reflect the species tree @xcite .
the discord of gene trees and the containing species tree can arise from horizontal gene transfer , lineage sorting , and gene duplication and loss .
the importance of these causes depends on the considered genes and species .
hence , inferring the species tree from gene trees has been investigated under various parsimony criteria . with lineage sorting ( also called deep coalescence ) , the problem is to find the tree minimizing extra gene lineages that had to coexist along species lineages @xcite ; with gene duplication , it becomes to find the tree minimizing gene duplications and/or losses @xcite .
inferring the species tree from a set of gene trees has often been studied under the gene duplication cost @xcite until very recently . in a seminal work @xcite ,
maddison addressed lineage sorting in the framework of coalescence theory .
coalescence theory is an active branch of population genetics concerned with tracing the genealogical history of a present - day gene copy . for a gene sampled from two individuals , one may ask : how deep in time do these two lineages coalesce ?
hence , the depth of this coalescence is a measure of the relationship between two sampled gene copies .
the more deep in time coalescence occurs , the more distantly related they are .
maddison proposed to use the total number of extra " gene lineages that fails to coalesce on a species tree to measure the inconsistence of a gene tree and species tree , called _ deep coalescence cost_. for the gene tree and species tree shown in figure [ dcexample ] , there are three gene lineages on a branch and two gene lineages on another branch that fail to coalesce , giving the deep coalescence cost of 3 . since coalescence theory provides the probability that a gene tree would exist in a species tree , it allows the inference problem to be studied in explicit statistical framework @xcite .
this seems to give the deep coalescence model an advantage over the other models .
the paper is a sequel of @xcite , which studies the complexity and algorithmic issues of inferring the species tree from a set of gene trees with the gene duplication / loss cost . in this work
, we present a relationship of the deep coalescence cost , the duplication cost , and the number of gene losses .
although deep coalescence and gene duplication are two different mechanisms responsible for the discord of gene trees and species trees , this relationship suggests that the deep coalescence cost and the duplication cost are closely related to each other as a similarity measure of trees .
we further show that inferring species tree from gene trees is also np - hard by minimizing the deep coalescence cost . , scaledwidth=60.0% ]
in this section , we shall introduce basic definitions and notations on gene duplication , gene loss and deep coalescence that are used in the following sections . for a set of @xmath0 taxa
, their evolutionary history is modeled as a rooted , full binary tree with @xmath0 leaves in which leaves are labeled with taxa , representing the labeling taxa , and internal nodes are unlabeled . here
, the fullness means that each internal node has exactly two children .
such a tree is called _
species tree_. in a species tree , each unlabeled internal node is considered as a _
taxon family _ which include as its members the subordinate species represented by the leaves below it
. thus , the evolutionary relation
@xmath1 is a descendant of @xmath0 " is expressed using the set - theoretic notation as @xmath2 " .
we also call an internal node an _ ancestor _ of the species below it . the model for gene relationship
is also a rooted , full binary tree with leaves representing genes , called a _
gene tree_. usually , a gene tree is reconstructed from a collection of gene family members sampled from the considered species .
we label the gene copies by the species from which they are sampled .
thus , leaf labels may not be unique in a gene tree as two or more gene copies might be found in a species .
an internal node @xmath3 corresponds to a multiset of leaf labels . finally , for a species or gene tree @xmath4
, we use @xmath5 to denote the set of leaf labels of it . for an internal node @xmath6 in @xmath4 , @xmath7 and @xmath8
are used to denote its two children .
let @xmath9 be a gene tree and @xmath10 a species tree such that @xmath11 .
for any nodes @xmath12 in @xmath10 , the _ least common ancestor _ of @xmath13 and @xmath14 is defined to be the smallest node @xmath15 in @xmath10 such that @xmath16 , which is denoted by @xmath17 . to reconcile the gene tree @xmath9 and the species tree @xmath10 , each node @xmath3 of @xmath9
is mapped to a unique node @xmath18 in @xmath10 as @xmath19 this mapping @xmath20 was first considered in @xcite and then formulated in @xcite .
we call @xmath20 the @xmath21 mapping or reconciliation of @xmath9 in @xmath10 . obviously , if @xmath22 , @xmath23 .
let @xmath3 be an internal node of @xmath9 .
if @xmath24 for some child @xmath25 of @xmath3 , then we say that a duplication occurs at @xmath18 ( or more exactly in the lineage entering @xmath18 ) in @xmath10 .
the total number of duplications arising in the lca reconciliation @xmath9 in @xmath10 is proposed to measure the discord of the gene tree and species tree and is called the _
duplication cost_. we use @xmath26 to denote the duplication cost for @xmath9 and @xmath10 .
note that the duplication cost is not symmetric .
a subset @xmath27 of ( internal and/or leaf ) nodes of a species tree @xmath10 is incompatible if @xmath28 for any @xmath29 . for an incompatible subset @xmath27 in @xmath10 ,
the restriction of @xmath10 on @xmath27 is the smallest subtree of @xmath10 containing @xmath27 as its leaf set , denoted by @xmath30 .
it is easy to see that the root of @xmath30 is the least common ancestor of the nodes from @xmath27 .
the homomorphic subtree @xmath31 of @xmath10 induced by @xmath27 is a tree obtained from @xmath30 by contracting all degree-2 nodes except for the root of @xmath30 .
let @xmath9 be a gene tree such that @xmath11 .
@xmath32 is well defined . to reconcile @xmath9 and @xmath10 in this general case , we consider the the lca mapping @xmath20 from @xmath9 to @xmath32 .
for any two nodes @xmath15 and @xmath13 of @xmath32 such that @xmath33 , we define @xmath34 that is , @xmath35 is the number of nodes on the path from @xmath13 to @xmath15 .
recall that @xmath36 and @xmath37 denote the children of @xmath3 .
the number of losses @xmath38 associated to @xmath3 is defined as @xmath39 this definition of @xmath38 is a generalization of the loss cost given in @xcite . when @xmath40 , our definition is then identical to the one given in @xcite .
the _ gene loss cost _ in the reconciliation of @xmath9 in @xmath10 is defined as the total number of losses @xmath41 .
we denoted this gene loss cost for @xmath9 and @xmath10 by @xmath42 . .
let @xmath9 be a gene tree and @xmath10 a species tree such that @xmath40 . under the lca mapping
@xmath43 , if a branch @xmath44 of @xmath10 is on the @xmath45 paths from @xmath46 to @xmath47 , @xmath48 ( @xmath49 ) , then we say that there are @xmath50 ` extra ' lineages on @xmath44 failing to coalesce on @xmath44 .
the _ deep coalescence_(dc ) _ cost _ is defined as the total number of the ` extra ' lineages on all branches of @xmath10 in the reconciliation @xmath20 of @xmath9 in @xmath10 ( see @xcite ) , which is denoted by @xmath51 .
note that the concept of deep coalescence is meaningful only if @xmath10 has 2 or not leaves .
we assume this throughout the paper . in general ,
if @xmath52 , the deep coalescence cost @xmath51 is defined as @xmath53 , where @xmath54 is the homomorphic subtree of @xmath10 induced by @xmath55 .
such a generalization will be used in the study of inferring the species tree from a set of gene trees
we have seen that deep coalescences , gene losses and duplications are inferred through the gene tree / speceis tree reconciliation .
actually , they are indeed closely related through a simple equation .
let @xmath9 be a gene tree and @xmath10 a species tree such that @xmath11 . under the lca mapping @xmath56 ,
an internal node @xmath57 is of * type-1 if @xmath58 for each child @xmath59 of @xmath3 ; * type-2 if there exists a unique child @xmath59 such that @xmath60 ; * type-3 if @xmath61 for each child @xmath59 of @xmath3 . note that type-2 or type-3 internal nodes correspond one - to - one with duplication events .
[ thm31 ] let @xmath9 be a uniquely leaf - labeled gene tree and @xmath10 a species tree such that @xmath40 .
then , @xmath62 * proof .
* let @xmath9 and @xmath10 have @xmath0 leaves .
assume that there are @xmath63 type-1 internal nodes @xmath64 @xmath65 type-2 internal nodes @xmath66 and @xmath67 type-3 internal nodes @xmath68 in @xmath9 under the lca mapping @xmath43 , respectively .
since @xmath9 is a full binary tree with @xmath0 leaves , @xmath9 has @xmath69 internal nodes and hence @xmath70 additionally , type-2 and type-3 nodes correspond one - to - one with duplication events , @xmath71 for simplicity , we assume that @xmath59 and @xmath72 are the children of @xmath3 for each type-1 internal node @xmath3 ; we also assume that @xmath36 is the unique child such that @xmath73 for each type-2 node @xmath3 .
since we use @xmath74 to denote the number of nodes on the path from @xmath18 to @xmath75 for a node @xmath3 and its child @xmath76 , the number of lineages contained in the path is @xmath77 .
therefore , by eqn .
( [ node_partition ] ) and ( [ dup_cost ] ) and the fact that @xmath78 , @xmath79 + \left[d\left(m(g''_{1j } ) , m(g_{1j})\right)+1\right]\right\}\\ & & + \sum ^{k_2 } _ { j=1 } \left[d\left(m(a(g_{1j } ) ) , m(g_{1j})\right)+1\right ] -|e(s)|\\ & = & c_{loss}(g , s ) + 2k_1 - ( 2n-2)\\ & = & c_{loss}(g ,
s ) -2(k_2+k_3 ) \\ & = & c_{loss } ( g , s ) - 2c_{dup}(g , s).\end{aligned}\ ] ] this concludes the proof . @xmath80 * remarks*. ( 1 ) following the proof of the equation in the above theorem , one can easily see that for an arbitrary gene tree @xmath9 in which there may be two or more gene copies are from the same species and a species tree @xmath10 such that @xmath40 , @xmath81 \(2 ) since the number of gene duplications and lossed can be calculated in linear time @xcite , the first remark implies that the deep coalescence cost can also be computed in linear time . by thm [ thm31 ] , @xmath82 for a species tree @xmath10 and a uniquely leaf labeled gene tree @xmath9 .
now we show that it is bounded below by the duplication cost for any arbitrary gene tree .
[ dc_duploss_thm ] let @xmath9 be a uniquely leaf - labeled gene tree and @xmath10 a species tree such that @xmath40 .
then , @xmath83 .
* proof . * denote the image node set of the lca mapping @xmath20 by @xmath84 , which is a subset of nodes in the species tree @xmath10 . for any internal node @xmath85 , we use @xmath86 to denote all internal nodes @xmath3 of the gene tree that are mapped to @xmath15 under @xmath20 . for any nodes @xmath87 and a descendant @xmath88 of @xmath87 in the gene tree @xmath9 , if @xmath89 , then @xmath90 for each node in the path from @xmath87 to @xmath88 .
since @xmath9 is uniquely leaf labeled , all internal nodes in @xmath86 form a rooted subtree of @xmath9 , denoted by @xmath91 , as illustrated in figure [ genetreepartition ] .
@xmath91 is not a full binary tree in general . in particular ,
its root might has degree 1 .
let @xmath92 denote the number of non - root degree-1 , degree-2 and degree-3 nodes in the subtree @xmath91 , respectively .
assume that @xmath91 has two or more nodes .
then , by definition , the root of @xmath91 corresponds with a gene duplication in the reconciliation of @xmath9 and @xmath10 ; each degree-2 or degree-3 node of @xmath91 also corresponds with a gene duplication .
therefore , there are @xmath93 duplication events at @xmath15 .
we now consider two cases .
the root of @xmath91 has degree 1 .
then @xmath91 has @xmath94 leaves , that is @xmath95 . for each leaf of @xmath91
, it has two children that are mapped to a node below @xmath15 in the species tree @xmath10 ; each non - root degree-2 node has exactly one child that is mapped to a node below @xmath15 and so is the root since it has degree 1 .
thus , there are @xmath96 image paths that contain one of the two lineages from @xmath15 to one of its children . case 2 .
the root has degree 2 . in this case , @xmath91 has @xmath97 leaves and there are @xmath98 image paths that contain one of the two lineages from @xmath15 to one of its children . by distributing the dc and duplication costs to each image node @xmath15 in @xmath84 , we obtain that @xmath99 this finishes the proof .
@xmath80 * remark * the fact @xmath100 holds even for arbitrary gene trees in which 2 or more leaves with the same label , which represent genes sampled from the same species . in the general case
, @xmath101 might be a forest a union of rooted trees .
however , the estimation ( [ lowerbound_est ] ) in the proof is still valid if the sum is over all the subtrees that are mapped to a node in the species tree , i.e. @xmath101 is replaced by a subtree of each resulting forest . of the gene tree in the species tree , @xmath102
is mapped to the green node , @xmath103 and @xmath104 to the red node , and @xmath3 to the purple node .
the nodes @xmath105 form a subtree of the gene tree.,scaledwidth=60.0% ]
parsimony criterion is often used for inference in biology . hence , inferring species tree from a set of gene trees is formulated as the following algorithmic problem * species tree problem * + input : a set of gene trees @xmath106 , @xmath107 .
+ solution : a species tree @xmath10 that minimizes the total cost @xmath108 , where @xmath109 is a cost function .
it is proved that the species tree problem is np - hard for the duplication and/or loss cost in @xcite . in this section ,
we prove the following theorem .
the species tree problem is np - hard under the dc cost .
* proof . * given a gene tree @xmath9 and a species tree @xmath10 , the dc cost @xmath51 can be computed in polynomial time since gene duplications and losses can be counted in linear time @xcite . therefore , the species tree problem is in np . to prove its np - hardness , we reduce the maximum cut problem to the decision version of the species tree problem . given an instance graph @xmath110 and a positive integer @xmath111 , the maximum cut problem is to partition the node set @xmath112 into two disjoint subsets @xmath113 and @xmath114 such that there are at least @xmath111 edges from @xmath115 that have one endpoint in @xmath113 and one endpoint in @xmath114 .
assume that @xmath116 and @xmath117 denotes the number of edges from @xmath115 , where @xmath118 .
we construct a corresponding instance of the species tree problem as follows .
.,scaledwidth=80.0% ] choose @xmath119 and @xmath120 .
for each node @xmath121 ( @xmath107 ) , we introduce a label with the same name @xmath121 .
we also introduce @xmath122 extra labels @xmath123 , @xmath124 and @xmath125 , @xmath126 . for each edge @xmath127 , we define two gene trees @xmath128 and @xmath129 as shown in figure [ edgetrees ] .
these two trees are same except that the leaf labels @xmath121 and @xmath130 are swapped .
let the trees shown in figure [ gaget_trees ] ( i)-(iii ) be written as @xmath131 $ ] , @xmath132 $ ] and @xmath133 $ ] , respectively . besides the ` edge ' gene trees @xmath128 and @xmath129 ( @xmath134 , the set @xmath135 of gene trees in the instance of the problem to be defined also contains @xmath136 , & 1\leq i < j \leq n,\;1\leq k\leq n , \ ; 1\leq m\leq m , & \\ g'_{(i , j , k , m ) } = l[y_i , y_j , x_k , z_m ] , & 1\leq i < j \leq
n,\;1\leq k\leq n , \ ; 1\leq m\leq m , & \\ g''_{m}=f[\{x_i\},\{y_i\ } , z_m ] , & 1\leq m\leq m. & \end{aligned}\ ] ] these three classes of gene trees are introduced to restrict the topology of the optimal species tree for the defined instance of the problem .
hence , we call them ` structural ' gene trees
. the np - completeness of the decision version of the species tree problem follows from the following two lemmas .
[ lemma1 ] if the graph @xmath137 has a cut of @xmath138 edges , there is a species tree @xmath139 having the dc cost @xmath140 * proof . *
assume that the node set @xmath112 of the graph @xmath137 divides into @xmath141 and @xmath142 such that there are exactly @xmath138 edges having one endpoint in @xmath113 and one endpoint in @xmath143 .
we define a species tree @xmath139 as shown in figure [ speciestree ] .
first , we observe that @xmath144 for each possible @xmath145 .
consider a non - cut edge @xmath146 ( @xmath147 ) . since @xmath148 , @xmath149 and @xmath150 .
if @xmath151 , we have that @xmath152 symmetrically , if @xmath153 , we have that @xmath154 for each cut edge @xmath146 ( @xmath147 ) with one endpoint in @xmath113 , say @xmath155 , and another in @xmath114 , we have that @xmath156 therefore , we have @xmath157 this finishes the proof of the lemma .
@xmath80 defined in lemma [ lemma1].,scaledwidth=80.0% ] $ ] .
( ii ) the resulting tree @xmath158 $ ] after replacing each leaf with a tree in a line tree.,scaledwidth=70.0% ] [ lemma2 ] if there is a species tree
@xmath10 having the dc cost @xmath159 , then the graph @xmath137 has a cut of at least @xmath160 edges .
* proof . * if @xmath161 , the fact is trivial .
hence , without loss of generality , we may assume that @xmath162 . here , we use @xmath163 $ ] to denote the line tree with leaves labeled by @xmath102 , @xmath164 , @xmath165 , @xmath166 , respectively , as shown in figure [ linetree ] ( i ) .
note that the leaf @xmath102 is a child of the root in @xmath167 $ ] .
for a set of trees
@xmath168 , @xmath169 , @xmath165 , @xmath170 , we use @xmath171\ ] ] to denote the tree obtained by replacing each leaf by a corresponding subtree in @xmath172 $ ] as shown in figure [ linetree ] ( ii ) .
let @xmath173 be a subset of leaves in the species tree @xmath10 and the least common ancestor of the leaves from @xmath173 be @xmath174 in @xmath10 . recall that the homomorphic subtree @xmath175 of @xmath10 induced by @xmath173 is the tree obtained from @xmath10 by removing all the nodes and edges that are not on a path from @xmath176 to a leaf from @xmath173 and then contracting all the degree-2 node except for the root @xmath174 .
for example , for @xmath139 defined in lemma[lemma1 ] , @xmath177 $ ] . set @xmath178 by replacing the children of a two - leaf root tree with @xmath179 and @xmath180
, we obtain a species tree @xmath181 $ ] from s. first , @xmath182 has the following property .
* fact 1 * @xmath183 . * proof .
* for each gene tree @xmath184 or @xmath129 , we use @xmath185 and @xmath186 to denote the lca mappings from @xmath4 to @xmath10 and @xmath182 , respectively . for each edge
@xmath187 in the spanning subtree over @xmath188s in @xmath4 , by the definition of @xmath180 , @xmath189 if and only if @xmath190 and @xmath191 . for each edge in the spanning subtree over @xmath192s , @xmath121 , @xmath193 and @xmath194s , the same property holds .
but , the edges incident to the root of @xmath4 may not satisfy the property discussed above .
let @xmath104 be the root of @xmath4 .
assume that @xmath195 is the left child of @xmath104 , which is the least common ancester of @xmath192s and @xmath194s , and @xmath196 the right child of @xmath104 .
it is possible that @xmath197 and/or @xmath198 .
however , @xmath199 , @xmath200 and @xmath201 , where @xmath202 is the root of @xmath182 , @xmath203 and @xmath204 the root of @xmath179 and @xmath180 respectively . since no other lineages fail to coalesce with @xmath205 on @xmath206 and with @xmath207 on @xmath208 respectively
, these two edges does not affect the deep coalescence cost .
thus , @xmath209 .
similarly , we also have the following three inequalities @xmath210 for any @xmath211 .
thus , the fact holds .
* fact 2 . * in @xmath179 , all the leaves @xmath192 must be below one child of the root and all the leaves @xmath194 must be below the other child of the root .
in other words , @xmath212 $ ] , where @xmath213 is a tree over @xmath192 and some @xmath121s and @xmath214 is a tree over @xmath194s and some @xmath193s .
* assume that the fact is false .
there are @xmath215 and @xmath216 such that @xmath217 $ ] , or there are @xmath218 and @xmath219 such that @xmath220 $ ] . if the former is true , then , @xmath221 this implies that @xmath222 contradicting to the fact that @xmath223
. if the latter is true , for any @xmath224 , @xmath225 .
again , we have that @xmath226 , leading to a contradiction . @xmath80 let @xmath227 and @xmath228
. then @xmath229 and @xmath230 .
* @xmath231 $ ] and @xmath232 $ ] .
* proof . *
note that @xmath233 $ ] and @xmath234 $ ] for any @xmath224 . if the claim is false , then , @xmath235 for any @xmath1 and hence @xmath236 a contradiction as in the proof of fact 2 .
@xmath80 let the least common ancestor of @xmath192s and @xmath194s be @xmath104 in @xmath182 .
we have shown that @xmath192s are below one child of @xmath104 , say @xmath237 , and @xmath194 s are below the other child of @xmath104 , say @xmath238 .
in addition , @xmath239 and @xmath240 are two line trees .
* for each edge @xmath146 ( @xmath147 ) such that @xmath121 and @xmath193 are in the same subtree as @xmath192s or as @xmath194s , then @xmath241 * proof . * without loss of generality
, we may assume that @xmath121 and @xmath193 are below @xmath237 in the same subtree as @xmath192s .
we consider the following cases .
@xmath242 $ ] for some @xmath243 . in this case
, we have that @xmath244 and @xmath245 hence , @xmath246\\ & \geq & n(n+1)+1\end{aligned}\ ] ] as the minimum value of @xmath247 is @xmath248 ( reaching at @xmath249 ) .
@xmath250 , x_{k+1 } , \cdots , x_{n-1 } , x_{n}]$ ] for some @xmath251 .
we have that @xmath252 and so @xmath253 as the minimum value of @xmath254 is 2n-2 ( reaching at @xmath255 ) . the fact is proved .
* for each edge @xmath146 such that @xmath121 is below @xmath237 in the same subtree as @xmath192 and @xmath193 is below @xmath238 in the subtree as @xmath194s .
then , @xmath256 * proof .
* let @xmath257\ ] ] and @xmath258.\ ] ] we have that all the internal nodes in @xmath129 are mapped onto the least common ancestor @xmath104 of @xmath192s and @xmath259s and thus @xmath260 since @xmath261 , the fact is proved .
let @xmath113 denote the subset of leaves @xmath121 below @xmath237 in the same subtree as @xmath192s and @xmath114 the subset of leaves @xmath193 below @xmath238 in the same subtree as @xmath194s .
then @xmath262 is a cut of the graph @xmath137 .
assume there are @xmath263 cut edges .
since there are @xmath264 non - cut edges , @xmath265 which implies that @xmath266 .
this finishes the proof of lemma [ lemma2 ] .
we conclude this paper by posing two related research problems . in this paper
, we have proved that species tree inference by minimizing deep coalescences is np - hard .
this justifies the effort from different groups in seeking efficient heuristic methods for the inference problem @xcite .
we have also discussed the relationship of the deep coalescence cost and the gene duplication cost .
is there any polynomial - time algorithm with constant approximation ratio for the species tree problem in the deep coalescence model ?
note that the heuristic method developed by than and nakhleh in @xcite seems to be effective . in @xcite , stege studied the parametric complexity of the species tree inference by minimizing gene duplications .
is is possible to develop efficient algorithm for parametric species tree inference under the deep coalescence model ?
this work was partially supported by arf r146 - 000 - 109 - 112 .
a preliminary version of this work was presented in the poster session of the recomb2000 .
goodman m , czelusniak j , moore gw , romero - herrera ae , matsuda g ( 1979 ) fitting the gene lineage into its species lineage , a parsimony strategy illustrated by cladograms constructed from globin sequences , _ syst .
_ 28:132 - 163 .
hey j , nielsen r ( 2004 ) multilocus methods for estimating population sizes , migration rates and divergence time , with applications to the divergence of _ drosophila pseudoobscura _ and
_ d. persimilis_. _ genetics _ 167:747 - 760 .
page r ( 1994 ) maps between trees and cladistic analysis of historical associations among genes , organisms , and areas .
_ 43:58 - 77 .
page r , m. charleston ma ( 1997 ) from gene to organismal phylogeny : reconciled trees and the gene tree / species tree problem .
phy . and evol .
_ 7:231 - 240 .
pamilo p , nei m ( 1988 ) relationship between gene trees and species trees , _
_ 5:568 - 583 .
roth c , rastogi a , arvestad l , dittmar k , light s , ekman d , david a. liberles da ( 2007 ) evolution after gene duplication : models , mechanisms , sequences , systems , and organisms _ j exp .
308b : 58 - 73 .
takahata n ( 1989 ) gene genealogy in three related population : consistency probability between gene and population trees . _ genetics _ 122:957 - 966 . than c , nakhlen l ( 2009 ) species tree inference by minimizing deep coalescences .
_ plos comput .
_ 5:e1000501.doi:10.1371/journal.pcbi.1000501 . | when gene copies are sampled from various species , the resulting gene tree might disagree with the containing species tree .
the primary causes of gene tree and species tree discord include lineage sorting , horizontal gene transfer , and gene duplication and loss .
each of these events yields a different parsimony criterion for inferring the ( containing ) species tree from gene trees . with lineage sorting ,
species tree inference is to find the tree minimizing extra gene lineages that had to coexist along species lineages ; with gene duplication , it becomes to find the tree minimizing gene duplications and/or losses . in this paper , we show the following results : \(i ) the deep coalescence cost is equal to the number of gene losses minus two times the gene duplication cost in the reconciliation of a uniquely leaf labeled gene tree and a species tree .
the deep coalescence cost can be computed in linear time for any arbitrary gene tree and species tree .
\(ii ) the deep coalescence cost is always no less than the gene duplication cost in the reconciliation of an arbitrary gene tree and a species tree .
\(iii ) species tree inference by minimizing deep coalescences is np - hard .
* index terms : * reconciliation of gene tree and species , deep coalescence , gene duplication and loss , parsimony criterion , np - hardness . |
the booms and the market crashes in financial markets have been an object of study in economics and a history of economy for a long time .
economists @xcite , and economic historians @xcite , @xcite , @xcite have often suggested the importance of psychological factors and irrational factors in explaining historical financial euphoria . as keynes @xcite , a famous economist and outstandingly successful investor , acutely pointed out in his book , _ the general theory of employment , interest and money _ , stock price changes have their origin in the collective crowd behaviour of many interacting agents rather than fundamental values which can be derived from careful analysis of present conditions and future prospects of firms . in a recent paper published in the economic journal , lux
@xcite modelled the idea explicitly and proposed a new theoretical framework to explain bubbles and subsequent crashes which links market crashes to the phase transitions studied in statistical physics . in his model
the emergence of bubbles and crashes is formalised as a self - organising process of infections among heterogeneous traders . in recent independent works , several groups of physicists @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite ,
@xcite proposed and demonstrated empirically that large stock market crashes , such as the 1929 and the 1987 crashes , are analogous to critical points .
they have claimed that financial crashes can be predicted using the idea of log - periodic oscillations or by other methods inspired by the physics of critical phenomena . in this paper
we present an interacting - agent model of speculative activity explaining bubbles and crashes in stock markets .
we describe stock markets through an infinite - range ising model to formulate the tendency of traders getting influenced by the investment attitude of other traders .
bubbles and crashes are understood and described qualitatively and quantitatively in terms of the classical phase transitions . although the interacting - agent hypothesis @xcite is advocated as an alternative approach to the efficient market hypothesis ( or rational expectation hypothesis )
@xcite , little attention has been given to the point how probabilistic rules , that agents switch their investment attitude , are connected with their decision - making or their expectation formations .
our interacting - agent model follows the line of lux @xcite , but differs from his work in the respect that we model speculative activity here from a viewpoint of traders decision - making .
the decision - making of interacting - agents will be formalised by _ minimum energy principle _ , and the stationary probability distribution on traders investment attitudes will be derived .
next , the stationary states of the system and speculative dynamics are analysed by using the mean field approximation .
it is suggested that the mean field approximation can be considered as a mathematical formularisation of _
keynes beauty contest_. there are three basic stationary states in the system : a bull market equilibrium , a bear market equilibrium , and a fundamental equilibrium .
we show that the variation of parameters like the bandwagon effect or the investment environment , which corresponds to the external field , can change the size of cluster of traders investment attitude or make the system jump to another market phase . when the bandwagon effect reaches some critical value , a second - order phase transition and critical behaviour can be observed .
there is a symmetry breaking at the fundamental equilibrium , and two stable equilibria , the bull market equilibrium and bear market equilibrium appear .
when the system stays in the bull market equilibrium , speculative bubble occurs in the stock market . for a certain range of the investment environment multistability and
hysteresis phenomena are observed .
when the investment environment reaches some critical values , the rapid changes ( the first - order phase transitions ) in the distribution of investment attitude are caused .
the phase transition from a bull market phase to a bear market phase is considered as a stock market crash .
then , we estimate the parameters of the interacting - agent model using an actual financial data . as an example of large crashes we will analyse the japan crisis ( bubble and crash in japanese stock market in 1987 - 1992 ) .
the estimated equation attempts to explain japan crisis over 6 year period 1987 - 1992 , and was constructed using monthly adjusted data for the first difference of topix and the investment environment which is defined below .
results of estimation suggest that the traders in the japanese stock market stayed the bull market equilibria , so that the speculative bubbles were caused by the strong bandwagon effect and betterment of the investment environment in 3 year period 1987 - 1989 , but a turn for the worse of the investment environment in 1990 gave cause to the first - order phase transition from a bull market phase to a bear market phase .
we will demonstrate that the market - phase transition occurred in march 1990 . in section 2
we construct the model . in section 3
we investigate the relationship between crashes and the phase transitions .
we implement an empirical study of japan crisis in section 4 .
we give some concluding remarks in section 5 .
we think of the stock market that large numbers of traders participate in trading .
the stock market consists of n traders ( members of a trader group ) .
traders are indexed by @xmath0 .
we assume that each of them can share one of two investment attitudes , buyer or seller , and buy or sell a fixed amount of stock ( q ) in a period .
@xmath1 denotes the investment attitude of trader @xmath2 at a period .
the investment attitude @xmath1 is defined as follows : if trader @xmath2 is the buyer of the stock at a period , then @xmath3 . if trader @xmath2 , in contrast , is the seller of the stock at a period , then @xmath4 . in the stock market
the price changes are subject to the law of demand and supply , that the price rises when there is excess demand , and the price falls when there is excess supply .
it seems natural to assume that the price raises if the number of the buyer exceeds the number of the seller because there may be excess demand , and the price falls if the number of seller exceeds the number of the seller because there may be excess supply .
thus a trader , who expects a certain exchange profit through trading , will predict every other traders behaviour , and will choose the same behaviour as the other traders behaviour as thoroughly as possible he could .
the decision - making of traders will be also influenced by changes of the firm s fundamental value , which can be derived from analysis of present conditions and future prospects of the firm , and the return on the alternative asset ( e.g. bonds ) .
for simplicity of an empirical analysis we will use the ratio of ordinary profits to total capital that is a typical measure of investment , as a proxy for changes of the fundamental value , and the long - term interest rate as a proxy for changes of the return on the alternative asset .
furthermore we define the investment environment as _ investment environment = ratio of ordinary profits to total capital - long - term interest rate_. when the investment environment increases ( decreases ) a trader may think that now is the time for him to buy ( sell ) the stock .
formally let us assume that the investment attitude of trader @xmath2 is determined by minimisation of the following _ disagreement _ function @xmath5 , @xmath6 where @xmath7 denotes the strength of trader @xmath8 s influence on trader @xmath2 , and @xmath9 denotes the strength of the reaction of trader @xmath2 upon the change of the investment environment @xmath10 which may be interpreted as an external field , and @xmath11 denotes the vector of investment attitude @xmath12 .
the optimisation problem that should be solved for every trader to achieve minimisation of their disagreement functions @xmath5 at the same time is formalised by @xmath13 now let us assume that trader s decision making is subject to a probabilistic rule .
the summation over all possible configurations of agents investment attitude @xmath14 is computationally explosive with size of the number of trader @xmath15.therefore under the circumstance that a large number of traders participates into trading , a probabilistic setting may be one of best means to analyse the collective behaviour of the many interacting traders .
let us introduce a random variable @xmath16 , @xmath17 .
the state of the agents investment attitude @xmath18 occur with probability @xmath19 with the requirement @xmath20 and @xmath21 .
we define the amount of uncertainty before the occurrence of the state @xmath18 with probability @xmath22 as the logarithmic function : @xmath23 . under these assumptions
the above optimisation problem is formalised by @xmath24 subject to @xmath25 , where @xmath26 .
@xmath18 is a state , and @xmath27 is information entropy .
@xmath22 is the relative frequency the occurrence of the state @xmath18 .
the well - known solutions of the above optimisation problem is @xmath28 where the parameter @xmath29 may be interested as a _ market temperature _ describing a degree of randomness in the behaviour of traders .
the probability distribution @xmath22 is called the _ boltzmann distribution _ where @xmath22 is the probability that the traders investment attitude is in the state @xmath30 with the function @xmath31 , and @xmath32 is the partition function .
we call the optimising behaviour of the traders with interaction among the other traders a _ relative expectation formation_. the trading volume should depends upon the investment attitudes of all traders . since traders are supposed to either buy or sell a fixed amount of stock ( q ) in a period , the aggregate excess demand for stock at a period is given by @xmath33 .
we assume the existence of a market - maker whose function is to adjust the price .
if the excess demand @xmath34 is positive ( negative ) , the market maker raises ( reduces ) the stock price for the next period .
precisely , the new price is calculated as the previous price @xmath35 plus some fraction of the excess demand of the previous period according : @xmath36 , where @xmath37 denotes the investment attitude of trader @xmath2 at period @xmath38 , @xmath39 the price change from the current period to the next period , i.e. @xmath40 , and the parameter @xmath41 represents the speed of adjustment of the market price . at the equilibrium prices that clear
the market there should exist an equal number of buyers or sellers , i.e. : @xmath42 . using the boltzmann distribution ( [ eqn : a4 ] ) the mean value of the price changes @xmath43
is given by @xmath44 .
we can derive analytically the stationary states of the traders investment attitude using a mean field approximation which is a well known technique in statistical physics .
let us replace the discrete summation of the investment attitude with the mean field variable , @xmath45 .
we additionally assume that the parameter @xmath7 is equal to @xmath46 @xmath47 and @xmath48 for every trader . in this case
our interacting - agent model reduces to the ising model with long - range interactions .
then the function ( [ eqn : a2 ] ) is approximated by @xmath49 .
since the mean value of the stationary distribution is given by @xmath50 , the mean field @xmath51 is @xmath52 the first term of the right side of eq .
( [ eqn : a5 ] ) represents that traders tend to adopt the same investment attitude as prediction of the average investment attitude and the second term represents the influence of the change of investment environment to traders investment attitude .
the first term may be interpreted as a mathematical formularisation of _
keynes beauty contest_. keynes @xcite argued that stock prices are not only determined by the firm s fundamental value , but in addition mass psychology and investors expectations influence financial markets significantly .
it was his opinion that professional investors prefer to devote their energies not to estimating fundamental values , but rather to analysing how the crowd of investors is likely to behave in the future .
as a result , he said , most persons are _ largely concerned , not with making superior long - term forecasts of the probable yield of an investment over its whole life , but with foreseeing changes in the conventional basis of valuation a short time ahead of the general public_. keynes used his famous _ beauty contest _ as a parable to stock markets . in order to predict the winner of a beauty contest ,
objective beauty is not much important , but knowledge or prediction of others predictions of beauty is much more relevant .
in keynes view , the optimal strategy is not to pick those faces the player thinks prettiest , but those the other players are likely to be about what the average opinion will be , or to proceed even further along this sequence .
we assume that the traders investment attitude changes simultaneously ( synchronous dynamics ) in discrete time steps .
we represent the dynamics of the investment attitude using a straightforward iteration of equation ( [ eqn : a5 ] ) , such that : @xmath53 we call @xmath54 the bandwagon coefficient and @xmath55 the investment environment coefficient below because @xmath54 may be interpreted as a parameter that denotes the strength that traders chase the price trend .
next , let us approximate the adjustment process of the price using the mean field @xmath56 .
we can get the following speculative price dynamics , @xmath57 inserting the mean field of the price change , @xmath58 into eq .
[ eqn : a6 ] , the dynamics of the mean field of @xmath56 can be rewritten as @xmath59 .
this equation shows that the traders result in basing their trading decisions on an analysis of the price trend @xmath60 as well as the change of the investment environment @xmath61 .
the stationary states of the mean field @xmath51 satisfy the following @xmath62 eq .
[ eqn : a8 ] has a unique solution @xmath51 for arbitrary @xmath63 , when @xmath54 is less than @xmath64 .
the stationary state @xmath65 with @xmath66 is called as the _ fundamental equilibrium _ which is stable and corresponds to the maximum of the stationary distribution of the investment attitude @xmath67 . in this stable equilibrium
there is equal numbers of traders sharing both investment attitudes on average . moving in the parameter space @xmath68 and starting from different configurations one
have several possible scenarios of market - phase transitions .
to begin with , let us consider the case that the investment environment @xmath10 is equal to zero . in theory , when the bandwagon coefficient @xmath54 is increased starting from the phase at @xmath69 , there is a symmetry breaking at critical point , @xmath70 , and two different market phases appear , that is , the _ bear market _ and the _ bull market_. in analogy to physical systems the transitions may be called the second - order phase transition . fig .
1 illustrates the two graphical solutions of the eq .
( [ eqn : a8 ] ) for @xmath71 and @xmath72 and for @xmath69 and @xmath72 .
the figure shows that for @xmath69 and @xmath72 the fundamental equilibrium is unique and stable , but for @xmath71 and @xmath72 the fundamental equilibrium @xmath73 becomes unstable , and the two new equilibria , the bull market equilibrium @xmath46 and the bear market equilibrium @xmath74 are stable . at the bull market equilibrium
more than half number of traders is buyer , so that the speculative bubbles occur in the stock market
. then , let us consider the effect of changes of the investment environment @xmath10 in the case with a weak bandwagon coefficient , @xmath69 and a positive investment environment coefficient , @xmath75 .
2 shows that as the investment environment @xmath10 changes for the better ( @xmath76 ) , the system shifts to the bull market phase .
by contrast , as the investment environment @xmath10 changes for the worse ( @xmath77 ) , the system shifts to the bear market phase .
therefore when the bandwagon effect is weak , the stock price go up or down slowly according rises or falls in the investment environment .
finally , let us consider the effect of changes of @xmath63 in the case with a strong bandwagon coefficient @xmath71 . in this case multistability and hysteresis phenomena in the distribution of investment attitude , as well as market - phase transitions
are observed .
the system has three equilibria , when @xmath71 and @xmath78 , where @xmath79 is determined by the equation , @xmath80 = \alpha .
\label{eqn : a9}\ ] ] two maximum of the stationary distribution on the investment attitude are found at @xmath81 and @xmath82 and one minimum at @xmath83 . for @xmath84 and @xmath71 ,
two of the three equilibria coincide at @xmath85 for @xmath86 and @xmath71 the two of three equilibria vanishes and only one equilibrium remains . in analogy to physical systems the transitions may be called first - order phase transitions . fig .
3 illustrates the effect of changes in the investment environment in the case with a strong bandwagon effect @xmath71 . in this case
speculative bubbles and market crashes occur .
when the investment environment changes for the better ( worse ) , the curve that denotes @xmath87 , shifts upward ( downward ) .
as the investment environment @xmath10 keeps on rising , and when it reach a critical value , that is , @xmath84 , the bear market equilibrium and the fundamental equilibrium vanish , and the bull market equilibrium becomes an unique equilibrium of the system , so that the speculative bubble occurs in the stock market . even though the investment environment changes for the worse
, the hysteresis phenomena are observed in a range of the investment environment that can be calculated by solving eq .
( [ eqn : a9 ] ) .
in other words the speculative bubble continues in a range of the investment environment .
however when the negative impact of the investment environment reaches a critical value , that is , @xmath84 , the bull market vanish
. further decrease of @xmath10 cause the market - phase transition from a bull market to a bear market . in a bear market
more than half number of trader is seller on average , so that the stock price continues to fall on average .
thus this market phase transition may be considered as the stock market crash .
perhaps the most spectacular boom and bust of the late twentieth century involved the japan s stock markets in 1987 - 1992 .
stock prices increased from 1982 to 1989 . at their peak in december 1989 , japanese stocks had a total market value of about 4 trillion , almost 1.5 times the value of all u.s .
equities and close to 45 percent of the world s equity market capitalisation .
the fall was almost as extreme as the u.s .
stock - market crash from the end of 1929 to mid-1932 .
the japanese ( nikkei ) stock - market index reached a high of almost 40,000 yen in the end of 1989 . by mid - august 1992 , the index had declined to 14,309 yen , a drop of about 63 percent . in this section
we will estimate the parameter vector @xmath68 for the japan s stock market in 1987 - 1992 .
we use monthly adjusted data for the first difference of tokyo stock price index ( topix ) and the investment environment over the period 1988 to 1992 . in order to get the mean value of the stock price changes
, topix is adjusted through the use of a simple centred 12 point moving average , and are normalised into the range of @xmath88 to @xmath89 by dividing by the maximum value of the absolute value of the price changes .
the normalised price change is defined as @xmath90 .
4 and fig . 5
show the normalised stock price changes and the investment environment respectively .
this normalised stock price change @xmath90 is used as a proxy of the mean field @xmath56 . inserting @xmath90 for eq .
( [ eqn : a8 ] ) , we get @xmath91 . the bandwagon coefficient @xmath54 and the investment environment coefficient @xmath55 should be estimated using any one of estimation technique .
since we can not get the analytical solution because the function @xmath92 is non - linear , we use the _ gradient - descent algorithm _ for the parameter estimation .
the gradient - descent algorithm is a stable and robust procedure for minimising the following one - step - prediction error function @xmath93 ^ 2 .
\label{eqn : a33}\ ] ] more specifically , the gradient - descent algorithm changes the parameter vector @xmath94 to satisfy the following condition : @xmath95 where @xmath96 , @xmath97 , and @xmath98 . to accomplish this
the gradient - descent algorithm adjusts each parameter @xmath99 and @xmath100 by amounts @xmath101 and @xmath102 proportional to the negative of the gradient of @xmath103 at the current location : @xmath104 where @xmath105 is a learning rate .
the gradient - descent rule necessarily decreases the error with a small value of @xmath105 . if the error function ( [ eqn : a33 ] )
, thus , has a single minimum at @xmath106 , then the parameters @xmath94 approaches the optimal values with enough iterations .
the error function @xmath107 , however , is non - linear , and hence it is possible that the error function ( [ eqn : a33 ] ) may have _ local minima _ besides the global minimum at @xmath106 . in this case
the gradient - descent rule may become stuck at such a local minimum . to check convergence of the gradient - descent method to the global minimum we estimate the parameters for a variety of alternative start up parameters @xmath108 with @xmath109 . as a consequence the gradient - descent rule estimated the same values of the parameters , @xmath110 with enough iterations .
the fit of the estimated equation can be seen graphically in fig .
( 6 ) , which compare the actual and forecasted series over one period , respectively . the correlation coefficient that indicates goodness of fit is @xmath111 .
the capacity of the model to forecast adequately in comparison with competing models is an important element in the evaluation of its overall performance .
let us consider the linear regression model @xmath112 .
the regression results are @xmath113 .
the correlation coefficient of the linear regression model is equal to 0.93 , and is lower that that of the interacting - agent model .
root mean squared errors ( rmses ) of the interacting - agent model is equal to 0.05 , and that of the linear regression model is equal to 0.148 over the period 1987 - 1992 . the interacting - agent model has a smaller rmse than that of the linear regression model
. by the traditional rmse criterion
the interacting - agent model is , thus , superior to the linear regression model in term of their forecasting performance . in conclusion
we can say that the model not only explains the japan crisis but also correctly forecasts changing the stock price over the period 1987 - 1992 .
given that the set of parameters ( @xmath114 ) is equal to ( 1.04 , 0.5 ) , the critical point of the investment environment at which the phase transition are caused from the bull market to bear market are calculated from the theoretical results in the preceding section .
one can say that the phase transition from the bull market to the bear market occurs when the investment environment is below @xmath115 under @xmath116 .
by contrast the phase transition from the bear market to the bull market occurs when the investment environment is beyond @xmath117 under @xmath116 .
.the stock market crash in the japanese stock market [ cols="^,^,^,^",options="header " , ] from theoretical view point one may say that bursting speculative bubbles will begin when the investment environment is below @xmath115 , and the process of the collapse of the stock market continues till the investment environment is beyond @xmath117 . in the real world
the investment environment became below @xmath118 at march 1990 for the first time and continued to be negative values since then over the period 1987 - 1992 .
the fall of the actual price in the japan s stock market began at march 1990 , and the stock prices continued to fall over the period 1990 - 1992 . on these grounds
we have come to the conclusion that theory and practice are in a perfect harmony .
( see table [ tab1 ] . )
this paper presents an interacting - agent model of speculative activity explaining the bubbles and crashes in stock markets in terms of a mean field approximation .
we show theoretically that these phenomena are easy to understand as the market- phase transitions .
bubbles and crashes are corresponding to the second - order transitions and the first - order transitions respectively .
furthermore we estimate the parameters of the model for the japanese stock market in 1987 - 1992 .
the empirical results demonstrate that theory and practice are in a perfect harmony .
this fact justifies our model that bubbles and crashes have their origin in the collective crowd behaviour of many interacting agents .
i would like to thank masao mori and thomas lux for helpful comments and suggestions to an earlier version of this paper .
99 j. m. keynes , _ the general theory of employment , interest and money _ .
harcourt , 1934 . c. mackay , _ extraordinary popular delusions and the madness of crowds _ the noonday press , 1932 . c. p. kindlberger , _ manias , panics , and
crashes : a history of financial crises _ , macmillan , london , 1989 .
j. k. galbraith , _ a short history of financial euphoria _ .
whittle direct books , tennessee , u.s.a . ,
t. lux,_the economic journal _ , * 105 * ( 1995 ) 881 - 896 .
t. kaizoji , _ simulation _ ( in japanese ) , 12 , 2 ( 1993 ) 67 - 74 .
d. sornette , a. johansen , j.p .
bouchaud , _ j. phys
. i. france _ * 6 * ( 1996 ) 167 - 175 .
d. sornette , a. johansen , _ physica a _ * 245 * ( 1997 ) 411 .
d. sornette and a. johansen , _ physica a _ * 261 * ( nos . 3 - 4 ) ( 1998 ) a. johansen , d. sornette , _ eur.phys.j .
b _ * 1 * ( 1998 ) 141 - 143 .
a. johansen , d. sornette , _ int .
c _ * 10 * ( 1999 ) 563 - 575 .
a. johansen , d. sornette , _ int .
c _ * 11 * ( 2000 ) 359 - 364 .
j. a. feigenbaum , p.g.o .
freund , _ int .
b _ * 10 * ( 1996 ) 3737 - 3745 .
j. a. feigenbaum , p.g.o .
freund , _ int .
b _ * 12 * ( 1998 ) 57 - 60 .
n. vandewalle , ph .
boveroux , a. minguet , m. ausloos , _ physica a _ * 225 * ( 1998 ) 201 - 210 .
n. vandewalle , m. ausloos , ph .
boveroux , a. minguet , _ eur .
j. b _ * 4 * ( 1998 ) 139 - 141 .
s. gluzman , and v. i. yukalov , _ mod .
b _ * 12 * ( 1998 ) 75 - 84 .
l. laloux , m. potters , r. cont , j .-
aguilar and j .-
bouchaud , _ europhysics letters _ , *
45 * , 1 ( 1999 ) 1 - 5 .
d. chowhury , d. stauffer _ eur .
* 8 * , ( 1999 ) 477 .
j. a. holyst , k. kacperski , f. schweitzer , submetted in _
physica a _ , ( 2000 ) .
t. lux , and m. marchesi , _ nature _ , * 297 * ( 1999 ) 498 - 500 .
e. fama , _ journal of finance _ * 46 * ( 1970 ) 1575 - 1618 .
d. e. rumelhart , g. e. hinton , r. j. williams , in : _ parallel distributed processing _ * 1 * , chap . 8 , d. e. rumelhart , j. l. mcclelland , et al .
, eds . cambridge , ma : mit press , 1986 .
figure 1 . the second - order phase transition . numerical solution of eq.([eqn : a8 ] ) for @xmath72 .
the curve is the rhs of [ eqn : a8 ] plotted for different values of @xmath54 : the aa line - @xmath119 , the bb line - @xmath120 .
+ figure 2 .
numerical solution of eq.([eqn : a8 ] ) for @xmath121 and @xmath122 .
@xmath87 is the rhs of ( 8) plotted for different values of @xmath54 : the aa line - @xmath123 , the bb line - @xmath124 .
+ figure 3 .
the first - order transition : speculative bubble and crash . numerical solution of eq .
( [ eqn : a8 ] ) for @xmath125 and @xmath122 .
the curves is @xmath87 plotted for different values of the investment environment : the aa line - @xmath126 , the bb line - @xmath127 .
+ figure 4 . the adjusted stock price change , @xmath128 : january 1987 - december 1992 .
+ figure 5 .
the japan s investment environment @xmath61 : january 1987 - december 1992 .
+ figure 6 .
one - step - forecast of the stock price changes : january 1987 - december 1992 . | in this paper we present an interacting - agent model of speculative activity explaining bubbles and crashes in stock markets .
we describe stock markets through an infinite - range ising model to formulate the tendency of traders getting influenced by the investment attitude of other traders .
bubbles and crashes are understood and described qualitatively and quantitatively in terms of the classical phase transitions .
when the interactions among traders become stronger and reach some critical values , a second - order phase transition and critical behaviour can be observed , and a bull market phase and a bear market phase appear .
when the system stays at the bull market phase , speculative bubbles occur in the stock market . for a certain range of the external field that we shall call the investment environment , multistability and hysteresis phenomena
are observed .
when the investment environment reaches some critical values , the rapid changes in the distribution of investment attitude are caused .
the first - order phase transition from a bull market phase to a bear market phase is considered as a stock market crash .
furthermore we estimate the parameters of the model using the actual financial data . as an example of large crashes we analyse japan crisis ( the bubble and the subsequent crash in the japanese stock market in 1987 - 1992 ) , and
show that the good quality of the fits , as well as the consistency of the values of the parameters are obtained from japan crisis .
the results of the empirical study demonstrate that japan crisis can be explained quite naturally by the model that bubbles and crashes have their origin in the collective crowd behaviour of many interacting agents .
+ * keyword * speculative bubbles ; stock market crash ; phase transition ; mean field approximation ; japan crisis |
unification models of seyfert galaxies , at least in their simplest formulation , postulate the presence of a geometrically and optically thick gas and dust torus which , at a distance of several to several tens of parsecs from the nucleus ( krolik , madau , and zycki 1994 ) , hides the primary source and the broad line forming region ( blr ) .
the orientation of the torus is independent of the host galaxy , and the observed differences between seyfert 1 and seyfert 2 nuclei are thus to be ascribed simply to the angle formed by the line of sight direction and the axis of the torus .
after the seminal work by antonucci & miller ( 1985 ) various pieces of evidence have been accumulating in favour of the unification model of seyfert galaxies ( see antonucci 1993 for a review of the subject ) : detection of broadened emission lines in optical spectropolarimetry observations of several seyfert 2 galaxies ( antonucci & miller 1985 ; miller & goodrich 1990 ; tran 1995 ; young et al .
1996 ) interpreted as scattering of the blr emission by warm material placed above the torus ; biconical structures in the light of the forbidden line at 5007 @xmath4 ( e.g. tadhunter and tsvetanov 1989 ) which is expected if the ionizing radiation field from the nucleus is anisotropic ; large ( @xmath5 @xmath6 or greater ) absorbing column densities observed in the hard x - ray spectra of seyfert 2 galaxies ( awaki 1997 ) . in the optical and uv bands
the nuclear emission is visible only if the line of sight does not intercept the torus . in the x - ray band ,
the interaction between electromagnetic radiation and matter is dominated by photoelectric absorption up to @xmath7 kev and by compton scattering at higher energies .
when the line of sight intercepts the torus ( seyfert 2 ) primary x - rays are able to leak through it ( compton thin source ) if the column density n@xmath8 is less than @xmath9 @xmath6 .
when n@xmath10 @xmath6 the primary x - rays can not escape and are either promptly absorbed via photoelectric interaction , or first compton downscattered and eventually photoabsorbed making the nucleus invisible ( compton thick source ) in direct emission .
emission from compton thick seyfert 2 galaxies can be detected by means of the reflection caused by the visible inner surface of the torus and/or via scattering by the material responsible for producing the broad lines observed in polarized light .
the detection of these sources in x
rays is hampered because of their faintness ( @xmath11 erg @xmath6 s@xmath12 ) and only a few of them are currently known ( matt 1997 and references therein ) .
their expected observable features in x - rays are : a 210 kev continuum flatter than the canonical slope observed for seyfert 1 , a strong ( equivalent width ( ew ) @xmath0 1 kev or higher when measured against the reflected continuum ) fe k@xmath13 feature around 6.4 kev , and a reflection `` hump '' at energies @xmath14 kev ( e.g. ghisellini , haardt , and matt 1994 ) .
ngc 7674 is a @xmath15 seyfert 2 nucleus hosted in a face on spiral galaxy ( sab ) with asymmetrical arms and a tidal connection to a nearby compact elliptical galaxy .
broad h@xmath13 and h@xmath16 components in polarized flux were first observed by miller & goodrich ( 1990 ) , and then by young et al .
the spectropolarimetric observations suggest that ngc 7674 is the only known seyfert 2 galaxy with hidden broad line region for which dust scattering is the dominant cause of the observed nuclear polarization ( tran 1995 ) .
preliminary analysis of the ginga observation of ngc 7674 ( awaki et al .
1991 ) gave a flat ( @xmath17 ) 210 kev spectrum with no indication for a fe line ( ew @xmath18 110 ev ) .
however , subsequent analysis indicated that the observed count rate was below the 3@xmath2 rms fluctuation of the cosmic x - ray background between 2 and 10 kev in the eight detectors of the large area counter onboard ginga ( smith and done 1996 ) . as a consequence , the 210 kev flux of @xmath19 erg @xmath6 s@xmath12 previously reported by awaki et al .
( 1991 ) has to be considered only as a 2@xmath2 upper limit ( smith , private communication ) . in the following
, detailed measurement of the x - ray spectral characteristics of ngc 7674 obtained by bepposax is reported .
the bepposax x - ray observatory ( boella et al . 1997a ) , a major programme of the italian space agency with participation of the netherlands agency for aereospace programs , was launched on april 30th 1996 from cape canaveral in florida .
the payload instruments are characterized by a wide spectral coverage and consist of four co - aligned narrow field instruments ( nfi ) plus two wide field cameras perpendicular to the axis of the nfi and looking in opposite directions .
the nfi include a low energy concentrator spectrometer ( lecs ; parmar et al .
1997 ) , three medium energy concentrator spectrometers ( mecs ; boella et al .
1997b ) , a high pressure gas scintillation proportional counter ( hpgspc ; manzo et al .
1997 ) , and a phoswich detector system ( pds ; frontera et al .
lecs and mecs are instruments with imaging capabilities ( angular resolution @xmath0 1@xmath20 ) and operate in the 0.110 kev and 1.510 kev spectral bands respectively . in the overlapping energy interval
the effective area of the lecs is about one third of the three mecs ( which is around 150 @xmath21 at 6 kev ) .
hpgspc and pds are instruments functioning via rocking collimators and cover the 4120 kev and 15 - 300 kev energy bands respectively . in the overlapping energy band pds is the most sensitive ( by a factor @xmath22 ) while hpgspc is optimized for spectral resolution .
bepposax nfi pointed at ngc 7674 from november 25th to november 26th 1996 for about @xmath23 s of effective observing time .
the on - source time for the lecs was @xmath040 ksec since this instrument is switched on only during satellite night time , while the on - source time for pds and hpgspc was @xmath2450% of the effective time due to rocking collimator functioning .
lecs and mecs good time intervals were selected imposing the minimum elevation angle above earth s limb to be @xmath25 and the instrument to be in its nominal configuration .
given the weak 210 kev source flux ( see next section ) lecs and mecs spectra were extracted from a 2@xmath20 radius region around the centroid of the source and the appropriate response matrices were generated .
the same region was used to extract background spectra from blank sky observation files .
the use of local instead of blank sky backgrounds did not make significant difference .
the pds consists of four phoswich units .
they operate in collimator mode , with two of them pointing to the source while the other two point @xmath26 away . the two pairs switch on and off source every 96 seconds .
the net source spectra have been obtained by subtracting the ` off ' to the ` on ' counts .
pds source data have been multiplied by a factor 1.42 to account for the mismatch in the absolute normalization with the mecs ( matt et al .
1997 , and references therein ) . leaving
this number free to vary would only slightly change the best fit parameters , without affecting the main results of the present analysis .
the net source count rate in the pds was @xmath27 cts / s between 13 and 60 kev , which gives a significant detection also after conservative subtraction of the systematic residuals which are currently evaluated @xmath28 cts / s in the 13200 kev band ( guainazzi and matteuzzi 1997 ) .
the net source count rate ( in units of @xmath29 cts / s ) was @xmath30 in the lecs ( 0.19 kev ) , @xmath31 in the three mecs ( 210 kev ) , while it was below detection limit in the hpgspc .
neither the soft ( @xmath00.12 kev ) nor the hard ( 210 kev ) energy band data show significant variability . statistical @xmath32 tests against constancy gave a probability greater than 25% , thus consistent with a constant source .
the spectra were thus extracted from the overall observing period and were rebinned to have a minimum of 20 counts in each channel .
the spectral analysis has been performed by means of the xspec 9.0 package , and using the instrument response matrices released by the bepposax science data centre in january 1997 .
figure [ res ] shows the ratio of the broad band spectrum between 0.1 and 100 kev to a simple power law ( @xmath33 ) model .
residuals indicate the presence of a significant soft excess below 2 kev , a strong line emission around 6.4 kev , and excess emission in the 1330 kev band .
these clear spectral features make ngc 7674 an interesting case for investigating and testing reflection models .
all the quoted errors correspond to 90% confidence intervals for one interesting parameter ( @xmath34 of 2.71 ) .
the mecs spectral data in the 210 kev energy band were first fitted with an absorbed power law model , plus a gaussian narrow ( @xmath35 fixed ) line to take into account the 6.4 kev feature , with all other parameters left free to vary .
the best fit ( model # 1 in table 1 ) parameters are a photon index @xmath36 , an absorption column density @xmath37 @xmath6 , and a gaussian line centered ( in the reference frame of the emitting source ) around 6.4 kev with an equivalent width of @xmath01 kev .
the fit improves significantly ( @xmath3898% via f - test ) if the line is allowed to broaden and gives a width of @xmath39 kev and an equivalent width of @xmath02 kev ( model # 2 ) . unless extreme iron abundances are assumed ( about an order of magnitude larger than the cosmic value ) ,
the observed iron emission line intensity can not be explained by transmission through the measured absorption column density which would predict an ew@xmath18100 ev ( makishima 1986 , ptak et al .
such a strong line , together with the unusual flatness of the continuum , is therefore a very strong evidence that the observed hard x - ray spectrum is reflection - dominated .
c ccc lll l + + + + model # & & n@xmath8 & e(fek ) & @xmath2(fek ) & ew(fek ) & @xmath40 + & & & ( @xmath41 ) & ( kev ) & ( kev ) & ( ev ) & + + 1 & & @xmath42 & @xmath43 & 0@xmath44 & @xmath45 & 32.4/28 + + + 2 & & @xmath46 & @xmath47 & @xmath48&@xmath49 & 26.1/27 + + + + + model # & & @xmath50 & e(fek ) & @xmath2(fek ) & ew(fek ) & @xmath40 + & & & ( kev ) & ( kev ) & ( ev ) & + + + 3a & & @xmath51 & @xmath52&@xmath53&900@xmath54 & 36.5/38 + + 3b & kev & @xmath55&@xmath56&@xmath57&757@xmath58 & 37.3/38 + + + 4 & & @xmath59 & @xmath60 & @xmath61 & @xmath62 & 39.5/39 + + + 5 & & @xmath63 & @xmath64 & 0@xmath44 & @xmath65 & 35.9/37 + & & & & @xmath66 & 0@xmath44 & @xmath67 & + + + 6 & & @xmath68 & 6.40@xmath44 & 0@xmath44 & 831@xmath69 & 36.4/39 + & & & & 7.06@xmath44 & 0@xmath44 & 103@xmath70 & + + + the low statistics at energies below @xmath242 kev do not allow to firmly discriminate among different models tested .
in fact , the excess detected in the lecs data at @xmath71 kev is well modeled both by a steep ( @xmath72 ) power law absorbed by the galactic column density in the direction of the source ( n@xmath73 @xmath6 ; dickey and lockman [ 1990 ] ) and by a raymond - smith model with solar abundance and [email protected] kev ( models # 3a and # 3b ) .
moreover , an acceptable fit is also obtained assuming an electron scattering model with @xmath74 ( model # 4 ) .
the possible implications of the scattering model are discussed in section 4 . the broad band ( 0.1100 kev ) spectrum has been fitted with a soft component ( see section 3.2 ) plus a pure reflection continuum resulting from a power law illumination of cold and thick material plus a gaussian line .
the use of this model is justified by the fact that the emission line centroid energy is consistent with k shell fluorescence from neutral iron and has an ew of the order of what is expected in the case of a pure reflection continuum ( e.g. matt , perola , and piro 1991 ; ghisellini , haardt , and matt 1994 ; krolik , madau , and zycki 1994 ) . in this reflection model ( plrefl in xspec )
the only free parameter is the relative normalization of the reflected component to the direct one . if this parameter is left free to vary , only a lower limit ( @xmath75 ) is obtained , thus confirming that the observed spectrum is dominated by reflection
. therefore a purely reflected spectrum ( i.e. no direct component ) was chosen .
the best fit gives in this case ( model # 3a ) an intrinsic photon index @xmath76 consistent with the average values of seyfert 1 galaxies ( nandra and pounds 1994 ) .
the @xmath77 kev observed flux is @xmath78 erg @xmath6 s@xmath12 corresponding to an observed luminosity of @xmath79 erg s@xmath12 . the observed @xmath80 kev emission feature can also be fitted with a more complex model .
an acceptable fit ( model # 5 ) is in fact obtained by adding to the fe neutral line a second emission feature .
the best - fit energy centroid of this second line ( 6.90 kev ) is consistent both with h - like fe ( 6.97 kev ) or with a blend of he- ( 6.70 kev ) plus h - like fe .
this indicates that , certainly , reflection from thick cold matter plays an important role .
nonetheless , contribution from reflection caused by ionized matter can not be ruled out on the basis of the present data .
finally , an acceptable fit ( model # 6 ) is also obtained assuming both k@xmath13 and k@xmath16 emission to be present in the form of narrow lines .
the best - fit equivalent widths are consistent with the k@xmath16/k@xmath13 ratio expected for neutral iron ( 0.135 [ weast 1987 ] ) .
the 2 - 10 kev flat spectrum together with the high equivalent width of the fe line strongly suggest that the observed spectrum of the seyfert 2 galaxy ngc 7674 is reflection dominated .
the observed luminosity , @xmath81 erg s@xmath12 is , therefore , expected to be much lower than the intrinsic one . assuming a scattering model ( model # 4 ) , the best fit luminosity of the soft component extrapolated in the 210 kev band is @xmath82 erg s@xmath12 . if the material that electron - scatters the nuclear radiation is confined in a cone - shaped region above the torus , then @xmath83 is linked to the nuclear intrinsic luminosity @xmath84 by the relation @xmath85 , where @xmath86 and @xmath87 are the mirror optical depth and subtended solid angle respectively . assuming ngc 1068 values of @xmath88 and 0.25 for @xmath86 and @xmath89 respectively ( iwasawa , fabian , and matt 1997 ) , then @xmath84 is estimated to be @xmath90 erg s@xmath12 .
it is interesting to note that this value is very close to the intrinsic luminosity of @xmath91 erg s@xmath12 evaluated from the observed flux corrected for the reddening in the narrow line region ( maiolino , private communication ) .
the fact that also the far infrared integrated @xmath92 @xmath93 m luminosity l@xmath94 erg s@xmath12 ( spinoglio and malkan 1989 ) is of the same order of magnitude suggests that both l@xmath95}$ ] and and l@xmath96 can be good indicators of the intrinsic x - ray luminosity .
if the reflection is caused by cold materials at the inner surface of the torus ( ghisellini , haardt , and matt 1994 ; krolik , madau , and zycki 1994 ) , then only a fraction @xmath97 of the reflecting surface is visible .
this fraction can be estimated from the intrinsic luminosity . using the notation of iwasawa , fabian , and matt ( 1997 )
, the observed luminosity can be written as @xmath98 where @xmath99 is the albedo for a isotropic illumination . using the above values for @xmath83 and @xmath84 and assuming @xmath100
, @xmath97 becomes equal to @xmath00.1 , which is consistent within a factor of 2 of what found for ngc 1068 .
pure reflection spectra have been observed in x - rays by the asca satellite in at least few other seyfert 2 galaxies such as ngc 6552 ( fukazawa et al .
1994 ; reynolds et al . 1994 ) , ngc 6240 ( kii et al . 1997 ) , the circinus galaxy ( matt et al . 1996 ) , and ngc 1068 which has also been confirmed by bepposax ( matt et al . 1997 , and references therein ) . in ngc 7674 , even if some contribution from a warm reflection component can not in principle be excluded , as demonstrated by the acceptable fit obtained with a blend of ionized lines ( model # 5 ) , the observed x - ray emission is certainly dominated by reflection from cold matter . in fact , on statistical basis ( better @xmath101 , see table 1 ) and self - consistency arguments , the most convincing explanation for the possible higher energy line is k@xmath102 emission from neutral iron ( model # 6 ) .
this is contrary to what observed in ngc 1068 and ngc 6240 where the ionized reflection component is much stronger , but similar to what observed in the circinus galaxy and ngc 6552 .
such differences may have different physical interpretations .
they could be ascribed to an orientation effect ( more face - on than ngc 1068 and ngc 6240 ) with respect to the line of sight of the torus , or to the condition ( e.g. low scattering efficiency ) of the warm mirror itself . in conclusion ,
the case of ngc 7674 adds to the increasing number of compton - thick seyfert 2 galaxies , thus suggesting that a dedicated study of a complete sample of optically selected seyfert 2 galaxies may allow the discovery of a significant number of compton - thick objects as indicated by recent bepposax results ( salvati et al .
1997 ; maiolino et al .
1997 ) .
we thank dr .
g. matt for helpful discussions , and dr .
d. smith who provided us with the ginga data .
this research has made use of saxdas linearized and cleaned event files produced at the bepposax science data centre .
g.m . , g.g.c.p .
, a.c . , and
acknowledge financial support from the italian space agency .
+ awaki , h. et al .
1991 , pasj , 43 , 195 awaki , h. 1997 , in emission lines in active galaxies : new methods and techniques , asp conference series , vol .
113 , b.m .
peterson , f .- z .
cheng , and a.s .
wilson eds , p. 44
antonucci , r.r.j 1993 , ara&a , 31 , 473 boella , g. , butler , r.c . ,
perola , g.c . ,
piro , l. , scarsi , l. , and bleeker , j. 1997a , a&as , 122 , 299 boella g. , chiappetti , l. , conte , g. et al .
1997b , a&as , 122 , 327 dickey , j.m . , and lockman , f.j .
1990 , araa , 28 , 215 frontera , f. , costa , e. , dal fiume , d. et al .
1997 , a&as , 122 , 357 fukazawa , y. , et al .
1994 , pasj , 46 , l141 ghisellini , g. , haardt , f. , and matt , g. 1994 , mnras , 267 , 743 guainazzi , m. , and matteuzzi , l. 1997 , bepposax sdc technical report , tr-011 iwasawa , k. , fabian , a.c . , and matt , g. 1997 , mnras , 289 , 443 kii , t. , nakawaga , t. , fujimoto , r. , ogasaka , t. , miyazaki , t. , kawabe , r. , terashima , y. 1997 in `` proc . of x - ray imaging and spectroscopy of cosmic hot plasma '' , eds
. f. makino , and k. mitsuda , universal academy press , tokyo , p. 161 krolik , j.h .
, madau , p. , and zycki , p.t .
1994 , apj , 420 , l57 maiolino , r. et al .
1997 , in preparation makishima , k. 1986 , in physics of accretion onto compact objects , eds .
k.o . mason , m.g .
watson , n.e .
white ( springer - verlag , berlin ) , 249 manzo , g. , giarrusso , s. , santangelo a. et al .
1997 , a&as , 122 , 341 matt , g. , perola , g. c. , and piro , l. 1991 , a&a , 245 , 75 matt , g. et al .
1996 , mnras , 281 , l69 matt , g. 1997 , proc . of the 2nd agn national meeting
`` from micro- to mega - parsec '' , eds .
a. comastri , t. venturi , m. bellazzini , mem soc .
astr . it .
68 , p. 127
, g. et al . 1997 , a&a , 325 , l13 miller , j.s . , and goodrich , r.w .
1990 , apj , 355 , 456 nandra , k. , pounds , k.a . 1994 , mnras , 268 , 405 parmar , a.n . , martin , d.d.e . ,
bavdaz , m. et al .
1997 , a&as , 122 , 309 ptak , a. , yaqoob , t. , serlemitsos , p.j .
, kunieda , h. , and terashima , y. 1996 , apj , 459 , 542 reynolds , c.s . ,
fabian , a.c . ,
makishima , k. , fukazawa , y. , and tamura , t. 1994 , mnras , 268 , l55 salvati , m. , bassani , l. , della ceca , r. , maiolino , r. , matt , g. , and zamorani , g. 1997 , a&a , in press smith , d. a. , and done , c. 1996 , mnras , 280 , 355 spinoglio , l. , and malkan , m.a .
1989 , apj , 342 , 83 ueno , s. , mushotzky , r.f . , koyama , k. et al .
1994 , pasj , 46 , l71 tadhunter , c. , and tsvetanov , z. 1989 , nature , 341 , 422 tran , h.d .
1995 , apj , 440 , 597 weast , r.c .
1987 handbook of chemistry and physics , ed .
weast , 67th edition , crc press 1987 , p.
e-187 young , s. , hough , j.h . ,
efstathiou , a. , wills , b.j . ,
bailey , j.a . ,
ward , m.j . , and axon , d.j .
1996 , mnras , 281 , 1206 | the seyfert 2 galaxy ngc 7674 has been observed within the bepposax core programme with the onboard narrow field instruments between 0.1 and 100 kev .
the broad - band spectrum shows four most relevant spectral components : a ) a soft excess below @xmath02 kev ; b ) a prominent ( ew@xmath01 kev ) fe line ; c ) a flat ( @xmath1 ) 210 kev continuum ; d ) a 4.5@xmath2 detection above 13 kev
. the flat power law spectrum can be very well explained within the current agn unified models assuming a steep ( @xmath3 ) intrinsic spectrum scattered by warm gas for the low energy band and totally reflected by optically thick cold matter ( plausibly a molecular torus ) for the high energy band .
the case of ngc 7674 adds to the increasing number of so called `` compton - thick '' seyfert 2 galaxies in which the direct emission is totally absorbed and the x - ray luminosity is thus at least one or two orders of magnitude larger than what inferred from the observed flux . |
over the last several decades algorithmic advances coupled with the development of high speed supercomputers , have made _ ab initio _ quantum mechanical simulations possible for a wide range of physical systems .
these methods have been used to provide a theoretical framework for interpreting experimental results and even to accurately predict the material properties before experimental data were available .
however , the calculations are currently restricted to systems containing a few hundred atoms.@xcite this limitation is set by the available computer power , and the scaling of the computational work with the number of atoms .
one of the most successful of the recent techniques is the car - parrinello method@xcite in which the electronic orbitals are expanded in plane - wave basis functions and the resulting hamiltonian is iteratively diagonalized .
the practical and efficient extension of _ ab initio _ quantum methods to larger and more difficult systems may be accomplished by the refinement and improvement of traditional methods or by the development of new techniques . although highly successful , traditional plane - wave methods encounter considerable difficulties when they are applied to physical systems with large length scales , or containing first - row or transition - metal atoms .
these difficulties may be partially eliminated by the use of preconditioned conjugate - gradient techniques,@xcite optimized pseudopotentials,@xcite augmented - wave methods,@xcite or plane waves in adaptive coordinates.@xcite however , these methods are still constrained by the use of a plane - wave basis set , and the necessity of performing fast fourier transforms ( fft ) between the real and reciprocal spaces . while fft s may be implemented in a highly efficient manner on traditional vector supercomputers , the current trend in supercomputer design is massively parallel architectures .
it is difficult to implement efficient fft algorithms on such machines , due to the required long range communications .
real - space methods are inherently local and therefore do not lead to a large communication overhead .
the scaling of several critical parts of large calculations is improved from @xmath1 in a plane - wave representation to @xmath2 , where @xmath3 is the number of atoms .
furthermore , preconditioning and convergence acceleration are most effectively carried out in real space ( q.v . section iii ) .
a real - space formulation is also required for efficient implementations of @xmath4 electronic structure methods , in which the computational work required scales linearly with the number of atoms .
these methods impose a localization constraint on the electronic orbitals@xcite or the electron charge density,@xcite which eliminates the @xmath5 orthogonalization step .
orbital - based real - space approaches , _
e.g. _ atom - centered or floating gaussians , are very well established . recently ,
however , there has been substantial interest in developing real - space orbital - independent methods , which permit systematic studies of convergence in the spirit of plane - wave methods .
these methods include finite elements,@xcite grids , and wavelets.@xcite the finite - element method was applied by white _
@xcite to one - electron systems .
they used both conjugate - gradient and multigrid acceleration@xcite to find the ground - state wavefunction .
two of the present authors @xcite used a basis with a high density of grid points in the regions where the ions are located , and a lower density of points in the vacuum regions , in conjunction with multigrid acceleration , to calculate the electronic properties of atomic and diatomic systems .
the core electrons were explicitly included and nearly singular pseudopotentials were used .
the non - uniform grid led to order of magnitude savings in the basis size and total computational effort .
the multigrid iterations , which provide automatic preconditioning on all length scales , reduced the number of iterations needed to converge the electronic wavefunctions by an order of magnitude in these multi - length - scale systems .
et al._@xcite used a similar method to solve for the ground state of h@xmath6 .
wavelet bases@xcite were used to solve the lda equations for atoms and the o@xmath7 molecule .
et al._@xcite have used high - order finite - difference methods and soft nonlocal pseudopotentials on uniform grids to calculate the electronic structure , geometry , and short - time dynamics of small si clusters and of an isolated so@xmath7 molecule .
et al._@xcite have used uniform grids and a smeared nuclear potential to examine the energetics and structures of atoms and small molecules .
they employed multigrid iteration techniques to improve the convergence rates of the kohn - sham functional .
real - space grids in curvilinear coordinates have been used by gygi and galli@xcite to compute the properties of atoms and co@xmath7 , and zumbach _
et al._@xcite tested it on o@xmath7 .
et al._@xcite used a uniform grid approach with pseudopotentials and a conjugate gradients scheme to calculate the electronic structure of p@xmath7 , and to study a positron trapped by a cd vacancy in cdte . in a previous communication@xcite
the present authors have outlined a multigrid - based approach suitable for large scale calculations , together with a number of test applications .
these included calculations for a vacancy in a 64-atom diamond supercell , an _ isolated _
@xmath8 molecule using non - periodic boundary conditions , a highly elongated diamond supercell , and a 32-atom supercell of gan that included the ga @xmath9 electrons in valence .
uniformly spaced grids were used ; in this case an effective `` cut - off '' may be defined , which is equal to that of the plane - wave calculation that uses the same real - space grid for the fft s .
whenever feasible , the corresponding calculations were also carried out using plane - wave techniques , and the two sets of results were in excellent agreement with each other .
this paper provides a comprehensive description of the real - space multigrid method , and reports extensions to non - uniform grids , non - cubic grids , and to molecular dynamics simulations with highly accurate forces .
several computational issues absent from plane - wave and orbital - based methods arise when using a real - space grid approach . in the plane - wave basis
the action of the kinetic energy operator on the basis functions can be computed exactly , and the wavefunctions , potentials , and the electron charge density can be trivially expanded in the basis . every basis - set integral , except those involving the lda exchange - correlation functional , can be computed exactly .
the computational errors in the calculations are mainly due to the truncation of the basis . in a real - space grid implementation
, the kohn - sham equations must be discretized explicitly , which presents important trade - offs between accuracy and computational efficiency .
furthermore , the quantum - mechanical operators are known only at a discrete set of grid points , which can introduce a spurious systematic dependence of the kohn - sham eigenvalues , the total energy , and the ionic forces on the relative position of the atoms and the grid .
we have developed a set of techniques that overcome these difficulties and have been used to compute accurate static and dynamical properties of large physical systems , while taking advantage of the rapid convergence rates afforded by multigrid methods .
this paper is organized as follows : in section ii a method for the accurate and efficient real - space discretization of the kohn - sham equations for cubic , orthorhombic , and hexagonal symmetries is described .
section iii focuses on the multigrid algorithms , which greatly accelerate convergence of the electronic wavefunctions and of the hartree potential .
tests of the convergence acceleration are described in section iv .
the calculation of ionic forces that are sufficiently accurate for large step molecular dynamics requires special methods , which are described in section v. section vi discusses performance issues for massively parallel supercomputers , and describes a highly scalable and efficient implementation on the cray - t3d , which has been tested on up to 512 processors .
the summary in section vii is followed by several technical appendices .
electronic structure calculations that use a real - space mesh to represent the wavefunctions , charge density , and ionic pseudopotentials must address a new set of technical difficulties when compared with plane - wave methods . in a plane wave representation
the form of the kinetic energy operator is obvious .
in contrast , the representation of the kinetic energy operator on a real - space grid is approximated by some type of finite differencing , the accuracy of which must be carefully tested .
below , we describe real - space discretizations for uniform cubic , orthorhombic , and hexagonal grids , as well as nonuniform scaled cubic grids that increase the resolution locally . for uniform cubic grids
we also describe and test the extension to periodic systems with arbitrary sampling of the brillioun zone .
previously,@xcite we described a real - space approach that uses uniform cubic grids with @xmath10 point k - space sampling . in direct comparisons with plane - wave calculations , we found nearly perfect agreement between the two methods for several test systems .
we now provide further details of our method . in real - space the wavefunctions , the electron charge density , and the potentials
are directly represented on a uniform three - dimensional real - space grid of @xmath11 points with linear spacing @xmath12 .
the physical coordinates of each point are @xmath13 the ions are described by norm - conserving pseudopotentials@xcite in the kleinman - bylander nonlocal form.@xcite these potentials are interpolated onto the grid from their radial representation . exchange and correlation effects are treated using the local density approximation ( lda ) of density functional theory , in which the total electronic energy of a system of electrons and ions may be expressed as @xmath14 - \mu_{xc}[\rho({\bf r } ) ] \ } \nonumber\\ & & -{1 \over 2 } \int \ , d{\bf r } \ , \rho({\bf r } ) \ , v_{hartree}({\bf r } ) + e_{ion - ion}. \label{lda - energy}\end{aligned}\ ] ] the minimization of this functional requires the solution of the kohn - sham equations @xmath15
= -{1 \over 2}\nabla^2 \psi_n + v_{eff } \psi_n = \epsilon_n \psi_n , \label{ksequations}\ ] ] subject to the orthonormality constraint on the eigenfunctions @xmath16 .
the accurate discretization of these equations on the grid structure described by eq .
( [ grid1 ] ) requires appropriate numerical representations of the integral and differential operators .
all integrations are performed using the three - dimensional trapezoidal rule : @xmath17 we have found that for high accuracy it is essential that the integrand @xmath18 be band - limited in the sense that its fourier transform should have minimal magnitude in the frequency range @xmath19 .
this is explicit in a plane - wave calculation since the basis is cut - off at a specific plane - wave energy .
the discrete real - space grid also provides a kinetic - energy cut - off of approximately @xmath20 . unlike the plane - wave basis
, high - frequency components above this cut - off can nonetheless manifest themselves on the grid .
this high - frequency behavior , which can introduce unphysical variation in the total energy or the electron charge density , is perhaps best seen when the ions , and hence their pseudopotentials , shift relative to the grid points.@xcite if the pseudopotentials contain significant high - frequency components near or above @xmath21 then , as the ions shift , the high frequency components are aliased to lower frequency components in an unpredictable manner
. this effect can be decreased by _
explicitly _ eliminating the high - frequency components in the pseudopotentials by fourier filtering . in the context of plane - wave calculations , king - smith
_ et al._@xcite recognized that the real - space integration of the nonlocal pseudopotentials could differ significantly from the exact result computed in momentum space , unless the potentials were modified so that fourier components near @xmath21 were removed .
fourier filtering of the pseudopotentials is thus required in real - space calculations for accurate results .
it is , of course , possible to use unfiltered potentials on real - space grids provided the grid spacing is sufficiently small , but our experience shows that the total energy and the electron charge density are often sufficiently well converged for significantly larger grid spacings provided that explicit pseudopotential filtering is used .
we use a somewhat different fourier filtering method than that proposed by king - smith _ et al .
_ , but it produces essentially the same effect ( see appendix b ) .
the differential operator in the kohn - sham equations is approximated using a _ generalized _ eigenvalue form : @xmath22 = & & \nonumber\\ { 1 \over 2 } { \bf a}_{mehr}[\psi_n ] + { \bf b}_{mehr}[v_{eff } \psi_n ] = & & \epsilon_n { \bf b}_{mehr}[\psi_n ] , \label{compact - implicit}\end{aligned}\ ] ] where @xmath23 and @xmath24 are the components of the _ mehrstellen _ discretization,@xcite which is based on hermite s generalization of taylor s theorem .
it uses a weighted sum of the wavefunction and potential values to improve the accuracy of the discretization of the _ entire _ differential equation , not just the kinetic energy operator .
in contrast to the central finite - differencing method , this discretization uses more _ local _ information ( next - nearest neighbors , for example ) .
the definition of the fourth - order _ mehrstellen _ operator used in the present work is specified by the weights listed in table [ mehrstellen - ortho ] , which pertain to both cubic and orthorhombic grids ( see below ) . a more detailed analysis of the _ mehrstellen _ operator and eq . ( [ compact - implicit ] )
is given in appendix a. this representation of the kohn - sham hamiltonian is short - ranged in real space in the sense that the operator can be applied to any orbital in @xmath25 operations . specifically , the application of the @xmath23 operator at a point involves a sum over 19 orbital values while the application of the @xmath24 operator requires a sum over 7 points .
the local potential multiplies the orbital pointwise , and the short - ranged nonlocal projectors require one integration over a fixed volume around each ion and a pointwise multiplication .
this sparseness permits the use of iterative diagonalization techniques , and the short - ranged representation of the hamiltonian leads to an efficient implementation on massively parallel computers .
the discussion up to this point has been restricted to uniform cubic grids , but the extension to a general orthorhombic grid is straightforward .
there are now three separate grid spacings @xmath26 , @xmath27 , and @xmath28 with the coordinates of each grid point given by : @xmath29 the orthorhombic _ mehrstellen _ operator described in table [ mehrstellen - ortho ] is used to discretize the kohn - sham equations and numerical integration is performed according to @xmath30 .
\label{integration1}\ ] ] in the preceding section the wavefunctions were assumed to be real , with the brillioun - zone sampling restricted to the @xmath10 point .
when these restrictions are lifted the kohn - sham equations become @xmath31 where the @xmath32 are now the periodic parts of the bloch functions .
they are complex - valued , which presents no additional difficulties in discretization .
the nonlocal projectors in @xmath33 have been multiplied by the phase factor @xmath34 .
the gradient term @xmath35 is computed using a central finite difference expression , which in one dimension has the form : @xmath36 where @xmath37 , @xmath38 , @xmath39 , and @xmath40 .
for a cubic grid structure the three dimensional generalization of this is the sum of the individual expressions for each coordinate axis . denoting this finite - difference operator by @xmath41 ,
the discretization of the kohn - sham equation becomes @xmath42 & & = { 1 \over 2 } { \bf a}_{mehr}[\psi_{nk } ] + \nonumber\\ & & { \bf b}_{mehr}[i{\bf k } \cdot \tilde { \bf \nabla } \psi_{nk } + { 1 \over 2 } { \bf k}^2\psi_{nk } + v_{eff}({\bf k } ) \psi_{nk } ] \nonumber\\ & & = \epsilon_n { \bf b}_{mehr}[\psi_{nk } ] , \label{complexcompact - implicit}\end{aligned}\ ] ] where @xmath23 and @xmath24 are again the components of the _ mehrstellen _ operator .
the accuracy of the discretization was tested by calculating the lattice constant and bulk modulus for an 8-atom si supercell .
a 26-ry equivalent cut - off was used with k - space sampling restricted to the baldereschi point.@xcite the calculated lattice constant was 5.38 with a bulk modulus of 0.922 mbar .
these are in good agreement with the corresponding plane - wave calculation with the cut - off of 26 ry : 5.39 and 0.960 mbar , respectively.@xcite bulk aluminum was selected as an additional test case .
a 4-atom cell was used with a 23 ry equivalent cut - off and k - space sampling of 35 special points in the irreducible part of the brillouin zone .
since al has partially occupied orbitals , the bands near the fermi level were occupied using a fermi - dirac broadening function of width 0.1 ev . the convergence of the total energy with respect to energy cut - off and the number of k - points was tested by increasing the cut - off from 23 to 49 ry , which produced a change of only 10 mev / atom , and by increasing the number of k - points to 56 , which changed the total energy by 3 mev / atom . the calculated lattice constant and bulk modulus of 4.02 and 0.734 mbar are in excellent agreement with the experimental values of 4.02 and 0.722 mbar , and with previous theoretical results of lam and cohen@xcite of 4.01 and 0.715 mbar . unlike plane - wave methods , where different symmetry groups can be handled easily
, an _ efficient _ real - space implementation for periodic systems with non - orthogonal lattice translation vectors requires considerable modifications to the orthorhombic - symmetry implementation .
the hexagonal grid describing the unit cell or supercell is generated by @xmath43 where the @xmath44 are the hexagonal bravais - lattice vectors.@xcite the @xmath45 ratio can be chosen arbitrarily by varying the two independent grid spacings , @xmath46 and @xmath28 , and the number of grid points , @xmath47 and @xmath48 .
however , one should use @xmath49 in order to maintain similar resolution both in the _ xy _ plane and along the _ z_-axis . because the indexing of this hexagonal grid is isomorphic to the cubic one , the computer representation of potentials and wavefunctions does not change .
the most important difference is in the discretization of the kohn - sham equations .
we have implemented a mixed sixth - order kinetic energy operator .
this discretization is described in appendix c , as well as the modifications required in the multigrid restriction and interpolation procedures ( q.v . section iii ) .
the above implementation has been tested on a 32-atom supercell of aln in the wurtzite phase .
the accuracy of the results was confirmed by comparison with plane - wave car - parrinello calculations on the same supercell .
generalized norm - conserving@xcite al and n pseudopotentials were used for both calculations with k - space sampling restricted to the @xmath10 point .
the cohesive energy from the real - space calculation was 11.5 ev per aln unit , which compares well with the value of 11.6 ev obtained in the car - parrinello calculations .
the eigenvalue degeneracies were identical in both calculations and the maximum difference in any eigenvalue was 0.04 ev .
the extension of the real - space grid representation to other bravais lattices proceeds in a similar manner , the only requirement being the existence of an accurate finite difference discretization . in real - space calculations it is possible to add resolution locally .
this is especially valuable for systems with a wide range of length scales such as surface or cluster calculations .
a high density of grid points can be used near the ions , with a low density in the vacuum regions .
other possible applications are simulations of impurities in bulk materials , where the impurity ions may require higher resolutions to be accurately represented . by using locally enhanced regions the required resolution may be added only where needed , thereby greatly reducing the total number of points required .
local enhancement of the grid resolution may be achieved by adding small high - resolution grids@xcite onto a uniform global grid , or by using a coordinate transform to warp the grid structure . our focus here will be on the second approach , which was first proposed by gygi@xcite for plane - wave basis sets and recently extended to real space by several workers.@xcite in the real - space approach , a continuous coordinate transform is applied to a uniform grid . in general the transformation
is non - separable , but we prefer a separable coordinate transformation in order to avoid mixed derivatives in the kinetic - energy operator . as a test of the utility and accuracy of this scaled - grid approach , we examined an interstitial oxygen impurity in si@xcite for two grid layouts : a dense uniform grid with a 76-ry cut - off and a scaled grid with a cut - off that varied from 18 ry to 76 ry .
the scaling - transformation that maps the fictitious computational grid @xmath50 to the physical warped grid @xmath51 is @xmath52 where @xmath53 is the @xmath50-coordinate of the oxygen atom , @xmath54 is the length of the supercell in the @xmath50-direction , and @xmath55 is an adjustable parameter between zero and one that controls the degree of scaling .
the @xmath56 and @xmath57 coordinates are scaled analogously .
the scaled grid required four times fewer points than the uniform grid to achieve the same convergence of the total energy .
since the coordinate transforms are continuous functions , the integration weights and the coefficients of the discretized kinetic - energy operator may be generated from the uniform grid values using the metric tensor of the transform in the manner outlined by gygi.@xcite with these modifications , the calculations proceed as for the uniform grids , but a sixth - order central finite - difference operator is used to discretize the first and second derivatives because we have not constructed a _
operator for the scaled grid . in each case
( uniform and scaled grids ) , a 64-atom supercell was used and the silicon atoms were fully relaxed .
the oxygen atom was held fixed in order to avoid pulay corrections@xcite to the ionic forces , which would have been required if it and the scaled grid were free to move .
the uniform and scaled - grid calculations are in very good agreement : the maximum difference in kohn - sham eigenvalues was 40 mev and the maximum difference in ionic coordinates was 0.03 .
to efficiently solve eq .
( [ compact - implicit ] ) , we have used multigrid - iteration techniques that accelerate convergence by employing a sequence of grids of varying resolutions .
the solution is obtained on a grid fine enough to accurately represent the pseudopotentials and the electronic wavefunctions .
if the solution error is expanded in a fourier series , it may be shown that iterations on any given grid level will quickly reduce the components of the error with wavelengths comparable to the grid spacing but are ineffective in reducing the components with wavelengths large relative to the grid spacing.@xcite the solution is to treat the lower frequency components on a sequence of auxiliary grids with progressively larger grid spacings , where the remaining errors appear as high frequency components .
this procedure provides excellent preconditioning for all length scales present in a system and leads to very rapid convergence rates . the operation count to converge one wavefunction with a fixed potential
is @xmath25 , compared to @xmath58 for fft - based approaches.@xcite there is no one multigrid algorithm but rather a collection of algorithms that share certain common features . in order to describe the implementation used in this work , we start with a description of a multigrid solver for poisson s equation .
this will then be used as a building block for the more sophisticated algorithms actually employed . a standard numerical problem that illustrates the multigrid algorithm is the poisson equation , @xmath59 , defined on a rectangular cell of dimensions @xmath60 with periodic boundary conditions .
a standard method of solving for @xmath61 is to discretize the equation on a uniform three dimensional grid with spacing @xmath12 and @xmath11 total points . the differential operator @xmath62 is represented by some form of finite differencing .
this produces a set of linear algebraic equations @xmath63 where @xmath64 and @xmath65 are the discretized forms of @xmath66 and @xmath67 , respectively , and @xmath68 is the finite - difference representation of @xmath62 .
if the system is small , direct matrix methods are an acceptable means of solving the equations ; however , for large systems the work required scales as @xmath69 , which is prohibitive .
an alternative approach is to use an _ iterative _ relaxation scheme such as the jacobi method.@xcite in this technique , the solution is iteratively improved .
first , define @xmath70 as an approximate solution of eq .
( [ mg1 ] ) , and the residual @xmath71 , a measure of the solution error , as @xmath72 an improved @xmath70 is generated using @xmath73 where @xmath74 is a pseudo time step and k is the inverse of the diagonal component of @xmath68 .
this approach will always converge to a solution for some value of @xmath74 if @xmath68 is diagonally dominant.@xcite however , the number of iterations required to reduce the magnitude of the residual to a specified accuracy is proportional to @xmath75 , so that the algorithmic cost to converge , @xmath76 , is too great.@xcite while more sophisticated relaxation methods such as gauss - seidel , successive overrelaxation ( sor ) , or the alternating direction implicit method ( adi)@xcite have improved convergence rates , they still scale as @xmath77 with @xmath78 , and are too slow for the grid sizes required in electronic structure calculations .
the slow convergence of the jacobi method can be qualitatively understood by noting that because @xmath68 is a short - ranged operator , the updated approximate solution , eq .
( [ res2 ] ) , is a linear combination of nearby values .
if the error in the current estimate of @xmath70 is decomposed into fourier components , it can be shown that one jacobi iteration considerably reduces the high frequency components of the error , but many jacobi iterations are needed to affect the longest wavelength components of the error .
the overall convergence rate is then limited by that of the lowest frequency components .
as the problem becomes larger , the lowest frequency representable on the grid becomes smaller and the convergence rate decreases .
the essence of the multigrid approach is the observation that the individual frequency components of the error are best reduced on a grid where the resolution is of the same order of magnitude as the wavelength of the error component .
this approach will maintain a high convergence rate for all frequency components of the error even when the problem size ( and the grid ) becomes very large .
we first describe a multigrid algorithm to solve poisson s equation that uses two grids ; a fine grid of spacing @xmath79 and a coarser auxiliary one of spacing @xmath80 . in this work
, we use a coarse - to - fine grid ratio @xmath81 of 2 , but other ratios are possible .
the solution is generated as follows : the high - frequency components of the solution error with wavelength @xmath82 are reduced by one or two jacobi iterations .
the residual @xmath83 , which should be devoid of high - frequency variation , is computed and transferred to the coarse grid by restriction ( see below ) .
next , poisson s equation on the coarse grid with the residual as a source term is solved by using the same iteration procedure as in eq .
( [ res2 ] ) , but with an initial estimate of zero .
the jacobi iteration on this level removes error components with wavelength @xmath84 . finally , the coarse grid solution is interpolated to the fine grid and added to the fine grid solution .
this process is referred to as a coarse - grid - correction scheme ( cgc ) .
a few applications of the cgc cycle are generally sufficient to solve poisson s equation to machine precision even for extremely large systems .
an obvious question is how the solution is obtained on the coarse grid .
if the total number of grid points in the coarse grid is small , a direct matrix method will be sufficient . if this number is so large that a direct method is impractical , then a second , coarser grid level is introduced and the two - grid algorithm is repeated in a recursive manner . when multiple grid levels are used , the pattern of cycling through the grids also needs to be considered .
we use a simple progression from the finest to the coarsest grid level and then back to the finest level , which is referred to as a v - cycle .
more complicated cycling schemes exist , but we have found that the v - cycle works as well in electronic structure calculations as the more sophisticated approaches .
another consequence of the multigrid approach is the reduction in the size of the grid ( and consequently in the work required ) on each level . for a uniform grid in three dimensions
a doubling of @xmath12 with each level leads to a factor of eight reduction in @xmath11 , so that the addition of extra coarse grid levels is computationally inexpensive .
the simplest choice for the restriction operator is to copy every other point in the fine grid directly to the coarse grid .
this so - called straight injection , while easy to implement , does not always yield good convergence rates .
a better choice is a weighted restriction , in which each coarse - grid value is the average of the 27 fine - grid values surrounding it . in our work ,
the weight assigned to each fine - grid point is proportional to the volume it occupies at a given coarse - grid point .
a good choice for interpolating from the coarse grid to the fine grid is the adjoint of the restriction operator , which in this case becomes simple tri - linear interpolation.@xcite the final accuracy of the solution is determined by the finite - difference representation of @xmath62 on the finest grid level .
it is neither necessary nor desirable to use the same representation of @xmath62 on all grid levels ; i.e. , @xmath85 may differ from @xmath86 in the form of the discretization as well as in the grid spacing .
the technique of changing discretization on different levels is referred to as deferred defect correction ( ddc ) or double discretization.@xcite it is especially valuable for problems where the operator used on the finest grid level is numerically unstable on the coarser grids , or when it is inconvenient to apply .
the accuracy of the fine - grid solution does not depend on the choice of this coarse - grid operator , which effects only the convergence rate . in solving poisson s equation , we use the _ mehrstellen _ operator on the finest grid level for high accuracy . however , it is unsuitable for convergence acceleration on the coarser grids because of stability problems .
thus , on the coarser grids a 7-point central finite - difference operator is used , which provides excellent stability and rapid high - frequency attenuation .
the extension of multigrid concepts to the solution of the kohn - sham equations introduces several complications .
first , the equations are non - linear since the eigenvalues and the orbitals must be computed simultaneously .
second , when the kleinman - bylander form of the nonlocal pseudopotentials is used , the equations become a set of integro - differential eigenvalue equations . finally , the hamiltonian depends upon the density , and must be solved self - consistently .
brandt _ et al._@xcite have given a multigrid algorithm for the standard eigenvalue problem .
below we describe an alternative means of linearizing the eigenvalue equations .
the multigrid technique recommended in the literature for non - linear integro - differential equations is the full approximation storage ( fas ) method.@xcite in fas the _ entire _ problem is discretized and solved on _ all _ grid levels .
in contrast , the cgc method outlined above generates the full solution only on the finest level . while the theoretical performance of fas on this problem is superior , its implementation is significantly more complex . in addition , it is difficult to obtain an accurate representation of the nonlocal pseudopotentials on the coarser grid levels .
furthermore , the kohn - sham equations need not be converged to maximum accuracy at every iteration , because the electron charge density ( and therefore the hamiltonian ) change after each multigrid step . for these reasons ,
a modification of the double discretization approach described above was used .
the discretized operator on the finest grid is the _ mehrstellen _ approximation of the kohn - sham hamiltonian , while a 7-point central finite - difference representation of @xmath62 alone is used on the coarse grids .
effectively , the coarse - grid equation for the wavefunction residual becomes poisson s equation , where the source term is the residual generated by the _ mehrstellen _
operator on the fine grid , eq.([mehrres ] ) .
our multigrid procedure begins with the selection of some initial wavefunctions and electron charge density .
we postpone discussion of initialization techniques until later and assume that an adequate start has been generated .
the following steps are then performed for each individual wavefunction : first , an estimate of the eigenvalue is calculated from the rayleigh quotient of the generalized eigenvalue equation , eq.([compact - implicit ] ) : @xmath87\rangle \over \langle\psi_n| { \bf b}_{mehr } [ \psi_n]\rangle } .
\label{rd1a}\ ] ] in the case of complex orbitals , the eigenvalue is calculated using the rayleigh quotient of the real ( or imaginary ) part of the generalized eigenvalue equation , eq .
( [ complexcompact - implicit ] ) : @xmath88| re[{\bf h}_{bloch } [ \psi_{nk}]]\rangle \over \langle re[\psi_{nk}]| re[{\bf b}_{mehr } [ \psi_{nk}]]\rangle } .
\label{rd1b}\ ] ] next , several jacobi iterations are applied to the orbital on the finest grid using eqs . ( [ res1 ] ) and ( [ res2 ] ) , where the residual is computed as @xmath89 - { \bf h}_{mehr}[\psi_n ] .
\label{mehrres}\ ] ] the fictitious time step @xmath90 used in the jacobi iteration is typically chosen between @xmath91 and @xmath92 a.u . in the case of complex orbitals ,
the real and imaginary components of the orbital are updated separately , using the appropriate generalization of eq .
( [ mehrres ] ) .
next , the residual is restricted to the first coarse grid .
a ddc coarse - grid cycle begins using the 7-point central finite - difference representation of @xmath62 instead of @xmath93 .
several auxiliary coarse grids can be used .
when the coarse - grid correction is interpolated onto the finest grid , only a fraction @xmath94 of it is added to the orbital , for reasons of stability .
a value of @xmath95 has been found to work for almost all systems .
( larger values may produce much higher convergence rates on some systems while being unstable for others , so some experimentation is necessary . ) before transferring the residual to the coarse grid , it is essential that enough jacobi iterations be performed to eliminate the high frequency components from the residual .
since the residual is used as the right hand side of a cgc correction cycle , any high frequency components will eventually be transferred to a coarser grid where they can not be represented correctly , greatly reducing the effectiveness of the multigrid cycle . in some cases
they may even make the process numerically unstable . in the above approach ,
the difficulties of discretizing the nonlocal pseudopotentials on the coarse grid levels are avoided because the potential term is computed on the finest grid and frozen thereafter .
the steps outlined above in the ddc apply only to a single wavefunction .
the full solution process also requires the application of the orthonormality constraints and an update of the electron charge density .
the full solution process ( one scf step ) consists of the following cycle .
first , the ddc is applied to all of the wavefunctions .
next , the orthonormality constraints are applied using the gram - schmidt procedure : @xmath96 the new electron charge density is generated by linear mixing : @xmath97 where @xmath98 is the occupation of the @xmath99 state and @xmath100 is a mixing parameter , generally set to a value between 0.5 and 0.9 .
next , the hartree potential is recomputed for the new charge density using a _
ddc cycle , and a new exchange - correlation potential is generated .
finally , a subspace diagonalization may be performed at this point .
this need only be done occasionally ( every 10 - 20 scf steps is generally adequate ) in order to unmix eigenstates that may be close in energy . because the _ mehrstellen _ hamiltonian leads to a non - hermitian generalized eigenvalue equation ( see appendix a )
, subspace diagonalization requires a brief discussion : we look for a unitary transformation of the current wavefunctions that better represents the eigenvectors of the hamiltonian , and are led to the following eigenvalue equation for the subspace : @xmath101 where @xmath102\rangle \\ b^{sub}_{m ,
n } = & & \langle\psi_m| { \bf b}_{mehr } [ \psi_n]\rangle,\end{aligned}\ ] ] and @xmath103 is the matrix of coefficients of the unitary transformation for the @xmath104 state . because @xmath105 is invertible ( see appendix a ) , the subspace equations are equivalent to @xmath106 where @xmath107 .
the matrix @xmath108 is not hermitian except when the subspace is a subset of the space of eigenvectors .
thus , we do _ not _ diagonalize @xmath108 because its eigenvectors are not necessarily orthogonal , which would spoil the orthogonality of the electronic orbitals .
instead , we discard the anti - hermitian part of @xmath108 , which is smaller than the hermitian part of @xmath108 by @xmath109 , and diagonalize the hermitian part .
this approximation works well in practice , and is exact at convergence
. the hermitian approximation does not affect the final accuracy of the solution because the multigrid - assisted jacobi iterations ultimately converge the orbitals . nonetheless , accurate subspace rotations are essential for good convergence : compare the convergence rates in figs.[siliconconvergence]-[diamondconvergence ] . as a test , we compared subspace diagonalizations with the _ mehrstellen _ and the hermitian sixth - order discretizations , and found that the convergence is significantly improved with the former .
the cycle described above is repeated until the electronic system converges to the desired tolerance , which may be monitored by computing the rms value of the residual vector for each wavefunction ( see eq .
( [ mehrres ] ) ) . when this reaches a value of @xmath110 a.u . for all wavefunctions in the occupied subspace ,
the convergence is sufficient for the computation of forces that are accurate enough for large step molecular dynamics with excellent energy conservation .
as was mentioned previously , the convergence rates depend on the choice of the initial wavefunctions and electron charge density .
a poor choice can lead to slow convergence rates or in some cases the system will not converge at all . apart from random initial wavefunctions or an approximate solution that is generated using an lcao basis set
, one can also use a double - grid scheme . in the latter method
the initial solutions are generated on a grid with a spacing twice as large as that used for the final grid .
the computational work on this coarse grid is eight times smaller than what is needed on the fine grid .
the approximate coarse - grids wavefunctions are then interpolated to the fine grid and used as the initial guess .
this process can reduce the number of scf cycles needed on the finest grid level by a factor of two to three , thereby achieving significant savings in the computational effort .
the theoretical convergence rates of multigrid methods may , in principle , be calculated exactly for certain types of problems . for well - behaved partial differential equations such as poisson s equation
discretized on @xmath11 points , @xmath25 total operations are required to obtain a solution accurate to the the grid - truncation error .
this compares well with fft based methods which require @xmath111 $ ] operations . for the kohn - sham equations ,
an exact theoretical bound on multigrid convergence rates is difficult to obtain due to self - consistency effects , and to the best of our knowledge this analysis does not yet exist . we have therefore elected to study convergence properties in an empirical fashion by performing tests on physical systems typical of the problems normally examined with density functional theory . in previous work@xcite
the present authors examined convergence rates for 8-atom supercells of perfect diamond as a function of the effective kinetic - energy cut - off determined by the grid resolution , for a 32-atom supercell of gan that included the ga @xmath9 electrons in valence , and for a highly elongated 96-atom diamond supercell .
it was found that multigrid convergence rates were largely independent of energy cut - off and cell geometries .
while promising , these results were obtained for perfect crystal configurations of semiconductor compounds , which are generally fairly easy to converge . in this article
we present the results of a more systematic study that includes disordered systems .
the first system selected was a 64-atom supercell of bulk silicon .
the ions were represented by a generalized norm - conserving pseudopotential@xcite and the grid spacing used corresponded to an energy cut - off of 12 ry .
the ionic positions were generated by a molecular dynamics simulation at a temperature of 1000 k. because the work required to converge to the ground state depends on the quality of the initial wavefunctions and charge density , we used random initial wavefunctions and a constant initial electron charge density to minimize any possible bias from the choice of a starting configuration . a small number ( 10% of the total ) of conduction band states
was included in the calculations .
fig.[siliconconvergence ] shows the convergence rate defined as the @xmath112 , plotted as a function of iteration number , where each iteration represents a single scf step .
results are shown for calculations performed with and without multigrid acceleration , where the latter used a steepest - descents algorithm .
in addition , the two calculations were repeated with , and without subspace diagonalizations of the orbitals . for the calculations that included subspace diagonalizations ,
the procedure was applied every 8 scf steps , which led to small discontinuities in the smooth evolution of the total energy .
the results show that maximum convergence rates are obtained when multigrid iterations are combined with subspace diagonalization .
the slowest convergence occurs for steepest descents with no subspace diagonalization .
for the two runs where subspace diagonalizations were performed , the multigrid run converged at roughly 2.5 times the rate of the steepest descents approach . while these results are encouraging , bulk silicon is a relatively straight - forward test and is well handled by standard plane - wave methods . as an example of a more difficult system we have considered a 64-atom diamond supercell with a substitutional nitrogen impurity .
standard pseudopotentials@xcite were used for both c and n. the strong n @xmath113 potential required an energy cut - off of 63 ry .
the presence of a localized nitrogen donor level together with the 63 ry cut - off makes the system more difficult to converge .
random initial wavefunctions were used , and fig .
[ diamondconvergence ] shows the observed convergence rates .
the convergence rates for the two runs that use subspace diagonalization are a factor of 4 better for multigrid than for steepest descents .
this relative improvement is considerably greater than that observed for the silicon cell and is the consequence of the automatic preconditioning provided by multigrid techniques for all of the length and energy scales present in the problem .
the multigrid convergence rates are largely independent of the grid spacing , which roughly corresponds to the kinetic energy cut - off in plane - wave approaches .
this is not true of the steepest descents algorithm , where the maximum stable time step that may be used decreases as the energy cut - off increases . when comparing the convergence rates of the multigrid and steepest descents approaches , the computational workload involved in each technique
must also be considered . a particular advantage of multigrid methods , when compared to other convergence acceleration schemes , is their low computational cost .
this is due to the factor of 8 reduction in the number of grid points on each successive multigrid level .
the computational time per scf step in the silicon and diamond runs described above increased by less than 10% when multigrid was used instead of steepest descents . for bigger systems , where the costs of orthogonalizing the orbitals and applying the nonlocal pseudopotentials begin to dominate the total computational time
, the extra work needed for the multigrid accelerations becomes negligible . in terms of computational time , the 64-atom si supercell described above required 1.6 seconds per scf step on 64 processors of a cray - t3d .
efficient structure optimizations and the calculation of dynamical quantities such as phonon frequencies and thermodynamic properties require accurate ionic forces . in plane - wave methods
the ionic forces are computed by applying the hellmann - feynman theorem.@xcite since the derivative of the pseudopotentials may be expressed exactly within the plane - wave basis , the accuracy of the ionic forces is limited only by machine precision and the degree of convergence to the born - oppenheimer surface . for the grid - based approach the accuracy of hellmann - feynman forces is limited by the numerical error in computing the integrals of the derivatives of the pseudopotentials .
this error decreases with grid spacing .
the differentiation of the radial potentials and projectors must be performed with care to include the effects of the fourier filtering .
alternatively , a derivative - free implementation of the hellmann - feynman forces can be used , which we term virtual displacements . in this scheme ,
the ionic pseudopotentials are numerically differentiated directly on the real - space grid .
the ions are moved through a set of small displacements , while the electron charge density and the wavefunctions are held fixed .
the potential energy is calculated for each displacement and finite - differenced to form the derivative .
the forces computed by the two methods agree well .
most of the forces we have calculated to date have been computed using virtual displacements .
a stringent test of the accuracy of the ionic forces is a constant - energy molecular dynamics simulation .
over the course of the simulation any systematic errors in the forces will manifest themselves as poor energy conservation .
a distinction has to be made between small random errors that appear as bounded oscillations in the total energy and errors that increase in magnitude with simulation time .
the small random errors are expected in the real - space approach because the energy of an ion varies by a small amount as its position changes relative to the grid points.@xcite this is of no particular concern as long as the magnitude of the variation is small and oscillatory in nature .
of greater concern are errors that are unbounded .
these could arise from errors in the forces , errors in integrating the equations of motion of the ions , and lack of self - consistency due to inadequate convergence of the electronic wavefunctions .
the first source of error was minimized by fourier filtering of the ionic pseudopotentials .
the second is generally not a problem unless the ionic time step is too large . for small time steps even a simple integrator such as the verlet algorithm is sufficient , and larger time steps may be handled by using higher order integrators , such as the beeman - verlet method.@xcite the last source of error is the most significant because hellmann - feynman forces are only accurate to first order in the error of the wavefunctions .
a high degree of self - consistency is thus necessary to obtain good energy conservation .
a 64-atom silicon supercell was selected to test energy conservation on a typical system .
the ions were given random initial displacements from the perfect crystal configuration , and several velocity rescaling steps were performed in order to attain an average ionic temperature of 1100 k. a constant - energy molecular - dynamics simulation over 1 ps was then carried out , using 80 a.u .
time steps and third - order beeman - verlet@xcite integration of the ionic equations of motion .
the potential , kinetic , and total energy of the system vs. simulation time are plotted in fig.[simd ] .
we observed good energy conservation : the maximum variation in the total energy was 1.75 mev , which corresponds to 27 @xmath0ev per atom .
the performance of a given algorithm when solving complicated problems depends not only on the theoretical efficiency , which may be quite high , but also on how adaptable the algorithm is to modern computer architectures .
one example are certain classical molecular dynamics algorithms , which perform only slightly better on vector supercomputers than on low cost engineering workstations , even though the supercomputer s theoretical peak performance may be an order of magnitude larger .
a particular strength of the car - parrinello method has been its efficient implementation on vector supercomputers , such as the cray - ymp .
however , vector performance , while improving steadily , is unlikely to increase by several orders of magnitude per decade as has occurred in the past . at the same time , the development of powerful , low cost microprocessors and memory , has led to massively parallel architectures consisting of a large number of microprocessors , linked by a high speed communication network .
although efficient implementations of plane - wave - based methods on massively parallel architectures exist , the fft - based algorithms do not scale well with the number of processors because the fft is a global operation .
below , we will describe a massively parallel implementation of the multigrid method .
although some of the code - optimization issues are architecture - specific , most are generic and thus applicable to any massively parallel computation .
the target machine is the cray - t3d , which uses up to 2048 dec - alpha microprocessors , each with a peak performance of 150 mflops .
each processor has 8 kb direct - mapped data and instruction caches and 8 mw of local memory .
the processors are linked together in a three - dimensional torus arrangement for data communication .
three issues have to be addressed in order to write an efficient code for this type of machine : minimizing communication costs between processors , balancing the work - load on each processor , and code optimization on the individual processors .
the majority of the data storage in the multigrid method consists of the wavefunction values on the real - space grid .
we will consider the case where the points are distributed on a uniform three - dimensional rectangular grid . if @xmath114 is the total number of wavefunctions , then @xmath115 total storage is required .
the simplest possible decomposition of data is to store complete wavefunctions on each processing element ( pe ) , where each pe stores @xmath116 orbitals . while conceptually simple , this approach will perform poorly for large systems with many wavefunctions , because orthogonalizing wavefunctions residing on different pe s requires sending large amounts of data between processors .
an alternative approach , and the one adopted by us , is to use real - space data decomposition . in this method ,
each pe is mapped to a specific region of space .
the electron charge density , hartree potential , and each wavefunction are distributed by regions over the processors . with this approach interprocessor communication
is restricted to two areas : the computation of integrals on the real space grid ( see eq.([integration ] ) ) , and the application of the finite - differencing operators . for integration ,
the ideal optimization strategy is to organize the calculation so that as many integrals as possible are computed at once .
this can be understood by considering the time required for interprocessor communication , which consists of a latency period and a transfer phase .
the latency period is significant and is the same whether 1 or 1000 words of data is transferred .
our integration procedure is as follows : first , calculate the intra - processor contributions to the integral ( i.e. , integrate over the subdomains ) ; second , store as many of these local integrals as possible ; finally , transfer them between processors in blocks and complete the integration by summing the local integrals .
it was straightforward to implement the above procedure in most cases , but the orthogonalization step required significant modifications . in a standard implementation of the gram - schmidt orthogonalization algorithm , wavefunction overlaps and updates are computed sequentially , and the integrals can not be computed in parallel . to reduce the number of data transfers , the following implementation of gram - schmidt orthogonalization was adopted .
first , the overlap matrix @xmath117 is computed as above : the local parts of the overlap integrals are computed and stored on each processor ; the integration is then completed by transferring them in blocks to the other processors .
second , the cholesky factorization@xcite of the overlap matrix is computed : @xmath118 .
the cholesky factor , @xmath119 , is relevant because its components are the overlaps between the new orthogonal wavefunctions and the original ones.@xcite finally , the diagonal components of the cholesky factor are used to normalize the wavefunctions , and the off - diagonal ones are used to complete the orthogonalization : @xmath120 for simplicity , the cholesky factorization of @xmath121 is currently performed on each processor .
the computational time to factorize scales as @xmath122 , but has not yet become a bottleneck . however , for very large systems ( greater than 800 orbitals ) , a parallelized cholesky factorization will save significant computer time and memory .
the second area where interprocessor communication is required is the finite differencing of the wavefunctions and hartree potential , since finite differencing is nonlocal .
however , in the _ mehrstellen _ discretization , the nonlocality is restricted to points within one grid unit in each cartesian direction .
interprocessor communication is thus always limited to nearest neighbor pe s regardless of the size of the system .
this low communication cost is a particular advantage of a _ mehrstellen _ type approach as opposed to a central finite - difference approach , which requires a higher degree of nonlocality to achieve the same level of accuracy .
load balancing and the efficient use of all pe s is a major concern for any parallel algorithm . with the method described above
, the load balancing is essentially perfect for all parts of the calculation except for the application of the nonlocal pseudopotentials .
these are applied to the wavefunctions in localized volumes around each ion .
if the distribution of ions in space is nonuniform , then some of the pe s will be idle for a fraction of each scf step .
however , actual calculations on many systems have shown that the application of the nonlocal potentials typically requires less than 10% of the total computational time on any pe , so that processor utilization will always exceed 90% .
the efficiency of the massively parallel implementation described here is illustrated in fig .
[ speedup ] , which shows the speedup in execution time per step for a given problem as the number of pe s is increased .
the graph indicates a superlinear relationship , which is an artifact due to single processor cache effects .
there are two competing factors here .
the first is the increased communication cost as the number of processors increases , which tends to reduce the speedup .
the second is the reduction in the amount of data stored on each processor and a consequent increase in the number of cache hits . as was discussed earlier ,
the communication costs are relatively small with the data model being used ; since the cache is relatively small , cache hit effects outweigh these . an apparent superlinear speedup is observed .
we have described the development of a multigrid - based method that uses a real - space grid as a basis .
the multigrid techniques provide preconditioning and convergence acceleration at all length scales , and therefore lead to particularly efficient algorithms .
a specific implementation of multigrid methodology in the context of density functional theory was described and illustrated with several applications .
the salient points of our implementation include : ( i ) the development of new compact discretization schemes in real space for systems with cubic , orthorhombic , and hexagonal symmetry , and ( ii ) the development of new multilevel algorithms for the iterative solution of kohn - sham and poisson equations .
the accuracy of the discretizations was tested by direct comparison with plane - wave calculations when possible and were found to be in excellent agreement in all cases .
these algorithms are very suitable for use on massively parallel computers and in @xmath4 methods .
we described an implementation on the cray - t3d massively parallel computer that led to a linear speedup in the calculations with the number of processors .
the above methodology was tested on a large number of systems .
a prior communication described tests on c@xmath123 molecule , and diamond and gan supercells .
the present article examined convergence properties in detail for a supercell of disordered si and the n impurity in diamond .
the multigrid techniques increased the convergence rates by factors of 2 and 4 , respectively , when compared to the steepest descents algorithm .
an extension to non - uniform grids that uses a separable coordinate transform to change grid resolution locally , e.g. , at the surface or near an impurity , was developed and tested on the o interstitial in si .
this extension results in only minor changes in methodology and coding , while the reduction in basis set size and thus in computational effort can be significant .
a complex version of multigrid code , capable of an arbitrary sampling of the brillouin zone , was also was developed and tested on bulk al .
large time - step molecular dynamics simulations require very accurate forces , which can potentially lead to difficulties in real space methods as the atoms move relative to the grid points .
we have described a set of techniques based on fourier filtering of pseudopotentials that eliminate these difficulties for grid spacings of sizes similar to those used in plane - wave calculations . a 1 ps test simulation of bulk si at 1100 k conserved the total energy to within 27 @xmath0ev per atom , and illustrated the high quality of these forces .
further applications of this methodology are in progress , including a simulation of surface melting of si,@xcite structural properties of large biomolecules that contain over 400 atoms,@xcite and electronic and structural properties of in@xmath124ga@xmath125n quantum wells.@xcite the multigrid methodology is also very suitable for @xmath4 implementations , and tests results for a 216-atom cell of bulk si were described recently.@xcite
the authors wish to thank c. j. brabec for designing some of the parallel multigrid routines ; m. buongiorno nardelli for his help with the complex - wavefunction and brillioun zone sampling routines ; and m. g. wensell for supplying his molecular dynamics routines .
this work was supported by onr grant number n00014 - 91-j-1516 , nsf grant number dmr-9408437 , and onr grant number n00014 - 96 - 1 - 0161 .
the calculations were performed at the pittsburgh supercomputer center and the north carolina supercomputing center .
the _ mehrstellen _ discretization differs from central finite - differencing in two important respects : first , higher accuracy in the discretization is achieved by using more local information , but this accuracy is fully realized only at convergence ; and second , the discretized kohn - sham eigenvalue equation eq .
( [ compact - implicit ] ) is non - hermitian because the operator @xmath24 does not commute with the potential operator . in this appendix
, we examine the accuracy of the _ mehrstellen _ discretization , and prove that the non - hermitian nature of @xmath126 does not change the nature of the wavefunctions : they remain orthogonal . for simplicity
, we analyze only the _ mehrstellen _ discretization of the orthorhombic lattice . the fourth - order _ mehrstellen _ discretization ( see table [ mehrstellen - ortho ] ) samples the hamilton and the wavefunction at 19 points @xmath127 = & & a f({\bf x } ) + \sum_{n=1}^3 b_n f({\bf x } \pm h_n \hat { \bf x}_n ) \nonumber\\ & & + \sum_{n < m } c_{n , m } f({\bf x } \pm h_n \hat { \bf x}_n \pm h_m \hat { \bf x}_m ) \\ { \bf b}_{mehr}[f({\bf x } ) ] = & & a ' f({\bf x } ) + \sum_{n=1}^3 b'_n f({\bf x } \pm
h_n \hat { \bf x}_n ) . \label{appa - cubic}\end{aligned}\ ] ] the accuracy of the _ mehrstellen _ discretization is one order higher than the corresponding central finite - differencing one , but this accuracy is achieved only at convergence.@xcite the small @xmath79 expansions of the @xmath23 and @xmath24 demonstrate this principle : @xmath128 note that by construction , @xmath129 to @xmath130 .
thus , the _ mehrstellen _ discretization of the kohn - sham equations is equivalent to @xmath131 - { \bf b}_{mehr } [ \epsilon_n \psi_n ] = \nonumber\\ & & { \bf b}_{mehr}[{\bf h}_{ks}\psi_n -\epsilon_n\psi_n ] + o(h^4 ) .
\label{appa-2}\end{aligned}\ ] ] the @xmath132 terms , implicit in the right - hand side , vanish at convergence , when @xmath133 .
a similar analysis applies to the discretization of the poisson equation : @xmath134 - { \bf b}_{mehr}[4\pi\rho ] = \nonumber\\ & & { \bf b}_{mehr}[-\nabla^2 v_h - 4\pi \rho ] + o(h^4 ) .
\label{appa-3}\end{aligned}\ ] ] unlike a plane - wave or central finite - differencing representation of the kohn - sham equations , the _ mehrstellen _ discretization eq.([compact - implicit ] ) leads to a _
non_-hermitian , _ generalized _ eigenvalue equation .
nonetheless , we prove that the right eigenvectors of the discretized operator , _ i.e. _ the electronic orbitals , are orthogonal because they are also eigenvectors of a _ hermitian _ hamiltonian .
the generalized eigenvalue equation can be recast into a more familiar form by multiplication by @xmath135 ( the invertibility of @xmath24 is discussed below ) : @xmath136 where @xmath137 is a non - compact discretization of @xmath62 of the same order as @xmath23 .
the solutions of this equation , the @xmath138 and @xmath139 , are the solutions of the original equation . because @xmath140 and @xmath24 are finite - differencing operators with constant coefficients , they are translationally invariant and thus commute .
they are also hermitian .
thus , @xmath141 is hermitian , and the wavefunctions of eq.([compact - implicit ] ) are orthogonal .
( [ compact - implicit ] ) is the preferred discretization for computation , and the equivalent eq .
( [ newmehr ] ) is of formal interest only because the operators @xmath135 and hence @xmath137 are long - ranged and therefore computationally expensive to apply .
finally , we consider the invertibility of the @xmath24 operator .
we show that under reasonable conditions @xmath24 has no _ zero _ eigenvalues ( in fact , it is a positive definite operator ) by arguing that its null space is empty .
it is straightforward to show that the null space of @xmath24 is comprised only of plane waves of maximum kinetic energy ; namely , @xmath142 ( or @xmath143 , etc . )
; see eq .
( [ bofg ] ) below .
thus , the null space of @xmath144 is empty whenever these plane waves can not be represented on the real - space mesh .
this condition can be realized in two ways : choice of grid size , or explicit projection . for periodic boundary conditions , when one or more of the linear dimensions @xmath145 , @xmath146 , or @xmath48 is odd , the maximum g - vector along that dimension is @xmath147 .
second , if the grid discretization can not be chosen to meet the formal invertibility condition , the pseudo inverse@xcite of @xmath24 exists and can be used ; that is , the few vectors in the null space of @xmath24 are projected out from the wavefunctions . on physical grounds
any orbital of such rapid variation should be excluded from the calculation because it is marginally representable on the mesh .
the pseudo inverse of @xmath24 is @xmath148 where the discrete fourier transform of @xmath24 is @xmath149
the pseudopotentials are short - ranged : the coulomb tail of the local potential is explicitly canceled and added to the madelung summation of the electrostatic energy , and by construction the nonlocal projectors have no coulomb tail . the nonlocal projectors and short - ranged local pseudopotentials are fourier filtered only once , when the appropriate potentials and grid spacing are selected .
the filtering procedure attenuates the high - frequency components while maintaining the localization of the projectors and potentials .
the unfiltered potentials or projectors are defined on a real - space radial grid , and are transformed to momentum space in order to filter the high - frequency components : @xmath150 where the cut - off function @xmath151 smoothly attenuates the radial fourier transform beyond @xmath152 .
the cut - off wave vector is determined by the grid spacing : @xmath153 .
the cut - off function is unity for @xmath154 and equals @xmath155 for @xmath152 .
the parameters @xmath100 and @xmath156 depend on the atomic species and are carefully adjusted to achieve the best results .
after the momentum - space filtering , the back - transformed potentials and projectors will extend beyond the original core radius .
for computational efficiency , it is important that the nonlocal pseudopotentials be short - ranged .
accordingly , a second filtering in real space is applied to reduce the large - radius oscillations beyond an empirically - determined radius @xmath157 .
the second filtering function is unity below the cut - off radius and equals @xmath158 above it .
example values for a carbon generalized norm - conserving pseudopotential with @xmath159 and @xmath113 nonlocalities are @xmath160 and @xmath161 , @xmath162 bohr , and @xmath163 . since the filtering procedure modifies the pseudopotentials , it is necessary to determine whether the modified potentials affect the system s physical properties .
because the degree of filtering is set by the real - space grid spacing @xmath12 , the effect is similar to performing an under - converged plane - wave calculation .
the last effects are well understood and can be measured quantitatively by progressively increasing the plane - wave cut - off . in particular , the main results of plane - wave calculations remain valid , even if they are significantly underconverged .
this is due to the uniform convergence properties of plane waves , which form a translationally invariant basis set .
similarly , the convergence effects may be monitored for a real - space calculation by decreasing the grid spacing . in our tests we found that the total energy of the system converges to an asymptotic value in a manner similar to that observed with plane waves .
the hexagonal grid structure described in eq .
( [ hexgrid ] ) is a simple hexagonal lattice .
because the @xmath57 axis is orthogonal to the @xmath164 plane , the @xmath62 operator may be written in separable form @xmath165 along the @xmath57 direction a sixth - order central finite - difference operator was selected @xmath166 where @xmath167 and the @xmath168 are given in table [ hexoperatorcoefficients ] . for the @xmath164 plane
the the lattice translational vectors are not orthogonal , and a central finite - difference expression is not applicable . instead a composite form was selected @xmath169 + o(h_{xy}^6),\end{aligned}\ ] ] where @xmath170 and the @xmath171 are given in table [ hexoperatorcoefficients ] . in the multigrid solution process
these sixth - order operators are only used on the finest grid level to compute the kinetic energy and the residual on coarser grid levels , a second - order operator is used to represent @xmath62 ; viz . , @xmath172 and @xmath173 + o(h_{xy}^2),\end{aligned}\ ] ] where the discretization weights are listed in table [ hexoperatorcoefficients ] .
the multigrid restriction operator uses a volume weighting scheme with the weights adjusted for the hexagonal grid , and similarly , the hexagonal generalization of tri - linear interpolation is used to transfer the coarse - grid correction to the fine grid .
the wavefunctions and hartree potential are generated using multigrid iterations in exactly the same manner as was described in section iii except for the modifications described here .
see , for example , i. stich , m. c. payne , r. d. king - smith , and j .- s .
lin , , 1351 ( 1992 ) ; k. d. brommer , m. needels , b. e. larson , and j. d. joannopoulos , _ ibid .
_ * 68 * , 1355 ( 1992 ) ; p. bogusawski , q .- m .
zhang , z. zhang , and j. bernholc , _ ibid . _ * 72 * , 3694 ( 1994 ) .
r. car and m. parrinello , , 2471 ( 1985 ) .
m. p. teter , m. c. payne , and d. c. allan , , 12255 ( 1989 ) .
t. a. arias , m. c. payne , and j. d. joannopoulos , , 1077 ( 1992 ) ; , 1538 ( 1992 ) .
d. vanderbilt , , 7892 ( 1990 ) .
a. m. rappe , k. m. rabe , e. kaxiras , and j. d. joannopoulos , , 1227 ( 1990 ) .
lin , a. qteish , m. c. payne , v. heine , , 4174 ( 1993 ) .
g. li and s. rabii ( 1992 ) , unpublished .
p. e. blchl , , 5414 ( 1990 ) .
f. gygi , europhys .
* 19 * , 6617 ( 1992 ) ; f. gygi , , 11692 ( 1993 ) .
d. r. hamann , , 7337 ( 1995 ) ; _ ibid . _
* 51 * , 9508 ( 1995 ) .
w. yang , , 1438 ( 1991 ) ; g. galli and m. parrinello , _ ibid .
_ * 69 * , 3547 ( 1992 ) ; f. mauri , g. galli , and r. car , , 9973 ( 1993 ) ; p. ordejn , d. a. drabold , m. p. grumbach , and r. m. martin , _ ibid .
_ * 48 * , 14646 ( 1993 ) ; j. kim , f. mauri , and g. galli , , 1640 ( 1995 ) . s. baroni and p. giannozzi , europhys
. lett .
* 17 * , 547 ( 1992 ) ; x .- p .
li , r. nunes , and d. vanderbilt , , 10891 ( 1993 ) ; m. daw , _ ibid .
_ * 47 * , 10895 ( 1993 ) ; w. hierse and e. stechel , _ ibid . _ * 50 * , 17811 ( 1994 ) ; s. goedecker and l. colombo , , 122 ( 1994 ) ; e. hernandez and m. j. gillan , , 10157 ( 1995 ) ; e. hernandez , c. m. goringe , and m. j. gillan , _ ibid _ * 53 * , 7147 ( 1996 ) .
s. r. white , j. w. wilkins , and m. p. teter , , 5819 ( 1989 ) .
j. bernholc , j .- y .
yi , and d. j. sullivan , faraday disc .
soc . * 92 * , 217 ( 1991 ) . j. r. chelikowsky , n. troullier , and y. saad , , 1240 ( 1994 ) ; j. r. chelikowsky , n. troullier , k. wu , and y. saad , , 11355 ( 1994 ) ; x. jing , n. troullier , d. dean , n. binggeli , j. r. chelikowsky , k. wu , and y. saad , _ ibid .
_ * 50 * , 12234 ( 1994 ) ; j. r. chelikowsky , x. jing , k. wu , and y. saad , preprint ( 1995 ) .
e. l. briggs , d. j. sullivan , and j. bernholc , , r5471 ( 1995 ) . k. a. iyer , m. p. merrick , and t. l. beck , j. chem . phys .
* 103 * , 227 ( 1995 ) ; t. l. beck , k. a. iyer , and m. p. merrick , proc .
sixth international conference on density functional theory , paris ( 1995 ) .
e. j. bylaska , s. r. kohn , s. b. baden , a. edelman , r. kawai , m. elizabeth , g. ong , and j. h. weare , presented at the sixth siam conference on parallel processing for scientific computing , san francisco ( 1995 ) . f. gygi and g. galli , , r2229 ( 1995 ) .
g. zumbach , n. a. modine , and e. kaxiras , preprint ( 1995 ) .
k. cho , t. a. arias , j. d. joannopoulos , and p. k. lam , , 1808 ( 1993 ) .
s. wei and m. y. chou , preprint ( 1995 ) .
a. brandt , math .
comp . * 31 * , 333 ( 1977 ) ; gmd studien , * 85 * , 1 ( 1984 ) .
a. p. seitssonen , m. j. puska , and r. m. nieminen , , 14057 ( 1995 ) .
d. r. hamann , , 2980 ( 1989 ) .
g. b. bachelet , d. r. hamann , and m. schluter , , 4199 ( 1982 ) .
d. r. hamann , m. schluter , and c. chang , , 1494 ( 1979 ) .
l. kleinman and d. m. bylander , , 1425 ( 1982 ) .
r. d. king - smith , m. c. payne , and j .- s .
lin , , 13063 ( 1991 ) .
l. collatz , _ the numerical treatment of differential equations _ , 3rd ed .
( springer - verlag , berlin , 1960 ) , p. 164 .
a. baldereschi , , 5212 ( 1972 ) .
m. buongiorno nardelli , private communication ( 1996 ) .
p. k. lam , m. l. cohen , , 4224 ( 1981 ) . n. w. ashcroft and n. d. mermin , _ solid state physics _ , ( holt , rinehart and winston , philadelphia , 1976 ) , p. 77 .
m. needels , j. d. joannopoulos , y. bar - yam , and s. t. pantelides , , 4208 ( 1991 ) .
pulay , mol .
* 17 * , 197 ( 1969 ) .
w. l. briggs , _ a multigrid tutorial _ , ( siam books , philadelphia , 1987 ) .
for a recent review , see m. c. payne , d. c. allan , m. p. teter , t. a. arias , and j. d. joannopoulos , , 1045 ( 1992 ) . w. h. press , s. a. teukolsky , w. t. vetterling , and b. p. flannery , _ numerical recipes in fortran _
( cambridge university press , cambridge , 1992 ) , sec . 19.5 .
g. m. philips and p. j. taylor , _ theory and applications of numerical analysis _ , ( academic press , london and new york , 1980 ) , p. 256 .
w. h. press , s. a. teukolsky , w. t. vetterling , and b. p. flannery , _ loc .
w. hackbusch , _ multi - grid methods and applications _ , ( springer - verlag , berlin , 1985 ) .
a. brandt , s. f. mccormick , and j. w. ruge , siam j. sci .
comput . * 4 * , 244 ( 1983 ) .
h. hellmann , _ einfhrung in die quantumchemie _ , ( deuticke , leipzig , 1937 ) .
r. p. feynman , phys .
rev . * 56 * , 340 ( 1939 ) .
d. beeman , j. comp .
20 * , 130 ( 1976 ) .
g. h. golub and c. f. van loan , _ matrix computations _
( the johns hopkins university press , baltimore , 1989 ) , p. 141 .
d. j. sullivan , unpublished ( 1995 ) .
m. g. wensell , e. l. briggs , c. roland , and j. bernholc , bull .
* 41 * , no . 1 , 258 ( 1996 ) .
c. j. brabec , e. l. briggs , d. j. sullivan , and j. bernholc , bull .
soc . * 41 * , no . 1 , 443 ( 1996 ) . m. buongiorno nardelli , k. rapcewicz , e. l. briggs , and j. bernholc , `` bulk and interfacial properties of nitride semiconductors '' , poster presented at the 1996 march meeting of the american physical society , st .
louis ( 1996 ) .
d. j. sullivan , e. l. briggs , c. j. brabec , and j. bernholc , bull .
soc . * 41 * , no . 1 , 443 ( 1996 ) . g. h. golub and c. f. van loan , _ loc .
_ , p. 243 .
.discretization weights for the fourth - order orthorhombic mehrstellen operators for the central , nearest - neighbor , and next nearest - neighbor grid points .
( a12 ) and ( a2 ) for the definitions of @xmath174 , @xmath175 , and @xmath176 .
the cubic - grid operator corresponds to @xmath177 . [ cols="<,^,^,^ " , ] | we describe a set of techniques for performing large scale _ ab initio _ calculations using multigrid accelerations and a real - space grid as a basis .
the multigrid methods provide effective convergence acceleration and preconditioning on all length scales , thereby permitting efficient calculations for ill - conditioned systems with long length scales or high energy cut - offs .
we discuss specific implementations of multigrid and real - space algorithms for electronic structure calculations , including an efficient multigrid - accelerated solver for kohn - sham equations , compact yet accurate discretization schemes for the kohn - sham and poisson equations , optimized pseudopotentials for real - space calculations , efficacious computation of ionic forces , and a complex - wavefunction implementation for arbitrary sampling of the brillioun zone .
a particular strength of a real - space multigrid approach is its ready adaptability to massively parallel computer architectures , and we present an implementation for the cray - t3d with essentially linear scaling of the execution time with the number of processors .
the method has been applied to a variety of periodic and non - periodic systems , including disordered si , a n impurity in diamond , aln in the wurtzite structure , and bulk al . the high accuracy of the atomic forces allows for large step molecular dynamics ; e.g. , in a 1 ps simulation of si at 1100 k with an ionic step of 80 a.u .
, the total energy was conserved within 27 @xmath0ev per atom . |
uavs are becoming increasingly present in our everyday lives , their extensive use recently jumped from military to hobby and professional applications .
the consumer market is growing and now it offers a wide range of micro and mini uavs at affordable costs . but this popularity induces some dangerous behavior , most people do not realize that a simple mistake can cause severe injuries to themselves or others.in the united - states , the faa has taken measures to inform hobbyists and encourage them to follow a code of conduct to prevent accidents .
the only form available is the advisory circular ac 91 - 57 from june 9th 1981 , it advises pilots to keep their uavs within their line of sight , below 400 feet above ground level , further than 5 miles from an airport ( or warn them ) , and to avoid flying above people.even for the vast majority of uav users that are responsible and careful in their use , there is no automated means to fly safely in regards to the uav s environment .
our work aims to provide such functionality to micro and mini uavs that are operated in urban areas . in this paper
, we propose a novel method for autonomous navigation for low - altitude uavs in urban areas . for a given mission
our method computes safe waypoints , which dynamically adapt the flight plan to the uav s surroundings by avoiding objects such as cars and pedestrians .
we take advantage of satellite and georegistered data to adapt the uav s mission layout by computing a weighted shortest path instead of flying in a straight line .
weights in our cost function for computing the flight path are defined using land - use summarized in three classes : most dangerous areas are roads and paths where people are prone to the danger the uav represents , safest are buildings and water , and the rest is in between ( fig . [ map ] ) . for increased safety , our method also adapts dynamically to moving objects while in flight by adding new local weight to the global weight map . in our general scenario
, we assume a uav with video camera flying over a given geographical region , for which geodetically accurate reference image , gis data of buildings and road networks are available .
captured videos are georegistered with the reference image in order to transfer pixel coordinates to gps coordinates . moving objects , e.g. vehicles and pedestrians ,
are detected and tracked from frame to frame . given the tracks , gis and reference image , the optimal uav path is dynamically computed . for simplicity in this paper , we employ ground truth tracks available from wpafb and pvlabs datasets providing geo - registered images and ground truth for moving objects . finally , we simulate a real flight by complying with the ac 91 - 57form and using parameters of compatible hardware .
many different topics are studied to enhance the usability and to develop new functions to make drones more capable and autonomous .
there are several subfields which are related to this work including video geo - registration , detection and tracking of moving objects in videos , detection of roads , buildings , water bodies from satellite imagery and flight path planning .
the most popular trend in uav video analysis has been moving object detection and tracking from aerial images , many approaches have been proposed with or without using gis data and geo - registration steps .
@xcite use epipolar constraint and flow vector bound to detect moving objects , teutsch _ et al . _
@xcite employ explicit segmentation of images , xiao _ et al . _
@xcite restrain the search on the road network , and lin _ et al .
_ @xcite use a motion model in geo - coordinates . moving object detection and tracking
are mainly used to follow targets , for surveillance as quigley _
_ @xcite and rafi _ et al . _
@xcite describe with their flight path adaptation solutions , or for consumer applications at very low - altitude as in @xcite and @xcite .
another area that has been getting a lot of attention is autonomous navigation .
different subproblems have been studied , path planning in dynamic environment @xcite , gis - assisted and vision - based localization using either road detection @xcite , buildings layout @xcite or dem ( digital elevation map ) @xcite .
various methods have been proposed for uav navigation , using optical flow with @xcite or without dem @xcite , or using inertial sensors @xcite .
+ obstacle avoidance is also a big concern for automating uav operation , but research has mostly been focused on ground robots @xcite , even if there has been adaptations for uavs as israelsen _
s intuitive solution for operators @xcite .
the approaches for autonomously navigating uavs have been studied , but previous work focus on target following or keeping the uav s integrity . however , in this paper we propose an autonomous uav navigation method in order to increase public safety in regards to drones operation , and also to prevent uavs finding themselves in difficult situations .
our contribution towards safe integration of small uavs into the airspace has two main steps .
the first step , described in section a and b , takes into account the physical surroundings of the uav by computing , as part of the mission preparation , a global path between the user - given start and end locations .
this path is represented as a succession of waypoints , exactly as users are accustomed to in mission planner softwares . before takeoff
the user is able to validate the automated path , and he can modify the waypoints if needed .
the second step , described in section c , runs in online fashion during the flight and takes into account the environment of the uav by dynamically adapting its behavior in regards to moving objects that need to be avoided .
a convenient approach to gain awareness of the uav s surroundings is to use satellite imagery and the meta - data provided by https://www.google.com/maps[google maps ] , https://www.digitalglobe.com/[digitalglobe ] , https://www.planet.com/[planet labs ] or others . to jointly use geo - registered data and aerial imagery
obtained from uav , there has to be a common representation and space .
public tools providing satellite images are very popular and well integrated in third party uav software such as http://planner.ardupilot.com/[mission planner ] . for simplicity and compliance
, the solution is then to register video images onto a geo - registered satellite image of the area of interest .
uav and world coordinates systems are related with , as described in @xcite .
@xmath0 with @xmath1 the world coordinates system , @xmath2 is the translation matrix derived from the vehicle s latitude , longitude and altitude , and @xmath3 , @xmath4 , @xmath5 , @xmath6 and @xmath7 are rotation matrices regarding respectively camera elevation angle , camera scan angle , vehicle pitch angle , vehicle roll angle and vehicle heading angle .
+ we chose to use google maps api for it s convenience and for the quality of the data provided .
this free api allows anyone to request satellite images and roadmaps displaying buildings and roads .
these three links , http://maps.googleapis.com/maps/api/staticmap?center=28.600558,-81.197722&zoom=18&format=png32&sensor=false&size=640x640&scale=2&maptype=satellite&style=feature:poi|visibility:off&style=feature:road|element:labels|visibility:off&style=feature:transit|visibility:off&style=feature:landscape|element:labels|visibility:off&style=element:labels|visibility:off&style=feature:landscape|element:geometry.stroke|visibility:off&style=feature:poi|element:geometry.stroke|visibility:off[1 ] , http://maps.googleapis.com/maps/api/staticmap?center=28.600558,-81.197722&zoom=18&format=png32&sensor=false&size=640x640&scale=2&maptype=roadmap&style=visibility:off&style=feature:road|visibility:on&style=element:labels|visibility:off&style=element:geometry|color:0x000000[2 ] and http://maps.googleapis.com/maps/api/staticmap?center=28.600558,-81.197722&zoom=18&format=png32&sensor=false&size=640x640&scale=2&maptype=roadmap&style=visibility:off&style=feature:landscape|visibility:on&style=feature:landscape|element:labels|visibility:off&style=feature:water|visibility:on|color:0x000000&style=feature:landscape.natural|visibility:off[3 ] , give example commands to request satellite , road map and building images .
we assume that flying above buildings represents less risk than doing so above other environmental elements such as roads or crowded streets .
the resulting weight map for the wright - patterson air force base ( wpafb ) area , shown in fig . [ map ] , displays three categories : * red is to be avoided , for roads and paths . *
green is to be preferred , for buildings and water . *
transparent is in between , for other land - use . to extract the map ,
only two gps locations are needed as input from the operator : top left and bottom right gps coordinates .
a grid of image gps locations is then computed based on google maps camera parameters and resolution level , in other words the ground sampling distance ( gsd ) , to ensure sufficient overlap between images for stitching .
we have defined the gps grid in a way that successive images have pure translations between them .
we thus can stitch them together using straight forward normalized cross correlation , which is a robust and fast method given that we manipulate large images and avoid scale change and rotation .
this process allows us to minimize the error while creating the geo - registered map . as a result
we obtain 3 images for any given area ( fig .
[ layers ] ) .
given the image center gps location , @xmath8 and @xmath9 , the corresponding gps location ( @xmath10 and @xmath11 ) of all other pixels location at @xmath12 from the center , can be determined as follows : @xmath13 with @xmath8 , @xmath10 , @xmath9 , @xmath11 representing latitudes and longitudes of the start and end points , @xmath14 the mean radius of the earth , @xmath15 the pixel ratio in meters per pixel depending on the ground altitude , and on the requested image scale and resolution , @xmath16 and @xmath17 are the difference in pixels on the map between the two points
. the vast majority of uas ( unmanned aerial systems ) can be used with a ground control station ( gcs ) , for example apm : copter ( previously known as arducopter ) has its own mission planner with all the necessary tools .
the conventional ways of controlling uavs are either with a manual radio controller or by using a gcs that defines successive gps waypoints ( specifying the gps location , altitude , and velocity ) to which the uav will fly autonomously . despite their efficiency and convenience , there is a crucial flaw with waypoints ; they are defined by the user and do not take into account the surroundings of the uav .
this is precisely what we want to tackle with our global path computation . by using the three types of data shown in fig .
[ layers ] , we define the optimal path ( example in fig . [ path ] ) between two points and thus add a safety parameter to mission planning .
we find the safest route between two gps coordinates by converting them into image pixels and computing a weighted shortest path algorithm using a * algorithm .
the segments lengths between two adjacent pixels are the euclidian distance multiplied by the weight defined by the map class .
pixels that are in red have a weight of @xmath18 , green is at @xmath19 , and the rest is at @xmath20 .
those values have been determined empirically .
this process ensures that the red areas are avoided but also makes sure the uav would nt take a long detour to reach its destination , thus keeping the loss of flight time to a minimum .
as the map is geo - registered , the outputted path can easily be converted into gps coordinates using and , and put in kml an txt files to be readable by mapping software and ground control stations ( fig .
[ pathmp ] ) .
this global weight map considers the static environment that the uav will encounter such as roads and buildings . in order to ensure a higher level of safety in all stages of the flight
, we also adapt the path locally during the flight in regards to moving objects as explained in the next section . for increased safety
, the path needs to be adapted dynamically during the flight to avoid moving objects detected in the field of view of the uav s embedded camera . in order to ensure a sufficient distance margin between each object and the uav , the weight map used for shortest path
is modified according to the objects location , trajectory and velocity .
we compute the new weights of the map by applying at chosen locations a multivariate normal probability density function .
the variance @xmath21 and @xmath22 for each distribution are dependent on the object s characteristics .
the @xmath23 term is proportional to the width of the object in pixels , and @xmath24 term is proportional to the object s velocity .
@xmath25 where @xmath26 is the current velocity of the object and @xmath27 the safety margin to avoid collision in seconds .
the resulting distribution is normalized , rotated to align with the object s trajectory , and centered on the chosen location ( [ weightequ ] ) .
the weight map is then multiplied with the distribution instead of being swapped in order to keep the global environment based information .
the locations where the distribution is applied to are defined given two criteria .
one is whether the object collides with the uav s path , and the other is how this collision happens . the object and uav
will take respectively @xmath28 and @xmath29 seconds to the collision point , if @xmath30 ( @xmath31 is set to 5 s in our experiments ) , the distribution is applied on the collision point and also on a projected location to avoid re - planing a path that will create a similar situation . the projected location
( [ t ] , [ d ] ) is estimated as follows : the time for the uav to travel to the current object s location is computed , the projected location is where the object will be at that time given constant velocity and trajectory for the object . for objects that will not collide or that do meet the requirement of @xmath32 , the distribution is applied at the next and projected locations .
@xmath33 \circ \phi\\ \label{weightequ}\ ] ] where @xmath34 is the weight map , @xmath35 $ ] the affine transformation applied to the multivariate normal probability density function @xmath36 to allocate costs to @xmath34 .
@xmath36 is centered at the wanted location @xmath37 and oriented given the rotation matrix @xmath38 dependent on the object s trajectory and @xmath39 is the translation component of the affine transformation , @xmath40 and @xmath40 are the image coordinates of the object , and @xmath41 and @xmath42 are respectively the distances in @xmath43 and @xmath44 to the desired location .
@xmath45 with @xmath46 the distance used in [ t ] , @xmath26 and @xmath47 the current velocities of the object and uav , @xmath48 the distance between object and uav , and @xmath49 the frame - rate or computation time . +
this method will ensure that the resulting path will leave sufficient ground distance between objects and the uav , and if multiple objects are close together , it will create a barrier and encourage the uav to find a safer path , thus preventing it to fly above any moving objects ( fig .
[ avoid ] ) .
in order to simulate a real world scenario as accurately as possible , our method uses dataset images , and typical uavs specifications and camera parameters .
we made sure to comply with the latest regulations and advice regarding uav operation , and used the following flight and hardware parameters : * altitude above ground level : @xmath50 .
* velocity : @xmath51 . *
camera horizontal field of view ( hfov ) : @xmath52 . * horizontal ground sampling resolution : @xmath53 .
+ the principles used to build the simulation scheme are the following : * uav videos are registered in the geo - referenced space , we can thus work in pixels coordinates , and convert back to gps anytime . *
the datasets ground truth gives the moving objects location for every frame ( motion vectors in fig .
[ snap ] ) .
* the uav will follow the global path ( blue in fig .
[ snap ] ) .
* for every frame the uav s displacement in the image is dependent on it s velocity and direction . *
the considered objects are only the ones visible in the field of view of the embedded camera ( exterior red dotted line around the uav in fig . [ snap ] ) . * for convenience we call collision the situation where the uav will fly over an object . * a collision is detected if the direction of an object s motion vector intersects the path in front of the uav . * a danger area
is computed , and is visible as the smallest red dotted rectangle in fig .
[ snap ] , for every frame depending on uav s velocity so that the uav will reach the boundary in 5 seconds at current and constant velocity .
@xmath54 where @xmath55 is the number of pixels to advance along the path , @xmath56 is the velocity of the uav , @xmath57 is the framerate of the dataset , and @xmath58 represents the ground sampling distance of the geo - registered map .
we use two datasets to run our safe navigation pipeline , wright - patterson air force base ( https://www.sdms.afrl.af.mil/index.php?collection=wpafb2009[wpafb ] ) @xcite and http://www.pv-labs.com/[pvlabs ] .
they are wide - area motion imagery ( wami ) , and provide ground truth for moving objects on ortho - rectified images captured by uavs .
both of those datasets have been captured at high altitude with embedded sensors and a matrix of multiple cameras .
we use the provided regions of interest outputted by a geo - registration step , described in @xcite . for each dataset
we run the different steps of the pipeline .
we first create the weight map using the process described in section [ sec3a ] .
videos are then precisely geo - registered onto the map via homography transformation . the global path ( fig .
[ paths1 ] ) is generated before the simulated flight and adapted dynamically on the way .
for both wpafb and pvlabs datasets , we defined 9 different pairs of start and end gps coordinates ( fig .
[ paths1 ] ) based on the environment and busyness of the roads to create challenging situations that will require global path adaptation . and
each path is executed at three different uav velocities : 5 , 8 , and 11 m / s .
the total traveled distance by using the global path compared to the classic straight line path , for each dataset executed for all paths at three above velocities , is @xmath59 higher or 5:13 min longer for wapafb and @xmath60 or 32 s for pvlabs , making our safety increased path an affordable measure in term of autonomy . to quantify the performance of the proposed method we introduce a metric assimilated to safety .
we consider the uav to object proximity , the closer the uav is to an object the more danger it represents for it , we therefore compute a total cost for each dataset as in .
@xmath61 with @xmath62 the global cost for the considered dataset , @xmath63 the ground distance between the uav and each object detected in the fov during the experiment , and @xmath64 a constant .
note that , for the same start and end locations , when different paths are compared , the uav will not encounter the same situations .
this is why , for clarity , we include , with the results in table [ tabres ] , the number of objects seen by the uav s camera throughout the simulation for each dataset .
.safety estimation results for wpafb and pvlabs [ cols="^,^,^,^",options="header " , ] we can clearly see in table [ tabres ] that our proposed method encounters more objects in the fov , but it has the means to keep the uav afar from them .
objects which are over 20 m away are not in danger , but having a car or pedestrian closer than 5 m to the uav represents a very concerning situation in terms of safety for people .
this is why we have chosen to compute the global cost with a negative exponential weight function , that way the shorter the distance , the more cost is applied to the global metric .
the proposed method encounters over twice the amount of moving objects but safely keeps away from them ( fig .
[ curves1 ] ) , making the resulting safety parameter much better than global path and , most of all , better than classic straight line path .
in this paper we introduced an environment and safety based path planning for low altitude uav operating in urban areas .
we compute a global path , for any mission given a pair of start and end gps locations , by using a weighted shortest path .
the weight map is defined using ground classification data summarized in three classes : highest cost is for roads and paths because of the high probability of presence of people for which the uav represents a safety threat , safest are buildings and water , and neutral areas are the rest .
additionally , we included a dynamic path planning that will modify locally the flight plan while in flight to avoid being close to moving objects such as vehicles and pedestrians .
our proposed method has been tested in simulation using geo - registered data and images from two wami datasets , wpafb and pvlabs , and it showed significant improvement compared to the current and manual mission planning solution in terms of a safety metric quantifying threat in function of uav - to - object distance .
our safety planning and navigation scheme can be implemented on - board a uav and will consist in the following steps : 1- before takeoff , acquire necessary gis data for the mission area , and generate mission waypoints using global weighted path planning , 2- during the flight , geo - register the embedded camera s images using a sensor model and gimbal readings , detect moving objects ( as in @xcite ) or any other type of objects to avoid , and generate new local path and waypoints to stay clear of the detected objects .
the research was supported by a dga - mris scholarship .
kimura et al . , `` automatic extraction of moving objects from uav - borne monocular images using multi - view geometric constraints , '' imav , 2014 .
teutsch et al . , `` evaluation of object segmentation to improve moving vehicle detection in aerial videos , '' avss , 2014 .
xiao et al . , `` vehicle detection and tracking in wide field - of - view aerial video , '' cvpr , 2010 . pritt et al . , `` georegistration of multiple - camera wide area motion imagery , '' igarss , 2012 .
habbecke et al .
, `` automatic registration of oblique aerial images with cadastral maps , '' trends and topics in computer vision , 2010 .
tchernykh , `` optical flow navigation for an outdoor uav using a wide angle mono camera and dem matching , '' ifac , 2006 .
hrabar et al . ,
`` combined optic - flow and stereo - based navigation of urban canyons for a uav , '' iros , 2005 .
pestana et al . , `` computer vision based general object following for gps - denied multirotor unmanned vehicles , '' acc , 2014 .
pestana et al . , `` vision based gps - denied object tracking and following for unmanned aerial vehicles , '' ssrr , 2013 .
ess et al .
, `` object detection and tracking for autonomous navigation in dynamic environments , '' international journal of robotics , 2010 .
israelsen et al .
, `` automatic collision avoidance for manually tele - operated unmanned aerial vehicles , '' icra , 2014 .
gonzalez et al . , `` using state dominance for path planning in dynamic environments with moving obstacles , '' icra , 2012 . | in recent years , consumer unmanned aerial vehicles have become very popular , everyone can buy and fly a drone without previous experience , which raises concern in regards to regulations and public safety . in this paper , we present a novel approach towards enabling safe operation of such vehicles in urban areas .
our method uses geodetically accurate dataset images with geographical information system ( gis ) data of road networks and buildings provided by google maps , to compute a weighted a * shortest path from start to end locations of a mission .
weights represent the potential risk of injuries for individuals in all categories of land - use , i.e. flying over buildings is considered safer than above roads .
we enable safe uav operation in regards to 1- land - use by computing a static global path dependent on environmental structures , and 2- avoiding flying over moving objects such as cars and pedestrians by dynamically optimizing the path locally during the flight .
as all input sources are first geo - registered , pixels and gps coordinates are equivalent , it therefore allows us to generate an automated and user - friendly mission with gps waypoints readable by consumer drones autopilots .
we simulated 54 missions and show significant improvement in maximizing uav s standoff distance to moving objects with a quantified safety parameter over 40 times better than the naive straight line navigation . |
quantum state transfer from one location to another is a significant problem for quantum information processing systems .
a quantum computer , which consists of different processing units , requires the quantum states to be transferred between its parts .
therefore , quantum state transfer will be an important part of quantum computer design .
there are various ways of achieving this task depending on the technology at hand @xcite . in this article , we consider two related fields of research , quantum state transfer and quantum walks on one - dimensional lattices .
quantum communication through a spin chain was first considered by bose @xcite and since then it has been studied in depth @xcite .
this procedure consists of interacting spins on a chain , whose dynamics is governed by heisenberg , xx or xy hamiltonians .
perfect state transfer ( pst ) through a spin chain , in which adjacent spins are coupled by equal strength , can be achieved only over short distances @xcite .
quantum walks ( qws ) have been introduced as a quantum analogue of classical random walks .
the continuous - time qw has been suggested by farhi and gutmann @xcite as a quantum algorithm to reach the @xmath0th level of a decision tree faster than the classical random walk .
the discrete - time qw has been introduced by aharonov @xcite where the walker has a larger average path length than its classical counterpart .
these properties of qws have allowed development of new quantum algorithms @xcite .
many experimental systems for qws have been implemented .
@xcite .
the time - evolution of qubit state transfer through a spin chain can be interpreted as a continuous - time qw and pst is possible over a spin chain of any length with pre - engineered couplings @xcite .
furthermore , this interpretation can be extended to discrete - time qw with a position - dependent coin operator @xcite .
pst in quantum walks on various graphs has been studied more specifically for the continuous - time model @xcite .
high fidelity transfer of specific quantum states on variants of cycles has been reported for the discrete - time qw @xcite without considering the internal coin state . in this article
, we show the perfect transfer of an unknown qubit state from one site ( @xmath1 ) to another ( @xmath2 ) on one - dimensional lattices in discrete - time qw architecture .
we treat the coin as our qubit whose state we aim to transfer .
the coin is an internal degree of freedom of the walker , e.g. polarization , 2-energy levels , angular momenta or spin , which moves on discrete lattice sites . at the end of the walk ,
we apply one more coin operator ( _ recovery operator _ ) to achieve pst .
the recovery operator is independent of the initial coin state and it can be determined before the walk once a coin operator is chosen .
we study the periodicity of each case where pst occurs and show that for all pst cases the quantum walk is also periodic .
moreover , we show that redefinition of the shift operator which amounts a change in the directions in which the walker can move , may lead to pst with appropriate choices of the coin operator .
this article is organized as follows . in section [ sec : dtqw ] , we present a brief review of discrete - time qw and introduce spatial and local approaches to the definition of directions for the walker .
we define the boundary conditions ( n - lines and n - cycles ) of the walk . in section [ sec : wtrans ] , we discuss the transfer of the walker between sites @xmath1 and @xmath2 without considering the coin state . in section [ sec : recpst ] , we introduce the recovery operator and give a precise definition of pst for discrete - time qws . in sections
[ sec : nline ] and [ sec : ncycle ] , we obtain the cases where pst occurs for n - lines and n - cycles . in conclusion section ,
we summarize our results .
one - dimensional discrete - time qw involves two discrete hilbert spaces .
one of them is the position space @xmath3 spanned by the basis states @xmath4 , and the other one is the coin space @xmath5 spanned by basis states @xmath6 .
these spaces correspond to the position and the internal states of the _ walker_. the total quantum state of the walker is determined by both its coin and position degrees of freedom . in other words , the whole space , @xmath7 , is spanned by the tensor product of base states which are denoted by @xmath8 .
time evolution of the walk is governed by a unitary operator which is applied in discrete time intervals ( so - called discrete - time ) to form the steps of the walk .
_ one step _ is defined by two subsequent unitary operators , _ coin operator _ which only affects the coin space and _ shift operator _ which affect both the coin and the position spaces .
thus , one step is given by the unitary operator @xmath9 , where @xmath10 , @xmath11 and @xmath12 are shift , coin and identity operators , respectively .
the most general unitary coin operator can be written as @xmath13 where @xmath14 gives the bias of the coin , i.e. , where @xmath14 and @xmath15 are probabilities for moving left and right , respectively . here ,
@xmath16 and @xmath17 are the parameters defining the most general unitary operator up to a @xmath18 phase .
the parameters @xmath19 are chosen arbitrarily at the beginning of the walk and they remain constant during the walk . for @xmath20 and @xmath21 , we obtain the well - known hadamard coin operator ( unbiased case ) .
the shift operator is given as @xmath22 where the sum is taken over all discrete positions of the space .
the shift operator forces the walker to move in a direction determined by its coin state .
the position hilbert space does not necessarily be infinite and it can be restricted to a finite number of sites @xmath23 by choosing appropriate conditions for boundaries . in this case , a different definition for the shift operator is required . in figure
[ fig : latt ] , two boundary conditions for the walk are presented and these are the ones that we will use throughout the article . in figure
[ fig : latt](a ) , the lattice with @xmath23 sites and reflecting boundaries ( _ n - line _ ) is represented .
self loops at the boundaries indicate that wave function is reflected after the shift operator is applied , similar to the approach used by romanelli for the broken links model @xcite .
the shift operator is of the form @xmath24 where the left ( right)-going part at the first ( last ) site is diverted to the right ( left)-going part at the same site to keep the flux conserved .
thus , the shift operator remains unitary . in figure
[ fig : latt](b ) , the lattice with even @xmath23 sites and periodic boundaries ( _ n - cycle _ ) is represented . here
, we simply connect the first and the last sites with one more edge .
[ cols= " < , < " , ]
in table [ tb:2line ] , the cases where the walker is found with probability @xmath25 on the 2-line are given .
the cases with coin state @xmath26 manifest periodicity . to achieve a pst , we consider the other cases where the total state is @xmath27\otimes{\mbox{$|2\rangle$}}. \label{eq : wf2line}\ ] ] @cccc @xmath14 & steps ( t ) & site ( x ) & coin state + & @xmath28 & 2 & @xmath29 + & @xmath30 & 1 & @xmath26 + & @xmath31 & 2 & @xmath29 + & @xmath32 & 1 & @xmath26 + @xmath33 & @xmath28 & 1 & @xmath26 + after t - steps
, we apply appropriate recovery operator , @xmath34 , on ( [ eq : wf2line ] ) . in this way
, we obtain the initial coin state and hence pst .
overall process can be written as @xmath35 thus , any coin state can be transferred on 2-line perfectly with appropriate @xmath36 values given in table [ tb:2line ] .
we note that recovery operator is constant for a given coin operator and it provides pst for all initial coin states . in each pst case
, the qw is periodic .
for example , after 4 steps of the walk with @xmath20 , if the walker proceeds @xmath31 more steps , the initial quantum state is recovered .
there is also a case with @xmath37 , where 2-line is periodic but it does not lead to pst .
pst requires the total state to localize more than one sites in turn and this process naturally gives rise to periodicity .
after applying the recovery operator , we initialize the walker with the initial coin state at a different site . for example , when @xmath38 , if the walker is acted on by sequence of operations , such as @xmath39 , it will be initialized on sites @xmath25 and @xmath40 alternatingly .
the sequence of initializations which keeps the initial coin state unchanged , suggest us to define a new classification for discrete - time qws which we call _ n - periodicity_. we can define one step of the walk for the example above as @xmath41 .
then , after each step , coin state will be conserved and the only change will occur in the position space .
in other words , @xmath42 is same as that of @xmath43 .
since the walker is localized on two sites in an alternating manner , the qw under consideration becomes 2-periodic .
in general , the number @xmath0 gives the total number of sites where initial coin state is localized during the time evolution .
if qw is periodic , we will call it 1-periodic , i.e. , well - known periodicity concept becomes a member of the general n - periodicity class .
thus , n - line or n - cycle allow maximum n - periodicity .
this definition is useful because it generalizes the periodicity definition so that it includes the pst too . for @xmath44 ,
reflecting boundaries ensure that there will always be a non - zero probability for finding the walker at @xmath1 , independent of @xmath45 , if there is no destructive interference .
however , the dimension of the position space for 2-line allows the wave function to vanish at @xmath1 and gives rise to the cases given in table [ tb:2line ] .
when we restrict the coin operators to @xmath46 , independent of the initial coin state , the walker is transferred from @xmath1 to @xmath2 and @xmath2 to @xmath1 at intervals of @xmath23 steps for all n - lines . in general , the walker is at the position @xmath2 or @xmath1 if @xmath47 or @xmath48 steps are taken , respectively . here
, @xmath49 specifies the number of `` round trips '' of the walker within the lattice . to find the coin state of the walker at @xmath45 ,
we have derived the total quantum states @xmath50\otimes{\mbox{$|n\rangle$ } } , \cr { \mbox{$|\psi_{2nl}\rangle$ } } & = e^{il\theta}[\alpha{\mbox{$|\ua\rangle$}}+\beta { \mbox{$|\da\rangle$}}]\otimes{\mbox{$|1\rangle$ } } , } \label{eq : nline1}\ ] ] where @xmath51 and @xmath52 are the parameters of the coin operator .
here , @xmath53 is a function which adds the phase @xmath54 for odd @xmath23 and it can be defined as @xmath55/2 $ ] .
it is shown in ( [ eq : nline1 ] ) that the total state is periodically localized at opposite sites which agrees with the numerical results .
furthermore , the walk is periodic with a period of @xmath56 steps up to an overall phase .
after @xmath23 steps , we apply recovery operator @xmath57 for pst .
recovery operator is a function of @xmath16 and @xmath17 which means that for all coin operators with @xmath46 , there is always a corresponding recovery operator .
hence , step operator @xmath58 makes n - line 2-periodic for @xmath46 .
for 2-cycle , full evolution can simply be written in matrix form as @xmath59 in ( [ eq:2cycleevol ] ) , we see that , shift operator swaps the position of the walker independent of its coin state . at @xmath60 ,
the total state becomes ( [ eq : coineffect ] ) with @xmath61 .
since the coin operator is unitary , @xmath62 leads to pst after first step .
if we define one - step as @xmath63 , qw becomes 2-periodic and it keeps the initial coin state unchanged . in other words , the initial coin state bounces back and forth between two sites .
in contrast to 2-line , 2-cycle allows pst for all coin operators with the aid of appropriate recovery operators .
we note that , if we choose @xmath21 , without any recovery operator , the walk is naturally periodic with a period of @xmath40 steps for @xmath64 $ ] which generalizes the @xmath20 condition in @xcite .
a special case for pst on circles is the 4-cycle . in this case , we achieve pst by using well - known hadamard coin operator ( in @xmath31 steps ) or the biased coin operator @xmath65 ( in @xmath28 steps ) without any recovery operators . these results are given in table [ tb:4cycle ] and figure [ fig:4cycle ] .
we see that for each pst case the walk is also periodic . in @xcite
, it has been already shown that 4-cycle has a period of @xmath32 steps for @xmath20 .
we extend this result by showing that it also has a period of @xmath30 steps for @xmath65 and period of @xmath28 steps for @xmath37 .
@cccc @xmath14 & step(t ) & site(x ) & coin state + & @xmath28 & 3 & @xmath26 + & @xmath30 & 1 & @xmath26 + & @xmath31 & 3 & @xmath66 + & @xmath32 & 1 & @xmath26 + @xmath33&6&1&@xmath26 + , without any recovery operators .
black ( hollow ) dots indicate @xmath67 ( @xmath68 ) . ] now , we consider the n - cycles with even @xmath23 and @xmath69 . since the coin operator is diagonal , @xmath70 and @xmath71 terms do not mix , and generate propagations in opposite directions .
after @xmath72 steps , we find the walker at @xmath2 with probability @xmath25 .
we note that the coin operator adds the phase @xmath73 to the coefficient of @xmath71 in each step .
thus , after @xmath72 steps , the total state becomes @xmath74 where @xmath75 . without loss of generality
, one can choose @xmath76 and use the recovery operator @xmath77+\pi)$ ] to achieve pst .
the step operator @xmath78 makes the walk 2-periodic .
if @xmath23 is odd , wave function does not localize at any site except the initial one .
the total state after @xmath23 steps is @xmath79 appropriate choice for the recovery operator can be given as @xmath80 at @xmath81 .
the step operator @xmath58 makes the walk 1-periodic .
we have shown that the coin operator which is restricted to @xmath46 allows pst on n - cycles .
although , it has not been indicated in the discussion about n - cycles above , spatial approach has been used intrinsically , i.e. , clockwise rotations correspond to @xmath70 .
if the walk is driven by the coin operator @xmath82 ( the flip coin operator ) , we define the directions with the local approach for pst ( see figure [ fig : dirappr ] ) .
when @xmath23 is odd , n - cycle is ill - defined since we have to label at least two edges with the same basis state .
therefore , we consider n - cycles with even @xmath23 only . if we label all edges as in fig .
[ fig : dirappr](b ) , after @xmath72 steps , the total state becomes @xmath83 both case have the overall phase @xmath84 where @xmath85 is the floor function .
the first case shows that pst is achieved after @xmath72 steps .
also , it is clear that we can use @xmath86 to recover the second case and make the walk periodic .
however , n - lines do not have the same property , i.e. , pst is not possible if we use flip coin operator with local approach .
we can demonstrate this fact by evaluating the first two steps as follows : @xmath87 where @xmath88 involves the flip coin operator .
we see that after each step , the first term in the summation is stuck at site @xmath25 because of the periodicity condition . for an n - line , after @xmath23 steps
, the second term will be stuck at site @xmath23 as well .
thus , neither pst nor periodicity is possible .
we have proposed a pst scheme by introducing recovery operators in discrete - time qw architecture on n - lines and n - cycles .
we have shown that by using identity or flip coin operator , an unknown qubit state can be transferred to an arbitrary distance perfectly with the aid of appropriate recovery operator .
the 2-cycle is the only lattice which allows pst for all coin operators up to n=10 .
also , the hadamard coin and biased coin @xmath65 allow pst on 2-line .
we have shown that the 4-cycle is a special case where pst occurs if the walk is driven by the hadamard coin operator or the biased coin operator @xmath65 , without any recovery operators .
moreover , we have introduced new periodic discrete - time qws on n - lines and also extended periodicity cases which has already been known for 2-cycle and 4-cycle @xcite .
we have shown the strong relation between the periodicity and pst .
since recovery operators are just additional coin operators and pst occurs after small number of steps ( which is comparable with the lattice size ) , it seems that the experimental realization of our scheme is quite feasible with today s technology .
we would like to thank b. pekerten , g. karpat and b. akmak for helpful discussions .
this work has been partially supported by the scientific and technological research council of turkey ( tubitak ) under grant no . | we propose a scheme for perfect transfer of an unknown qubit state via the discrete - time quantum walk on a line or a circle . for this purpose
, we introduce an additional coin operator which is applied at the end of the walk .
this operator does not depend on the state to be transferred .
we show that perfect state transfer over an arbitrary distance can be achieved only if the walk is driven by an identity or a flip coin operator .
other biased coin operators and hadamard coin allow perfect state transfer over finite distances only .
furthermore , we show that quantum walks ending with a perfect state transfer are periodic .
january 2015 _ keywords _ : quantum walk , quantum state transfer , quantum state revival |
relativistic a+a collisions are an important source of experimental information about the qcd phase diagram and the strongly interacting matter properties . the last stage of such collisions is traditionally analyzed within the statistical approach which gives us an excellent opportunity to reveal the parameters of chemical freeze - out .
this approach is based on the assumption of the thermal equilibrium existence during the last stage of reaction .
such an equilibrium can be reached due to intensive particle scattering .
the stage of the system evolution when the inelastic reactions between hadrons stop is referred to as a chemical freeze - out ( fo ) .
particle yields are determined by the parameters of fo , namely by chemical potentials and temperature .
this general picture is a basis of the hadron resonance gas model ( hrgm ) @xcite which is the most successful one in describing the hadronic yields measured in heavy - ion experiments for energies from ags to lhc . despite a significant success of the hrgm in the experimental data analysis
there are a few unresolved problems . in general
they are related to the description of hadron yields which contain ( anti)strange quarks .
especially the energy dependence of @xmath5 and @xmath6 ratios was out of high quality description .
excess of strange hadrons yields within the hrgm led physical community to ponder over strangeness suppression .
the first receipt to resolve this problem was to introduce the strangeness suppression factor @xmath0 which should be fitted in order to describe the experimental data @xcite .
however , such an approach is not supported by any underlying physical model and the physical meaning of @xmath0 remains unclear @xcite .
in addition the strangeness suppression approach in its original form does not contain a hard - core repulsion between hadrons , while the latter is an important feature of the hrgm .
a significant role of the hard - core repulsion was demonstrated once more in refs . @xcite where the global fit of hadron yield ratios was essentially improved ( to @xmath1 1.2 ) compared to all previous analyses .
the most advanced way to account for the hard - core repulsion between hadrons is to consider a hadron gas as a multi - component mixture of particles with different radii @xcite . within this approach all baryons and mesons except for the kaons and the pions
are endowed by the common hard - core radii @xmath7 and @xmath8 , respectively . at the same time
the kaon and the pion radii @xmath9 and @xmath10 are fitted independently in order to provide the best description of @xmath5 ratio @xcite .
this is an important finding since the non - monotonic energy dependence of @xmath5 ratio may indicate some qualitative changes of the system properties and may serve as a signal of the deconfinement onset .
this is a reason why this ratio known as the strangeness horn is of a special interest .
note , that the multi - component approach substantially increased the strangeness horn description quality , without spoiling the other ratios including @xmath6 one .
however , even this advanced approach does not reproduce the topmost point of the strangeness horn indicating that the data description is still not ideal . in order to resolve this problem in ref .
@xcite the @xmath0 factor was considered as a free parameter within the hrgm with multi - component repulsion .
although the @xmath0 data fit improves the strangeness horn description quality sizably , it does not seem to be useful for the description of other hadron multiplicities @xcite .
furthermore , in contrast to the claims established on the low - quality fit @xcite , at low energies it was found @xcite that within the error bars in heavy ion collisions there is an enhancement of strangeness and not a suppression .
however , the effect of apparent strangeness non - equilibration can be more successfully explained by the hypothesis of separate chemical fo for all strange hadrons .
since all the hadrons made of @xmath11 and @xmath12 quarks are under thermal equilibration whereas the hadrons containing @xmath13 quark are not , then it is reasonable to assume two different fos for these two kinds of particles . following this conclusion in ref .
@xcite a separate strangeness fo ( sfo ) was introduced .
note , that according to @xcite both fo and sfo parameters are connected by the conservation laws of entropy , baryonic charge and isospin projection , while the net strangeness is explicitly set to zero at fo and at sfo .
these conservation laws are crucial elements of the concept of separate sfo developed in @xcite which allows one to essentially reduce the number of independent fitting parameters .
another principal element that differs the hrgm of @xcite from the ideal gas treatment used in @xcite is the presence of multi - component hard - core repulsion . using the hrgm of @xcite it was possible to successfully describe all hadron multiplicities measured in a+a collisions at ags , sps and rhic energies with @xmath1 1.06 .
the concept of separate sfo led to a systematic improvement of all experimental data description .
however , the topmost point of the strangeness horn again was not fitted even within the experimental error .
note , however , that the general description of @xmath14 ratio energy dependence was rather good except for the upper point .
since an introduction of the @xmath15 factor demonstrated a remarkable description of all points of the strangeness horn , whereas the separate sfo led to a systematic improvement of all hadron yields description , we decided to combine these elements in order to describe an experimental data with the highest possible quality .
this ambitious task is the main aim of the present paper .
in addition , the problem of residual strangeness non - equilibration should also be clarified due to its importance from the academic point of view .
evidently , the best tool for such a purpose is the most successful version of the hrgm , i.e. the hrgm with the multi - component hadronic repulsion and sfo . as it will be shown below , such an approach makes it possible to describe 111 hadron yield ratios measured for 14 values of the center of mass collision energy @xmath16 in the interval from 2.7 gev to 200 gev with the highest quality ever achieved .
the paper is organized as follows .
the basic features of the developed model are outlined in section 3 . in section 4
we present and discuss the new fit of hadronic multiplicity ratios with two chemical freeze - outs and @xmath0 factor , while section 5 contains our conclusions .
in what follows we treat a hadronic system as a multi - component boltzmann gas of hard spheres . the effects of quantum statistics are negligible for typical temperatures of the hadronic gas whereas the hard - core repulsion between the particles significantly affects a corresponding equation of state @xcite .
the present model is dealing with the grand canonical treatment .
hence a thermodynamical state of system under consideration is fixed by the volume @xmath17 , the temperature @xmath18 , the baryonic chemical potential @xmath19 , the strange chemical potential @xmath20 and the chemical potential of the isospin third component @xmath21 .
these parameters control the pressure @xmath22 of the system .
in addition they define the densities @xmath23 of corresponding charges @xmath24 ( @xmath25 ) . introducing the symmetric matrix of the second virial coefficients @xmath26 with the elements @xmath27
, we can obtain the parametric equation of state of the present model in a compact form @xmath28 the equation of state is written in terms of the solutions @xmath29 of the following system @xmath30^{-1 } } \biggr ] \
, , \quad \quad \\ & & \hspace*{-4mm}\phi_i ( t ) = \frac{g_i}{(2\pi)^3}\int \exp\left(-\frac{\sqrt{k^2+m_i^2}}{t } \right)d^3k \ , .
% \label{eqiv } % \end{aligned}\ ] ] it is worth to note , that quantities @xmath31 have a meaning of @xmath32 sort of hadrons partial pressure .
each @xmath32 sort is characterized by its full chemical potential @xmath33 , mass @xmath34 and degeneracy @xmath35 .
function @xmath36 denotes the corresponding particle thermal density in case of ideal gas . finally , the superscript @xmath18 here is the symbolic notation for operation of a column transposition which yields a row of quantities @xmath29 .
the obtained model parameters for two freeze - outs and their dependence on the collision energy are shown in figs .
[ fig : saguni]-[fig : saguniii ] . in order to account for the possible strangeness non - equilibration
we introduce the @xmath15 factor in a conventional way by replacing @xmath37 in eq .
( [ eqii ] ) as @xmath38 where @xmath39 is a number of strange valence quarks plus number of strange valence anti - quarks .
the principal difference of the present model from the traditional approaches is that we employ an independent chemical fo of strange particles .
let us consider this in some detail .
the independent freeze - out of strangeness means that inelastic reactions ( except for the decays ) with hadrons made of s quarks are switched off at the temperature @xmath40 , the baryonic chemical potential @xmath41 , the strange chemical potential @xmath42 , the isospin third projection chemical potential @xmath43 and the three - dimensional emission volume @xmath44 . in general case
these parameters of sfo do not coincide with the temperature @xmath40 , the chemical potentials @xmath41 , @xmath42 , @xmath43 and the volume @xmath44 which characterize the freeze - out of non - strange hadrons .
the particle yields are given by the charge density @xmath45 in ( 1 ) and the corresponding volume at fo and at sfo .
at the first glance a model with independent sfo contains four extra fitting parameters for each energy value compared to the traditional approach ( temperature , three chemical potentials and the volume at sfo instead of strangeness suppression / enhancement factor @xmath15 ) .
however , this is not the case due to the conservation laws . indeed , since the entropy , the baryonic charge and the isospin third projection are conserved , then the parameters of fo and sfo are connected by the following equations @xmath46 the effective volumes can be excluded , if these equations are rewritten as @xmath47 thus , the baryonic @xmath41 and the isospin third projection @xmath43 chemical potentials at sfo are now defined by eqs .
( [ eq : fo_sfo1 ] ) .
note , that the strange chemical potentials @xmath48 and @xmath42 are found from the condition of vanishing net strangeness at fo and sfo , respectively .
therefore , the concept of independent sfo leads to an appearance of one independently fitting parameter @xmath40 .
hence , the independent fitting parameters are the following : the baryonic chemical potential @xmath19 , the chemical potential of the third projection of isospin @xmath21 , the chemical freeze - out temperature for strange hadrons @xmath40 , the chemical freeze - out temperature for all non - strange hadrons @xmath49 and the @xmath15 factor .
an inclusion of the width @xmath50 of hadronic states is an important element of the present model .
it is due to the fact that the thermodynamical properties of the hadronic system are sensitive to the width @xcite . in order to account for the finite width of resonances we perform the usual modification of the thermal particle density @xmath37 .
namely , we convolute the boltzmann exponent under the integral over momentum with the normalized breit - wigner mass distribution . as a result , the modified thermal particle density of @xmath32 sort hadron acquires the form @xmath51 here @xmath34 denotes the mean mass of hadron and @xmath52 stands for the threshold in the dominant decay channel .
the main advantages of this approximation is a simplicity of its realization and a clear way to account for the finite width of hadrons .
the observed hadronic multiplicities contain the thermal and decay contributions .
for example , a large part of pions is produced by the decays of heavier hadrons .
therefore , the total multiplicity is obtained as a sum of thermal and decay multiplicities , exactly as it is done in a conventional model . however , writing the formula for final particle densities , we have to take into account that volumes of fo and sfo can be different : @xmath53 here the first term on the right hand side is due to decays after fo whereas the second one accounts for the strange resonances decayed after sfo .
the factor @xmath54 can be replaced by @xmath55 due to baryonic charge conservation .
@xmath56 denotes the branching ratio of the y - th hadron decay into the x - th hadron , with the definition @xmath57 = 1 used for the sake of convenience .
the input parameters of the present model ( masses @xmath34 , widths @xmath50 , degeneracies @xmath35 and branching ratios of all strong decays ) were taken from the particle tables of the thermodynamical code thermus @xcite .
* data sets and fit procedure*. the present model is applied to fit the data .
we take the ratios of particle multiplicities at midrapidity as the data points .
in contrast to fitting multiplicities themselves such an approach allows us to cancel the possible experimental biases . in this paper
we use the data set almost identical to ref . @xcite . at the ags energies ( @xmath58 agev or @xmath59 agev ) the data are available with a good energy resolution above 2 agev . however , for the beam energies 2 , 4 , 6 and 8 agev only a few data points are available .
they corresponds to the yields for pions @xcite , for protons @xcite , for kaons @xcite ( expect for 2 agev ) .
the integrated over @xmath60 data are also available for @xmath61 hyperons @xcite and for @xmath62 hyperons ( for 6 agev only ) @xcite . however , as was argued in ref .
@xcite , the data for @xmath61 and @xmath62 should be recalculated for midrapidity .
therefore , instead of raw experimental data we used the corrected values from @xcite . next comes the data set at the highest ags energy ( @xmath63 agev or @xmath64 agev ) .
similarly to @xcite , here we analyzed only the na49 mid - rapidity data @xcite .
since the rhic high energy data of different collaborations agree with each other , we analyzed the star results for @xmath65 gev @xcite , @xmath66 gev @xcite , @xmath67 gev @xcite and 200 gev @xcite .
the criterion to define the fitting parameters of the present model is a minimization of @xmath68 , where @xmath69 and @xmath70 are , respectively , the theoretical and the experimental values of particle yields ratios , @xmath71 stands for the corresponding experimental error and a summation is performed over all available experimental points .
* combined fit with sfo and @xmath0 factor*. recently performed comprehensive data analysis @xcite for two alternative approaches , i.e the first one with @xmath0 as a free parameter and the second one with separate fo and sfo , showed the advantages and disadvantages of both methods .
thus , the @xmath0 fit provides one with an opportunity to noticeably improve the strangeness horn description with @xmath72 , comparably to the previous result @xmath73 @xcite , but there are only slight improvements of the ratios with strange baryons ( global @xmath74 ) .
the obtained results for the sfo approach demonstrate a nice fit quality for the most problematic ratios for the hrgm , especially for @xmath75 , @xmath76 , @xmath77 and @xmath78 .
although the overall @xmath79 is notably better than with the @xmath0 factor @xcite , but the highest point fitting of the horn got worse .
these results led us to an idea to investigate the combination of these two approaches in order to get the high - quality strangeness horn description without spoiling the quality of other particle ratios . for 14 values of collision energy
@xmath80 2.7 , 3.3 , 3.8 , 4.3 , 4.9 , 6.3 , 7.6 , 8.8 , 9.2 , 12 , 17 , 62.4 , 130 , 200 gev the best description with two separate freeze - outs and the @xmath15 fit gives @xmath81 = 42.96/41 @xmath82 1.05 , which is only a very slight improvement compared to the previously obtained results @xmath81 = 58.5/55@xmath82 1.06 found for two freeze - outs ( strange and all other particles ) , which are connected by the conservation laws .
note , however , that the value of @xmath83 itself , not divided by number of degrees of freedom , has improved notably , although the deviation of the @xmath15 factor from 1 does not exceed 27 % ( see fig . [
fig : saguniii ] ) .
these findings motivate us to study what ratios and at what energies are improved .
as we mentioned earlier , at each collision energy there are five independent fitting parameters in the considered model with the simultaneous sfo and the @xmath0 fit , while for some collision energies the number of experimental ratios is lower or equal to the number of parameters .
for example , for the energies @xmath16 = 2.7 , 3.3 , 3.8 , 4.3 , 9.2 , 62.4 gev the number of available ratios is small ( 4 , 5 , 5 , 5 , 5 , 5 , respectively ) from which only kaons and @xmath61 contain strange quarks .
therefore , for these energies we obtained a perfect data description since we had to solve the above equations . as a result for these energies
the relative deviation of the fit is almost a zero , but it gives us somewhat larger uncertainties for the fitting parameters .
for the energies @xmath80 17 , 130 gev we observed that the resulting fit quality became better compared to the work @xcite .
the most significant improvements correspond to the collision energies @xmath80 6.3 , 7.6 , and 12 gev , that are plotted in fig .
[ fig : sagunv ] . fig .
[ fig : sagunv ] demonstrates very nice fit quality , especially for such traditionally problematic ratios as @xmath14 , @xmath84 , @xmath85 and @xmath86 . for @xmath80 7.6
gev the seven ratios out of ten are improved .
a special attention in our consideration was paid to the strangeness horn , i.e. to @xmath14 ratio .
another reason for a through study of the strangeness horn is a traditional problem of the hrgm to fit it .
as one can see from fig .
[ fig : saguniv ] , the remarkable @xmath14 fit improvement for @xmath80 2.7 , 3.3 , 4.3 , 4.9 , 6.3 , 7.6 , 12 gev justifies the usage of the present model .
quantitatively , we found that @xmath81 improvement due to sfo+@xmath15 introduction is @xmath81=1.5/14 , i.e. even better than it was done in @xcite with @xmath81=3.3/14 for the @xmath15 fitting approach and @xmath81=6.3/14 for sfo and @xmath87 .
in addition , in fig . [
fig : sagunvi ] we give the @xmath88 and @xmath89 ratios to show that two separate freeze - outs inclusion with the @xmath15 fit still does not improve these ratios .
the @xmath88 fit quality , for instance , is ( @xmath81=10/8 ) .
hence , up to now the best fit of the @xmath88 ratio was obtained within the sfo approach with @xmath87 . as it was mentioned in @xcite
a too slow decrease of model results for @xmath88 ratio compared to the experimental data is typical for almost all statistical models .
evidently , the too steep rise in @xmath88 behavior is a consequence of the @xmath90 anomaly @xcite .
similar results are reported in refs .
@xcite as the @xmath2 , @xmath3 and @xmath4 selective suppression .
since even an introduction of the separate strangeness freeze - out with the strangeness enhancement factor does not allow us to better describe these ratios , we believe that there is an unclarified physical reason which is responsible for it . within the present model we also found a selective improvement and a certain degradation of the fit quality of various ratios for different collision energies .
for instance , the @xmath84 ratio is slightly increased for @xmath80 6.3 and 7.6 gev , but the situation drastically changes for @xmath80 12 gev .
the same tendency is typical for @xmath91 . on the contrary , for @xmath92 ratio
there is a noticeably worse data description within sfo+@xmath15 approach at @xmath80 6.3 , 7.6 gev , but for @xmath80 12 gev the fit quality is sizably better compared to all previous approaches .
thus , within the present model we reveal a noticeable change in the trend of some ratios at @xmath80 7.6 - 12 gev .
we have performed an elaborate fit of the data measured at ags , sps and rhic energies within the multi - component hadron resonance gas model . the suggested approach
to separately treat the freeze - outs of strange and non - strange hadrons with the simultaneous @xmath15 fitting gives rise for the top - notch strangeness horn description with @xmath81=1.5/14 .
the developed model clearly demonstrates that the successful fit of hadronic multiplicities includes all the advantages of these two approaches discussed in @xcite . as a result for @xmath80 6.3 , 7.6 , 12 ,
130 gev we found a significant data fit quality improvement .
the achieved total value of @xmath81 is @xmath93 1.05 , while the @xmath15 values are consistent with the conclusion @xmath94 ( within the error bars ) .
a possible exception is the topmost point of the strangeness horn , at which the mean value of the strangeness enhancement factor is @xmath95 , but with the large error bars .
in addition , the description of ratios containing the non - strange particles , especially such as @xmath84 and @xmath91 , gets better compared to previously reported results @xcite . at the same time
the lack of available data at @xmath16 = 2.7 , 3.3 , 3.8 , 4.3 , 9.2 , 62.4 gev forced us to solve the corresponding equations which in combination with the large experimental error bars led to rather large uncertainties of the fitting parameters . from a significant improvement of the data description
we conclude that the concept of separate chemical freeze - out of strange particles is an essential part of heavy ion collision phenomenology which should be taken into account in further studies of strongly interacting matter properties . however , the remaining problem with @xmath2 , @xmath96 , @xmath4 ratios led us to a conclusion that there is an unclarified physical reason which is responsible for them .
the residual non - equilibration of strange particles found here seems to be weak and , perhaps , the better experimental data will help us to reduce it further .
the obtained description of the hadron multiplicity ratios reached the highest quality ever achieved and this fact demonstrates that the suggested approach is almost a precise tool to elucidate the thermodynamics properties of hadron matter at two chemical freeze - outs . the fresh illustrations to this statement
can be found in @xcite . however , to get more reliable conclusions from this approach we need more experimental data with an essentially higher accuracy .
* acknowledgments .
* we would like to thank a. andronic for providing an access to well - structured experimental data .
the authors are thankful to i. n. mishustin , d. h. rischke and l. m. satarov for valuable comments .
, a.i.i . and g.m.z .
acknowledge a support of the fundamental research state fund of ukraine , project no f58/04 .
k.a.b . acknowledges also a partial support provided by the helmholtz international center for fair within the framework of the loewe program launched by the state of hesse . | we propose an elaborate version of the hadron resonance gas model with the combined treatment of separate chemical freeze - outs for strange and non - strange hadrons and with an additional @xmath0 factor which accounts for the remaining strange particle non - equilibration .
two sets of chemical freeze - outs parameters are connected by the conservation laws of entropy , baryonic charge , isospin projection and strangeness .
the developed approach enables us to perform a high - quality fit of the hadron multiplicity ratios for ags , sps and rhic energies with total @xmath1 1.05 .
a special attention is paid to a complete description of the strangeness horn .
a well - known @xmath2 , @xmath3 and @xmath4 selective suppression problem is also discussed . |
* theoretical framework for a minimal signaling circuit . * to obtain the central results , we start with an example which illustrates the efficacy of the wk theory , and suggests a way to a more detailed , realistic model of the enzymatic push - pull loop . consider a small portion of a signaling pathway [ fig .
[ f1]c ] , involving two chemical species : one with time - varying population @xmath0 ( the `` input '' ) , and another one with population @xmath1 ( the `` output '' ) whose production depends on @xmath0 .
these could be , for example , the active , phosphorylated forms of two kinases within a signaling cascade , with @xmath2 downstream of @xmath3 .
the upstream part of the pathway contributes an effective production rate @xmath4 for species @xmath3 , which in general can be time - dependent , though for now we will make @xmath4 constant .
the output @xmath2 is produced by a reaction , @xmath5 , with a rate @xmath6 that depends on the input .
the species are deactivated with respective rates @xmath7 and @xmath8 , mimicking the role of the phosphatases .
the input @xmath0 will vary over a characteristic time scale @xmath9 , fluctuating around the mean @xmath10 .
the output deactivation rate sets the response time scale @xmath11 over which @xmath1 can react to changes in the input .
the dynamical equations , within a continuum , chemical langevin ( cl ) description @xcite , are given by : @xmath12 where the additive noise contribution @xmath13 , with @xmath14 and @xmath15 denoting the mean of population @xmath16 .
the function @xmath17 is gaussian white noise with correlation @xmath18 .
the @xmath19 brackets denote an average over the ensemble of all possible noise realizations . for small deviations @xmath20 from the mean populations @xmath15 , eq .
can be solved using a linear approximation , where we expand the rate function to first order , @xmath21 , with coefficients @xmath22 , @xmath23 .
( we will return later to the issues of nonlinearity and discrete populations . )
the result is : @xmath24,\nonumber\end{aligned}\ ] ] where in the second line we have introduced an arbitrary scaling factor @xmath25 ( to be defined below ) inside the brackets , and divided through by @xmath26 outside the brackets .
the solution for @xmath27 has the structure of a linear noise filter equation : @xmath28 , with @xmath29 . in this analogy
, we have a signal @xmath30 together with a noise term @xmath31 forming a corrupted signal @xmath32 .
the output @xmath33 is produced by convolving @xmath32 with a linear filter kernel @xmath34 . as a consequence of causality , the integrals in eq .
run over @xmath35 , so the filtered output @xmath36 at any time @xmath37 depends only on @xmath38 from the past .
the utility of mapping the push - pull system onto a noise filter comes from the application of wk theory , which is designed to solve a key optimization problem : out of all possible causal , linear filters @xmath39 , what is the optimal function @xmath40 that minimizes the differences between the output @xmath36 and input @xmath41 time series . in our example , this means having @xmath27 reproduce as accurately as possible the scaled input signal @xmath42 .
specifically , we would like to minimize the relative mean - squared error @xmath43 . for a particular @xmath44 and @xmath27 , the value of @xmath45 is smallest when @xmath46 , which we will use to define the gain @xmath26 . in this case
@xmath45 reduces to @xmath47 .
the great achievement of wiener @xcite and kolmogorov @xcite was to show that @xmath48 satisfies the following wiener - hopf equation : @xmath49 where @xmath50 is the correlation between points in time series @xmath51 and @xmath52 , assumed to depend only on the time difference @xmath37 . given @xmath53 and @xmath54 , which are properties of the signal @xmath41 and noise @xmath55 , it is possible to solve eq . for @xmath48 .
the corresponding minimum value of the error @xmath45 is : @xmath56 the solution of the wiener - hopf equation requires the following correlation functions , which can be derived from eq . : @xmath57 , @xmath58 , and @xmath59 , where the parameter @xmath60 . plugging these into eq .
, we can solve for the optimal filter function by assuming a generic ansatz @xmath61 , finding the unknown coefficients @xmath62 and rate constants @xmath63 by comparing the left and right sides of the equation . in our case , a single exponential ( @xmath64 ) is sufficient to exactly satisfy eq .
( see details in appendix [ a : wk ] ) , and we get @xmath65 .
the conditions for achieving wk optimality , @xmath66 , are then : @xmath67 from eq .
the minimum relative error is : @xmath68 the fidelity between output and input is described through a single dimensionless optimality control parameter , @xmath69
. it can be broken up into two multiplicative factors , reflecting two physical contributions : @xmath70 .
the first term , @xmath71 , is a burst factor , measuring the mean number of output molecules produced per input molecule during the active lifetime of the input molecule .
the second term , @xmath72 , is a sensitivity factor , reflecting the local response of the production function @xmath73 near @xmath74 ( controlled by the slope @xmath75 ) relative to the production rate per input molecule @xmath76 .
note that @xmath77 only if @xmath73 is globally nonlinear , since physical production functions satisfy @xmath78 for all @xmath79 . if @xmath73 is perfectly linear , @xmath80 , then @xmath81 , and @xmath82
. thus the limit of efficient noise suppression , @xmath83 , where @xmath84 becomes small , can be achieved by making the burst factor @xmath85 and/or enhancing the sensitivity @xmath86 , at the cost of introducing nonlinear effects ( discussed in detail below ) . for optimality to be realized
, we additionally need an appropriate separation of scales [ eq . ] between the characteristic time of variations in the input signal , @xmath9 , and the response time of the output , @xmath11 .
the latter should be faster by a factor of @xmath87 .
the scaling @xmath88 for large @xmath69 is the same as the burst factor scaling of the target population variance in biochemical negative feedback networks intended to maintain homeostasis and suppress fluctuations @xcite .
the slow @xmath89 decay in both cases , compared to the more typical scaling of variance with @xmath90 ( inversely proportional to the number of signaling molecules produced ) reflects the same underlying physical challenge : the difficulty of suppressing or filtering noise in stochastic reaction networks .
the error @xmath45 defined above is based on the instantaneous difference between the input @xmath41 and output @xmath36 time series .
one of the powerful features of the wk formalism is that it naturally extends error minimization to cases where the goal is extrapolating the future signal , where we seek to minimize the difference between @xmath36 and @xmath91 for some @xmath92 @xcite .
given the time delays inherent in many biological responses , particularly where feedback is involved , such predictive noise filtering has significant applications @xcite , which we will explore in subsequent work . for now
, we confine ourselves to the instantaneous error , which is sufficient to treat the kinase - phosphatase push - pull loop .
we also note that there is no unique measure of signal fidelity .
besides @xmath45 , one can optimize the mutual information between the output and input species in the cascade @xcite .
for example , in the two - component cascade with nonlinear regulation , considered below , a spectral expansion of the master equation allows for efficient numerical optimization of the system parameters for particular forms of the rate function , maximizing the mutual information @xcite .
* effects of nonlinearity and discrete populations .
* for the subclass of gaussian - distributed signal @xmath41 and noise @xmath55 time series ( as is the case within the cl picture ) , the wk filter derived above , based on the linearization of the cl , is optimal among all possible linear or nonlinear filters @xcite .
if the system fluctuates around a single stable state , and the copy numbers of the species are large enough that their poisson distributions converge to gaussians ( mean populations @xmath93 ) , the signal and noise are usually approximately gaussian
. however , the rate function @xmath73 will never be perfectly linear in practice , and thus one needs to consider how nonlinearities in @xmath73 will affect the minimal @xmath45 .
in addition , the discrete nature of population changes , which becomes important at lower copy numbers , has to be explicitly taken into consideration .
surprisingly , the wk result of eq . can be generalized even to cases where the linear , continuum assumptions underlying wk theory no longer hold .
starting from the exact master equation , valid for discrete populations and arbitrary @xmath73 , we have rigorously solved the general optimization problem for the error @xmath45 between output and input using the principles of umbral calculus @xcite .
the detailed proof is in appendix [ a : non ] , but the main results are as follows .
any function @xmath73 can be expanded in terms of a set of polynomials @xmath94 as @xmath95 .
the @xmath94 are polynomials of degree @xmath96 , given by @xmath97 and the coefficients @xmath98 are related to moments of @xmath73 , @xmath99 .
the average is taken with respect to the poisson distribution @xmath100 .
the first two polynomials are @xmath101 and @xmath102 , giving @xmath103 and @xmath104 .
remarkably , the relative error @xmath45 has an exact analytical form in terms of the @xmath98 , @xmath105^{-1}.\ ] ] this expression is bounded from below by @xmath106 where @xmath107 .
the equality is only reached when @xmath108 and @xmath73 has an optimal linear form , @xmath109 , with all @xmath110 for @xmath111 . in this optimal case , @xmath112 and @xmath113 , and hence @xmath114 , @xmath115 from eq . .
a : numerical optimization results for the hill production function @xmath73 that minimizes relative error @xmath45 between input and output , with each color corresponding to different values of the parameter @xmath116 ( see text for other parameters ) .
the input probability distribution @xmath117 is superimposed in black ( the height scale is arbitrary ) .
b : for each value of @xmath118 from panel a , circles show the minimal @xmath45 .
the lower bound @xmath119 [ eq . ] is drawn as a blue curve .
c : analogous to panel b , but showing the ratio @xmath120 at which the minimum @xmath45 is achieved .
the blue curve shows the wk prediction for this ratio , @xmath121 . ]
making @xmath122 large , for example by increasing @xmath123 , is desirable for better signal transduction , but with a caveat .
we can keep @xmath45 near @xmath119 even for a globally nonlinear @xmath73 so long as @xmath73 remains approximately linear in the vicinity of the mean @xmath74 , and the nonlinear corrections @xmath98 for @xmath124 are negligible .
large @xmath123 can be achieved through a highly sigmoidal input - output response , known as ultrasensitivity , which is biologically realizable in certain regimes of signaling cascades @xcite .
however , our theory predicts that as @xmath73 goes to the extreme limit of a step - like profile around @xmath74 , @xmath45 should become significantly higher than @xmath119 , and the benefits of ultrasensitivity vanish .
the reason for this is that letting @xmath123 become arbitrarily large ( making the step sharper ) necessarily implies that @xmath73 eventually deviates substantially from @xmath125 .
we know that any physically sensible @xmath73 satisfies the constraint @xmath78 for @xmath126 .
if @xmath127 and @xmath128 for @xmath124 , the function @xmath73 would be negative for @xmath129 , violating the physical constraint .
hence the coefficients @xmath98 for @xmath124 must be non - negligible when @xmath123 is sufficiently large , leading to @xmath130 .
we can illustrate this result numerically for @xmath73 that have the form of a hill function , @xmath131 , defined by the three parameters @xmath132 , @xmath133 , and @xmath134 .
this represents a typical sigmoidal behavior in biochemical systems , with a small production rate for @xmath135 switching over to a saturation level @xmath132 for @xmath136 .
we performed a numerical minimization of @xmath45 ( evaluated using eq . ) over the parameter space , at fixed @xmath4 , @xmath7 , @xmath137 , and @xmath118 . using eq .
is numerically extremely efficient , since the coefficients @xmath98 typically decay quite rapidly , allowing the infinite sum to converge after a small ( @xmath138 ) number of terms .
fixing @xmath137 and @xmath118 is equivalent to specifying the first two moments of @xmath73 , which in turn defines a curve in the three - dimensional parameter space of @xmath132 , @xmath133 , and @xmath134 . after numerically solving for this curve ,
the minimization procedure consists of searching along the curve ( and varying the free system parameter @xmath8 ) to find the parameter set that yields the smallest @xmath45 .
[ f2]a shows optimization results for @xmath139 s@xmath140 , @xmath141 s@xmath140 , @xmath142 s@xmath140 , and varying @xmath118 , with the optimal hill function @xmath73 ( the one with smallest @xmath45 ) at each @xmath118 drawn in a different color .
the corresponding minimal values of @xmath45 are shown in fig .
[ f2]b as circles in the same colors , with @xmath119 using eq . drawn as a blue curve for comparison .
larger values of @xmath118 have optimal @xmath73 profiles that are increasingly step - like , with steeper slopes near @xmath74 . for the range @xmath143 the maximum slope ( @xmath144 )
is still small enough that @xmath73 remains approximately linear across the entire @xmath3 range where @xmath117 is non - negligible ( the distribution is superimposed in fig . [ f2]a ) .
hence minimal @xmath45 values are very close to @xmath119 , decreasing with @xmath118 .
the ratios @xmath120 at which these minimal @xmath45 values occur , shown in fig . [ f2]c , are nearly equal to the predicted value @xmath145 ( blue curve ) .
we can estimate that this near - optimality will persist up to @xmath146 , since that is roughly the slope of an @xmath73 that rises from zero near the left edge of @xmath147 ( at @xmath148 ) to a value of @xmath137 at @xmath149 . for @xmath150 , or equivalently @xmath151 ,
the nonlinearity of @xmath73 becomes appreciable around @xmath74 , distorting the output signal and leading to minimal @xmath45 noticeably larger than @xmath119 , and actually increasing with @xmath118 .
thus moving towards the ultrasensitive limit @xmath152 is initially beneficial for noise filtering , but only up to a point : @xmath73 does not have to be globally linear , but local linearity of @xmath73 near @xmath74 , which can be satisfied readily , is best for accurate signal transduction . * enzymatic push - pull loop can act as an optimal wk filter .
* the system considered so far is the simplest realization of a signaling circuit , in the sense that it involves only two species , related through a single phenomenological production function , @xmath73 . in reality ,
an enzymatic push - pull loop involves intermediates complexes of the substrate with the kinase or phosphatase whose binding , unbinding , and catalytic reactions all contribute to the stochastic nature of signal transmission .
can the wk theoretical framework be used to describe optimality in this complicated context ?
let us consider a more microscopic model of the loop reaction network [ fig .
the active kinase is either free ( @xmath153 ) or bound to substrate ( @xmath154 ) .
the input @xmath3 is defined as the total active kinase population @xmath155 .
upstream modules control kinase activation and deactivation , described by rates @xmath4 and @xmath156 respectively .
the kinase can phosphorylate the substrate , converting it from inactive ( @xmath157 ) to active ( @xmath158 ) form .
analogously , in the reverse direction , free phosphatases ( @xmath159 ) form complexes with the active substrate ( @xmath160 ) , which lead to dephosphorylation , returning the substrate to inactive form .
the output @xmath2 is the total active substrate population @xmath161 .
the reactions for substrate modification , with corresponding rate constants , are : @xmath162{\kappa_\text{b } } s_k \xrightarrow{\kappa_\text{r } } k + s^\ast\\ p + s^\ast & \xrightleftharpoons[\rho_\text{u}]{\rho_\text{b } } s^\ast_p \xrightarrow{\rho_\text{r } } p + s. \end{split}\ ] ] we chose representative rate values based on a model of the map kinase cascade @xcite ( all units are in s@xmath140 ) : @xmath163 , @xmath164 , @xmath165 , @xmath166 , @xmath167 , @xmath168 .
the rate @xmath156 in the model controls the characteristic time scale over which the input signal varies .
we let @xmath169 s@xmath140 , which sets this scale to minutes .
mean free substrate and phosphatase populations ( which together with the rates determine all equilibrium population values ) are in the ranges : @xmath170 , @xmath171 molecules / cell .
we simulated the dynamics of this system numerically using kinetic monte carlo ( kmc ) @xcite , with sample input and output trajectories shown in fig .
[ f3]d - f for @xmath172 and three values of @xmath173 .
as the free phosphatase population is varied , we see different degrees of signal fidelity , with the closest match between @xmath27 and @xmath174 for the intermediate case in fig .
are we seeing behavior similar to an optimal wk filter ?
as detailed in appendix [ a : map ] , we can approximately map the phosphorylation cycle to a noise filter using the same method as in our first example : starting from the full dynamical equations in the linear cl approximation , we derive the correlation functions required to solve the wiener - hopf relation , eq . .
the effective parameters resulting from the mapping are : @xmath175 where @xmath176 , @xmath177 , @xmath178 , @xmath179 , @xmath180 , and @xmath181 .
. is valid in the regime @xmath182 , with corrections of order @xmath183 and @xmath184 shown in appendix [ a : map ] .
such a mapping allows us to use wk results in eqs . and
to predict the conditions for optimality and the minimal possible @xmath45 .
[ f3]a - b show the left ( solid lines ) and right - hand ( dashed lines ) sides of both conditions in eq . as a function of @xmath173 for @xmath185 ( @xmath172 ) .
the @xmath173 value at the intersections , where the conditions are fulfilled , is marked by a diamond .
[ f3]c shows that exactly at this value @xmath45 achieves a minimum , given by @xmath84 from eq .
( dashed line ) . the cl approximation ( solid curves ) and kmc simulations ( circles ) are in excellent agreement .
thus , the phosphorylation cycle can indeed be tuned to behave like an optimal wk noise filter , even for a realistic signaling model . in light of the mapping in eq .
, we can now understand the behavior of the trajectories in fig .
[ f3]d - f , which correspond to @xmath185 . in panel d , where @xmath186 , we have @xmath187 [ fig .
[ f3]a ] , and the output @xmath27 becomes excessively smooth , since it can not respond quickly enough to changes in the input signal @xmath174 .
the corresponding power spectral density ( psd ) of the output , shown in panel g , is smaller at high frequencies compared to the psd of the input . in panel f
, we have the opposite situation of @xmath188 at @xmath189 .
the output response is too rapid , generating additional noise that obscures the signal . in this case
the output psd ( panel i ) has an extra high frequency contribution relative to the input psd .
panel e represents the optimal intermediate @xmath190 , where @xmath191 and the wk conditions are fulfilled .
the input and output psds ( panel h ) are similar at all frequencies . the minimum of @xmath45 in fig .
[ f3]c is shallow , meaning that near - optimal filtering persists even when the phosphatase population is not precisely tuned to the wk condition . for @xmath173 values that vary nearly five - fold between @xmath192 ,
the error @xmath45 remains within 5% of the minimum value @xmath84 .
another aspect of the filter s robustness can be highlighted by perturbing the enzymatic parameters @xmath193 , @xmath194 , @xmath195 , @xmath196 , @xmath197 , and @xmath198 .
if we randomly vary all these parameters within a range between 0.1 and 10 times the values listed above after eq . , and calculate the resulting conditions for wk optimality [ eq . ] for each new parameter set , we obtain the results in fig .
[ f4 ] . for a given @xmath173 , the shaded intervals in the figure correspond to the 68% confidence intervals on the input kinase frequency scale @xmath156 and the mean substrate population @xmath199 at optimality .
thus , for a broad range of biologically relevant enzymatic parameters , we get a sense of how the populations of @xmath173 and @xmath199 must complement each other , and an associated time scale @xmath200 reflecting how quickly the input signal can vary and still be accurately transduced . from the trends in fig .
[ f4 ] , we see that to get the system to respond to more rapidly varying signals , we need larger populations of @xmath173 and @xmath199 . as a concrete example , for the hyperosmolar glycerol ( hog ) signaling pathway in yeast , discussed further in the next section , kinase substrates have cell copy numbers of between @xmath201 and @xmath202 , while the ptp and ptc phosphatases that have been identified as targeting the pathway are present in cell copy numbers between @xmath203 and @xmath204 @xcite . using these population scales as a rough guide for @xmath199 and @xmath173 ( ignoring complications like multiple phosphorylation steps and sharing of phosphatases between different pathways )
we see from fig .
[ f4 ] that the corresponding @xmath205 s@xmath140 .
this range of optimal time scales is consistent with the experimental observation that the hog pathway can faithfully transduce osmolyte signals at frequencies @xmath206 s@xmath140 @xcite . and
the characteristic frequency scale @xmath156 over which the active kinase input signal varies .
this is at wk optimality [ eq . ] , using the mapping of eq . and the parameter values @xmath193 ,
@xmath194 , @xmath195 , @xmath196 , @xmath197 , @xmath198 listed in the text after eq . .
the shaded region between the dashed curves shows the 68% confidence interval for achieving wk optimality , resulting from randomly perturbing all the parameter values so that they can be up to ten - fold smaller or larger .
b : analogous to panel a , but showing the relation between @xmath173 and substrate population @xmath199 at wk optimality . ]
( solid blue ) and @xmath1 ( solid purple ) in a @xmath207 system driven by an oscillatory upstream flux @xmath208 ( see text for parameters ) .
dashed lines are local means @xmath209 and @xmath210 .
b : for trajectories in a , the deviations from local means , @xmath211 ( blue ) and @xmath212 ( purple ) .
c - e : results calculated from kmc for a system with @xmath207 and oscillatory @xmath208 at varying driving periods @xmath213 .
@xmath9 is marked by a vertical dashed line .
c : the minimum errors @xmath45 ( circles ) and @xmath214 ( squares ) .
@xmath84 is marked by a horizontal dashed line .
d : the minimum local coefficient of variation @xmath215 .
e : the mean phosphatase population values @xmath173 at which the minima shown in panels c and d are achieved ( @xmath45 : circles , @xmath214 : squares , @xmath216 : crosses ) .
the @xmath173 value for wk optimality is marked by a horizontal dashed line . ] * noise filtration in a push - pull loop driven by oscillatory input .
* remarkably , since eq .
is independent of @xmath4 , the system can serve as an optimal filter for a range of @xmath4 values , so long as the condition @xmath217 is satisfied .
this regime , involving saturated kinases and unsaturated phosphatases , has been previously identified as a candidate for efficient signal transmission by gomez - uribe _ et al . _
@xcite . to check the filter operation with varying upstream flux
, we used a time - dependent @xmath208 , driving the system with oscillatory input .
this is motivated by microfluidic experimental setups @xcite , where the hog pathway of yeast was probed by exposing the cells to periodic osmolyte pulses . in the experiments ,
the input signal is the extracellular osmolyte concentration , and the output is the degree to which the activated kinase hog1 localizes in the nucleus , where it initiates a transcriptional response to the osmolar shock .
though the biochemical network relating the output to input consists of a complex series of enzymatic push - pull loops , the overall behavior was quantified through response functions in terms of input signal frequency @xcite , related to the fourier transforms of the input - output correlation functions .
such correlation functions are the basic ingredients in assessing filter optimality in the wk theory . here
, we will focus only on a single push - pull loop , and use input at varying frequencies to determine whether @xmath84 remains a meaningful constraint on filter performance even for non - stationary signals . in fig .
[ f5]a we show a sample @xmath0 and @xmath1 kmc trajectory at optimality for @xmath218 , with @xmath219 s@xmath140 , @xmath220 , and @xmath221 s. the input has two characteristic time scales , @xmath213 and @xmath222 s. for @xmath223 , we define relative error in terms of deviations from local , time - dependent means : @xmath214 defined using @xmath224 and @xmath225 [ fig .
[ f5]b ] , where @xmath226 , @xmath227 are shown as dashed curves in fig .
[ f5]c shows kmc results for minimum @xmath45 and minimum @xmath214 as a function of @xmath213 for a system tuned to optimality with @xmath207 .
the values of @xmath173 at which these minima are achieved are shown in fig .
[ f5]e . at @xmath228
we find @xmath229 , since both the input and output have time to adjust to the slowly varying local means .
in fact , the minimum @xmath214 approaches @xmath84 for @xmath223 , as optimality is unaffected by the slow oscillation in @xmath208 .
the @xmath173 where the minimum @xmath214 occurs also approaches the value predicted by wk theory ( fig .
the filter transduces the signal with high fidelity . in the opposite limit of small @xmath230 ,
the rapidly varying @xmath208 essentially averages out , since neither the input nor the output have time to respond to the sharp changes in @xmath208 .
thus the system sees an effective constant flux @xmath231 .
here @xmath45 , the error estimate with respect to the global mean , is more relevant than @xmath214 . in this regime
the minimum @xmath232 , @xmath45 approaches @xmath84 for @xmath233 , and the @xmath173 value where @xmath45 is minimized agrees with the wk prediction .
the two regimes in system behavior , with a changeover at the time scale @xmath9 , reflect the fact that the enzymatic loop acts an effective low - pass filter @xcite : it can accurately transmit the low frequency component of @xmath208 , but integrates over the high - frequency portion above a certain bandwidth
. the overall bandwidth of a cascade of push - pull loops has been experimentally characterized for the yeast hog pathway , yielding a value of @xmath234 s@xmath140 @xcite . using this as a rough estimate of the bandwidth scale @xmath7 in individual loops
, we could expect to see a changeover between the two regimes depending on whether the driving frequency is much slower or faster than @xmath235 .
regardless of the magnitude of the driving frequency , both @xmath45 and @xmath214 always remain greater than @xmath84 , so the latter remains a bound on noise filter efficiency even for dynamic input .
more generally , the low - pass filtering property of the enzymatic loop can be fine - tuned to optimize other signal transmission characteristics besides @xmath45 and @xmath214 .
these two errors are minimized when the output fluctuations ( @xmath236 or @xmath237 ) closely follow the scaled input fluctuations ( @xmath238 or @xmath239 ) .
however one could imagine biological scenarios where the desired outcome was a smoothed output that mirrored the oscillatory driving signal .
in other words we could demand that @xmath1 , as shown for example in fig .
[ f5]a ( purple trajectory ) , deviates minimally from the oscillatory local mean @xmath240 ( superimposed dashed line ) . in this case
, the natural quantity to minimize would be a local coefficient of variation , @xmath241 . from the oscillatory kmc simulations described above , we calculate @xmath216 , and find that it can be made small in the slow oscillation regime @xmath242 , as shown in fig .
[ f5]d , which plots the minimum @xmath216 as a function of @xmath213 for @xmath243 s. from fig .
[ f5]e , which shows the @xmath173 values at which the minimum @xmath216 occurs ( crosses ) , we see that in the large @xmath213 limit this @xmath173 value is smaller than the wk prediction .
this makes sense , since as we know from the case of a constant driving function ( @xmath244 ) , illustrated in fig .
[ f3]d , keeping @xmath173 below the wk optimum smooths the output . for systems more complex than the enzymatic loop , smoothed output ( homeostasis around a constant mean , or tracking of a driven , time - varying local mean )
can be enhanced by introducing some negative feedback mechanism from the output back to the input @xcite . for such negative feedback systems
it turns out there exists a mapping onto a different wk filter @xcite .
we have demonstrated the usefulness of a generalized wk filter theory as a way of characterizing signal fidelity in an enzymatic push - pull loop .
this basic motif of biological signal transduction can effectively realize an optimal wk noise filter . through a novel analytical approach ,
we have generalized wk ideas beyond their original linear context , thus providing fidelity bounds in strongly nonlinear cases , including ultrasensitive production and oscillatory input driving . even for a complex kinase - phosphatase reaction network with multiple intermediates ,
the theory predicts the conditions for accurate signal transduction , yielding a bound on the error in terms of a single dimensionless optimality control parameter @xmath69 .
the results highlight how physics and engineering concepts can be use to understand how biology robustly tunes push - pull loops to optimality by setting the copy numbers of phosphatase and substrate molecules .
we can relate the wide range of cellular signaling protein copy numbers observed experimentally to optimal time scales on which the cell can accurately transduce the signal , and thus yield an effective physiological response . since our approach is formulated in terms of correlation functions of signal and noise ,
quantities readily accessible from both theory and simulation , the current work can be generalized to other complex signaling networks .
the ultimate goal is to give insights into the design principles underlying the large , intertwined biochemical pathways that determine how the cell can process and respond to diverse sources of external stimuli .
this work was supported by a grant from the national science foundation ( che13 - 61946 ) .
given the correlation functions , @xmath245 we would like to find the optimal filter function @xmath246 that satisfies the wiener - hopf equation , @xmath247 since @xmath248 and @xmath249 consist of exponential terms and dirac delta functions , a reasonable ansatz for @xmath246 is a sum of @xmath250 exponentials , @xmath251 , with parameters @xmath62 , @xmath63 , @xmath252 . plugging this into eq . , along with the correlation functions from eq . , and
carrying out the integral , we find @xmath253 , \quad t>0 . \end{split}\ ] ] comparing the left - hand and right - hand sides of eq .
, we see that the coefficients of the linearly independent exponential terms on both sides must match , giving @xmath254 equations : @xmath250 coefficients of @xmath255 , plus one for @xmath256 .
since there are @xmath257 unknown parameters in the ansatz , the only value of @xmath250 that gives a closed set of equations is @xmath64 . with this choice of @xmath250 , the resulting two equations are @xmath258 the only physically sensible solution of eq . for @xmath259 and @xmath260 ( where @xmath261 as @xmath262 ) is @xmath263
thus the optimal filter is @xmath264
to obtain results for the general signal pathway model , where we assume neither linearity of the production function @xmath73 or a continuum description , we start with an exact equation for the stationary joint distribution @xmath265 of the input and output . using this , we will derive expressions for various moments of the distribution which enter into the relative mean - squared error @xmath266 from the master equation , @xmath265 satisfies @xmath267\\ & \qquad + f \left [ { \cal p}(i-1,o)-{\cal p}(i , o)\right]\\ & \qquad + \gamma_o \left[(o+1){\cal p}(i , o+1)-o { \cal p}(i , o)\right]\\ & \qquad + r(i)\left[{\cal p}(i , o-1)-{\cal p}(i , o)\right]=0 . \end{split}\ ] ] let us define a generating function @xmath268 . by multiplying eq . by @xmath269 and then summing over @xmath2 , we can derive the following equation for @xmath270 , @xmath271 + f \left [ h_{i-1}(z)-h_i(z)\right]\\ & \qquad+ \gamma_o ( 1-z)h^\prime_i(z ) + r(i)(z-1)h_i(z)=0 .
\end{split}\ ] ] plugging in @xmath272 , eq .
can be solved for @xmath273 , the marginal probability distribution of the input .
the result is @xmath274 , the poisson distribution .
this implies that the first and second input moments are given by @xmath275 moments involving the output @xmath2 can be obtained by manipulation of eq . .
taking its first derivative with respect to @xmath276 , and then setting @xmath272 , we find @xmath277 + f \left [ h^\prime_{i-1}(1)-h_i^\prime(1)\right]\\ & \qquad - \gamma_o h^\prime_i(1 ) + r(i)h_i(1)=0 . \end{split}\ ] ] similarly , taking the second derivative of eq
. with respect to @xmath276 , and setting @xmath272 , gives @xmath278 + f \left [ h^{\prime\prime}_{i-1}(1)-h_i^{\prime\prime}(1)\right]\\ & \qquad- 2\gamma_o h^{\prime\prime}_i(1 ) + 2 r(i)h^\prime_i(1)=0 .
\end{split}\ ] ] from the definition of the generating function , @xmath279 and @xmath280 .
summing eqs . and over all @xmath3 yields the following moment relations , @xmath281 evaluating @xmath282 involves finding the mean of @xmath73 over the known input distribution @xmath283 .
however , finding @xmath284 involves the unknown distribution @xmath285 .
moreover , the last remaining moment in eq . for the mean - squared error , @xmath286 , can also be expressed in terms of this distribution , @xmath287 .
thus it is crucial to have additional information about @xmath285 .
we know that @xmath285 satisfies eq . , and let us assume an ansatz for @xmath285 of the form @xmath288 for some function @xmath289 . plugging this into eq . , and using the fact that @xmath290 is the poisson distribution , we find @xmath291=0,\ ] ] where @xmath292 is an operator acting on @xmath289 , defined as @xmath293 here @xmath294 is the finite difference operator , which acts on a function @xmath295 as @xmath296 .
thus the function @xmath289 which solves eq . is @xmath297 , where the operator @xmath298 .
thus @xmath299 , and @xmath300 note that the terms on the right - hand sides inside the @xmath301 brackets are solely functions of @xmath3 , and hence the averages depend on @xmath117 .
plugging eqs . , , and into eq .
gives an expression for the relative error , @xmath302 } , \\
{ \cal m}[r(i ) ] & \equiv \langle r(i ) { \cal l } r(i ) \rangle - \langle r(i ) \rangle^2 . \end{split}\ ] ] to make further progress on the evaluation of @xmath45 , it would be helpful to express @xmath73 in terms of eigenfunctions of @xmath292 ( which would also be eigenfunctions of @xmath303 ) .
to do this , we employ a set of techniques known as umbral calculus @xcite , which starts with the observation that the function @xmath73 can be expanded in a newton series ( the finite difference analogue of the taylor series ) , @xmath304 where @xmath305 is the @xmath306th falling factorial of @xmath3 ( with @xmath307 ) .
the newton series expansion exists assuming @xmath73 fulfills certain analyticity and growth conditions @xcite , which are satisfied for all physically realistic production functions .
finite difference operators acting on @xmath308 result in linear combinations of falling factorials . in particular , @xmath309 and @xmath310 .
thus the operator @xmath292 acting on @xmath308 gives @xmath311.\ ] ] if we consider functions like @xmath73 as vectors in the basis of falling factorials @xmath312 , with components @xmath313 , then from eq .
the operator @xmath292 is a simple bidiagonal matrix in this basis , with elements @xmath314 the eigenvalues of @xmath315 of @xmath292 , labeled by @xmath316 in decreasing order , are just the diagonal matrix components , @xmath317 .
the corresponding eigenfunctions are @xmath318 the @xmath96th eigenfunction @xmath94 is a polynomial in @xmath3 of degree @xmath96 , with the first few given by @xmath319 the eigenfunctions @xmath94 are mathematically related to expansions of the master equation through alternative approaches , for example the spectral method of refs . @xcite .
in fact , @xmath320 , where @xmath321 is the mixed product defined in eq .
a8 of ref .
@xcite ( with @xmath74 substituted for the rate parameter @xmath322 ) . since eq . can be inverted to express @xmath308 in terms of the eigenfunctions , @xmath323 we can write @xmath73 as in terms of the eigenfunctions by plugging eq . into eq . ,
@xmath324 where we have used the property that @xmath325 for @xmath326 .
the operator @xmath327 acting on @xmath73 is then @xmath328 since the quantities in eq . for @xmath45 involve averages with respect to @xmath117 , it is useful to derive the first and second moments of the eigenfunctions . from the fact that the falling factorials have
very simple averages in the poisson distribution , @xmath329 , we find using eq . that @xmath330 .
this implies that @xmath331 . to find @xmath332 , we start from the chu - vandermonde identity @xcite , the umbral analogue of the binomial theorem , @xmath333 for @xmath334 and @xmath335 this gives @xmath336 where we have used the fact that @xmath337 . multiplying both sides by @xmath338
, we find @xmath339 the second equality is based on the relation @xmath340 , which follows from the definition of the falling factorial . taking the average of both sides of eq .
yields @xmath341 an alternative expression for @xmath342 can be derived by substituting the eigenfunction expansion of eq .
for both @xmath338 and @xmath308 , @xmath343 comparing the right - hand sides of eqs .
and we see that @xmath344 . together with eqs . and
this allows us to calculate @xmath345 & = \langle r(i ) { \cal l } r(i ) \rangle - \langle r(i ) \rangle^2\\ & = \sum_{n^\prime=0}^\infty \sum_{n=0}^\infty \sigma_{n^\prime}\sigma_n \frac{\gamma_o}{\gamma_o+n\gamma_i } \langle v_{n^\prime}(i ) v_n(i)\rangle - \sigma_0 ^ 2\\ & = \sum_{n=1}^\infty \sigma_n^2 \frac{\gamma_o n !
\bar{i}^n}{\gamma_o+n\gamma_i}. \end{split}\ ] ] using the fact that @xmath346 , we can similarly evaluate @xmath347 -\bar{i } \sigma_0\\ & = \frac{\gamma_o \bar{i } \sigma_1}{\gamma_o+\gamma_i}. \end{split}\ ] ] plugging eqs . and into eq .
, we obtain our final expression for the relative error , @xmath348^{-1}.\ ] ] this expression can be readily calculated numerically for any given @xmath73 , as was done in the main text for the family of hill function production rates . to facilitate evaluation
, we express the coefficients @xmath98 as moments with respect to the poisson distribution @xmath117 in the following manner , using the expansion of eq .
, @xmath349 from the definition of @xmath94 in eq .
, the coefficients @xmath98 can be written @xmath350 using eq .
the @xmath98 can be numerically calculated for any @xmath73 .
the sum in eq .
converges quickly because the @xmath98 decrease rapidly with @xmath96 , so typically only @xmath98 for @xmath351 are needed to get accurate results for @xmath45 . the expression in eq . also allows us to determine under what conditions the relative error @xmath45 becomes minimal . for this to occur we need @xmath352 , since otherwise @xmath45 takes its maximum value of 1 .
the sum within the brackets in eq .
is composed of only non - negative terms , and @xmath45 is smallest when this sum is minimal .
this can be achieved by setting @xmath353 for all @xmath124 .
thus @xmath45 is bounded from below by @xmath354^{-1},\ ] ] where the equality is only reached when @xmath73 has an optimal linear form , @xmath355 .
the right - hand side of eq . is minimized with respect to @xmath8 when @xmath356 , with @xmath357 . at this optimal @xmath8 , the inequality in eq .
the full set of reactions for the enzymatic push - pull loop is given by @xmath359{f } k\\ k + s & \xrightleftharpoons[\kappa_\text{u}]{\kappa_\text{b } } s_k \xrightarrow{\kappa_\text{r } } k + s^\ast,\\ p + s^\ast & \xrightleftharpoons[\rho_\text{u}]{\rho_\text{b } } s^\ast_p \xrightarrow{\rho_\text{r } } p + s. \end{split}\ ] ] the corresponding steady - states populations are @xmath360 where @xmath176 , @xmath177 , @xmath178 , @xmath179 , @xmath180 , and @xmath181 . for the system in eq . , the associated set of chemical langevin equations is @xmath361 where the equations on the last line come from the assumptions that the total populations of free / bound phosphatase ( @xmath362 ) and free / bound substrate in all its forms ( @xmath363 ) remain constant .
the noise terms @xmath364 , where the @xmath365 are gaussian white noise functions with correlations @xmath366 .
the constants @xmath367 are the power spectra of the noise terms , given by @xmath368 we are interested in how the kinase input signal @xmath369 is transduced into the active substrate output @xmath370 , and in particular whether the system can be approximately mapped onto a wk noise filter of the form given in the main text ( eq . 2 ) .
( recall that @xmath371 for any time series @xmath372 . ) since the wk description hinges on the form of the correlation functions of input and output , we will need to calculate such correlations for the dynamical equations in eq . .
after linearizing these equations , it will be easier to work in fourier space , where the fourier - transformed correlation functions correspond to power spectra : @xmath373 for a given @xmath374 .
hence it will useful , before proceeding further , to recast main text eq . 2 , the time - domain noise filter , as a fourier - space relation in terms of the power spectra .
the result is @xmath375 . \end{split}\ ] ] our goal in this section is to show that @xmath376 and @xmath377 calculated for the enzymatic push - pull loop in eq .
have the approximate form of eq . , with effective values for @xmath7 , @xmath8 , @xmath378 , and @xmath69 expressed in terms of the loop reaction rate parameters .
the equilibrium populations @xmath379 and @xmath380 scale with @xmath74 as @xmath381 and @xmath382 . similarly , @xmath383 and @xmath384 .
each deviation from the mean@xmath385 , @xmath386 , @xmath387 , and @xmath388we will explicitly divide into a component that scales with @xmath389 or @xmath236 like the mean population ( the `` slowly '' varying component ) , and the remainder ( the `` quickly '' varying component , denoted with subscript @xmath390 ) : @xmath391 we can interpret eq . as defining a change of variables from the set @xmath385 , @xmath386 , @xmath387 , and @xmath388 to the set @xmath236 , @xmath392 , @xmath389 , @xmath393 .
the nomenclature of slow and quick components comes from the fact that if the enzymatic reaction rates ( @xmath394 , @xmath395 , @xmath396 , @xmath397 ) are made extremely rapid , the characteristic time scales for the @xmath398 and @xmath392 fluctuations become so small that the quick components can be neglected , since there would be nearly instantaneous equilibration between the free and bound enzyme populations . in general , however , we can not assume this limiting case always holds , so we will take into account both the slow and quick components in our analysis . where @xmath400 denotes the fourier transform of @xmath372 .
eq . can be solved analytically for @xmath401 , @xmath402 , @xmath403 , @xmath404 , though for simplicity we will not write out the full solutions , since these would take up too much space . rather we will sketch out the basic approach to calculating and approximating the associated power spectra .
the structure of the solutions to eq . ,
for example @xmath403 , is a linear combination of the the noise functions , @xmath405 , with coefficients @xmath406 .
the corresponding power spectrum is @xmath407 , with @xmath367 given by eq . .
the function @xmath408 can be written out in the form of a rational function with even powers of @xmath409 in the numerator and denominator , @xmath410 where @xmath411 and @xmath412 are coefficients independent of @xmath409 , and @xmath413 , @xmath414 for the case of @xmath415 . in order to simplify eq .
further , we will make two assumptions : ( i ) the characteristic time scale over which the input signal varies , @xmath200 , is much longer than the characteristic time scales of the enzymatic reactions , @xmath416 and @xmath417 , where @xmath16 denotes the various subscripts @xmath418 , @xmath419 , and @xmath420 . for the parameters in the main text , @xmath421 , while @xmath416 , @xmath422 .
this the physically interesting regime , since we can expect the system to efficiently transduce signals that vary more slowly than the intrinsic reactions that carry out the transduction . limiting our focus to frequencies @xmath423 , @xmath417 , it turns out that the higher order powers of @xmath409 in both the numerator and denominator of eq . are negligible , and the power spectrum can be approximated by @xmath424 ( ii ) we assume that the system is in the regime where @xmath425 .
thus we will expand the coefficients @xmath426 and @xmath427 in eq . up to first order in @xmath183 and @xmath184 , resulting in a @xmath428 that has the form of eq . .
namely , @xmath429 and @xmath430 , where the effective @xmath431 is given by @xmath432 31ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) @noop _ _ ( , ) | cellular signaling involves the transmission of environmental information through cascades of stochastic biochemical reactions , inevitably introducing noise that compromises signal fidelity .
each stage of the cascade often takes the form of a kinase - phosphatase push - pull network , a basic unit of signaling pathways whose malfunction is linked with a host of cancers .
we show this ubiquitous enzymatic network motif effectively behaves as a wiener - kolmogorov ( wk ) optimal noise filter . using concepts from umbral calculus ,
we generalize the linear wk theory , originally introduced in the context of communication and control engineering , to take nonlinear signal transduction and discrete molecule populations into account .
this allows us to derive rigorous constraints for efficient noise reduction in this biochemical system .
our mathematical formalism yields bounds on filter performance in cases important to cellular function like ultrasensitive response to stimuli .
we highlight features of the system relevant for optimizing filter efficiency , encoded in a single , measurable , dimensionless parameter .
our theory , which describes noise control in a large class of signal transduction networks , is also useful both for the design of synthetic biochemical signaling pathways , and the manipulation of pathways through experimental probes like oscillatory input .
extracting signals from time series corrupted by noise is a challenge in a number of seemingly unrelated areas .
minimizing the effects of noise is a critical consideration in designing communication and navigation systems , and analyzing data in diverse fields like medical and astronomical imaging .
more recently , a number of studies have focused on how biological circuits , comprised of chemical signaling pathways mediated by genes , proteins , and rna , cope with noise @xcite .
one of the key discoveries in the past decade is that the naturally occurring systems that control all aspects of cellular processes undergo substantial stochastic fluctuations both in their expression levels and activities .
noise may even have a functional role @xcite , providing coordination between multiple interacting chemical partners in typical circuits . because of the variety of ways noise influences cellular functions , it is important to develop a practical and general theoretical framework for describing how biological systems cope with and control the inevitable presence of noise arising from stochastic fluctuations . in the context of communication theory ,
the optimal noise - reduction filter , discovered independently by wiener @xcite and kolmogorov @xcite in the 1940 s , inaugurated the modern era of signal processing , providing the first general solution to the problem of extracting useful information from corrupted signals .
we show that this classic result of wartime mathematics , developed to guide radar - assisted anti - aircraft guns , yields insights into the efficiency limits of generic biochemical signaling networks . dealing with noise in biological signal transduction
is at first glance even more daunting than in engineered systems . in order to survive , cells must process information about their external environment @xcite , which is transmitted and amplified from stimulated receptors on the cell surface through elaborate pathways of post - translational covalent modifications of proteins .
a typical example is phosphorylation by protein kinases of target proteins , which then become activated to modify targets further downstream . signaling occurs through cascades involving multiple stages of such activation [ fig . [ f1]a ] .
since each enzymatic reaction is stochastic , noise inevitably propagates through the cascade , potentially corrupting the signal @xcite .
our work focuses on a basic signaling circuit : a `` push - pull loop '' where a substrate is activated by one enzyme ( i.e. phosphorylation by a kinase ) and deactivated by another ( i.e. dephosphorylation by a phosphatase ) @xcite [ fig . [ f1]b ] .
since cascades have a modular structure , formed through many such loops in series and parallel , understanding the stochastic properties at the single loop level is a prerequisite to addressing the complex behavior of entire pathways @xcite .
the push - pull loop can act like an amplifier , taking the input signal
the time - varying population of kinase and approximately reproducing it at larger amplitude through the output the population of active , phosphorylated substrate @xcite . depending on the parameters , small changes in the input can be translated into large ( but noise - corrupted ) output variations .
the amplification is essential for sensitive response to external stimuli , but it must also preserve signal content to be useful for downstream processes .
thus , the signaling circuit , despite operating in a noisy environment , needs to maintain a high fidelity between output and amplified input . from a design perspective
, the natural question that arises is what are the general constraints on filter efficiency ?
are there rigorous bounds , which depend only on certain collective features of the underlying biochemical network architecture ?
discovering such bounds is important both to explain the metabolic costs of noise suppression in biological systems @xcite , and also for bioengineering purposes . in particular , for constructing synthetic signaling networks , we would like to make the most efficient communication pathway with a limited set of resources ( free energy costs ) . to answer these questions , using the enzymatic push - pull loop as an example , we introduce a new mathematical framework , inspired by the wiener - kolmogorov ( wk ) theory for optimal noise filtration .
the original wk theory has restrictions that make it of limited utility in the biological context
it assumes that the input and output are continuous variables describing stationary stochastic processes .
more critically , the filter is assumed to be linear .
exploiting the power of exact analytical techniques based on umbral calculus @xcite , we overcome these limitations , thus generalizing the wk approach .
this crucial theoretical development enables us to provide a rigorous solution to the filter optimization problem , taking into account discrete populations and nonlinearity .
we can thus understand constraints in biologically significant regimes of the push - pull loop behavior , for example highly nonlinear , `` ultrasensitive '' amplification @xcite .
our theory predicts that optimality can be realized by tuning phosphatase levels , which we verified through simulations of a microscopic model of the loop reaction network , including cases where the system is driven by an oscillatory input @xcite , which is relevant to recent experimental probes @xcite .
the optimality is robust , with the filter operating at near - optimal levels even when the wk conditions are only approximately fulfilled , over a broad range of realistic parameter values .
although illustrated using a push - pull loop , the theory is applicable to a large class of signaling networks , including more complex features such as negative feedback or multi - site phosphorylation of substrates . |
more than thirty years after the first numerical simulations of binary black hole dynamics , the numerical relativity community is finally ready to compare binary black hole simulations with experimental data .
thanks to a series of recent breakthroughs , long term evolutions of inspiralling binary black holes that last for more than one orbit have been obtained with several independent codes , and accurate gravitational wave signals have been computed @xcite .
the use of numerical waveforms as templates for gravitational wave detection requires large - scale parameter studies , and correspondingly large computational resources .
the main current technical problems in the field are the efficiency of the numerical simulations and the development of a `` data analysis pipeline '' , connecting numerical simulations with analytical calculations of the early inspiral and late ringdown phases , and ( eventually ) with gravitational wave searches in actual detector data . to build a common language between the numerical relativity and data analysis communities
we must develop a deeper understanding of the physical content of the simulations using analytical techniques , such as post - newtonian ( pn ) theory and black hole perturbation theory .
a better analytical understanding of the simulations is important for many reasons : * to determine which regions of the parameter space ( mass , spin magnitude and inclination , orbital separation , eccentricity ... ) _ must _ be explored by numerical simulations , and which regions can be covered by ( say ) analytically - inspired interpolations of the numerical waveforms .
this would obviously save a significant amount of computing time . * to develop optimal strategies for the construction of detection templates , using a combination of numerical and analytical techniques . * to understand details of the non - linear physics encoded in the strong - field merger gravitational waveforms , and extract as much science as possible from a detection .
in this paper we focus on point ( 3 ) , and we try to develop a general framework to quantitatively compare analytical calculations of the inspiral and ringdown waveforms with the `` full '' waveforms produced by numerical simulations , extending from late inspiral through merger and ringdown . our work can be considered an extension of the recent analysis by buonanno , cook and pretorius ( @xcite , henceforth bcp ) .
bcp studied simulations of non - spinning , equal mass black hole binaries starting out at three different initial separations . in this work
we examine a larger set of simulations performed using the bam code @xcite and the moving puncture method .
we consider seven different mass ratios ( @xmath8 to @xmath9 in steps of @xmath10 ) with initial coordinate separation @xmath11 , roughly corresponding to @xmath12 orbits before merger .
for each mass ratio , the simulations were carried out at three different resolutions . to explore the effect of initial separation on the physical parameters of the remnant , we also consider two runs at separation @xmath13 ( for @xmath14 and @xmath15 ) , and one run at separation @xmath16 ( for @xmath17 ) .
we typically use an extraction radius @xmath18 , with the exception of the @xmath17 run with @xmath16 , in which case we extract gravitational waves at @xmath19 and @xmath20 .
section [ setup ] contains details of our numerical setup . in section [ memory ]
we study in some detail a well - known issue with the extraction of gravitational waveforms from numerical simulations : the problem of fixing integration constants when we integrate the weyl scalar @xmath21 twice in time to obtain the gravitational wave amplitude @xmath22 . fixing the integration constants to zero
produces a systematic drift in @xmath22 and in its first time derivative .
this drift is sometimes referred to in the literature as a `` memory effect '' , but this is somewhat misleading .
the so - called `` memory effect '' is really due to numerical errors , wrong initial conditions and limitations of wave extraction techniques , and it should _ not _ be confused with the christodoulou memory , which is a true ( if typically small ) physical effect due to the non - linearity of general relativity @xcite . we find that the extraction radius is critical to reduce the amplitude drift , and that resolution only seems to affect the drift for low - amplitude components of the wave . in section [ sec : transition ] we study the inspiral - merger transition .
we start by projecting the 2.5pn gravitational wave amplitude for quasi - circular , non - spinning binaries @xcite onto spin - weighted spherical harmonics . in this way
we obtain the spin - weighted spherical harmonic components of the weyl scalar as pn series in the binary s orbital frequency : @xmath23 .
we refer to this analytical expression of the gravitational wave amplitudes as the _ post - newtonian quasi - circular _ ( pnqc ) approximation ( see section [ sec : pnqc ] for details , and appendix [ app : multipoles ] for a complete list of all the multipolar components ) .
the pnqc approximation can be used in two ways .
first , given the orbital frequency evolution @xmath24 , we can compute ( an approximation to ) the multipolar components @xmath25 .
conversely , given the modulus of the wave amplitude @xmath26 , we can numerically invert the pn expansions to obtain a pnqc estimate of the orbital frequency : @xmath27 . in sections [ freq - ests ]
we compare @xmath28 with two alternative estimates of the orbital frequency , first introduced in bcp : @xmath29 ( an estimate obtained from the gravitational wave frequency ) and @xmath30 ( computed from the punctures coordinate motion ) . using these three different estimates of the orbital frequency , we study the convergence of the pnqc approximation .
we find that , as in the point particle case @xcite , the convergence of the pn series is not monotonic .
we also study the effect of resolution and wave extraction on the agreement between pnqc results and numerical results .
we find that low resolution increases numerical noise in the frequencies and amplitudes at late times .
a small extraction radius produces systematic errors at large separations , where gravitational wavelengths are longer , but it does not sensibly affect the ringdown phase . in section [ energy - j ] we study in detail the total radiated energy @xmath31 and the final angular momentum @xmath32 as functions of the mass ratio , providing fitting formulas for each of these quantities .
we also compare the energy and angular momentum fluxes with their pnqc estimates , and we study ( both analytically and numerically ) the multipolar distribution of the radiation . to leading order , we find that @xmath33 and @xmath34 , where @xmath35 is the so - called symmetric mass ratio , and we provide fitting formulas for these quantities .
as predicted by the pnqc approximation , odd-@xmath6 multipoles of the radiation are suppressed in the equal mass limit . as the mass ratio increases , higher multipoles ( with @xmath36 ) carry a larger fraction of the total energy : for @xmath37 , @xmath38 typically carries @xmath39 of the total energy ( see table [ tab : summary ] below ) . in section [ fitmerger ]
we turn our attention to the merger - ringdown transition . during ringdown
the waveform can be described as a superposition of complex exponentials , the quasinormal modes ( qnms ) . in @xcite
we argued that prony methods ( which are well - known in signal processing ) are in many ways `` optimal '' methods to extract qnm frequencies from a numerical signal .
after explaining our choice of the fitting window , in section [ wrjm ] we use prony methods and standard , non - linear least - squares fits to look at the time dependence of the final black hole s parameters .
we find resolution - dependent deviations in these parameters from the values predicted by linear black hole perturbation theory .
these effects may be due to non - linearities and/or to rotational mode coupling , but at present we can not exclude the possibility that they are , more trivially , an artifact of finite numerical resolution . in section [ sec : crosscorrelate ]
we show that , by cross - correlating information from different multipolar components of the ringdown waves , we can find an empirical `` best guess '' for the optimal time to estimate the final black hole s mass and angular momentum .
we argue that , because of the no - hair theorem , this best guess corresponds to the last time when the angular momenta ( or masses ) obtained by fitting the dominant multipoles agree with each other . in support of this argument
, we also show that estimates of the mass and spin of the final black hole based on qnm fits are in remarkable agreement with wave extraction methods .
black hole qnms do not form a complete set , and for this reason it is not possible to define unambiguously the beginning of the ringdown phase .
in section [ sec : rdstart ] we consider three different definitions of the ringdown starting time , two of which have already appeared in the literature ( but not in the context of binary black hole simulations ) .
the first definition is based on looking for the time at which a qnm expansion provides the best fit to the actual numerical waveform , in the sense of a suitably defined norm @xcite .
unfortunately , when applied to our numerical waveforms , this method is not particularly useful .
the reason is that the norm is quite flat ( and even worse , has some oscillations ) over a wide range of starting times around the minimum . a second
, more useful definition looks for the time maximizing the _ energy content _ of the qnm component of the waveform . for this reason , following nollert @xcite , we call it the energy maximized orthogonal projection , or emop .
we find that the `` emop time '' @xmath40 and the maximum fraction of energy carried by ringdown ( @xmath41 ) are remarkably independent of the mass ratio @xmath7 .
this is an indication that the ringdown waveform is in some sense `` universal '' : it does not depend too much on the details of the pre - merger phase .
to our knowledge , the third definition of the ringdown starting time has not been introduced before .
it uses a detection - based criterion , maximizing the `` effective energy '' deposited in a matched filter . in the conclusions we present a list of open problems and directions for future research . to improve readability ,
some lengthy equations and technical material are presented in the appendices .
appendix [ app : multipoles ] lists the spin - weighted spherical harmonic components of the weyl scalar , up to and including 2.5pn terms in a pn expansion of the waveforms .
appendix [ app : postplunge ] provides fits for the energy , angular momentum and linear momentum radiated after the estimated time of formation of a common apparent horizon ( cah ) .
since the total energy radiated in a simulation depends on the initial separation of the binary , in this appendix we also try to provide estimates for the energy , angular momentum and linear momentum radiated `` after plunge '' .
a problem here is that the innermost stable circular orbit ( isco ) is a controversial concept for comparable - mass binaries , and there is no unique way to define the beginning of the plunge phase . given these intrinsic ambiguities , we estimate the starting time of the plunge , @xmath42 , as the time when the orbital frequency @xmath43 becomes larger than the isco frequency computed in pn theory ( at 2pn or 3pn order , to bracket uncertainties )
. we also present a comparison of our results with pn estimates of the post - plunge radiation by blanchet _
_ @xcite .
computational resources and resolution limitations reduce the accuracy of numerical simulations for large mass ratio . unfortunately , many astrophysical black hole binaries could have @xmath44 or larger ( see eg .
@xcite and references therein ) .
it is important to determine the maximum value of @xmath7 that should be simulated in numerical relativity , or equivalently , the smallest value of @xmath7 for which black hole perturbation theory can be considered adequate for detection and/or parameter estimation .
appendix [ app : pointparticles ] collects some results from perturbation theory that may be useful in this context .
we point out that , for large mass ratio , our numerical simulations seem to be in reasonable agreement with perturbative calculations of particles plunging with large angular momentum into a schwarzschild black hole .
finally , in appendix [ app : polarization ] we introduce quantitative measures of the polarization state of the waveform .
we show that the polarization of the wave ( as viewed from the normal to the orbital plane ) is circular for both inspiral and ringdown , with the exception of the unphysical portions of the wave : the initial data burst and the final , noise - dominated part of the ringdown waveform . in all of this paper we adopt geometrical units ( @xmath45 ) .
unless otherwise indicated , physical quantities are usually normalized to the total arnowitt - deser - misner ( adm ) mass of the system @xmath46 .
the adm masses of all configurations presented in this study have been calculated using ansorg s @xcite spectral solver for binary black hole puncture data . due to the spectral accuracy and the compactification of the coordinates , which facilitates evaluation of the adm mass at infinite radius , the uncertainties in this quantity are negligible relative to those arising out of the numerical time evolution .
.[tab : summary ] summary of the main results of this paper ( see text ) . to convert from radiated momenta to kick velocities in km s@xmath47 ,
the numbers in this table must be multiplied by @xmath48 km s@xmath47 .
further details ( including estimates of the uncertainties ) are given in the bulk of the paper . [ cols="^,^,^,^,^,^,^,^,^,^,^,^,^,^,^",options="header " , ] one of the most prominent features borne out of binary black hole simulations seems to be the absence of strong non - linearities : the potential barrier close to the black hole horizon acts as a very effective cloak , filtering out many non - linear features of the dynamics @xcite . for this reason results from perturbation theory ( i.e. , @xmath49 )
can usually be extrapolated to the equal mass ratio case , yielding very good agreement with full blown numerical simulations . to quote smarr , ``
the agreement is so remarkable that something deep must be at work '' @xcite .
this was also found to be the case for the head - on collision of two black holes .
the perturbation theory result @xmath50 , with @xmath51 @xcite can be compared to the full numerical result for @xmath17 : @xmath52 ( see @xcite ) .
simple scaling arguments applied to perturbative results work surprisingly well . on this basis , a natural conjecture is that particles plunging with large but sub - critical orbital angular momentum should describe reasonably well the final stages of a binary black hole inspiral @xcite .
indeed , some of the results discussed in the main text suggest that the merger of two equal mass black holes can be described by extrapolating results from perturbation theory .
for instance , the final spin of the black hole can be predicted by using the small mass ratio approximation : see eq .
( [ iscofinalj ] ) and the related discussion . for inspiralling binaries evolving through quasi - circular orbits ,
the plunge ( `` merger '' ) happens when the smaller body crosses the isco , even though this notion is not well defined for comparable - mass bodies .
we argue that the merger phase should be reasonably well described by a particle plunging with an orbital angular momentum @xmath53 only slightly smaller than the marginal value @xmath54 ( corresponding to a radius equal to @xmath55 ) . in this case
the trajectory resembles that of a particle going through the merger phase : a quasi - circular orbit followed by a plunge .
it is important to specify how close @xmath53 should be to the marginal value .
we look for orbits that complete barely less than one lap before plunging .
this guarantees that the energy output is due only to a plunge trajectory ; it also guarantees that the orbit was quasi - circular before the plunge .
we therefore argue that , as seen from the table , @xmath56 is a near - optimal value .
an obvious objection is that the isco for point particles is not at @xmath57 ( the location of the marginally bound orbit ) , but rather at @xmath58 .
fortunately , for more massive bodies an approximate isco can be defined , and it is usually closer to @xmath57 . we can thus try to extend these perturbative results to understand the merger phase . extrapolating the point particle results presented in the previous subsections we get , for non - rotating holes : = 0.485,=0.15,v_recoil=232kms^-1 . for instance , the value @xmath59 for the energy is found extrapolating the last row in table [ tab : energyplungenonrotating ] by the substitution @xmath60 . from fits of the corresponding numerical quantities at the 3pn isco ( which should work as a good reference point , for lack of a better guess )
we get : = 0.421,=0.19,v_recoil^isco=120kms^-1 , the two sets of values are reasonably consistent .
the largest disagreement refers to the linear momentum radiated , and therefore to the recoil velocity of the final hole .
getting information about the polarization content of a waveform is simple in the presence of a monochromatic wave . in more general settings ,
making statements about the polarization state is easier in fourier space .
methods to compute the polarization content of a waveform solely from a time - domain analysis were presented in @xcite . here
we shall adapt these techniques for the case at hand . from the ( real ) polarization components @xmath61 , in order to avoid problems with the integration constants and to sidestep the memory effect discussed in section [ memory ] .
this should have no influence on the final results .
] we can define the so - called _ analytic signal _
@xmath62 in the following way : h_+ & & h_+(t)+ih_+(t ) , + h _ & & h_(t)+ih_(t).the imaginary part of the analytic signal is the hilbert transform @xmath63 of the signal @xmath64 , defined as ( t)=_-^+ d , where the integral is taken as the cauchy principal value . for reference
we note that the hilbert transform of @xmath65 is @xmath66 and the transform of @xmath67 is @xmath65 . from the analytic signal
we define a covariance matrix @xmath68 as @xmath69 where an asterisk stands for complex conjugation .
we note that the covariance matrix is hermitian , and thus its eigenvalues @xmath70 are real and positive . without loss of generality , we assume @xmath71 . it can be shown that the normalized eigenvector @xmath72 ( @xmath73 ) associated with @xmath74 points in the direction of the largest amount of polarization @xcite .
one can define an elliptical component of polarization @xmath75 as p_e= , where x=_. the quantity @xmath76 for circular polarization , and @xmath77 for linear polarization . for illustration , consider the waveform @xmath78 .
this is a good approximation to a typical inspiral waveform at large orbital separation , as viewed from the normal to the orbital plane , and it is obviously circularly polarized .
for this waveform we have @xmath79 and @xmath80 .
this implies @xmath81 and @xmath76 , as expected . for @xmath82 ( linear polarization ) we would have @xmath83 and @xmath77 .
thus @xmath75 is a good indicator of the degree of circular or linear polarization .
we can also define a polarization strength as follows : p_s=1- .
@xmath84 means that the waveform is entirely linearly polarized or circularly polarized ( there is only one polarization component ) , and @xmath85 means that the two polarization states have comparable magnitude . in fig .
[ wf1-pol ] we show the result of computing @xmath75 using the dominant ( @xmath86 ) component of a binary black hole merger waveform with @xmath87 .
this plot clearly shows that the polarization is circular for both inspiral and ringdown , with the exception of the unphysical portions of the wave : the initial data burst and the final , noise - dominated part of the ringdown waveform . 99 a. buonanno , g. b. cook and f. pretorius , arxiv : gr - qc/0610122 .
b. bruegmann , j. a. gonzalez , m. hannam , s. husa , u. sperhake and w. tichy , arxiv : gr - qc/0610128 .
j. a. gonzalez , u. sperhake , b. bruegmann , m. hannam and s. husa , phys .
lett . * 98 * , 091101 ( 2007 ) [ arxiv : gr - qc/0610154 ] . f. pretorius , phys . rev .
lett . * 95 * , 121101 ( 2005 ) [ arxiv : gr - qc/0507014 ] .
j. g. baker , j. centrella , d. i. choi , m. koppitz and j. van meter , phys .
d * 73 * , 104002 ( 2006 ) .
m. campanelli , c. o. lousto and y. zlochower , phys .
d * 73 * , 061501 ( 2006 ) .
f. herrmann , d. shoemaker and p. laguna , arxiv : gr - qc/0601026 ; u. sperhake , arxiv : gr - qc/0606079 . b. szilagyi , d. pollney , l. rezzolla , j. thornburg and j. winicour , arxiv : gr - qc/0612150 . h. p. pfeiffer , d. a. brown , l. e. kidder , l. lindblom , g. lovelace and m. a. scheel , arxiv : gr - qc/0702106 .
j. g. baker , j. r. van meter , s. t. mcwilliams , j. centrella and b. j. kelly , arxiv : gr - qc/0612024 .
d. christodoulou , phys .
lett . * 67 * , 1486 ( 1991 ) ; a. g. wiseman and c. m. will , phys .
d * 44 * , r2945 ( 1991 ) ; k. s. thorne , phys .
d * 45 * , 520 ( 1992 ) .
k. g. arun , l. blanchet , b. r. iyer and m. s. s. qusailah , class .
* 21 * , 3771 ( 2004 ) [ erratum - ibid . *
22 * , 3115 ( 2005 ) ] [ arxiv : gr - qc/0404085 ] .
l. e. kidder , l. blanchet and b. r. iyer , arxiv:0706.0726 [ gr - qc ] .
e. poisson , phys .
d * 52 * , 5719 ( 1995 ) [ addendum - ibid .
d * 55 * , 7980 ( 1997 ) ] .
e. berti , v. cardoso , j. a. gonzalez and u. sperhake , arxiv : gr - qc/0701086 .
e. n. dorband , e. berti , p. diener , e. schnetter and m. tiglio , phys .
d * 74 * , 084028 ( 2006 ) .
nollert , _ characteristic oscillations of black holes and neutron stars : > from mathematical background to astrophysical applications _ , unpublished habilitationsschrift ( 2000 ) .
l. blanchet , m. s. s. qusailah and c. m. will , astrophys .
j. * 635 * , 508 ( 2005 ) .
e. berti , class .
* 23 * , s785 ( 2006 ) [ arxiv : astro - ph/0602470 ] .
m. ansorg , b. brgmann and w. tichy , phys .
d * 70 * , 064011 ( 2004 ) .
a. buonanno and t. damour , phys
. rev .
d * 62 * , 064015 ( 2000 ) .
j. g. baker , m. campanelli , c. o. lousto and r. takahashi , phys .
d * 65 * , 124012 ( 2002 ) .
m. koppitz , d. pollney , c. reisswig , l. rezzolla , j. thornburg , p. diener and e. schnetter , arxiv : gr - qc/0701163 .
r. j. gleiser and a. e. dominguez , phys .
d * 68 * , 104018 ( 2003 ) e. berti , s. iyer and c. m. will , phys .
d * 74 * , 061503 ( 2006 ) .
m. a. miller , phys .
d * 69 * , 124013 ( 2004 ) .
p. c. peters , phys .
rev . * 136 * , b1224 ( 1964 ) .
l. blanchet , t. damour , g. esposito - farese and b. r. iyer , phys .
lett . * 93 * , 091101 ( 2004 ) .
e. poisson , phys .
d * 47 * , 1497 ( 1993 ) ; h. tagoshi and m. sasaki , prog .
phys . * 92 * , 745 ( 1994 ) .
p. papadopoulos , phys .
d * 65 * , 084016 ( 2002 ) [ arxiv : gr - qc/0104024 ] .
y. zlochower , r. gomez , s. husa , l. lehner and j. winicour , phys .
d * 68 * , 084014 ( 2003 ) .
e. berti and v. cardoso , phys .
d * 74 * , 104020 ( 2006 ) .
r. h. price , phys .
d * 5 * , 2419 ( 1972 ) .
j. g. baker and m. campanelli , phys .
d * 62 * , 127501 ( 2000 ) .
e. e. flanagan and s. a. hughes , phys .
d * 57 * , 4535 ( 1998 ) .
t. baumgarte , p. brady , j. d. e. creighton , l. lehner , f. pretorius and r. devoe , arxiv : gr - qc/0612100 .
k. a. arnaud _ et al .
_ , arxiv : gr - qc/0701139 ; k. a. arnaud _ et al .
_ , arxiv : gr - qc/0701170 . j. p. a. clark and d. m. eardley , astrophys . j. * 215 * , 311 ( 1977 ) . | we study the inspiral , merger and ringdown of unequal mass black hole binaries by analyzing a catalogue of numerical simulations for seven different values of the mass ratio ( from @xmath0 to @xmath1 ) .
we compare numerical and post - newtonian results by projecting the waveforms onto spin - weighted spherical harmonics , characterized by angular indices @xmath2 .
we find that the post - newtonian equations predict remarkably well the relation between the wave amplitude and the orbital frequency for each @xmath2 , and that the convergence of the post - newtonian series to the numerical results is non - monotonic . to leading order
the total energy emitted in the merger phase scales like @xmath3 and the spin of the final black hole scales like @xmath4 , where @xmath5 is the symmetric mass ratio .
we study the multipolar distribution of the radiation , finding that odd-@xmath6 multipoles are suppressed in the equal mass limit .
higher multipoles carry a larger fraction of the total energy as @xmath7 increases .
we introduce and compare three different definitions for the ringdown starting time . applying linear estimation methods ( the so - called prony methods ) to the ringdown phase , we find resolution - dependent time variations in the fitted parameters of the final black hole . by cross - correlating information from different multipoles
we show that ringdown fits can be used to obtain precise estimates of the mass and spin of the final black hole , which are in remarkable agreement with energy and angular momentum balance calculations . |
algol ( @xmath5 perseus ) is the eponymous eclipsing binary system which consists of a primary early - type main sequence star and a roche lobe - filling secondary late - type giant or subgiant star .
these systems have undergone a period of mass transfer during which material was transferred from the initially more massive present day late - type star to its initially less massive early - type companion .
algol is the brightest and nearest example of this type .
the primary star ( algol a ) is a b8 v main sequence star , while the secondary ( algol b ) is a k2 iv subgiant that has lost about half of its original mass to the present day primary ( e.g. , * ? ? ?
* ) . since algol has a short period orbit of 2.87 days ( e.g. , * ? ? ?
* ) the two stars are tidally locked , and their orbital and rotational periods are synchronized .
algol was first confirmed as an x - ray source by @xcite , who suggested that the x - ray emission might be explained by a mass - transfer model , where the more massive b star accretes material from the less massive k star , either through roche lobe overflow or a stellar wind . in both cases ,
the infalling material is shock - heated to x - ray temperatures when it hits the b star . according to this mass - transfer model
, we would expect to observe an x - ray eclipse during the optical primary eclipse .
however , observations made with the solid state spectrometer ( sss ) of the _ einstein _ observatory revealed no sign of an x - ray eclipse at the time of optical primary eclipse @xcite . at present , it is widely believed that the x - ray emission from algol arises mostly , if not completely , from the corona of algol b @xcite .
the two stars are tidally locked and rapidly rotating , and within the rotationally - excited dynamo paradigm , it is expected that the convection zone of the k star would experience increased dynamo activity whose resulting magnetic energy is subsequently dissipated at the stellar surface in the form of a hot , x - ray emitting corona @xcite .
however , the nature and structure of this corona remains a topic of debate .
more than four orders of magnitude in x - ray luminosity and an order of magnitude in plasma temperature separate the solar corona from coronae of most active stars such as the algol - type secondaries ( e.g. * ? ? ?
we currently have very little idea of the physical and spatial characteristics and appearance of these much more active coronae , and whether or not magnetic structures are restricted to the stellar surface or whether significant x - ray emitting plasma resides in larger structures , including inter - binary loops , with magnetic structures linking the two stars @xcite .
@xcite analyzed exosat observations of algol and found no clear x - ray eclipse during the optical secondary minimum .
if this x - ray emission was indeed from the secondary star , his study suggests that the coronal extent of algol b must be comparable to or greater than the size of the star itself .
however , @xcite did detect a secondary x - ray eclipse in rosat observations , and concluded that the scale height of the algol b corona is @xmath6 .
similarly , a shallow eclipse at optical secondary minimum was seen by asca @xcite .
a dramatic eclipse of a large flare observed by bepposax enabled the size of the flaring structure to be estimated as being less than 0.6 stellar radii and pinpointed the location of the structure to polar regions @xcite .
based on these deductions and on observations of flares on algol seen by ginga , exosat , rosat , and xmm - newton , @xcite suggest that the corona is essentially concentrated onto the polar regions of the k star , with a more compact ( smaller than the star ) flaring component and a perhaps somewhat more extended ( comparable in size to the star ) quiescent corona .
such a picture of polar - dominated activity has long been suspected based on optical doppler imaging techniques , which consistently see dark polar spots on active stars ( e.g. * ? ? ?
@xcite also suggested polar emission was responsible for the lack of rotational modulation in the observed flux in a _
low energy transmission grating spectrograph ( letgs ) observation of algol .
however , an xmm - newton observation of an algol flare studied by @xcite was interpreted as lying at lower latitudes . while this interpretation is somewhat subjective ,
it does suggest that even if polar emission dominates , significant activity at all latitudes is likely present .
similar conclusions regarding the corona of the k star might be drawn based on radio detections of algol .
in particular , @xcite presents results from 8.4 and 15 ghz observations of algol obtained using the very long baseline array ( vlba ) in 1995 .
the vlba maps show a double - lobed structure in the radio corona of the k star , which appear to originate in or near the polar caps .
the quiescent radio emission does not show signs of orbital modulation , again suggestive of a corona concentrated on the poles of the k star .
caution is warranted in the interpretation of the spatial distributions of radio and x - ray emission , however , since the former is produced by gyrosynchrotron emission from relativistic electrons , and this population might not be co - spatial with the thermal electrons responsible for x - ray emission .
direct x - ray observations present the only means to probe the spatial distribution of this thermal population . in this paper
, we primarily use the _ chandra _ high energy transmission grating spectrograph ( hetgs ) x - ray spectrum of algol to test this emerging picture of coronal activity .
we compare observed doppler shifts with the theoretical orbital velocities of the primary and secondary stars of algol , and compare the observed line widths with theoretical expectations based on instrumental , thermal , and rotational broadening .
based on these comparisons , we are able to place the first observational limits on the contribution of the b8 dwarf to the x - ray emission of algol , and to place constraints on the scale height , and any non - thermal motions , characterizing the corona of algol b. finally , we show that lines of o vii seen from quadrature to primary eclipse in a low energy transmission grating spectrograph ( letgs ) observation of algol support our conclusions that the corona of algol b at temperatures of a few @xmath7 k must be radially extended to at least one stellar radius .
the _ chandra _ hetgs employed for the observations used in this analysis consists of two gratings the high energy grating ( heg ) and the medium energy grating ( meg ) .
the heg covers a wavelength range of 1.2 - 15 ( 10.0 - 0.8 kev ) , with a typical line width of @xmath80.012 .
the meg covers a wider wavelength range of 2.5 - 31 ( 5.0 - 0.4 kev ) , with a typical line width of @xmath80.023 .
a more thorough description of this instrumentation can be found in @xcite .
we made use of both heg and meg spectra .
the hetgs observation of algol ( obsid 604 ) studied here was obtained as part of the guest observer ( go ) program and was undertaken in the standard instrument configuration using the acis - s detector between ut 02:20 and 17:31 on 2000 april 1 , for a total effective exposure time of 54660 s. in order to investigate the accuracy of the instrument line response function , we also analyzed an observation of capella ( obsid 2583 ) obtained for routine calibration .
this observation was also taken in the standard instrument configuration using the acis - s detector between ut 17:46 and 02:03 on 2002 april 29 to april 30 , for a total exposure time of 29700 s. we chose this particular observation of capella because it was obtained at an orbital phase of @xmath8 0.5 , when the two giants were near conjunction and spectral features were minimally broadened by orbital motion .
we also analyzed an letgs observation of algol ( obsid 2 ) , in order to examine the ratio of the o vii forbidden to intercombination lines as a function of orbital phase .
the data were observed between ut 18:22 and 17:11 on 2000 march 12 to march 13 , for a total exposure time of 82200 s. for details of the observation itself we refer to @xcite .
the standard ciao pipeline - processed ( version 2.2 ) data were downloaded from the chandra public data archive .
subsequent analysis was done using the idl software package pintofale @xcite .
the first order meg and heg spectra are illustrated in figure [ f : spec ] , together with identifications of prominent spectral lines .
emission lines used in our analyses are labeled with a larger font .
prior to spectral analysis , we examined the lightcurves of both algol and capella in order to determine their level of variability and whether any significant events ( such as large flares ) occurred during the observations that may be relevant for subsequent interpretation of doppler shifts and line broadening .
the lightcurves were derived by extracting all _ dispersed _ events in the standard ciao spectral extraction region , then binning the events at 100 second time intervals .
we emphasize here that we did not use the 0th order events since these are strongly affected by pile - up .
the observed lightcurve of capella is relatively flat , with no obvious signs of flaring , or other variability .
the lightcurve of algol on the other hand , shows a definite and significant flare in the beginning of the observation ( figure [ f : lc ] ) .
note that the orbital phase of the algol observation begins at @xmath9 = 0.48 , just before algol b starts to come out of eclipse .
it is possible that the impulsive phase of the flare has been affected by geometric occultation and we return to this in
[ s : results ] .
the flare has decayed by @xmath10 , soon after algol b passes conjunction .
algol b appears to be quiescent for the remainder of the observation , which ends at @xmath11 . in this section
, we investigate the algol spectra for evidence of doppler shifts resulting from orbital motion .
the following analysis is restricted to the emission lines listed in table [ t : whichlines ] .
we have chosen these particular lines because they have high s / n ratios and are not significantly blended with lines of other atomic species .
[ h ] c c c c wvl [ ] & ion & grating & diffraction order + 8.42 & mg xii & meg & 1,3 + 12.13 & ne x & meg , heg & 1 + 15.01 & fe xvii & meg & 1 + 16.78 & fe xvii & meg & 1 + 18.98 & o viii & meg & 1 + 24.78 & n vii & meg & 1 + [ t : whichlines ] orbital velocity as a function of orbital phase is derived separately for each emission line listed in table [ t : whichlines ] . for each line , we first bin the events into time intervals ( which are later converted to orbital phase ) , where the bin size is proportional to the total observed counts of the emission line , thereby maintaining an approximately constant s / n ratio for each bin .
negative and positive order events ( in both meg and heg ) were combined to obtain higher signal to noise ( s / n ) ratios . since we are trying to accurately measure the wavelength centroids of emission lines ,
what is important here is the s / n ratio , rather than the resolution .
although the heg has better resolution than the meg , most of even the brightest lines proved too faint for useful analysis in heg spectra .
most of the lines we analyze here are therefore meg lines . a modified lorentzian function ( ` beta - profile ' ) described by the relation @xmath12 where @xmath13 is the amplitude and @xmath14 is a characteristic line width
, is then fit to the events along the wavelength axis for each time interval . with @xmath15 ,
this function has been found to be a good match to observed hetgs line profiles , better than gaussians ( * ? ? ? * in preparation ) . in this way
, we obtain the observed wavelength centroids as a function of time and orbital phase .
the line - of - sight orbital velocities were derived for all emission lines by subtracting the rest wavelengths of lines from their observed centroids .
we expect the dominant component of the doppler shifts to result from the orbital motion of algol b , which amounts to @xmath16 km s@xmath2 at quadrature .
the predicted orbital velocity is described by @xmath17 where @xmath18 is the radius of orbit , @xmath19 is the inclination , @xmath9 is the orbital phase , and @xmath20 is the orbital period .
we used an inclination of 81.4 degrees @xcite and a period of 2.87 days .
we fitted a sine model to the observed line wavelengths using the ciao fitting engine sherpa .
the model used to fit the data is given by @xmath21 ) \label{e : sinemodel}\ ] ] where @xmath22 is a constant y - offset [ ] , @xmath23 is the rest wavelength of the emission lines [ ] , @xmath24 is the speed of light [ km s@xmath2 ] , @xmath26 is the amplitude of the final fit which represents the line - of - sight orbital velocity [ km s@xmath2 ] ( given by 2@xmath27r @xmath28/p ) , @xmath9 is the orbital phase , r is the effective orbital radius of x - ray emitting material [ km ] , @xmath19 is the inclination of orbit [ radians ] , and p is the orbital period [ s ] .
the parameters for all emission lines were forced to be the same , and the orbital period of algol b was fixed at the value noted above .
the only parameters allowed to vary were @xmath22 and @xmath26 .
we allowed @xmath22 to be a free parameter because the absolute wavelength calibration of hetg+acis - s observations is @xmath8 0.011 for meg and @xmath8 0.006 for heg .
therefore , each emission line may need a different @xmath29 offset in order to account for wavelength calibration uncertainties . because @xmath22 was a free parameter , errors on the reference wavelengths , though extremely small for the h- and he - like ions , are not relevant for our analysis .
the data for all spectral lines were fit simultaneously .
the amplitude of the best fit was a=187.5 km s @xmath2 . from the amplitude
, we can compute the effective orbital radius of x - ray emitting material , @xmath18 ( equation [ e : vorb ] ) .
we obtain an effective orbital radius of 10.74 @xmath30 0.93 r@xmath31 .
this result will be further discussed in
[ s : results ] .
figure [ f : orbit_all ] illustrates the observed line of sight orbital velocity as a function of orbital phase , derived from various emission lines .
figure [ f : orbit_all ] also shows the theoretical orbital velocities of algol a and algol b , calculated under the assumption that the x - ray emission is centered on one of the two stars ( i.e.the expected curve if _ all _ of the x - ray emission had been located at the poles of algol b or algol a ) .
in addition to measuring individual emission lines , we also utilized a cross - correlation technique to obtain doppler shifts as a function of orbital phase .
events were binned into 5 time intervals corresponding to different orbital phases , and spectra were extracted for each phase bin .
spectra from each phase bin were cross - correlated with a reference spectrum whose orbital velocity is known .
we used capella ( obsid 2583 ) as our reference spectrum because , as noted in
[ s : observations ] , this particular observation was taken when the two giants were near conjunction in orbital phase .
this ensures that the orbital velocity of the spectrum is nearly zero .
other systematic velocities involved can all be determined with high precision and subtracted : these are the radial velocities of the capella and algol systems , and the velocity of the _ chandra _ satellite in the direction of these objects during the observations .
algol has a radial velocity of only 4 km s@xmath2 @xcite , while capella has a radial velocity of 30 km s@xmath2 @xcite , therefore we must redshift the algol velocities by 26 km s@xmath2 .
the orbital velocity of _ chandra _ during these observations was less than @xmath82 km s@xmath2 .
since this velocity is small compared to the errors incurred during the cross - correlation analysis , we ignored this orbital motion .
the algol and capella spectra were cross - correlated by shifting the algol spectrum with respect to the capella spectrum in velocity steps of 25 km s@xmath2 , ranging from -200 to 500 km s@xmath2 . at each velocity shift , we computed the @xmath32 value as follows : @xmath33 ^ 2}{\sigma_{ref}(\lambda)^{2}+\sigma_{2}(\lambda(1+\frac{v}{c}))^2 } ; \label{e : chi2}\ ] ] where @xmath34 and @xmath35 are the two spectra being compared ( capella and velocity shifted algol ) , and @xmath36 and @xmath37 are their respective errors .
these quantities are then summed over the entire spectral region .
the velocity shift at which minimum @xmath32 is achieved represents the doppler shift in that particular phase bin of the algol spectrum relative to the capella spectrum . in order to better understand the errors incurred from cross - correlating the two spectra and computing @xmath32
, we utilized a monte carlo method by randomizing the counts within errors on the algol spectrum 25 times , and repeating cross - correlation for each randomized data set .
parabolae were then fit to @xmath32 versus velocity shift for each of the 25 simulations in order to find the velocity that corresponds to the minimum @xmath32 of each simulation .
the mean velocity is then adopted as the final doppler shift for that given interval of phase , and the standard deviation of the mean is the @xmath38 error on the final doppler shift .
figure [ f : orbit_crosscor ] illustrates the line - of - sight velocity of algol as a function of orbital phase , as obtained by the cross - correlation / monte - carlo analysis method described above .
also illustrated are the theoretical orbital velocities of algol a and algol b. once again , we used the sherpa fitting engine to fit a sine model to the observed doppler shifts obtained via the cross - correlation analysis .
doppler shifts for the positive and negative orders , as well as both meg and heg , were fit simultaneously with the model described by equation [ e : sinemodel ] .
the amplitude of the sine model is a=173.4 @xmath30 5.1 km s@xmath2 .
assuming an orbital period of 2.87 days , and an inclination angle of 81.4 degrees , this amplitude corresponds to an orbital radius of 9.93 @xmath30 0.29@xmath39 .
this radius is consistent , within 1@xmath40 uncertainties , with our previous result of 10.74 @xmath30 0.93@xmath39 obtained by measuring wavelength centroids of individual emission lines .
we also applied the cross - correlation / monte - carlo analysis to the algol spectrum for regions greater than and less than 10 , in order to compare the line - of - sight velocities for lines formed at cooler versus hotter temperatures .
doppler shifts for both regions of the spectrum were consistent , though uncertainties for the hotter region of the spectrum were considerably larger ( @xmath850 km s@xmath2 ) than the cooler region ( @xmath810 km s@xmath2 ) , since most of the stronger emission lines lie at wavelengths greater than 10 . in order to test for the presence of non - thermal broadening in the observed algol line profiles
, we compare these with carefully - computed theoretical profiles .
since the dominant source of line - broadening in the observed spectra is the instrumental profile , it is important that we confirm that our understanding of the instrumental broadening is correct , before proceeding with the comparison . in a far ultra - violet study of capella
, @xcite found that the observed profile of fe xviii originated largely from the g8 component , and that only instrumental , rotational ( 3 km s@xmath2 ) , and thermal broadening were required to match the observed profile .
this gives us some confidence that the coronal spectra of the capella giants are free from significant additional sources of non - thermal broadening .
it is not certain which of the two capella stars dominated during the time of the observation analyzed here ; for example , @xcite found that , in 1999 september 12 , the g1 component dominated in the light of fe xxi 1354 seen by the space telescope imaging spectrograph ( stis ) .
the dominance of one component over the other is not important here . by constructing two theoretical profiles ( described below ) , one with rotational broadening of 3 km s@xmath2 ( g8 component ) and another with rotational
broadening of 36 km s@xmath2 ( g1 component ; @xcite ) , we find that the difference in fwhm of the two profiles are negligible of order 0.0001 and two orders of magnitude smaller than typical observed line widths .
theoretical line profiles were synthesized by convolving the predicted rotational , thermal , and instrumental line broadening , where rotational profiles were derived by assuming solid body rotation at the stellar surface ( e.g. * ? ? ?
in our final synthesis of capella theoretical line profiles , we have adopted a rotational velocity value of the g1 star ( 36 km s@xmath2 ; @xcite ) .
the width of the thermal profile is calculated as the weighted average of the thermal widths computed at various temperatures where the weighting function is given by the product of the emission measure and the line emissivity at each temperature .
the fwhm of the thermal width for a given ion , at a given temperature , is then described by : @xmath41 where @xmath42 is the emission measure , @xmath43 is the emissivity , @xmath44 is the electron temperature , @xmath45 is the boltzmann constant , and @xmath46 is the ion mass .
we derived an emission measure @xmath42 for capella based on the measured line strength ratios of h - like and he - like ions , as described by ( * ? ? ? * in preparation ) .
this method of using line strength ratios to derive an emission measure is similar to what is described in ( * ? ? ? * in press ) .
we defer a detailed discussion of this analysis to a future paper , but emphasize that the final thermal profile does not depend critically on the exact details of the emission measure distribution : choosing a single isothermal emission measure at a temperature of @xmath47 produced very similar results .
line profiles for the h - like lyman-@xmath48 doublets of o viii , ne x , and mg xii were each modeled with the doublet wavelength separation fixed at @xmath49 for ne x and mg xii , and @xmath50 for o viii @xcite .
both components of a given line were forced to have the same line width , and relative fluxes of the two components were fixed at the theoretical ratio of 2:1 .
the final theoretical line profiles are then created by convolving the rotational and thermal profiles with the expected instrumental broadening .
the instrumental line profile was derived from high - fidelity ray - traces of the _ chandra _ instruments ( * ? ? ? * in press ) . the predicted and observed line widths
are compared in table [ t : linewidths ] and figure [ f : order_comp ] .
these comparisons show that the theoretical capella line widths are in good agreement with the observed line widths of the data .
we are therefore confident that the current model of the instrumental line broadening is sufficient to yield theoretical profiles that are accurate to within errors of the observed algol line widths . prior to measuring the algol line widths
, we eliminated line broadening due to the orbital velocity of algol b by shifting the event positions by the predicted amount ( @xmath51 ) . here , we adopt the true radius of orbit of algol b , @xmath52 ; we also performed the same line width analysis assuming the smaller of the two effective radii , @xmath53 found from the doppler shifts above , but found no significant differences between the two .
the negative and positive order spectra were then coadded .
great care has to be taken here .
bright sources such as algol suffer from significant pile - up in the 0th order , resulting in a distorted source profile .
this distortion can confuse the standard pipeline 0th order location algorithm , resulting in a computed centroid that does not correspond exactly to the true centroid .
if the 0th order centroid location is not precise , there will be a systematic wavelength shift between opposite orders .
co - addition of the opposite orders will then produce an artificially broadened line profile . in order to avoid this problem , we therefore ensured that the line centroids in opposite spectral orders were identical prior to co - addition , on a line by line basis .
the widths of lines listed in table [ t : whichlines ] were then measured by fitting with a modified lorentzian function ( equation [ e : lorentz ] ) . since the measured line widths are sensitive to the adopted level of the continuum , we have computed a model continuum based on the derived emission measure distribution of algol and have used this to guide continuum placement . in this way
, the continuum level adopted makes use of the signal in the spectrum over a broad wavelength range and accurate placement can be achieved such that final uncertainties in measured line quantities do not have a significant contribution from continuum placement errors .
the final continuum placement was done by eye , using effectively `` line free '' regions as a guide . in order to verify that this continuum placement did not introduce significant errors in our line parameters , we performed a sensitivity test by changing the adopted continuum level by different amounts that stretched plausibility regarding the true continuum level , and re - measuring lines for the different levels .
although the continuum placement is partly subjective because of the pseudo - continuum due to the presence of weak lines and broad wings of closely spaced lines , in the worst case the uncertainty of the fwhm of any of our lines resulting from uncertainties in continuum placement is not more than @xmath80.002 .
for some of the stronger lines , such as ne x , it is unlikely that the line width uncertainty is at most @xmath80.001 .
theoretical profiles for algol were generated in the same way as described for capella in
[ s : capella ] .
the theoretical fwhm were then compared to the measured line widths of the observed data ; these comparisons are listed in table [ t : linewidths ] and illustrated in figure [ f : order_comp ] .
we find some evidence for moderate excess line broadening in several of the emission lines . in the meg spectra ,
all lines but n vii , o viii , and mg xii show signs of significant excess line widths .
the two fe xvii lines , ne x , and mg xii show significant excess line widths , ranging from 3@xmath40 to almost 6@xmath40 . in the heg data ,
ne x shows an excess width of 3@xmath40 , though mg xii shows no sign of excess line broadening .
in order to determine whether any of the observed excess broadening could be due to the flare activity in the first part of the observation we also performed the analysis described above for flare and quiescent periods .
line widths for both periods were found to be consistent within statistical uncertainties .
the excess widths we have detected in algol may be attributed to either or both of two plausible sources of non - thermal velocities : turbulence or flows within coronal structures , and `` excess '' rotational broadening above what is expected from surface emission .
figure [ f : profile ] is an example comparison between observed and theoretical line profiles .
this figure shows that a theoretical profile with an additional gaussian velocity of 125 km s@xmath2 is a reasonably good match to the fe xvii line , while an additional velocity of 300 km s@xmath2 is too wide , and a velocity of 0 km s@xmath2 is too narrow .
we obtain a crude estimate of the turbulent velocities involved by comparison of observed and theoretical line profiles , computed as described in the previous section .
we convolved the latter with an additional gaussian broadening corresponding to a range of velocities from 0 to 300 km s@xmath2 .
figure [ f : turbulent ] illustrates the fwhm of these synthesized profiles as a function of the additional non - thermal velocity , and also indicates where the observed fwhm fall on these curves .
most of the lines we have analyzed show observed fwhm that correspond to additional velocities of @xmath8 50 - 150 km s@xmath2 .
we are able to place an upper - limit on additional turbulent velocities at 170 km s@xmath2 .
excess rotational broadening can occur if the corona on algol b is significantly radially extended .
line widths can therefore be used to place direct spectroscopic constraints on this radial extension .
limits on the radial extent of the corona of algol b were determined by comparison of observed and theoretical profiles computed for different scale heights .
we varied the scale height of the theoretical rotational profiles from within the stellar surface ( zero rotation ) to 8.75@xmath54 , where the scale height is defined as zero at the stellar surface .
then we convolved these with the thermal and instrumental broadening , as before .
the final profiles are compared with the fe xvii 15.01 resonance line in figure [ f : profile ] .
this figure shows good qualitative agreement with a profile computed for coronal scale height of 3@xmath54 ; a scale height of 8@xmath54 is clearly too broad , while purely surface rotational broadening is too narrow .
the upper two panels of figure [ f : scale ] show the fwhm of the theoretical profiles as a function of coronal scale height , together with the observed fwhm overplotted in thicker lines
. more than half of the observed fwhm fall approximately at a scale height of @xmath8 2 to 3@xmath54 , which corresponds to an excess rotational velocity of @xmath8 125 to 185 km s@xmath2 , in addition to the rotational velocity at the stellar surface ( @xmath8 62 km s@xmath2 ) .
the bottom panel of figure [ f : scale ] shows a distribution of scale heights which were derived by randomly generating a set of line widths that are distributed in a gaussian manner , centered on the measured line width of each emission line .
the number of monte - carlo realizations for each emission line was weighted by the measured flux of that line .
we generated a total of 30087 realizations for all lines combined .
we then obtained the scale height that corresponds to each realization of a line width by interpolating along the theoretical curves shown in the top two panels of figure [ f : scale ] .
the lowest panel of figure [ f : scale ] shows the distribution of the scale heights for all realizations of line widths , for both meg and heg .
vertical lines indicate the median scale height , and the 1@xmath40 and 3@xmath40 limits .
the median indicates a scale height of 3.1@xmath54 , with 1@xmath40 and 3@xmath40 upper limits occurring at scale heights of 3.8 and 4.6 stellar radii , respectively .
the lower 1@xmath40 limit occurs at 0.9@xmath54 . in the case of line width realizations that are smaller than what is physically possible , these scale heights have been set equal to -1@xmath54 , which accounts for the large peak at -1@xmath54 .
the uv excitation of the @xmath56 term by algol a provides a possible test of coronal geometric extent . in the spherically - symmetric case
, a very compact corona should exhibit modulation of the ratio of the forbidden and intercombination lines , @xmath55 , with orbital phase : at phase @xmath57primary optical eclipse the visible hemisphere of algol b is not illuminated by algol a and the @xmath55 ratio should revert very closely to its pressure - dominated value . the active star survey of ( * ? ? ? * in preparation ) indicates that this ratio should be @xmath58 - 3 , instead of 0.94 , as observed by @xcite .
there is no evidence for orbital modulation of the @xmath55 ratio in the hetgs observation ; this is not surprising because the observation covered only those phases during which the visible k star corona was substantially illuminated by algol a. we have therefore re - examined the letgs observation ( obsid 2 ) that covers phases @xmath59 - 1.03 to search for evidence of orbital - modulated changes in the @xmath55 ratio .
the end of this observation corresponds to primary eclipse when the visible hemisphere of the k star is not irradiated by algol a. we split the algol letgs data into three phase bins ( lacking the signal in the o vii lines to make more fine divisions ) and examined the relative strengths of the intercombination and forbidden lines at 21.80 and 22.10 .
the spectral region containing these lines is illustrated for the three phase bins in figure [ f : o7letg ] : there is no perceptible change in the ratio between these different bins .
by modeling observed doppler shifts in terms of orbital motion , we were able to estimate the effective orbital radius of the x - ray emitting material .
we found this effective radius to be @xmath60 @xmath30 0.93@xmath39 from the analysis of individual emission lines , and @xmath61 @xmath30 0.29@xmath39 from the cross - correlation analysis ; these results are consistent within @xmath62 uncertainties .
the statistical uncertainties in the cross - correlation method are considerably lower because all the information in the spectrum is used , rather than just information in bright lines .
the 1@xmath40 error bar of the line - of - sight velocities obtained via the cross - correlation analysis is @xmath88 km s@xmath2 for the meg data and @xmath811 km s@xmath2 for the heg data .
however , limitations due to uncertainties in the calibration of the _ chandra _ instruments cause systematic uncertainties which appear somewhat larger than what is implied by the statistical uncertainties . for example , a few of the data points in figure [ f : orbit_crosscor ] show discrepancies between the meg and heg results , by @xmath82@xmath40 . in comparison with the x - ray doppler radius , the orbital radius of algol b around the system center of mass
is @xmath63 @xcite .
since the effective radius we have derived is smaller than @xmath64 , we can infer that the x - ray emitting material is in fact not perfectly centered on algol b , but is shifted slightly inward toward the primary star .
the effective orbital radius of the x - ray emission allows us to place a constraint on the possible contribution from algol a emission that would skew the apparent x - ray doppler radius toward the center of gravity . as mentioned in
[ s : intro ] , it has been argued that accretion onto the primary star could be a source of x - rays .
if this were true , a small x - ray contribution from the accretion activity of algol a could explain the inward shift of the effective radius of x - ray emitting material .
assuming that the emission of the k star corona is centered on the star itself , an effective radius of @xmath65 for the x - ray emission would indicate that @xmath885% of the total x - ray emission is from algol b , and the remaining 15% is from algol a. @xcite considered the issue of accretion - driven x - ray emission in algol - type binaries in their comparison between algols and rs cvn - type binaries .
the latter are comprised of two late - type stars in which neither component filled their roche lobes .
@xcite found that the algol - type binaries are in fact slightly x - ray deficient relative to their rs cvn cousins , suggesting strongly that the accretion activity of algol - type binaries is not a significant source of x - rays .
the 15% effect we are seeking here would , however , be quite inconspicuous in this type of statistical study .
the possibility of accretion giving rise to significant x - ray flux in algol was reviewed by @xcite .
two key parameters are the shock temperature of the accreting gas , and the mass transfer rate .
it seems unlikely that the accretion shock can exceed @xmath66 k , and pustylnik suggests a maximum x - ray luminosity from accretion of about @xmath67 erg s@xmath2 based on a mass transfer estimate of order @xmath68 yr@xmath2 .
the x - ray luminosity of algol as measured from the hetg observation analyzed here is about @xmath69 erg s@xmath2 ( * ? ? ?
* in preparation ) ; accretion does therefore appear sufficient to account for @xmath70% of the total luminosity .
one test of this would be whether or not lines formed only above @xmath71 k lines that could only plausibly originate from coronal emission
show the same center of gravity as lines formed at cooler temperatures .
unfortunately , our data are of insufficient quality to perform this test since only weaker lines in the spectrum exclusively originate from such hot plasma .
the brightest line formed at these higher temperatures is h - like si xiv @xmath726.18 .
the line - of - sight velocity as a function of orbital phase indicated by this line is consistent within statistical uncertainties with the velocity seen in the other , cooler lines ( figure [ f : orbit_all ] )
. however , measurement errors are sufficiently large that we can not rule out differences at the 15% level .
evidence for plasma associated with accretion and with temperatures of at least @xmath73 k has , however , been found in the algol systems v356 sgr and tt hya from recent fuse observations ( * ? ? ?
* ; * ? ? ?
* respectively ) .
emission detected from o vi appeared to be associated with a bipolar flow that makes a large angle with the orbital plane .
however , in order for this plasma to contribute significantly at x - ray wavelengths , it must be comprised of components at least an order of magnitude hotter than the formation temperature of o vi .
another possible , and we suggest more likely , explanation for the inward shift of the effective radius is that the corona of algol b is not spherically symmetric or exactly centered on the center of mass of algol b , but rather has some asymmetry and structure on the side facing inward , toward algol a. the surface of algol b itself will be severely distended in this direction by the gravitational field of algol a the roche lobe filling factor for algol b is expected to be very close to unity , such that any equatorial corona could well lie beyond the l1 point and would not be gravitationally bound to algol b. we illustrate the approximate geometry of the system in figure [ f : geom ] .
this figure was produced using the _ nightfall _ program by r. wichmann , which accounts for the relative masses of the two stars , the roche lobe filling factors , and inclination of the orbit .
we also note that , in the x - ray lightcurve ( figure [ f : lc ] ) , we see a flare just as algol b comes out of eclipse .
this flare does not have a sharp rise , but a more gradual one that appears to start before eclipse egress .
it is possible that this flaring plasma is located on the hemisphere facing algol a , and that the characteristics of its intensity evolution over time are modulated by the eclipse .
if the flaring plasma is indeed located on the hemisphere facing algol a , this could also explain why the effective radius of the x - ray emission during this observation is shifted toward algol a. by comparing observed with theoretical line widths , we have some found evidence for moderate excess line widths that we attribute to the possible presence of non - thermal broadening . while not definitive , this result is supported by recent analyses of far ultraviolet spectra obtained by the hubble space telescope @xcite and the far ultraviolet explorer @xcite .
both studies report the need for excess line broadening to understand the profiles of the forbidden lines of fe xxi at 1354 and fe xviii at 976 in the more active stars of their sample with the largest projected rotational velocities ( @xmath74 ) .
one source of excess broadening might be turbulence or `` explosive events '' .
our line profile analysis ( e.g. , figures [ f : turbulent ] & [ f : profile ] ) suggests that a random velocity component of @xmath75 km s@xmath2 could explain the observed line widths .
this is similar to the sound speed at @xmath76 k. such a velocity is reminiscent of the non - thermal broadening of over 100 km s@xmath2 seen in transition region lines of stars of different activity as reviewed by @xcite .
it was suggested in this body of work that the broadening might be caused by the acceleration of plasma in magnetic reconnection events associated with microflaring .
it is possible that the broadening we see in hot x - ray lines is related to this .
other types of flows are also possible . @xcite observed flows with velocities up to 40 km s@xmath2 in solar coronal loops based on detailed transition region and coronal explorer ( _ trace _ ) time - resolved imaging and simultaneous solar and heliospheric observatory ( _ soho _ ) sumer spectra . while the velocities inferred here are three times this , it is not difficult to envisage faster flows when considering that the coronal energy deposition rate of algol b is @xmath77 times that of the sun .
however , as pointed out by @xcite , in both the above cases the broadening mechanisms should be ubiquitous among the more active stars and should not be seen only in those with the largest @xmath74 .
another possible source of excess line widths is rotational broadening originating in a corona with significantly radially - extended structure . as noted in
[ s : intro ] , the evidence from observations of algol during eclipse does not provide an unambiguous picture , with some observations requiring apparently extended coronae @xcite , and others being consistent with scale heights of a stellar radius or less @xcite .
the scale height for plasma of mean molecular weight @xmath78 at a temperature @xmath44 on a star of mass @xmath79 and radius @xmath80 is @xmath81 .
the parameters for algol b were summarized recently by @xcite , who adopted @xmath82 and @xmath83 ( see also * ? ? ? * ; * ? ? ? * ) . for a coronal temperature @xmath84
k on algol b , the scale height is then about @xmath85 in the absence of strong centrifugal forces .
our median scale height and 1@xmath40 upper limit are 3.1@xmath54 and 3.8@xmath54 , respectively .
while this is slightly larger than the thermal scale height in the absence of strong centrifugal forces arising from rapid rotation , it is important to keep in mind the following things .
firstly , the line profile analysis depends on subtraction of the dominant instrumental and thermal broadening components and is inherently prone to additional systematic error arising as a result of imperfections in the description of these dominant effects .
secondly , for a corona on algol b with a radial extent comparable to the stellar radius , the competition between gravitational and centrifugal forces becomes significant : centrifugal acceleration on a single star with the same rotation rate is equivalent to gravitational acceleration at a height of @xmath86 . while the binary nature of algol complicates the picture , it is possible that some regions of an extended corona comprise plasma bound by magnetic fields , as has been discussed in connection with other rapidly rotating stars , such as ab dor ( e.g. * ? ? ?
( see also @xcite who concludes that the 1032 o vi profile of ab dor , observed with orpheus , is not rotationally broadened beyond the photospheric surface @xmath74 value . )
we conclude that , in the interpretation that the tentative evidence for excess line widths is the result of rotational broadening , the extension of the corona would be similar to the expected thermal scale height and comparable to the stellar radius .
this interpretation is consistent with that of @xcite based on their fuse analysis of active stars . these authors suggested that their excess line widths could result from coronal structures with heights up to @xmath87 .
such a picture would also be consistent with the eclipse studies of @xcite and @xcite .
with regard to eclipses , it is important to note that extended structure at the stellar poles would give rise to no obvious x - ray eclipse .
however , pole - dominated emission also would not give rise to strong rotational broadening .
our results suggest that , if rotational broadening is the correct interpretation for our observed line widths , this emission must be distributed around the star and not concentrated at the poles .
the hetg observation does cover the secondary optical eclipse , and there is some evidence that we do indeed see some obscuration of the x - ray corona , despite the complications presented by the flare .
we have computed theoretical x - ray lightcurves for algol using a spherically - symmetric , optically - thin emitting shell coronal model .
the model assumes that algol a is 100% x - ray dark , and uses the following physical and orbital parameters : @xmath88 , @xmath89 , @xmath90 , @xmath91 , @xmath92 days , and @xmath93 degrees @xcite .
theoretical lightcurves were calculated for an array of algol b coronal scale heights , ranging from 0.1 to 4.0 stellar radii .
figure [ f : lc ] shows the observed lightcurve of algol with three theoretical lightcurves of various scale heights overplotted .
also plotted is a similar curve computed for polar emission restricted to high latitudes corresponding to the scale height predicted by quasi - static loop models ( see below ) .
if we assume that the count rate at phase @xmath8 0.5 represents the quiescent level of algol b , as does the level toward the end of the observation , we can see from the overplotted models that scale heights of @xmath94-@xmath95 represent the data reasonably well .
the polar emission model clearly does not represent the observed eclipse minimum ; we discuss this further below .
a scale height commensurate with the thermal scale height would appear to clash with the recent interpretation of density - sensitive line ratios seen in _
letgs observations of algol by @xcite .
the interpretation of helium - like line ratios in terms of plasma density is complicated in the case of algol by the radiation field of algol a , which is sufficiently strong to cause significant radiative excitation of electrons from the upper level of the forbidden line to the upper level of the intercombination line through the transition @xmath4 in the ions c , n and o. nevertheless , @xcite pointed out that , even with densities of order @xmath96 @xmath97similar to those found for other active stars in the extensive survey of ( * ? ? ?
* in preparation)simple quasi - static coronal loop model scaling laws ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) that predict that loop peak temperature @xmath44 , pressure @xmath20 and length @xmath98 are related by @xmath99 , suggest maximum loop lengths of order a few @xmath96 cm .
this is only a fraction of the radius of algol b. as mentioned in [ s : f / i_analysis ] , there was no detectable orbital modulation of the o vii @xmath55 ratio in the algol letgs data , which was observed during orbital phase @xmath59 - 1.03 .
@xcite also found no differences in the @xmath55 ratio of algol spectra which was extracted during the quiescent and flaring phases of an xmm - newton observation .
this result has important implications .
if we assume that the corona of algol is indeed similar to that of other active stars , such as its `` coronal twin '' hr 1099 @xcite , then we must conclude that _ the bulk of the o vii emission is always illuminated by algol a_. this can be achieved by having a fairly extended corona with a scale height @xmath100 , as our line profile analysis indicates . instead , _ if the o vii emitting coronal structures are compact compared to the stellar radius
, then they must reside predominantly at the poles _ where the algol a uv radiation field is not significantly shadowed at primary eclipse
. if scale heights are fairly large ( @xmath101 ) , the classical quasi - static loop scaling laws must then be quite inapplicable to these coronal structures . in the case of compact ( by necessity ) quasi - static loops with a plasma density of a few @xmath96 @xmath97 ,
@xcite estimate a coronal volumetric filling factor of up to 0.3unreasonably large for emission confined to polar regions . as noted earlier in this section ,
the synthetic lightcurve computed for the case of polar emission with a scale height of @xmath102 , similar to that predicted by quasi - static models , is a very poor match to the observed eclipse minimum light ( figure [ f : lc ] ) .
this model predicts very sharp eclipses which have never been observed in x - ray observations of algol . in summary , the combined evidence of a tentative detection of excess line widths , the shape of the x - ray lightcurve , and the lack of orbital modulation of the o vii @xmath55 ratio going into primary eclipse , all point to the plasma source at temperatures up to several @xmath7 k having a scale height of the same order as the stellar radius , and being distributed around the k star . such distributed coronal structure could be responsible for , e.g. , the flaring at lower latitudes , which was inferred by @xcite based on an xmm - newton x - ray lightcurve .
unfortunately , the x - ray data alone can not rule out other scenarios involving compact structures , though these would require special and fortuitous placement of dominant active regions in order to remain consistent with past and present observations of the coronal emission of algol .
@xcite drew similar conclusions based on the failure to observe a strong x - ray eclipse in the exosat observation , though with the benefit of not having seen data from subsequent observations that have led to strong ambiguities in interpretation .
our favored picture of the x - ray emission of algol is also consistent with the picture painted by @xcite , and with the interpretation of excess line widths seen in uv and fuv coronal forbidden lines by @xcite .
what we have not tackled in this article , however , is the very hot ( @xmath103 k ) plasma that often flares and seems to be located at the poles .
the recent plasma density survey of ( * ? ? ? * in preparation ) indicates that this very hot material is at much higher densities of @xmath104 @xmath97 , and so occupies quite different structures .
these are likely to be the structures responsible for pole - dominated flaring activity , such as that highlighted by @xcite .
a detailed analysis of high quality _ chandra _ hetgs spectra of the algol system provides new insights into the geometry of the x - ray emission in this system .
based on this analysis we draw the following conclusions : 1 .
our study clearly reveals doppler shifts of the x - ray emitting plasma corresponding to orbital motion with a velocity at quadrature of 150 km s@xmath2 .
these data thus provide the first definitive proof that the coronal activity of algol b dominates the x - ray emission of the system , as has long been suspected ( e.g. , * ? ? ?
the observed doppler motion of the x - ray plasma on algol b appears to be off - center relative to the stellar center of mass , and shifted toward algol a. we suggest that this occurs as a result of the tidal distortion of the surface of algol b. the hetgs light curve exhibits a flare near secondary eclipse which , if located on the hemisphere facing the primary b8 star as expected , might also suggest the presence of a dominant active region that could bias the emission toward a slightly smaller orbital radius .
3 . alternatively
, x - ray activity of algol a , possibly as a result of accretion from algol b , could be responsible for the smaller apparent orbital radius of the x - ray emission . in this case
, we estimate that such a contribution amounts to no more than 15% of the total emission .
such an x - ray flux would be consistent with current ideas as to the plausible range of accretion activity in algol .
we have found some evidence for excess line broadening in bright x - ray emission lines , above that expected from surface rotation and thermal motions .
if this effect is real , it suggests broadening by rotation of a radially - extended corona . the observed widths would require a coronal scale height of at least one stellar radius
this would be consistent with recent observations that detected excess broadening in the forbidden fe xviii 974 line , seen in fuse fuv spectra of rapidly rotating active stars .
5 . while turbulence , microflares or flows could also produce excess line broadening , the idea of a coronal plasma with significant radial extent is also supported by uv - sensitive lines of o vii seen in an letgs observation of algol .
no change in the uv - sensitive lines was seen going into primary eclipse , when the majority of the visible hemisphere of algol b was shadowed from the radiation field of algol a. 6 .
while it is possible that other physical effects are responsible for excess line widths and lack of modulation of the uv - sensitive o vii lines , coronal extension would appear to be the most likely .
a consistently high plasma density of a few @xmath105 @xmath97 , as would be indicated by the o vii lines in the absence of uv excitation , would be unlike _ any _ of the densities seen at o vii temperatures in the sample of active stars surveyed by ( * ? ? ? * in preparation ) .
this picture also appears to be consistent with conclusions drawn from recent lightcurve analyses which were summarized in [ s : intro ] . +
coupled with the results of the coronal density survey of ( * ? ? ?
* in preparation ) , we can summarize as follows .
coronal plasma with temperatures of up to several @xmath7 k appears to have a significant component lying in extended structures with a scale size similar to the stellar radius and densities typically of a few @xmath96 @xmath97similar to solar active regions .
very hot coronal plasma appears more exclusively in much more compact regions with densities of order @xmath104 @xmath97 , and in very active , rapidly rotating stars appears to be concentrated more toward stellar poles .
further spectroscopic studies of these apparently different coronal regimes , especially with regard to possible abundance anomaly differences between them , should prove very interesting .
24.78 & 0.0306 @xmath30 0.0033 & 0.0217 @xmath30 0.0055 & 0.0285 & 0.0267 & 0.0021 + 18.97 & 0.0256 @xmath30 0.0013 & 0.0216 @xmath30 0.0009 & 0.0249 & 0.0238 & 0.0006 + 16.78 & 0.0284 @xmath30 0.0025 & 0.0190 @xmath30 0.0007 & 0.0212 & 0.0211 & 0.0072 + 15.01 & 0.0266 @xmath30 0.0017 & 0.0196 @xmath30 0.0005 & 0.0199 & 0.0199 & 0.0067 + 12.13 & 0.0266 @xmath30 0.0008 & 0.0205 @xmath30 0.0006 & 0.0217 & 0.0212 & 0.0049 + 8.42 & 0.0211 @xmath30 0.0009 & 0.0191 @xmath30 0.0008 & 0.0197 & 0.0195 & 0.0014 + , t. h. , canizares , c. r. , dewey , d. , mcguirk , m. , pak , c. s. , & schattenburg , m. l. 1994 , in proc .
2280 , p. 168 - 180 , euv , x - ray , and gamma - ray instrumentation for astronomy
v , oswald h. siegmund ; john v. vallerga ; eds . , 168180 | in a study of _ chandra _ high energy transmission grating spectra of algol , we clearly detect doppler shifts caused by the orbital motion of algol b. these data provide the first definitive proof that the x - ray emission of algol is dominated by the secondary , in concordance with expectations that the primary b8 component should be x - ray dark
. however , the measured doppler shifts are slightly smaller than might be expected , implying an effective orbital semi - major axis of about @xmath0 instead of @xmath1 for the algol b center of mass .
this could be caused by a small contribution of algol a , possibly through accretion , to the observed x - ray flux , in which case such a contribution does not exceed 10 - 15% .
we suggest the more likely explanation is an asymmetric corona biased toward the system center of mass by the tidal distortion of the surface of algol b. a detailed analysis of the profiles of the strongest lines indicates the presence of excess line broadening amounting to approximately 150 km s@xmath2 above that expected from thermal motion and surface rotation .
possible explanations for this additional broadening include turbulence , flows or explosive events , or rotational broadening from a radially extended corona .
we favor the latter scenario and infer that a significant component of the corona at temperatures @xmath3 k has a scale height of order the stellar radius .
this interpretation is supported by the shape of the x - ray lightcurve and tentative detection of a shallow dip at secondary eclipse .
we also examine the o vii intercombination and forbidden lines in a low energy transmission grating spectrograph observation and find no change in their relative line fluxes as the system goes from quadrature to primary eclipse . since these lines appear to be strongly affected by uv irradiation from algol a through radiative excitation of the @xmath4 transition , this supports the conjecture that the corona of algol b at temperatures of several million k must be significantly extended and/or located toward the poles to avoid being shadowed from algol a during primary eclipse . |
i am indebted to l. m. simons who , inspired by the colella - overhauser - werner experiments , put forward the idea of measuring the gravitational interaction of antimatter with a mach - zehnder interferometer at a conference in the 1980s .
he has also repeatedly argued that muonium would be almost completely antimatter and thus be a reasonable target for an experimental test of its gravitational interaction - suffering only from insufficient source strength .
i am also thankful to d. taqqu for discussions and explanations concerning the production of a suitable muonium beam . | recently a new technique for the production of muon ( @xmath0 ) and muonium ( @xmath1 ) beams of unprecedented brightness has been proposed .
as one consequence and using a highly stable mach - zehnder type interferometer , a measurement of the gravitational acceleration @xmath2 of muonium atoms at the few percent level of precision appears feasible within 100days of running time .
the inertial mass of muonium is dominated by the mass of the positively charged - antimatter - muon .
the measurement of @xmath2 would be the first test of the gravitational interaction of antimatter , of a purely leptonic system , and of particles of the second generation .
the gravitational acceleration of antimatter has not been measured so far .
an experiment with antiprotons ( see @xcite and references therein ) did not succeed because of the extreme difficulty to sufficiently shield the interaction region from electromagnetic fields . for a similar reason , results of measurements with electrons @xcite are discussed very controversial and the plan to eventually compare with positrons was never realized . not affected by these problems are neutral systems like antihydrogen and , consequently , considerable effort today is devoted to the preparation of suitable samples of antihydrogen ( compare @xcite ) . a possibility to measure
the effect of gravitation on neutral particles is via a phase acquired in the gravitational potential in a suitably built interferometer , demonstrated in the classic colella
overhauser werner ( cow ) experiment @xcite . in case of limited source performance ,
when one has to deal with extended sources , comparatively large beam divergence and poor energy definition , mach - zehnder type interferometers have striking advantages @xcite
. their performance has been demonstrated , among others , with neutrons @xcite and atoms @xcite .
the idea to apply interferometry to the measurement of an antimatter system was inspired by the cow - experiments and dates back , as far as i know , to the 1980s @xcite but was put into print , with explicit mentioning of antihydrogen , positronium and antineutrons , first in 1997 @xcite .
common problems of the species are the quality of the particle beams and the availability of suitable , sufficiently large gratings .
the case of positronium was further elaborated suggesting the use of standing light waves as diffraction gratings @xcite but the realization of an experiment appears still very challenging . in the meantime also other experimental approaches to measure the gravitational interaction have been proposed for antihydrogen and positronium , see @xcite for an overview .
no discussion about a gravity experiment using muonium atoms ( m = @xmath1 ) appeared in the literature yet and the original idea of using m atoms for testing antimatter gravity is again by simons @xcite .
the suitability of m atoms comes from the fact that the inertial mass of the muon is some 207 times larger than the one of the electron , thus , muonium is almost completey , to 99.5% , antimatter .
an interesting feature is that m atoms are almost exclusively produced at thermal energies by stopping @xmath0 in matter which they often leave again as thermalized , hydrogen - like , m atom .
however , up to until recently , a gravitational experiment with muonium would have been science fiction .
the reason for this publication is , that there is now the real chance to perform such an experiment within the next few years .
an experiment with m atoms would constitute the first test of the gravitational interaction of antimatter with matter .
it would also be the first and probably unique test of particles of the second generation . while it would also be the first test in a purely leptonic system one should note that tests of the equivalence principle proving at a high level of precision that the gravitational interaction is independent of composition of test masses also in principle prove ( to still impressive precision ) that electrons fall in the same way as the rest of the material . for a recent review on tests of the equivalence principle see @xcite .
as a first measurement , the determination of the sign of interaction could be already interesting ( for a discussion of antigravity see @xcite , but also , e.g , @xcite ) , however , a reasonable first goal for such an experiment would be to determine @xmath2 to better than 10% .
one should add here , that it is not at all obvious that there could be a discrepancy between the gravitational interaction of matter and antimatter , see @xcite .
but the universality of @xcite has been disputed and possible scenarios have been sketched in @xcite .
anyhow , an experimentalist will probably favor the direct measurement ( and this , again , not only with respect to antimatter but also to a lepton of the second generation ) over the discussion of models .
the following quote from @xcite for _ antiprotons _ holds equally well for _ muonium atoms _ : `` it would be the first test of gravity , i.e. general relativity , in the realm of antimatter .
even if the experiment finds exactly what one expects , namely that antimatter falls toward the earth just as matter does , it would be , a classic , one for the text books. .... of course , if a new effect were found in the _ antiproton _ gravity experiment , then there would be no telling what exciting physics could follow . ''
the muonium experiment appears feasible now because of two recent inventions : ( i ) a new technique to stop , extract and compress a high intensity beam of positive muons , to reaccelerate the muons to 10kev and focus them into a beam spot of 100@xmath3 m diameter or even less @xcite ; and ( ii ) a new technique to efficiently convert the muons to m atoms in superfluid helium at or below 0.5k in which they thermalize and from which they get boosted by 270k perpendicular to the surface when they leave into vacuum @xcite .
assuming an existing surface muon beam of highest intensity as input , see e.g. @xcite , it should be possible to obtain an almost monochromatic beam of m atoms ( @xmath4 ) with a velocity of about 6300 m/s ( corresponding to 270k or a wavelength @xmath5 ) and a 1-dimensional divergence of @xmath6mrad at a rate of about @xmath7s@xmath8 m atoms @xcite .
this is a many orders of magnitude brighter beam than available up to now . following the approach of @xcite a mach - zehnder type interferometer
should be used in the muonium experiment . the principle with the source , the three grating interferometer and the detection region
is sketched in fig .
[ setup.eps ] .
we assume here three identical gratings and use the first two for setting up the interference pattern which is scanned by moving the third grating .
the setup is rather short , because the decay length of the m atoms is about 1.4 cm only ( @xmath9s ) .
the whole system from source to detection may be 4 decay lengths long , and without further collimation the source illuminates a cross section of less than 5 mm over the length of the interferometer .
the three free - standing gratings can be made sufficiently large with existing , proven technology with a period of 100 nm @xcite resulting in a diffraction angle @xmath10mrad .
the optimum distance @xmath11 between two gratings is slightly larger than one decay length ; however , for simplicity here @xmath12 cm . assuming another length @xmath11 each , for distances of the source and the detector to the nearest interferometer grating , results in 4 decay lenghts .
decay and transmission loss by the three 50% open ratio gratings reduces the initial m rate by a factor @xmath13 , yielding @xmath14s@xmath8 detected m. because only the indicated first order diffraction carries the desired information but essentially all transmitted m are detected , the interference pattern has a reduced contrast of somewhat below 4/9 . assuming a contrast of @xmath15 and using eqn .
( 3 ) of @xcite yields the statistical sensitivity of the experiment : @xmath16 which means that the sign of @xmath2 is fixed after one day and 3% accuracy can be achieved after 100days of running .
scheme of the experimental setup : the m beam comes from the cryogenic @xmath0 beam target on the left hand side , enters and partially traverses the interferometer and reaches the detection region on the right hand side .
the dimensions are not to scale and the diffraction angle @xmath17 is in reality smaller than the divergence . ] with the quite satisfactory statistics , the next important issues are the alignment and stability of the interferometer .
the gravitational phase shift to be observed is ( using the notation of @xcite ) @xmath18 this is rather small but still an order of magnitude larger than the phase shift due to the acceleration induced by the rotation of the earth ( sagnac effect : @xmath19 ) .
other accelerations of the system as a whole , e.g. from environmental noise , mainly affect the contrast and must therefore be suppressed .
the same is true for misalignments of the gratings and their drifts .
the effects must be kept below the phase shift , for example , for an unwanted translation @xmath20 of the third ( scanning ) grating perpendicular to the m beam and the lines of the grating one requires @xmath21 and consequently @xmath22 rotational misalignment of the gratings around the m beam must be much less than the period over beam height ratio , 100nm/5 mm , or 20@xmath3rad and corresponding drifts must not exceed 20nrad . in a similar way , limits for all other static or dynamic deviations from the perfect alignment of the three identical , equidistant , parallel gratings can be obtained .
the relatively small size of the interferometer is a major advantage for the stabilization . as in previous matter interferometry experiments
@xcite the muonium experiment must use ( multiple ) laser interferometry for alignment , monitoring and feedback position stabilization .
the gratings for the laser interferometry are ideally integrated in the m atom gratings as perfect alignment is required .
state of the art piezo systems can be used for positioning the gratings and for scanning of the third grating with a step precision of 50pm or better . because it will not be possible to perform a comparative experiment with the ( @xmath23 ) system
, the gravitational effect on the m atoms will be calibrated with a laser interferometer : in case of a vertical separation of the beam paths in the inteferometer , as in fig .
[ setup.eps ] , the m fringes will be shifted by gravity with respect to the laser interference pattern . by a @xmath24 rotation of the setup around the beam axis
the gravity shift can be removed from the m interference while the laser stays unaffected ( the gravitational redshift effects can be safely neglected at the level of precision required here ) .
of course , the sagnac effect has to be taken into account , which does not affect the interference in the vertical but will do so in the horizontal position .
given the possible sensitivity of the experiment it might be just at the edge , but in principle can be measured by horizontal measurements at @xmath24 and @xmath25 rotations of the setup , respectively .
also a setup with a vertical m beam would allow studying the sagnac effect and the stability of the interferometer .
this could also be a first stage of the experiment because for practical reasons the m beam development would probably first produce a vertical beam @xcite .
the muonium detection itself is straight forward : the @xmath26 from the @xmath0 decay must be detected in a way that the position of the m atom can be determined to be behind the interferometer .
a coincidence signal from the detection system is desirable in order to suppress background coming from the 3 orders of magnitude more muon decays upstream of the third grating .
one way to achieve this is to detect both , the decay @xmath26 and the remaining atomic @xmath27 from the m atom decay in flight .
the @xmath26 ( energies up to 53mev ) could be detected in a scintillation detector which is segmented along the beam axis and surrounds the m beam .
the @xmath27 ( energies of 10ev ) will be electrostatically accelerated out of the decay region onto , e.g. , a microchannel plate . in conclusion
, a muonium gravity experiment appears feasible today .
the development of the required m beam could be done within a few years and the interferometer set up in parallel and tested with its lasers interferometry offline .
it appears possible to have the gravitational acceleration of m atoms measured to better than 10% within the next 5 years . |
chiral symmetry together with its spontaneous breaking is one of the key ingredients in the low - energy hadron or nuclear physics . due to its spontaneous
breaking , up and down quarks , whose current masses are of the order of a few mev , acquire the large constituent masses of a few hundreds mev , and are consequently responsible for about 99% of mass of the nucleon and hence that of our world .
one would thus consider that chiral condensate @xmath7 , the order parameter of the chiral phase transition , plays an essential role in the hadron - mass genesis in the light quark sector . on the other hand ,
chiral symmetry gets restored in systems where hard external energy scales such as high - momentum transfer , temperature(@xmath8 ) , baryon density and so on exist , owing to the asymptotic freedom of qcd .
then , several questions may arise : are all hadronic modes massless in such systems ? can hadrons be massive even without non - vanishing chiral condensate ?
an interesting possibility was suggested some years ago by detar and kunihiro @xcite , who showed that nucleons can be _ massive even without the help of chiral condensate _
due to the possible _ chirally invariant mass terms _ , which give _ degenerated _ finite masses to the members in the chiral multiplet ( a nucleon and its parity partner ) even when chiral condensate is set to zero : in order to demonstrate this possibility for a finite-@xmath8 case , they introduced a linear sigma model which has a nontrivial chiral structure in the baryon sector and a mass - generation mechanism essentially different from that by the spontaneous chiral symmetry breaking .
interestingly enough , their chiral doublet model has recently become a source of debate as a possible scenario of _ observed parity doubling in excited baryons _
@xcite , although their original work @xcite was intended for an application to finite-@xmath8 systems .
it is thus an intriguing problem to clarify the chiral structure of excited baryons in the light quark sector beyond model considerations .
one of the key observables which are sensitive to the chiral structure of the baryon sector is axial charges @xcite .
the axial charge of a nucleon @xmath9 is encoded in the three - point function @xmath10 u.\ ] ] here , @xmath11 is the isovector axial current .
the axial charge @xmath12 is defined by @xmath13 with the vanishing transferred momentum @xmath14 .
it is a well - known fact that the axial charge @xmath15 of @xmath16 is 1.26 .
though the axial charges in the chiral broken phase can be freely adjusted with higher - dimensional possible terms and can not be the crucial clues for the chiral structure @xcite , they would reflect the internal structure of baryons and would play an important role in the clarification of the low - energy hadron dynamics . in this paper , we show the first unquenched lattice qcd study @xcite of the axial charges @xmath0 of @xmath1 and @xmath2 , two lowest nucleon resonances in the negative parity channel .
we employ @xmath17 lattice with two flavors of dynamical quarks , generated by cp - pacs collaboration @xcite with the renormalization - group improved gauge action and the mean - field improved clover quark action .
we choose the gauge configurations at @xmath18 with the clover coefficient @xmath19 , whose lattice spacing @xmath20 is determined as 0.1555(17 ) fm .
we perform measurements with 590 , 680 , and 680 gauge configurations with three different hopping parameters for sea and valence quarks , @xmath21 and @xmath22 , which correspond to quark masses of @xmath23 150 , 100 , 65 mev and the related @xmath24-@xmath25 mass ratios are @xmath26 , @xmath27 and @xmath28 , respectively .
statistical errors are estimated by the jackknife method with the bin size of 10 configurations .
our main concern is the axial charges of the negative - parity nucleon resonances @xmath1 and @xmath2 in @xmath29channel .
we then have to construct an optimal operator which dominantly couples to @xmath1 or @xmath2 .
we employ the following two independent nucleon fields , @xmath30 and @xmath31 in order to construct @xmath32 correlation matrices and to separate signals of @xmath1 and @xmath2 . here ,
@xmath33 and @xmath34 are dirac spinors for u- and d- quark , respectively , and @xmath35 denote the color indices . even after the successful signal separations , there still remain several signal contaminations : signal contaminations by _ scattering states _ and _ wraparound effects_. due to the unquenched gauge configurations , the negative parity nucleon states could decay to @xmath24 and n , and their scattering states could inevitably get into the spectrum .
the sum of the pion mass @xmath36 and the nucleon mass @xmath37 is however in our setups heavier than the masses of the lowest two states ( would - be @xmath1 and @xmath2 ) in the negative parity channel .
we then do not suffer from any scattering - state signals .
the other possible contamination is wraparound effects @xcite .
since we perform unquenched calculations , the excited nucleon @xmath38 can decay into @xmath9 and @xmath24 , and even when we have no scattering state @xmath39 , we could have another type of `` scattering states '' .
the baryonic correlator @xmath40 can still accommodate , for example , the following term .
@xmath41 here , @xmath42 denotes the temporal extent of a lattice .
such a term is quite problematic and mimic a fake plateau at @xmath43 in the effective mass plot because it behaves as @xmath44 . in order to eliminate such contributions
, we impose the dirichlet condition on the temporal boundary for valence quarks , which prevents valence quarks from going over the boundary .
( wraparound effects can be found even in quenched calculations @xcite . )
we employ zero - momentum - projected point - type operators for the sinks , and employ wall - type operators in the coulomb gauge for the sources .
after we diagonalize the @xmath32 correlation matrices , we can construct optimized operators for n(1535 ) and n(1650 ) states .
once we construct optimized operators , we can easily compute the ( non - renormalized ) vector and axial charges @xmath45}}$ ] for the positive- and negative - parity nucleons via three - point functions with the so - called sequential - source method @xcite . in practice ,
we evaluate @xmath46}}(t)$ ] defined as @xmath47}}(t ) = \frac { { \rm tr}\ \gamma_{a } \langle b(t_{\rm snk } ) j_\mu^{a}(t ) \overline b(t_{\rm src } ) \rangle } { { \rm tr}\ \gamma_{a } \langle b(t_{\rm snk } ) \overline b(t_{\rm src } ) \rangle } , \ ] ] and extract @xmath46}}$ ] by the fit @xmath46}}=g_{a}^{\pm{\rm [ lat]}}(t)$ ] in the plateau region .
@xmath48 denotes the ( optimized ) interpolating field for nucleons .
@xmath49 is @xmath50 , and @xmath51 is an axial vector current inserted at @xmath52 .
we show in fig .
[ 3pointfunc ] the non - renormalized axial charge @xmath53}}(t)$ ] for @xmath1 as a function of the current insertion time @xmath52 .
they are rather stable around @xmath54 . the non - renormalized axial charge of @xmath1 , @xmath53}}(t)$ ] , as a function of the current insertion time @xmath52 .
] we finally reach the renormalized charges @xmath55}}_{a}$ ] with the prefactors @xmath56 , which are estimated with the values listed in ref .
@xcite .
the renormalized axial charges of the positive- and the negative - parity nucleons are plotted as a function of the squared pion mass @xmath57 .
the solid line is drawn at @xmath58 and the dashed line is drawn at @xmath59 . ]
we first take a stock of the axial charge @xmath60 of the ground - state positive - parity nucleon , which is well known and can be the references . the axial charge @xmath60 of the positive parity nucleon is shown in fig . [ axialvectorc ] as open squares .
the axial charge of the positive parity nucleon shows little quark - mass dependence , and they lie around the experimental value 1.26 .
we finally show the axial charges of the negative - parity nucleon resonances in fig .
[ axialvectorc ] .
one finds at a glance that the axial charge @xmath61 of @xmath1 takes quite small value , as @xmath62 and that even the sign is quark - mass dependent . while the wavy behavior might come from the sensitiveness of @xmath61 to quark masses
, this behavior may indicate that @xmath61 is rather consistent with zero .
these small values are not the consequence of the cancellation between u- and d - quark contributions .
the u- and d - quark contributions to @xmath61 are in fact individually small @xcite . on the other hand , the axial charge @xmath63 of @xmath2 is found to be about 0.55 , which has almost no quark - mass dependence .
the striking feature is that these axial charges , @xmath64 and @xmath65 , are consistent with naive nonrelativistic quark model calculations @xcite , @xmath66 and @xmath67 .
such values are obtained if we assume that the wave functions of @xmath1 and @xmath2 are @xmath68 and @xmath69 neglecting the possible state mixing .
( here , @xmath70 denotes the orbital angular momentum and @xmath71 the total spin . ) in the chiral doublet model @xcite , the small @xmath0 is realized when the system is decoupled from the chiral condensate @xmath7 .
the small @xmath61 of @xmath1 then does not contradict with the possible and attempting scenario , the _ chiral restoration scenario in excited hadrons _
@xcite . if this scenario is the case , the origin of mass of @xmath1 ( or excited nucleons ) is essentially different from that of the positive - parity ground - state nucleon @xmath16 , which mainly arises from the spontaneous chiral symmetry breaking .
however , the non - vanishing axial charge of @xmath2 unfortunately gives rise to doubts about the scenario .
we have performed the first lattice qcd study of the axial charges @xmath0 of @xmath1 and @xmath2 , with two flavors of dynamical quarks employing the renormalization - group improved gauge action at @xmath3=1.95 and the mean - field improved clover quark action with the hopping parameters , @xmath4=0.1375 , 0.1390 and 0.1400 .
we have found the small axial charge @xmath61 of @xmath1 , whose absolute value seems less than 0.2 and which is almost independent of quark mass , whereas the axial charge @xmath63 of @xmath2 is found to be about 0.55 .
these values are consistent with naive nonrelativistic quark model predictions , and could not be the favorable evidences for the chiral restoration scenario in ( low - lying ) excited hadrons .
further investigations on the axial charges of @xmath1 or other excited baryons will cast light on the chiral structure of the low - energy hadron dynamics and on where hadronic masses come from .
all the numerical calculations were performed on nec sx-8r at rcnp and cmc , osaka university , on sx-8 at yitp , kyoto university , and on bluegene at kek .
the unquenched gauge configurations employed in our analysis were all generated by cp - pacs collaboration @xcite .
we thank l. glozman , d. jido , s. sasaki , and h. suganuma for useful comments and discussions .
this work was supported by a grant - in - aid for scientific research by monbu - kagakusho ( no .
20028006 and 20540265 ) , the 21st century coe `` center for diversity and university in physics '' , kyoto university and yukawa international program for quark - hadron sciences ( yipqs ) . c. detar and t. kunihiro , phys .
d * 39 * , 2805 ( 1989 ) .
l. y. glozman , phys .
lett . * 99 * , 191602 ( 2007 ) [ arxiv:0706.3288 [ hep - ph ] ] .
r. l. jaffe , d. pirjol and a. scardicchio , phys .
lett . * 96 * , 121601 ( 2006 ) [ arxiv : hep - ph/0511081 ] .
r. l. jaffe , d. pirjol and a. scardicchio , phys .
rept . * 435 * , 157 ( 2006 ) d. jido , t. hatsuda and t. kunihiro , phys .
lett . * 84 * , 3252 ( 2000 ) d. jido , m. oka and a. hosaka , prog .
phys . * 106 * , 873 ( 2001 ) [ arxiv : hep - ph/0110005 ] .
b. w. lee , chiral dynamics , gordon and breach , new york , 1972 a. ali khan _ et al . _
[ cp - pacs collaboration ] , phys . rev .
d * 65 * , 054505 ( 2002 ) [ erratum - ibid .
d * 67 * , 059901 ( 2003 ) ] t. t. takahashi , t. umeda , t. onogi and t. kunihiro , phys .
d * 71 * , 114509 ( 2005 ) [ arxiv : hep - lat/0503019 ] .
s. sasaki , k. orginos , s. ohta and t. blum [ the riken - bnl - columbia - kek collaboration ] , phys .
d * 68 * , 054509 ( 2003 ) [ arxiv : hep - lat/0306007 ] .
j. c. nacher , a. parreno , e. oset , a. ramos , a. hosaka and m. oka , nucl .
a * 678 * , 187 ( 2000 ) [ arxiv : nucl - th/9906018 ] .
l. y. glozman and a. v. nefediev , arxiv:0801.4343 [ hep - ph ] . | we show the first lattice qcd results on the axial charges @xmath0 of @xmath1 and @xmath2 .
the measurements are performed with two flavors of dynamical quarks employing the renormalization - group improved gauge action at @xmath3=1.95 and the mean - field improved clover quark action with the hopping parameters , @xmath4=0.1375 , 0.1390 and 0.1400 .
we construct 2@xmath52 correlation matrices and diagonalize them so that the signals of @xmath1 and @xmath2 are properly separated .
wraparound contributions in the correlator , which can be another source of signal contaminations , are eliminated by imposing the dirichlet boundary condition in the temporal direction .
the axial charge of @xmath1 is found to take small values as @xmath6 , whereas that of @xmath2 is approximately 0.5 , which is almost independent of quark masses and consistent with the predictions by the naive nonrelativistic quark model . |
the slc linear electron - positron collider@xcite has long been a powerful facility for testing the standard model via measurements of electroweak couplings at the @xmath0 pole .
its unique attributes of a highly longitudinally polarized electron beam ( @xmath8% ) and small spot size ( 1.5 x 0.8 x 700 ) @xmath9m@xmath10 in ( x , y , z ) have recently been augmented by improved luminosity for the 1997 - 1998 sld run and these have proved advantageous for precision electroweak measurements .
the sld detector@xcite is designed to take advantage of these special attributes of the slc . the polarized differential cross section at born level for the process
, @xmath11 is given by : @xmath12 where @xmath13 is the final state coupling , @xmath14 is the electron beam polarization ( @xmath15 for right - handed beams ) , @xmath16 where @xmath17 is the angle of the final state fermion with respect to the electron beam axis , and @xmath18 and @xmath19 are the vector and axial vector couplings which specify the z - f coupling
. one can define the left right asymmetry which equals the initial state coupling asymmetry @xmath2 , @xmath20 } = a_{e } \label{eqn : alr } \end{aligned}\ ] ] although one can use all final states to build the asymmetry , in practice , our cuts select hadronic final states : leptonic final states are treated separately ( see section iv ) . if @xmath14 can be measured precisely , then @xmath21 provides us with an effective way to measure @xmath22 : @xmath23 where the evolution of @xmath24 signifies the convention of making eqn .
[ eqn : alreff ] the _ definition _ of the weak mixing angle
. thus higher order effects must be explicitly accounted for when comparing this value to non-@xmath25-pole measurements .
additionally , one can construct a forward - backward left - right asymmetry and extract the final state coupling @xmath26 , @xmath27 - [ \sigma_r^f(z ) - \sigma_r^f(-z ) ] \over \sigma^f_l(z ) + \sigma_l^f(-z ) + \sigma_r^f(z ) + \sigma_r^f(-z ) } \nonumber \\ & = & |p_e|a_f { 2z \over 1 + z^2}. \label{eqn : alrfb } \end{aligned}\ ] ] for b - quark final state fermions and assuming full detector acceptance , eqn .
[ eqn : alrfb ] yields @xmath28 .
it is interesting to note that while @xmath21 is very sensitive to the weak mixing angle , @xmath29 in insensitive to @xmath30 , and instead is sensitive to vertex effects which in the standard model are expected to be negligible at the @xmath31 vertex .
precision polarimetry of the slc electron beam is accomplished with the compton polarimeter @xcite , which employs compton scattering between the high - energy electron beam and a polarized nd : yag laser beam ( @xmath32 nm ) to probe the electron beam polarization .
the compton scattered electrons , which lose energy in the scattering process but emerge essentially undeflected , are momentum analyzed by the compton spectrometer downstream of the sld interaction point , beyond which they exit the beam - line vacuum through a thin window , and enter a threshold erenkov detector segmented transverse to the beam - line . a history of polarization performance is given elsewhere @xcite .
the preliminary measurement for 1998 is @xmath33(syst)% where the relative systematic uncertainty can be expected to improve substantially ( to below the 0.7% level ) when the analysis is complete .
@xmath21 is essentially measured with hadronic final states only , as the selection efficiency for leptonic final states is intentionally kept low . for the last three measurements ( 1996 , 1997 and 1998 ) ,
a 99.9% pure hadronic sample was selected by requiring that the absolute value of the energy imbalance ( ratio of vector to scalar energy sum in the calorimeter ) be less than 0.6 , that there be at least 22 gev of visible calorimetric energy , and that at least 4 charged tracks be reconstructed in the central tracker ; this selection was 92% efficient . after counting
the number of hadronic decays for left- and right - handed electron beams and forming a left - right asymmetry , a small experimental correction for backgrounds ( and negligible corrections for false asymmetries ) was applied ; for the 1998 dataset this correction was 0.06% .
this then yielded a value for the measured left - right asymmetry ( 1998 ) of @xmath34 where the systematic uncertainty is dominated by the uncertainty in the polarization scale .
the translation of this result to the @xmath0-pole asymmetry @xmath35 was a @xmath36% effect , where the uncertainty arises from the precision of the center - of - mass energy determination .
this small error due to beam energy uncertainty is slightly larger than seen previously ( it has been quoted as @xmath37% ) , and reflects the results of a scan of the z peak used to calibrate the energy spectrometers to lep data , which was performed for the first time during the 1998 run .
this correction yields the 1998 preliminary result of @xmath38 the six measurements @xmath35 performed by sld are shown in table [ tab:9298sin2tw ] , along with their translations to @xmath39 .
the values are consistent with each other yielding a @xmath40/@xmath41/@xmath42 for a straight line fit to the ( preliminary ) average .
the ( preliminary ) averaged results for 1992 - 98 are @xmath43
parity violation in the @xmath0-lepton couplings is measured by a maximum likelihood fit to the differential cross section ( eqn .
[ eqn : polcross ] ) separately for the three leptonic final states , including the effects of t - channel exchange for the @xmath44 final state . for the 1996 - 98 data sample ,
leptonic final states in the range @xmath45 are selected by requiring that events have between two and eight charged tracks in the cdc , at least one track with @xmath46 gev / c .
one of the two event hemispheres was required to have a net charge of -1 , while the other one had to have a net charge of + 1 .
events are identified as bhabha candidates if the total calorimetric energy associated with the two most energetic tracks is greater than 45 gev . on the other hand ,
if there is less than 10 gev of associated energy for each track , and there is a two - track combination with an invariant mass of greater than 70 gev / c@xmath47 , the event is classified as a muon candidate .
the selection criteria for tau candidates is somewhat more complex .
first , the @xmath48 pair invariant mass is required to be less than 70 gev / c@xmath47 , with a separation angle between the two tau momentum vectors of at least 160@xmath49 .
at least one track in the event must have a momentum greater than 3 gev / c .
each @xmath50 hemisphere must have an invariant mass less that 1.8 gev/@xmath51 ( to suppress @xmath0 hadronic decays ) , and the associated energy in the lac from each track must be less than 27.5(20.0 ) gev for tracks with polar angles less than ( greater than ) @xmath52 .
for the electron final state , t - channel effects are incorporated into the angular distribution , while for the tau final state , a @xmath53-dependent efficiency correction has been applied to account for the correlation between visible energy and net tau polarization .
note that all channels provide information about @xmath54 , which comes in through the left - right cross section asymmetry .
the preliminary combined result for the 1993 - 1998 datasets are @xmath2=@xmath55 , @xmath3=@xmath56 and @xmath4=@xmath57 .
the result for @xmath58 ( which is the only direct measurement ) reflects the best single measurement , and is competitive with the overall lep value @xmath59 @xcite . a combination of these values , assuming lepton universality , yields the value @xmath60
the determination of @xmath29 requires three ingredients : a measurement of the polarization , tag the @xmath61-quark ( i.e. , discriminate from @xmath62 flavored events ) and tag the @xmath61-quark charge ( i.e. , was the underlying quark a @xmath61 or @xmath63 ? ) .
the sld has three separate analyses for obtaining these quantities and measuring @xmath29 : 1 ) jet charge , 2 ) cascade kaon tag and 3 ) lepton tag .
although @xmath29 can be extracted from eqn .
[ eqn : alrfb ] , most of these analyses use a maximum likelihood treatment based on eqn . [ eqn : polcross ] .
@xmath64 } \label{eqn : maxlike}\end{aligned}\ ] ] here , @xmath65 is the fraction of tagged events of quark flavor @xmath66 , ( @xmath67 ) is a correction factor to account for asymmetry dilution due to @xmath68 mixing and @xmath69 is a cos@xmath17 dependent qcd correction for quark flavor @xmath66 .
* momentum - weighted jet - charge : * this analysis uses an inclusive vertex mass tag to select a sample of @xmath70 , and the net momentum - weighted jet - charge to identify the sign of the underlying quark .
the track charge sum and difference between the two hemispheres are used to extract the analyzing power from data , thereby reducing mc dependencies and lowering many systematic effects .
* cascade kaon : * the decay channel @xmath71 is exploited to tag the b charge in this analysis .
the gas - radiator data of the erenkov ring imaging detector ( crid ) is used to identify charged kaons with high - impact parameter tracks .
the charges of the kaon candidates are summed in each hemisphere and the difference between the two hemisphere charges is used to determine the polarity of the thrust axis for the b - quark direction .
* leptonic tag : * electrons and muons are identified in hadronic decays to enrich the @xmath72 event samples .
the lepton charge provides the quark - antiquark discrimination while the jet nearest in direction to the lepton approximates the quark direction .
the lepton total and transverse momentum ( with respect to the jet axis ) are used to classify each event by deriving probabilities for the decays @xmath73 , @xmath74 , @xmath75 , @xmath76 and misidentified electrons .
the preliminary results for the three analyses are given in table [ tab : ab ] , along with the corresponding statistical and systematic errors .
the jet - charge analysis uses the most amount of available data covering 1993 - 1998 partial data sample of approximately 400,000 @xmath25 decays ( from a total of 550k events available ) .
the lepton tag analysis uses data from 1993 - 1996 , or approximately 200,000 events .
finally , the kaon tag uses the 1993 - 1995 runs , or approximately 150k events .
the sld average is also given in the table along with the lep average for comparison .
combining @xmath21 and @xmath77 obtains the preliminary sld result for the effective weak mixing angle , @xmath78 these results are shown in figure 1 .
it is interesting to note that the world values for @xmath80 from lepton measurements ( the results presented here , the lep leptonic averages @xmath81 and @xmath54 , and the lep results of @xmath82 from @xmath50-polarization ) are in excellent agreement with each other , and differ by @xmath83 from those obtained from `` hadron - only '' measurements ( @xmath84 , @xmath85 , and @xmath86 ) .
these results for @xmath29 and @xmath21 have been transformed and displayed in figure 2 , following the scheme of a full @xmath88 coupling analysis proposed by takeuchi et al .
@xcite here , @xmath89 is a variable which isolates deviations in the right - handed @xmath61-quark neutral current couplings .
this is plotted versus @xmath90 .
since lep s @xmath91 is a mixture of @xmath29 and @xmath54 , it comes in at @xmath92 .
the standard model is the horizontal line below the 68% and 90% cl contours . the slac linear collider design handbook ( 1984 ) ; ) ; + m. woods , in _ aip conference proceedings 343 _ , 230 ( 1995 ) .
the sld design report , slac report 273 ( 1984 ) .
m. woods , in _
spin-96 conference proceedings _ , 843 ( 1997 ) ; + m. woods , slac - pub-7955 ( 1997 ) .
reinertsen , slac - pub-8029 ( 1998 ) ; + k.g .
baird , slac - pub-8017 ( 1998 ) .
lep electroweak working group .
r. takeuchi , in _ proc . of the 8th meeting of dpf _
, 1231 - 1240 ( 1994 ) . | we present the measurements of the leptonic asymmetries in @xmath0 decays measured by the sld experiment at slac .
a data sample of approximately 550,000 @xmath0 bosons are used for the preliminary measurement of @xmath1 and the leptonic final state measurements of @xmath2 , @xmath3 and @xmath4 , representing the entire set of 1992 - 1998 sld runs .
when combining all results , a preliminary value for the effective weak mixing angle is obtained : @xmath5 .
we also present the direct measurements of @xmath6 from left - right forward - backward asymmetries , using partial data samples from 1993 - 1998 sld runs .
the preliminary value combines jet charge , cascade kaon and semileptonic decay analyses to obtain @xmath7 . |
the final stages of the evolution of low and intermediate mass stars involve complex and poorly understood physical phenomena that lead to the formation of axially symmetric bipolar planetary nebulae ( pne ) .
the origin of these spectacular structures is generally associated to interactions between the material expelled from the evolved star and an orbiting companion ( lagagec & chesneau @xcite ; soker & livio @xcite ) . the presence of a circumstellar disk is also invoked to collimate mass loss and foster the appearance of the axial symmetry , but such disks have only been observed in a few evolved stars .
l@xmath0 pup ( hd56096 ) is one of the nearest ( @xmath3 , van leeuwen @xcite ) and brightest ( @xmath4 , @xmath5 ) asymptotic giant branch ( agb ) star that exhibits mira - like pulsations with a period of 141days ( bedding et al .
this star is particularly interesting as kervella et al .
( @xcite ) discovered a circumstellar disk around l@xmath0pup using adaptive optics in the near - infrared .
this discovery was recently confirmed from polarimetric imaging at visible wavelengths by kervella et al .
( @xcite ) , that also revealed the presence of a secondary source .
we present these two sets of observations in sect .
[ observations ] , and we discuss the spectral energy distribution and disk geometry in sect .
kervella et al . ( @xcite ) reported observations of l@xmath0 pup using the vlt / naco ( rousset et al .
@xcite ) adaptive optics system . using a `` lucky imaging '' approach
, they obtained diffraction limited images in a series of narrow - band filters ranging in wavelength from 1 to 4@xmath2 m .
these images show the presence of a dust band in front of l@xmath0 pup .
the band exhibits a high opacity in the @xmath6 band ( @xmath7 m ) and becomes translucent in the @xmath8 and @xmath9 bands ( @xmath10 m ) where the light scattering becomes less efficient ( fig .
[ naco ] ) . in the @xmath11 band ,
the thermal emission from the inner rim of the dust disk was also detected , as well as a large loop extending to more than 10au from the star .
new observations by lykou et al .
( @xcite ) using the sparse aperture masking mode of naco also showed the presence of a disk - like structure , and ohnaka et al .
( @xcite ) confirmed the presence of an elongated central emission in the east - west direction . the spectro - polarimetric high - contrast exoplanet research ( sphere , beuzit et al .
@xcite ) is a high performance adaptive optics ( fusco et al .
@xcite ) recently installed at the very large telescope .
kervella et al .
( @xcite ) observed l@xmath0 pup using sphere equipped with the imaging polarimeter zimpol ( roelfsema et al .
this instrument provides diffraction limited imaging at visible wavelengths , down to @xmath12 nm . the observed intensity image and the map of the degree of linear polarization in the @xmath13 band ( @xmath14 nm ) are presented in fig .
[ zimpol ] , together with a nomenclature of the observed features .
the right panel shows the observed features ( field of view 0.6 , kervella et al .
@xcite).,title="fig:",width=151 ] pup .
the right panel shows the observed features ( field of view 0.6 , kervella et al .
@xcite).,title="fig:",width=151 ] pup .
the right panel shows the observed features ( field of view 0.6 , kervella et al .
@xcite).,title="fig:",width=151 ] the intensity image shows a bright unresolved source at a separation of 0.032 from the agb star ( labeled b in fig .
[ zimpol ] , right panel ) .
although the nature of this source is unknown , its brightness and color are consistent with a late k giant star .
several spiral structures are also observed , as well as two thin plumes extending perpendicularly to the disk plane .
the map of the degree of linear polarization @xmath15 shows a maximum at a radius of 6au from the central star .
kervella et al .
( @xcite ) interpret this maximum as 90@xmath16 scattering occurring at the inner rim of the dust disk .
the radius is consistent with the observed extension of the thermal emission in the @xmath9 band ( fig .
[ naco ] , central panel ) .
the radiative transfer modeling ( using the radmc-3d code ; dullemond @xcite ) presented by kervella et al .
( @xcite , fig . [ sedmodel ] , right panel ) shows that the circumstellar disk is seen almost edge on , with an inclination of @xmath17 on the line of sight .
this model reproduces satisfactorily the observed spectral energy distribution ( fig .
[ sedmodel ] , left ) , that exhibits a remarkably flat flux density between 1 and @xmath18 m .
the total dust mass in the radmc-3d disk model is @xmath19 m@xmath20 . assuming a gas - to - dust ratio of 100 , this translates to a total mass including gas of @xmath21 to @xmath22 m@xmath20 .
pup s disk ( right , taken from kervella et al .
@xcite).,title="fig:",height=158 ] pup s disk ( right , taken from kervella et al .
the environment of l@xmath0 pup hosts the two major ingredients suspected to play a role in the shaping of pne : a close - in companion ( whose physical nature is still largely enigmatic ) and a circumstellar dust disk . thanks to its proximity and the availability of the sphere adaptive optics providing 20mas angular resolution in the visible , it is now possible to monitor in real time the dynamical evolution of the different components of the immediate environment of l@xmath0 pup ( companion , spirals , plumes , ... ) . considering their linear scales
, we expect a significant evolution over timescales of only a few years .
l@xmath0 pup thus exhibits a strong potential to improve our understanding of the mass loss of low and intermediate mass stars and the formation of bipolar pne .
_ j .- ph .
berger : _ if you have such a massive companion you should have strong tidal truncation and a cavity .
it is surprising not to see any sign of cavity in the spectral energy distribution ( sed ) .
+ _ answer : _ the sed of l@xmath0 pup is atypical due to its flatness over a major part of the infrared domain ( @xmath23 m ) .
the fact that the disk is seen almost edge - on affects the flux contribution from the inner rim of the disk through absorption and reddening , as shown by the radmc-3d model .
this may explain that the signature of the cavity is not immediately observable in the sed . _
j. groh : _ is this object representative of its class , or is it a peculiar object ?
would it be possible to observe a larger sample to investigate this phenomenology ? + _ answer : _ the agb star itself appears to be close in physical properties to the mira star r vir , whose pulsation period is 145.5 days ( see e.g. eisner et al .
more generally , l@xmath0 pup a could well be classified as a low amplitude mira ( kervella et al .
the presence of circumstellar disk has been proposed around the mira v cvn by neilson et al .
( @xcite ) based on polarimetry , but this is a rare example of such a detection around a mira .
the difficulty to extend the sample is the angular scale of the disks : l@xmath0 pup is very nearby , so the angular scale is resolvable using sphere , but for more distant stars , optical or millimeter interferometry would be the only options . _
a. lobel : _ when you overplot the @xmath18 m image showing the extended loop , it appears to be projected on the northern cone .
could it tell about the direction of the orbital motion on the sky ? + _ answer : _ there is unfortunately an ambiguity whether the loop is located on the nearby or far surface of the northern cone .
the difficulty is that the thermal emission is isotropic , while the scattering is preferentially in the forward direction .
we therefore very likely observe in fig .
[ zimpol ] the nearby side of the cones , but we can not be sure for the thermal emission of the loop . _
t. ueta : _
1 ) bright spots a and b in the zimpol image could still be just brightly lit surfaces of the top of the disk .
polarization vector maps should help pinpoint the location of the central bright source .
does the centrosymmetric structure of vector maps point to the location of a ?
+ 2 ) the distribution of scattering matter can be better probed by the polarized flux @xmath24 rather than the degree of polarization @xmath15 .
the disk edge might show differently in @xmath24 than in @xmath15 , or do they look the same ? + _ answer : _ 1 ) the nature of the secondary source is indeed uncertain .
but its flux contribution is considerable in the visible ( @xmath25% of the primary ) , pointing at a stellar origin .
regarding the polarization vector , the difficulty is that the secondary is located so close to the primary ( 32 mas ) , that its measurement is highly uncertain .
we see a slight deviation from the classical single central source vector map ( purely orthoradial ) , but not at a sufficiently significant level . + 2 ) the degree of polarization @xmath15 has the advantage to be a simple proxy for the scattering angle @xmath26 ( in the single scattering approximation ) , and therefore well suited to probe the overall geometry of the disk .
the polarized flux @xmath24 is sensitive to both the scattering angle @xmath26 and the scattering dust density @xmath27 , therefore relatively more complex to model . in l@xmath0 pup ,
the @xmath24 map is generally similar to the total intensity map , except close to the primary star where the polarization is low ( forward scattering ) .
the secondary source exhibits a relatively strong @xmath24 contribution , suggesting that it contains a significant fraction of scattered light from the primary .
but its overall degree of polarization @xmath15 is low , indicating that it is not purely scattering .
the presence of an accreting disk around a companion star is a possibility that we plan to investigate . _
j. milli : _ is the angle of polarization deduced from zimpol observations consistent with the morphological model ? + _ answer : _ for most of the field of view apart from the very central part , the polarization vector is orthoradial , consistent with a central source illuminating the dust envelope ( as in the radmc-3d model ) .
but very close to the star ( within @xmath28mas ) , the determination of the polarization angle is uncertain ( see also the answer to dr .
ueta s question 1 ) . _
j. hoffman : _ does your radiative transfer model predict your observed polarization results ?
+ _ answer : _ we have encountered some difficulties with the numerical computation of the polarization with the radmc-3d code .
but we indeed plan to compare the observed polarization maps to the predictions of the model . 99 bedding , t. r. , zijlstra , a. a. , jones , a. , et al .
2002 , mnras , 337 , 79 beuzit , j .- l . , feldt , m. , dohlen , k. , et al .
2008 , spie conference series , 7014 , 18 dullemond , c. p. 2012 , radmc-3d : a multi - purpose radiative transfer tool , astrophysics source code library eisner , j. a. , graham , j. r. , akeson , r. l. , et al .
2007 , apj , 654 , l77 fusco , t. , sauvage , j .- f . , petit , c. , et al .
2014 , spie conference series , 9148 , 1 kervella , p. , montargs , m. , ridgway , s. t. , et al .
2014 , a&a , 564 , a88 kervella , p. , montargs , m. , lagadec , e. , et al .
2015 , a&a , 578 , a77 lagadec , e. & chesneau , o. 2014 , arxiv:1410.3692 lykou , f. , klotz , d. , paladini , c. , et al .
2015 , a&a , 576 , a46 neilson , h. r. , ignace , r. , & henson , g. d. 2014 , iau symposium , vol .
301 , 463 ohnaka , k. , schertl , d. , hofmann , k. h. , & weigelt , g. 2015 , a&a , in press , arxiv:1507.06668 roelfsema , r. , bazzon , a. , schmid , h. m. , et al . 2014 ,
spie conference series , 9147 , 3 rousset , g. , lacombe , f. , puget , p. , et al .
2003 , spie conference series , 4839 , 140 soker , n. & livio , m. 1989 , apj , 339 , 268 van leeuwen , f. 2007 , a&a , 474 , 653 | adaptive optics observations in the infrared ( vlt / naco , kervella et al . 2014 ) and
visible ( vlt / sphere , kervella et al . 2015 ) domains revealed that the nearby agb star l@xmath0 pup ( @xmath1pc ) is surrounded by a dust disk seen almost edge - on .
thermal emission from a large dust `` loop '' is detected at 4@xmath2 m up to more than 10au from the star .
we also detect a secondary source at a separation of 32mas , whose nature is uncertain .
l@xmath0 pup is currently a relatively `` young '' agb star , so we may witness the formation of a planetary nebula .
the mechanism that breaks the spherical symmetry of mass loss is currently uncertain , but we propose that the dust disk and companion are key elements in the shaping of the bipolar structure .
l@xmath0 pup emerges as an important system to test this hypothesis . |
there are many cases in bayesian modeling where a certain choice of prior distribution allows for computationally simple or tractable inference .
for example , * conjugate priors yield posteriors with a known parametric form and therefore allow for non - iterative , exact inference @xcite . *
certain priors yield models with tractable conditional or marginal distributions , which allows efficient approximate inference algorithms to be applied ( e.g. gibbs sampling @xcite , sampling in collapsed models @xcite , or mean - field variational methods @xcite ) .
* simple parametric priors ( e.g. the normal distribution ) allow for computationally cheap density queries , maximization , and sampling @xcite , which can allow for easier use in iterative inference algorithms ( e.g. metropolis - hastings @xcite , gradient - based mcmc @xcite , or sequential monte carlo @xcite ) . *
certain priors mitigate issues of identifiability , and allow for simpler posteriors without multiple modes @xcite .
however , more sophisticated priors that provide a better depiction of observed or expert knowledge do not , in general , allow for the above inference techniques . instead , researchers must resort to more general and computationally expensive inference methods .
this can encourage the use of convenient priors in practice , rather than priors that might yield a more realistic or accurate inference , which is a criticism of bayesian methods @xcite . in this paper , we investigate the following question : for a given model ,
is it possible to use _ any convenient prior _ to infer a false posterior , and afterwards , given some true prior of interest , quickly transform this result into the true posterior ? intuitively , our strategy is the following : for a given model , we first choose any computationally convenient false prior , and perform inference , which returns the false posterior .
we then use our inferred false posterior , a true prior of interest , and the false prior , to efficiently produce samples from the true posterior via a method we call _ prior swapping_. one key attribute of prior swapping is that it runs without touching any data .
existing general inference algorithms are iterative and _ data - dependent _ : parameter updates at each iteration involve data , and the computational complexity or quality of each update depends on the amount of data used .
thus , inference on larger datasets takes more time .
furthermore , in practice , we often want to perform inference for a number of different priors or hyperparameter settings ; for each , some data - dependent inference algorithm must be run to compute a new result .
in contrast , prior swapping algorithms are _ data - independent_i.e .
parameter updates do not involve data , and neither the complexity nor quality of each update depends on data size .
we therefore advocate doing difficult inference in two steps : first , perform data - dependent inference using the most computationally convenient prior for a given model , and then , for all future priors of interest ( e.g. complex priors or a range of prior hyperparameter settings ) , perform quick , data - independent prior swapping .
we summarize the advantages of this approach : 1 .
there often exists some false posterior that can be computed more efficiently than the true posterior ; in some cases , we can therefore apply significantly less - costly inference procedures to more - sophisticated models than previously possible .
2 . the prior swapping procedure runs independently of the data ; this can allow for considerable speed - ups when doing inference on large datasets or ranges of hyperparameter values .
3 . we can often maintain theoretical guarantees of existing approximate inference algorithms ; e.g. when inferring the false posterior with an asymptotically - exact method ( such as sampling ) , prior swapping can return asymptotically - exact true posterior samples .
we can update our prior , or incorporate new prior information , in an online fashion , without redoing data - dependent inference .
this is a black box procedure , which can be run directly on the output of most existing inference methods to add richer prior information to inference results .
suppose we have a dataset of @xmath0 real , finite - dimensional vectors @xmath1 @xmath2 @xmath3 @xmath4 , and we are interested in a family of models specified by the likelihood function @xmath5 , parameterized by a real , d - dimensional vector @xmath6 .
suppose we have a prior distribution over the space of model parameters @xmath7 , with probability density function ( pdf ) @xmath8 .
the likelihood and prior define a joint model given by the pdf @xmath9 @xmath2 @xmath10 . in bayesian inference , we are interested in computing the posterior distribution
i.e . a conditional distribution of this joint model with pdf @xmath11 suppose we ve chosen a different prior distribution @xmath12 , which we refer to as a _ false prior _
( while we refer to @xmath8 as the _ true prior _ ) .
we can now can define a new joint model @xmath13 @xmath2 @xmath14 , with posterior ( conditional ) pdf @xmath15 .
we refer to this second posterior distribution as a _
false posterior_. the goal of our method is to first infer a false posterior and then leverage it to infer a true posterior . to carry out this transformation
, we use the _ prior swap function _
@xmath16 , which we define as the false posterior density multiplied by the true prior density and divided by the false prior density , i.e. @xmath17 note that @xmath16 is proportional to the true posterior density @xmath18 .
however , depending on how we represent the false posterior @xmath19 , @xmath16 can have a much simpler analytic representation than @xmath18 , which is typically defined via a likelihood function ( which is a function of the data ; see eq .
[ trueposterior ] ) .
it is therefore often possible to apply existing inference algorithms to @xmath16 , instead of @xmath18 , to perform quicker inference .
we will use @xmath20 to denote the _ prior swap density _
, i.e. @xmath20 @xmath2 @xmath21 .
our strategy is to first compute some expression for @xmath19 , and then infer the true posterior using @xmath20 as a target density .
this can be done with various types of inference for @xmath19 , including closed form ( exact ) computations , sampling - based methods ( such as markov chain monte carlo ) , and other approximate inference strategies such as variational inference and expectation propagation methods .
we illustrate prior swapping in fig .
[ fig : priorswappingillustration ] .
[ [ fixed - complexity - false - posterior . ] ] fixed - complexity false posterior .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + as a first case , suppose that we have computed an analytic expression @xmath22 for the false posterior density with a fixed - complexity parametric form ( i.e. the complexity of the expression does not grow as inference proceeds ) .
for example , this is the case if we * compute @xmath19 exactly ( in a closed form ) via a conjugate prior . *
generate samples from @xmath19 , via mcmc or other sampling methods , then use these samples to compute a parametric density estimate @xmath23 . *
apply optimization - based approximate inference methods , such as variational inference or expectation propagation , that return analytic approximate posteriors @xmath24 or @xmath25 . in each case , while we use some data - dependent algorithm to infer the false posterior , our resulting expression @xmath22 is not a function of the data . given this expression ,
consider the functions [ cols= " < , < " , ] we can use these to draw posterior samples extremely efficiently with a variety of standard mcmc algorithms . at each iteration in mcmc , to draw a new parameter , we must evaluate a target function associated with the posterior density .
for example , we evaluate a function proportional to @xmath26 in metropolis - hastings ( mh ) @xcite and @xmath27 in gradient - based mcmc methods ( such as langevin dynamics ( ld ) @xcite , hamiltonian monte carlo ( hmc ) @xcite , and their stochastic variants @xcite ) . due to the likelihood function in @xmath18 ,
these target functions are data - dependent . in prior swapping
, we instead evaluate @xmath28 ( eq . [ psfixedcomplexity01 ] ) in mh , or @xmath29 ( eq . [ psfixedcomplexity02 ] ) in ld , hmc , or gradient ascent optimization to a map point estimate ( see appendix for detailed algorithm pseudocode ) . here
, each iteration only requires evaluating a few simple analytic expressions , and thus has @xmath30 complexity with respect to data size .
[ [ consistent - false - posterior . ] ] consistent false posterior .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + suppose our false posterior inference procedure yields a set of samples @xmath31 . above
, we proposed the strategy of computing a fixed - complexity parametric density estimate @xmath23 , and plugging this into the prior swap function ( which we will denote @xmath32 ) .
this @xmath33 is , in general , an inconsistent estimate of @xmath19 , and using it in prior swapping yields asymptotically-_biased _ samples from @xmath18 . here
, we aim to answer the question : given samples from @xmath19 , can we develop an efficient method that returns asymptotically-_exact _ samples from @xmath18 ?
suppose we instead use a consistent false posterior density estimate , such as a nonparametric @xcite or semiparametric @xcite estimate .
we will prove ( sec .
[ theoreticalguarantees ] ) that plugging a consistent estimate for @xmath19 into the prior swap function yields asymptotically - exact samples . however , the cost of these consistent estimates grows with the number of samples @xmath34 ; typically , evaluating their pdf or gradient has a complexity of @xmath35 . using these estimates for @xmath19 in the above prior swapping procedure
therefore has a complexity of @xmath35 _ per iteration _ , which is costly for large @xmath34 .
we instead propose the following prior swapping method , which still yields asymptotically - exact samples from @xmath18 , yet does not require a significant increase in computation : first generate approximate posterior samples using @xmath32 ( as above ) , and then _ correct _ these samples via algorithms from the parallel mcmc literature ( described below ) , designed to sample from the product of densities . to motivate this method , we choose a a consistent semiparametric false posterior estimate @xmath36 ( see @xcite for background and consistency guarantees ) , which can be viewed as the product of a parametric density estimate @xmath23 and a nonparametric correction function .
this is written @xmath37 , \end{aligned}\ ] ] where we use @xmath38 to denote a probability density kernel , with bandwidth @xmath39 , where @xmath40 as @xmath41 ( see @xcite for details on probability density kernels and bandwidth selection ) .
a general parametric family for @xmath23 that we can use in nearly all cases is the family of false prior distributions .
these are typically simple parametric distributions over the correct parameter space , which contain parameterizations in a broad vicinity of @xmath19 , and for which there exist efficient parameter estimation algorithms .
given @xmath42 , we write the semiparametric prior swap function as @xmath43 \left [ \frac{1}{t } \sum_{t=1}^t \frac{k \left(\frac{\|\theta - \widetilde{\theta}_t\|}{h}\right ) } { \hat{f}_{\phi|x}^p(\widetilde{\theta}_t ) h^d } \right ] = \left [ \hat{f}_{\widetilde{ps}}^p ( \theta ) \right ] \left [ \frac{1}{t } \sum_{t=1}^t w_t k \left(\frac{\|\theta - \widetilde{\theta}_t\|}{h}\right ) \right ] .
\nonumber\end{aligned}\ ] ] hence , the prior swap function is proportional to the product of two densities : the parametric prior swap density , and a correction density .
we can easily generate samples from both of the densities that comprise @xmath44the former with ( fixed - complexity ) parametric prior swapping and the latter by sampling from components in the correction density with frequency proportional to the components weights .
r0.59 estimate @xmath23 using @xmath45 @xmath46 parametric estimate .
+ sample @xmath47 @xmath46 parametric prior swapping .
+ sample @xmath48 .
+ sample @xmath49 via @xmath50 .
we therefore turn to sample combination methods for efficiently generating samples from the product of densities @xcite . given two sets of samples
@xmath51 @xmath52 @xmath53 and @xmath54 @xmath52 @xmath55 , these algorithms aim to return @xmath56 @xmath2 @xmath57 .
these methods are efficient because each density product sample can be generated without iterating through either set of @xmath34 input samples ; typically , only a single sample from each input density is required
. each sample can therefore be produced with constant @xmath30 complexity .
we summarize the full asymptotically - exact prior swapping procedure in alg .
[ alg : priorswapping ] . in the appendix
, we give pseudocode for the density product sample combination algorithms , summarize the complexity of all methods , and also discuss how this semiparametric prior swapping method allows for easy incorporation of observed prior information .
this setting also has some relation to importance sampling ( is ) @xcite .
for example , one could use the false posterior as an is proposal ( i.e. simply reweight and then use samples from @xmath19 ) .
however , in practice , performance would only be adequate for false posteriors that are very similar to the true posterior ( and not for arbitrary @xmath58 , especially in high dimensions @xcite .
we show this empirically by comparing against this strategy in sec .
[ empiricalresults ] . here
we give theoretical guarantees about the correctness of the samples generated via prior swapping .
first , note that if we have an exact analytic expression @xmath59 for the false posterior , the prior swap function is proportional to the true posterior , i.e. @xmath60 . using @xmath61 in standard mcmc algorithms therefore carries out mcmc on an exact true posterior target and comes with existing guarantees , such as producing asymptotically - exact posterior samples @xcite . in the second setting
, to prove that we generate asymptotically - exact samples given a consistent false posterior estimator , we need to show , as @xmath41 , that @xmath62 is consistent for @xmath20 and that alg . [
alg : priorswapping ] indeed draws samples from @xmath62 .
given false posterior samples @xmath31 and @xmath63 , the estimator @xmath64 ( eq . [ eq : semiparest ] ) is consistent for @xmath16 , i.e. its mean - squared error satisfies @xmath65 < \frac{c}{t^{4/(4+d ) } } \nonumber\end{aligned}\ ] ] for some @xmath66 and @xmath67 .
the procedure given in alg .
[ alg : priorswapping ] generates samples from @xmath68 ( eq .
[ eq : semiparest ] ) as @xmath41 .
proofs for both theorems and definition of @xmath69 are given in the appendix .
we show empirical results on bayesian generalized linear models ( including linear and hierarchical logistic regression ) with sparsity and heavy tailed priors , on latent factor models ( including mixture models and topic models ) with relational priors over factors ( e.g. diversity - encouraging , agglomerate - encouraging , etc . ) , and feedforward neural networks , where we show hyperparameter tuning of weight - decay l2 regularization via prior swapping with normal priors .
we aim to demonstrate empirically that prior swapping allows us to apply less - costly inference algorithms to more - complex models than was previously possible , and that it efficiently yields correct samples .
we compare the following inference procedures : * * mcmc on the true posterior : * mcmc sampling algorithms run directly on @xmath18 .
* * vi on the true posterior : * variational inference algorithms run directly on @xmath18 . *
* false posterior : * using the inferred false posterior @xmath19 . * * importance sampling with false posterior proposal : * using @xmath19 as a proposal density and running importance sampling . * * prior swapping ( fixed - complexity ) : * prior swapping using a fixed - complexity expression @xmath22 , including exactly - computed @xmath19 , parametric estimates @xmath23 , and approximations @xmath24 . * * prior swapping ( consistent ) : * prior swapping via alg .
[ alg : priorswapping ] using a semiparametric @xmath36 . * * prior swapping to map : * optimization , using the prior swap function , to a map point estimate . *
* false - true hybrid prior * prior swapping on @xmath70 without dividing out the false prior . *
* normal approximation : * prior swapping using normal approximations for @xmath19 , @xmath71 , and @xmath72 . to assess performance , we use two metrics : posterior error and test error . to compute posterior error
, we run a single chain of mcmc on the true posterior for one million steps , and use these samples as groundtruth ( after removing the first quarter as burn in ) .
we then compare these groundtruth samples with those returned by our inference methods . specifically , we compute the euclidean distance between the means of both sets of samples , the euclidean distance between the variances of both sets of samples , and the estimated @xmath73 distance between the groundtruth density @xmath74 and a sampled posterior density @xmath75 : @xmath76 .
posterior error is the normalized average of these three computations .
test error , which we report for classification models , is the proportion of misclassified held - out ( test ) data . for timing plots , to assess performance at a given time point , we collect samples drawn before this time point , remove the first quarter as burn in , and compute the above metrics .
sparsity - encouraging regularizers have gained a high level of popularity over the past decade due to their ability to produce models with greater interpretability , predictive accuracy , and parsimony .
for example , the @xmath77 norm has been used to induce sparsity with great effect @xcite , and has been shown to be equivalent to a mean - zero independent laplace prior @xcite . in a bayesian setting , inference given a sparsity prior can be difficult , and researchers often resort to computationally intensive methods ( such as hamiltonian monte carlo ) or biased posterior approximations ( e.g. expectation propagation @xcite ) that make factorization or parametric assumptions @xcite .
we provide a cheap yet accurate solution : first use a more - tractable prior ( such as a normal prior , in which closed form inference is possible ) , and then use prior swapping to quickly convert the result to the posterior given a sparsity prior .
our first set of experiments are on sparse bayesian linear regression models @xcite , which we can write as @xmath78 . for @xmath71 , we compute results on on laplace , student s t , and verysparse ( with pdf @xmath79 @xmath2 @xmath80 @xcite ) priors . here , a normal @xmath72 is conjugate and allows for exact inference .
our second set of experiments are on a hierarchical bayesian logistic regression model @xcite , which we write as @xmath81 .
we will consider a hierarchical @xmath82 . here
, we also use a normal @xmath72 ( see @xcite for examples of convenient inference in bayesian logistic regression under normal priors ) . in these experiments
, we generate synthetic data from the models in order to show results under varying @xmath0 and @xmath83 . for comparison methods , we use mh for mcmc , and a mean field approximation @xcite for vi . in fig .
[ fig : lmall ] we show results for sparse linear models , with @xmath0=10,000 observations , and @xmath83=20 dimensions . in ( a ) and ( b )
we show convergence plots and see that prior swapping performs faster inference ( by a few orders of magnitude ) than mcmc .
we also see that semiparametric prior swapping ( alg .
[ alg : priorswapping ] ) achieves nearly identical performance as prior swapping on the exactly computed @xmath19 .
the other comparison methods yield posteriors that are incorrect or very slow to converge . in ( b )
we halve the variance of our false prior , which hurts performance of the comparison methods , but leaves prior swapping unchanged . in ( c )
we show 1-d density marginals as we increase the prior sparsity , and in ( d ) we show prior swapping results for different sparsity priors . in fig .
[ fig : lrall ] , we show results for hierarchical logistic regression . in ( a ) and
( b ) we vary the number of observations ( @xmath0=10 - 120,000 ) and see that prior swapping has a constant wall time while the wall times of both mcmc and vi increase with @xmath0 . in ( b ) we see that the prior swapping methods achieve the same test error as the standard inference methods . in ( c ) and ( d )
we vary the number of dimensions ( @xmath83=1 - 40 ) . in this case , all methods have increasing wall time , and again the test errors match . in ( e ) , ( f ) , and ( g ) , we vary the prior hyperparameter ( @xmath84=1 - 1.05 ) . for prior swapping ,
we infer a single @xmath19 ( using @xmath85 ) with both mcmc and vi , and compute _ all other _ hyperparameter results using this @xmath19 .
this demonstrates that prior swapping can quickly infer correct results over a range of hyperparameters . here
, the asymptotically - exact semiparametric prior swapping method matches the test error of mcmc slightly better than the parametric method .
many latent variable models in machine learning such as mixture models , topic models , probabilistic matrix factorization , and various others
involve a set of latent factors ( e.g. components or topics )
. often , we d like to use priors that encourage interesting behaviors among the factors .
for example , we might want dissimilar factors via a diversity - promoting prior @xcite or for the factors to show some sort of sparsity pattern @xcite .
inference in such models is often computationally expensive or designed on a case - by - case basis @xcite .
however , when conjugate priors are placed over the factor parameters , collapsed gibbs sampling can be applied . in this method ,
the factor parameters are integrated out , leaving only a subset of variables ; on these , the conditional distributions can be computed analytically , which allows for gibbs sampling over these variables .
afterwards , samples of the collapsed factor parameters can be computed .
hence , we propose the following strategy : first , assign a prior for the factor parameters that allows for collapsed gibbs sampling ; afterwards , reconstruct the factor samples and apply prior swapping for more complex relational priors over the factors .
we can thus perform convenient inference in the collapsed model , yet apply more - sophisticated priors to variables in the uncollapsed model .
we first show results on a gaussian mixture model ( gmm ) @xcite , written @xmath86 .
using a normal @xmath72 over @xmath87 allows for collapsed gibbs sampling .
we also show results on a topic model ( latent dirichlet allocation ( lda ) @xcite ) for text data ( for the form of this model , see @xcite ) . here , using a dirichlet @xmath72 over topics allows for collapsed gibbs sampling . for mixture models ,
we generate synthetic data from the above model ( @xmath0=10,000 , @xmath83=2 , @xmath88=9 ) , and for topic models , we use the simple english wikipedia corpus ( n=27,443 documents , vocab=10,192 words ) @xcite , and set @xmath88=400 topics . in fig .
[ fig : latentfactor ] we show results for mixture and topic models . in ( a )
we show inferred posteriors over gmm components for a number of relational priors , which we define in ( b ) . in ( c )
, we apply the diversity - promoting prior to lda , to separate redundant topics . here , we show two topic - clusters ( `` geography '' and `` family '' ) in @xmath19 , which are separated into thematically - similar yet distinct topics after prior swapping . in ( a ) and
( c ) we also show wall times of the inference methods . learning neural networks with weight decay ( l2 regularization )
can be viewed as finding the map point estimate of a bayesian neural network model with a normal prior @xcite .
since the prior swap function can be used as an objective for optimization to a map estimate , we aim to use prior swapping for quick hyperparameter tuning of weight decay in neural networks .
we will compare this to finding the optimal weight decay via stochastic gradient descent @xcite and stochastic gradient langevin dynamics @xcite , two popular methods for learning and inference in neural networks .
these stochastic gradient methods have only a weak dependence on data at each iteration ; however , their updates may be noisy or suboptimal @xcite , while prior swapping updates involve exact gradients ( of the prior swap function ) without any stochasticity .
furthermore , note that we can use stochastic gradient inference methods in conjunction with prior swapping ( i.e. to compute @xmath19 , which we will do here ) .
r0.7 we run our experiment on an eight - layer fully connected deep neural network with 400 nodes per layer , yielding a model with 1,280,410 dimensions . for data , we use the mnist handwritten digits classification dataset ( @xmath0=60,000 , @xmath83=784 ) @xcite .
we consider a family of true priors @xmath89 over neural network parameters ( for weight decay ) , and aim to compute map point estimates over a range of hyperparameters @xmath90 . due to the high dimensionality
, we make a parametric approximation @xmath33 for the false posterior , which we assume to be normal with a diagonal covariance matrix .
we use sgld to infer @xmath33 ( choosing a single @xmath90 to parameterize @xmath72 ) . for prior swapping ,
we use this @xmath33 to learn map estimates for all other @xmath90 values in the range . in fig [ fig : nnall ]
we show wall time and test error for the comparison methods over the set of weight decay parameters ( @xmath90=50 - 10,000 ) . for prior swapping ,
we perform gradient ascent optimization using the gradient log prior swap function ( eq . [ psfixedcomplexity02 ] ) .
we see in ( c ) that all methods yield the same optimal parameter ( @xmath90=550 ) with the lowest test error .
however , in ( a ) and ( b ) we see that prior swapping takes less time than sgd and sgld .
we have presented procedures to carry out the task of prior swapping : given any false posterior ( computed using some convenient false prior ) and a true prior of interest , our algorithms generate samples from the true posterior .
empirically , we have demonstrated prior swapping on a number of models and priors , and have shown that it can ( 1 ) quickly generate correct samples , ( 2 ) allow for less - costly data - dependent inference algorithms to be applied to more - complex models than previously possible , ( 3 ) allow for quick model selection or hyperparameter tuning , and ( 4 ) be used directly on top of existing inference algorithms to add richer prior information to pre - computed inference results , without having to revisit the data .
theoretically , we have shown that , given a stream of false posterior samples , this strategy can be used to generate asymptotically exact samples from the true posterior . we furthermore hope that prior swapping can be successfully implemented as a black box method , and paired with existing automatic inference engines or probabilistic programming languages @xcite . since prior swapping does not require tuning and can be applied directly to the output of inference yielded by these frameworks , it has potential to aid in general purpose automated inference with more - realistic and difficult priors .
* appendix for `` prior swapping for data - independent inference '' *
here , we prove the two theorems stated in sec .
[ theoreticalguarantees ] . throughout this analysis
, we assume that we have @xmath34 samples @xmath91 @xmath92 @xmath93 @xmath92 @xmath94 from the false - posterior @xmath19 , and that @xmath95 denotes the bandwidth of our semiparametric false - posterior density estimator @xmath36 .
let hlder class @xmath96 on @xmath93 be defined as the set of all @xmath97 times differentiable functions @xmath98 whose derivative @xmath99 satisfies @xmath100 we also define the class of densities @xmath101 to be @xmath102 we also assume that the false - posterior density @xmath19 is bounded , i.e. that there exists some @xmath103 such that @xmath104 for all @xmath6
. given false posterior samples @xmath105 and @xmath63 , the estimator @xmath64 ( eq . [ eq : semiparest ] ) is consistent , i.e. its mean - squared error satisfies @xmath65 < \frac{c}{t^{4/(4+d ) } } \nonumber\end{aligned}\ ] ] for some @xmath66 and @xmath67 .
to prove mean - square consistency of our semiparametric prior swap estimator @xmath106 , we give a bound on the mean - squared error ( mse ) , and show that it tends to zero as we increase the number of samples drawn from the false - posterior . to prove this , we bound the bias and variance of the estimator , and use this to bound the mse .
we first bound the bias of our semiparametric prior swap estimator . for any @xmath107 , we can write the bias as @xmath108 - f_{\widetilde{ps } } \right| & = \left| \mathbb{e } \left [ \hat{f}_{\phi|x}^s \frac{f_\theta}{f_\phi } \right ] - f_{\phi|x } \frac{f_\theta}{f_\phi } \right| \\ & = \left| \frac{f_\theta}{f_\phi } \mathbb{e } \left [ \hat{f}_{\phi|x}^s \right ] - f_{\phi|x } \right| \\ & = \tilde{c } \left| \mathbb{e } \left [ \hat{f}_{\phi|x}^s \right ] - f_{\phi|x } \right| \\ & \leq ch^2\end{aligned}\ ] ] for some @xmath66 , where we have used the fact that @xmath109 - f_{\phi|x } \right| \leq \tilde{c } h^2 $ ] for some @xmath110 ( given in @xcite ) .
we next bound the variance of our semiparametric prior swap estimator . for any @xmath107 , we can write the variance of our estimator as @xmath111 & = \text{var } \left [ \hat{f}_{\phi|x}^s \frac{f_\theta}{f_\phi } \right ] \\ & = \frac{f_\theta^2}{f_\phi^2 } \text{var } \left [ \hat{f}_{\phi|x}^s \right ] \\ & \leq \frac{c}{th^d}\end{aligned}\ ] ] for some @xmath66 , where we have used the facts that @xmath112 \leq \frac{c}{th^d}$ ] for some @xmath66 and @xmath113 ^ 2 \leq \tilde{c}$ ] for some @xmath114 ( given in @xcite ) .
next , we will use these two results to bound the mean - squared error of our semiparametric prior swap estimator , which shows that it is mean - square consistent .
we can write the mean - squared error as the sum of the variance and the bias - squared , and therefore , @xmath115 & \leq c_1 h^2 + \frac{c_2}{th^d } \\ & = \frac{c}{t^{4/(4+d)}}\end{aligned}\ ] ] for some @xmath66 , using the fact that @xmath63 . the procedure given in alg .
[ alg : priorswapping ] generates samples from @xmath68 ( eq . [ eq : semiparest ] ) as @xmath41 . in alg .
[ alg : priorswapping ] , note that line 2 is equivalent to standard mcmc ( such as metropolis - hastings or hamiltonian monte carlo ) run on a posterior target function proportional to @xmath116 , and thus yields samples from @xmath117 as @xmath41 ( see @xcite for mcmc guarantees ) . in line 3
, we can sample from @xmath118 exactly via the following procedure : for each sample @xmath119 , 1 .
draw an index @xmath120 : @xmath121 .
2 . draw the sample @xmath122 : @xmath123 .
we thus have two streams of asymptotically - exact samples , which is exactly the setting of the product density sample combination procedures @xcite , which run on the output of mcmc algorithms to generate asymptotically - exact samples from the distribution with pdf proportional to the product of the pdfs of the markov chains stationary distributions ( see @xcite , sec . 5 ) .
we give pseudocode for these sample combination procedures in the `` sample combination algorithm pseudocode '' section of the appendix .
therefore , as @xmath41 , line 4 of alg .
[ alg : priorswapping ] yields samples from @xmath124 \left [ \frac{1}{t } \sum_{t=1}^t \hat{f}_{\phi|x}^p(\widetilde{\theta}_t)^{-1 } k \left(\frac{\|\theta - \widetilde{\theta}_t\|}{h } \right ) \right ] $ ] .
here we summarize the complexity of our prior swapping algorithms . assume that we have run an existing inference algorithm to compute either a fixed - complexity analytic expression for the false - posterior density , or to draw @xmath34 false - posterior samples , given some dataset of @xmath0 observations .
suppose that we wish to compute @xmath125 samples from the true - posterior .
given a fixed - complexity analytic expression for the false - posterior , each step ( i.e. generating each sample ) in an mcmc algorithm requires a constant @xmath30 number of operations , and drawing @xmath125 samples via our procedure therefore requires @xmath126 operations
. given @xmath34 samples from the false - posterior , the asymptotically - exact ( semiparametric ) procedure in alg .
[ alg : priorswapping ] must perform the fixed - complexity prior swapping procedure on a parametric false posterior @xmath19 ( @xmath126 operations for @xmath125 samples ) , draw @xmath125 samples from the correction density ( @xmath35 operations to compute component weights and @xmath126 operations to draw @xmath125 samples , for a total complexity of @xmath127 ) , then apply a density product sample combination algorithm @xcite ( @xmath126 operations for @xmath125 samples ) .
therefore , the asymptotically - exact ( semiparametric ) procedure has a total complexity of @xmath127 to draw @xmath125 samples . in general ,
given the false - posterior inference result , none of the prior swapping algorithms depend on the number of data points @xmath0 .
here we give pseudocode for the prior swapping procedure , given a fixed - complexity ( parametric ) false posterior expression @xmath59 , using the prior swap functions @xmath61 and @xmath128 ( eqs .
[ psfixedcomplexity01 ] and [ psfixedcomplexity02 ] ) . in alg .
[ alg : ps_mh ] , we show prior swapping via the metropolis - hastings algorithm , which makes use of @xmath61 . in alg .
[ alg : ps_hmc ] we show prior swapping via hamiltonian monte carlo , which makes use of @xmath128 . a special case of alg .
[ alg : ps_hmc ] , which occurs when we set the number of simulation steps to @xmath129 ( in line 6 ) , is prior swapping via langevin dynamics .
initialize @xmath130 .
@xmath46 initialize markov chain .
+ initialize @xmath130 .
@xmath46 initialize markov chain .
here we give pseudocode for a product density sample combination algorithm @xcite .
we will write this algorithm for our setting ( i.e. for generating samples from the product of _ two _ densities ) though these algorithms are typically more general .
we mainly follow alg . 1 from @xcite .
the intuitive idea behind these algorithms is the following : given two sets of samples @xmath51 @xmath52 @xmath53 and @xmath54 @xmath52 @xmath55 , they aim to return @xmath56 @xmath2 @xmath57 . at each iteration in this algorithm , four steps are carried out : 1 .
choose one of the two input sample sets ( uniformly at random ) .
2 . re - draw a sample from the chosen sample set .
3 . accept this drawn sample , or reject it and keep the previously drawn sample .
4 . save a noisy average of the two current samples ( one from each sample set ) .
we give the pseudocode for this procedure in alg .
[ alg : samplecombination ] .
draw @xmath131 .
@xmath46 initialize sample indices . + set @xmath132
the asymptotically - exact method for prior swapping allows for a related way to more easily incorporate observed prior information .
often , instead of ( or in addition to ) an analytic expression for our prior beliefs , we have prior observations , such as previously observed outcomes of what we aim to infer .
incorporating information such as this can lead to better data - driven priors . here
, we will extend our asymptotically - exact prior swapping procedure to this setting .
suppose that we only have access to samples @xmath133 from the true - prior @xmath71 , in addition to false - posterior samples @xmath134 from @xmath19 . in this case
, we can apply a density product sample combination algorithm @xcite ( described in the `` sample combination algorithm pseudocode '' section of the appendix ) to generate samples @xmath135 from @xmath136 @xmath137 @xmath138 , and then construct a semiparametric estimate @xcite of this density ( the same one used in defining the asympotically - exact prior swapping method ) , written @xmath139 we then define the prior swap function for observed prior samples , @xmath140 , to be @xmath141 \left [ \frac{1}{t } \sum_{t=1}^t w'_t k \left(\frac{\|\theta - \widetilde{\theta}_t\|}{h}\right ) \right],\end{aligned}\ ] ] where we ve defined @xmath142 . as in the asymptotically - exact prior swapping procedure , @xmath140 is proportional to the product of two densities , both of which we can easily sample from ( note that we can sample from @xmath143 using the fixed - complexity parametric prior swapping method ) .
as before , we then apply the density product sample combination methods to generate samples from @xmath144 .
we summarize this full procedure in alg [ alg : priorswappingobserved ] .
sample @xmath145 via @xmath146
. + estimate @xmath147 using @xmath45 @xmath46 parametric estimate .
+ sample @xmath148 @xmath46 parametric prior swapping .
+ sample @xmath149 .
+ sample @xmath150 via @xmath151 . | while bayesian methods are praised for their ability to incorporate useful prior knowledge , in practice , priors that allow for computationally convenient or tractable inference are more commonly used . in this paper , we investigate the following question : _ for a given model , is it possible to use any convenient prior to infer a false posterior , and afterwards , given some true prior of interest , quickly transform this result into the true posterior ? _ we present a procedure to carry out this task : given an inferred false posterior and true prior , our algorithm generates samples from the true posterior .
this transformation procedure , which we call `` prior swapping '' works for arbitrary priors .
notably , its cost is independent of data size .
it therefore allows us , in some cases , to apply significantly less - costly inference procedures to more - sophisticated models than previously possible .
it also lets us quickly perform any additional inferences , such as with updated priors or for many different hyperparameter settings , without touching the data .
we prove that our method can generate asymptotically exact samples , and demonstrate it empirically on a number of models and priors . |
deuteron - gold ( ) collisions have been extensively studied at the relativistic heavy ion collider ( rhic ) , both for their intrinsic interest and as a control experiment @xcite to judge the suppression seen in central gold - gold ( ) collisions at sufficiently high transverse momenta ( @xmath7 ) @xcite .
unexpectedly , not only data , but recent extended @xmath7 coverage data also display a suppressed nuclear modification factor in central collisions@xcite .
this motivates a study of possible mechanisms that may result in a nuclear modification factor smaller than unity at sufficiently high transverse momenta ( @xmath8 gev / c ) in central collisions .
the nuclear modification factor @xmath9 compares the spectra of produced particles in collisions to a hypothetical scenario in which the nuclear collisions are assumed to be a superposition of the appropriate number of nucleon - nucleon collisions . in the transverse momentum window @xmath10
gev / c @xmath11 gev / c , the nuclear modification factors are dominated by the cronin peak @xcite .
several physical pictures of this enhancement have been proposed @xcite .
one family of models @xcite advances an explanation of the cronin effect in terms of the interplay between nuclear shadowing @xcite and the multiple scattering of particles propagating in the strongly - interacting medium ( multiscattering ) @xcite . using the hijing shadowing prescription @xcite , our model gave a reasonable description of the cronin effect in central collisions at midrapidity @xcite .
at the same time , we obtained nuclear modification factors close to unity at high transverse momenta ( @xmath12 gev / c @xmath13 gev / c ) . in the high-@xmath7 region
multiscattering no longer affects @xmath9 .
it is then natural to ask if nuclear shadowing can explain suppression effects at high @xmath7 . in particular , since at @xmath14 @xmath15gev we are in the emc region of the shadowing function for @xmath12 gev / c @xmath13 gev / c , we ask if the emc effect @xcite plays a role in understanding these experiments .
in this paper we first investigate collisions at the highest rhic energies and the role of the emc effect at transverse momenta up to @xmath16 gev / c .
we then examine whether recent shadowing parameterizations incorporating theoretical uncertainties for the first time@xcite can account for the experimental information .
we also display calculational results with a modest energy loss using energy - loss parameters applied earlier to data .
finally we extend our considerations to the energy range of the large hadron collider ( lhc ) .
figure [ fig1 ] displays recent phenix data ( triangles with error bars)@xcite for the most central collisions , where a high-@xmath7 suppression is clearly seen .
while incoherent multiscattering can only lead to enhancement in @xmath9 , nuclear shadowing displays two regions where an @xmath17 can be expected : ( i ) at small @xmath18 ( @xmath19 ) , and ( ii ) in the emc region ( @xmath20 ) @xcite . at rhic energies
the small-@xmath18 region is inconsequential at @xmath8 gev / c .
thus we focus attention on the emc effect as a possible mechanism for the measured suppression .
various shadowing parameterizations developed in the last 15 years @xcite show different behaviors at small-@xmath18 , but the emc region appears rather robust in most models . to see the effect of the emc region ,
we calculate pion production in a wide momentum range . for this purpose
we use a perturbative qcd improved parton model @xcite .
the model is based on the factorization theorem and generates the invariant cross section as a convolution of ( nuclear ) parton distribution functions @xmath21 , perturbative qcd cross sections @xmath22 , and fragmentation functions @xmath23 .
we perform the calculation in leading order , following refs .
@xcite : @xmath24 where @xmath25 and @xmath26 represent the factorization and fragmentation scales , respectively , @xmath27 , @xmath28 , and @xmath29 are momentum fractions , and @xmath30-s stand for two - dimensional transverse momentum vectors .
the initial state effects of shadowing and multiscattering are included following the treatment in refs .
@xcite . since the effects we investigate are on the @xmath31 level , it is customary to present the obtained results on a linear scale in terms of the nuclear modification factor @xmath32 here @xmath33 is the average number of binary collisions in the various impact - parameter bins . together with the data in central collisions , we display our results with several shadowing parameterizations in fig.[fig1 ] .
we use the hijing shadowing including nuclear multiscattering ( solid lines ) , the eks shadowing ( where multiscattering is represented by strong anti - shadowing ) ( dashed line ) , and the hkn parameterization ( with and without nuclear multiscattering , dotted and dash - dotted lines , respectively ) .
it can be seen in fig .
[ fig1 ] that the suppression associated with the emc effect shows up at transverse momenta @xmath34 gev / c in all models considered , and does not explain the suppression in the data at around @xmath2 gev / c .
the cronin peak at @xmath35 gev / c is best reproduced by the hijing parameterization .
the `` hkn+multiscattering '' model appears to overshoot the data at low @xmath7 , while it gives very similar results to hijing at @xmath36 gev / c .
since the uncertainty of the hkn nuclear parton distribution functions is also available , we next examine the theoretical uncertainty of the description . in fig .
[ fig2 ] the effect of the uncertainty given by the hkn shadowing parameterization is illustrated on the nuclear modification factor , @xmath9 .
the errors are calculated by the hessian method , using the original code of the hkn group .
the mean value as a function of @xmath7 is represented by a solid line , surrounded by an error band of approximately @xmath37 .
although the data fall within this band at low @xmath7 , for @xmath38 gev / c the observed suppression is stronger than allowed by this calculation . since we are not aware of any other initial - state suppression , we consider physics operating in the final state .
in particular , jet energy loss suggests itself @xcite .
this idea is supported by the presence of a minor energy loss effect in peripheral collisions @xcite .
assuming @xmath39 fm for the average static transverse size of the traversed medium , and the usual glv parameters of @xmath40 fm mean free path and @xmath41 gev screening mass , we calculate the effect of jet energy loss on the nuclear modification factor for central collisions .
this is displayed in fig .
it can be seen that inclusion of this energy loss results in a parallel down - shift of @xmath9 relative to curves in fig [ fig2 ] .
the slope of the data as a function of @xmath7 is still very different from that of the calculated results . nevertheless , taking into account all experimental and theoretical uncertainties
, one could consider the displayed result with @xmath42 to be an acceptable compromise .
jet energy loss depends on the transverse parton density , which relates to the measurable hadronic quantity @xmath43 , where @xmath44 is the transverse area of the deconfined region . assuming a realistic geometry for central and collisions and considering the experimental data on @xmath45
, we obtain only a factor of @xmath10 difference for the transverse parton densities between the and cases . on this basis
, one could expect the jet - quenching effect in to be even stronger than shown by the calculated band in fig .
however , jet energy loss in non - thermal matter is an open question , which we plan to investigate in a forthcoming paper @xcite .
we expect that the emc region will shift to higher and higher transverse momenta with increasing collision energy . at the same time , due to the increasing parton density , the energy loss should also become larger .
we repeat our calculation for @xmath46 collisions at @xmath47 gev and @xmath48 tev to display these tendencies .
the results are shown in fig .
4 . following the ordering of the lines on the left of the figure , the top curve represents the results with hijing shadowing from fig .
the second line corresponds to a similar calculation ( i.e. hijing shadowing , no energy loss ) at @xmath49 gev c.m .
the next line shows the result at @xmath48 tev without jet quenching , while the dotted and dashed lines illustrate the effect of energy loss .
first we took @xmath39 fm for the transverse size of the medium , and used the value of @xmath40 fm as earlier .
the introduction of a smaller @xmath50 value to represent the increase in the transverse density of colored scattering centers with increasing c.m .
energy gives @xmath51 .
it can be seen that the dip corresponding to the emc effect shifts to higher transverse momenta with increasing energy as expected . while @xmath9 is above unity at @xmath35 gev / c for 0.2 tev ( cronin effect ) , at @xmath52 and @xmath48 tev , where we are deep in the shadowing region at these transverse momenta , at most a small ripple appears on @xmath9 , which is rising towards one with increasing transverse momentum in this region .
the effect of varying @xmath50 at @xmath48 tev is minimal at transverse momenta @xmath53 gev / c , but quite significant at @xmath54 gev / c @xmath55 gev / c .
in conclusion , while this solution appears initially tempting , the emc effect does not explain the unexpected suppression seen at @xmath56 gev / c in @xmath9 at @xmath14 @xmath15gev .
taking into account experimental uncertainties as well as the uncertainties of the hkn nuclear parton distribution functions , the experimental and theoretical results can be brought into agreement using a non - negligible amount of final state parton energy loss ( with standard opacity parameters ) .
the non - thermal nature of the system casts some doubt on the parameter values .
the presence of jet quenching in the system is somewhat surprising at first sight , but is justified by the similarity ( within a factor of @xmath10 ) of the real transverse densities in the and systems . for lhc energies we do not have a fully reliable baseline calculation at present .
it is clearly seen , however , that shadowing ( i.e suppression ) will dominate the momentum region up to @xmath57 gev / c .
the effect of final state interactions ( jet energy loss ) may also reduce the nuclear modification factor , especially in this momentum region .
the emc dip in the nuclear modification factor moves toward higher and higher transverse momenta with increasing energy .
we thank shunzo kumano for providing the code and the data for calculating the error bars of the hkn parameterization of parton distribution functions . our work was supported in part by hungarian otka t043455 , t047050 , and
nk62044 , by the u.s .
department of energy under grant u.s .
de - fg02 - 86er40251 , and jointly by the u.s . and
hungary under mta - nsf - otka oise-0435701 .
et al . _ ( phenix ) , phys .
lett . * 91 * , 072303 ( 2003 ) ; phys . rev . *
c74 * , 024904 ( 2006 ) .
b.a . cole , nucl .
phys . * a774 * 225 ( 2006 )
. j. adams _ et al . _
( star ) , phys .
lett . * 91 * , 072304 ( 2003 ) ; phys . rev . *
c70 * , 064907 ( 2004 ) .
et al . _ ( phenix ) , nucl . phys . *
a757 * , 184 ( 2005 ) .
j. adams _ et al . _
( star ) , nucl . phys . *
a757 * , 102 ( 2005 ) .
et al . _ ( phenix ) , arxiv : nucl - ex/0610036 .
[ cp coll .
rev . * d11 * , 3105 ( 1975 ) .
d. antreasyan _ et al . _
[ cp coll . ] , phys .
* d19 * , 764 ( 1979 ) . m. gyulassy and p. lvai , phys . lett . *
b442 * , 1 ( 1998 ) .
kopeliovich , j. nemchik , a. schafer , a.v .
tarasov , phys .
lett . * 88 * , 232303 ( 2002 ) .
d. kharzeev , yu.v .
kovchegov , k. tuchin , phys . rev . *
d68 * , 094013 ( 2003 ) .
a. accardi , m. gyulassy , phys .
b586 * , 244 ( 2004 ) .
blaizot , f. gelis , r. venugopalan , nucl . phys . *
a743 * , 12 ( 2004 ) .
hwa , c.b .
yang , phys .
* 93 * , 082302 ( 2004 ) .
c.y . wong and h. wang ,
phys . rev .
* c58 * , 376 ( 1998 ) .
wang , phys . rev . *
c61 * , 064910 ( 2001 ) .
g. papp , p. lvai , and g. fai , phys .
* c61 * , 021902 ( 2000 ) . y. zhang _
et al . _ ,
phys . rev . * c65 * , 034903 ( 2002 ) .
g. papp , g.g .
barnafldi , g. fai , p. levai , and y. zhang , nucl .
a698 * , 627 ( 2002 ) .
barnafldi , p. lvai , g. papp , g. fai , and m. gyulassy , eur .
j. * c33 * , s609 ( 2004 ) .
d.f . geesaman _
et al . _ ,
nucl . part .
sci . * 45 * , 337 ( 1995 ) .
k. eskola , v.j .
kolhinen , and c.a .
salgado , eur .
j. * c9 * , 61 ( 1999 ) .
n . wang and m. gyulassy , phys . rev .
* d44 * , 3501 ( 1991 ) ; comput .
. commun . * 83 * , 307 ( 1994 ) .
s.j . li and x.n .
wang , phys .
b527 * , 85 ( 2002 ) . m. hirai , s. kumano , and m. miyama , phys . rev . *
d64 * , 034003 ( 2001 ) .
m. hirai , s. kumano , and t .- h .
nagai , phys . rev . *
d70 * , 044905 ( 2004 ) ; nucl .
. suppl . * 139 * , 21 ( 2005 ) .
p. lvai , g.g .
barnafldi , g. fai , and g. papp , + arxiv : nucl - th/0306019 .
p. lvai , g.g .
barnafldi , g. fai , and g. papp , + nucl . phys . *
a783 * , 101c ( 2007 ) .
* a698 * , 631 ( 2002 ) . i. vitev and m. gyulassy , phys .
lett . * 89 * , 252301 ( 2002 ) .
m. gyulassy , p. lvai , i. vitev , phys .
lett . * 85 * , 5535 ( 2000 ) ; nucl . phys . *
b571 * , 197 ( 2000 ) .
m. gyulassy , p. lvai , i. vitev , nucl .
b594 * , 371 ( 2001 ) .
g. g. barnafoldi , p. levai , g. papp and g. fai , arxiv : hep - ph/0609023 .
g. g. barnafoldi , p. levai , g. papp and g. fai , _ in preparation_. | we investigate the influence of modified nuclear parton distribution functions ( pdfs ) on high- hadron production at rhic and lhc energies using a pqcd - improved parton model . for application at rhic , we focus on the possible contribution of the emc modification of the nuclear pdfs in the @xmath0 region to the observed suppression of production at @xmath1 gev / c in collisions .
we study three different parameterizations of the nuclear pdf modifications and find that they give consistent results for for neutral pions in the region @xmath2 gev / c @xmath3 gev / c .
we find that the emc suppression of the parton distributions in the @xmath4 nucleus does not strongly influence the for in the region where the suppression is observed .
using the hkn parameterization , we evaluate systematic errors in the theoretical resulting from uncertainties in the nuclear pdfs .
the measured nuclear modification factor is inconsistent with the pqcd model result for @xmath1 gev / c even when the systematic uncertainties in the nuclear pdfs are accounted for .
the inclusion of a small final - state energy loss can reduce the discrepancy with the data , but we can not perfectly reproduce the dependence of the measured . for the lhc , we find that shadowing of the nuclear pdfs produces a large suppression in the yield of hadrons with @xmath5 gev / c in @xmath6 collisions . |
goal of heavy ion collision ( hic ) experiments is to reveal properties of the compressed hot nuclear matter created during the collision of heavy nuclei .
final momentum distributions of most abundant particles - hadrons are however influenced ( if not determined ) during the dilute and late freeze - out stage of the heavy ion collision .
interesting information about the early stage of the collision is in this case hidden by subsequent collective effects of strongly interacting hadrons . fortunately other particles - leptons and photons produced in all stages of hic process , are not influenced by the later hadronic phase .
thus distributions of leptons and photons can provide us with more or less direct information about the early stages of the collision process .
recent experimental results on invariant mass distribution of dilepton pairs @xcite have attracted increased attention due to the observed excess of dilepton yield at regions @xmath0gev ( na50 ) and @xmath1 gev ( ceres ) . theoretical analysis on dilepton pair production @xcite leads to the conclusion that unconventional production mechanisms ( e.g. modified @xmath2-meson mass @xcite or in - medium pion dispersion relation @xcite ) are necessary to explain present experimental data . in this contribution
specific results of the computer simulation of dilepton pairs production from expanding pion gas are presented . besides the invariant mass distribution of the produced dileptons also other properties of the dileptonic signal produced from the pion gas via @xmath3 annihilation channel are analyzed .
short description of the computer simulation is given in next section .
influence of the finite size of pion gas ( and its initial state ) on invariant mass spectrum of dileptons is discussed in section 3 using results of the computer simulation . in section 4 rapidity distribution of dilepton pairs obtained from the simulation
is presented .
non - equilibrium features of the pion gas expansion are discussed in this context .
second - order azimuthal asymmetry in the transversal momentum distribution of dileptons is predicted by the simulation of non - central pb+pb 160 gev / n collisions in section 5 .
possible consequences of the effects found and presented here on other areas of hic research are scetched at the end of this paper .
for the simulation of dilepton production from the expanding pion gas rescattering model @xcite ( inspired by the model of t.j.humanic @xcite ) was used .
initial stage of pion gas in pb+pb 160 gev / n collisions ( positions , momenta and time of creation of pions ) was taken from events generated by cascade program @xcite . in the rescattering program pions
move in small time steps @xmath4fm as free relativistic particles and their momenta are changed in mutual collisions . in each time step distance of pions
@xmath5 is compared to critical distance @xmath6 , where @xmath7 is total elastic @xmath8 cross section .
if condition @xmath9 is fulfilled collision procedure for a given pair of pions is called . more detailed description of the simulation can be found in @xcite .
two independent methods of the implementation of dilepton production via @xmath3 annihilation channel were used in the simulation .
first one is based on standard expression for the rate of dilepton production @xcite : @xmath10 \label{ee1}\ ] ] where @xmath11 is the cross section for @xmath12 at given @xmath13 , @xmath14 , @xmath15 is the relative velocity @xmath16 and @xmath17 is distribution function of pions in phase space , @xmath18 .
distribution function @xmath17 is calculated by rescattering program in our case .
it depends mainly on the isospin averaged cross section @xmath19 for elastic @xmath20 interaction which regulates scattering of pions and it is influenced also by formation time @xmath21 and interaction time @xmath22 parameters of the rescattering model used @xcite . after multiplying and dividing right - hand side of eq.([ee1 ] ) by @xmath23 and integrating over @xmath24 the rate of dileptons from the pion gas can be expressed as @xcite : @xmath25 where @xmath26 is distribution of @xmath27 collisions as a function of invariant mass @xmath28 of the colliding @xmath27 pairs and @xmath29 denotes ratio of the number of @xmath30 collisions to all @xmath20 interactions .
distribution @xmath26 is calculated by the rescattering program .
0.15 cm = 10.7 cm 1.7pt 0.15 cm at the same time another method for the simulation of @xmath3 annihilation in the expanding pion gas was used . cross section @xmath31 for the process @xmath32
was multiplied by factor 13000 in the rescattering program and pairs of pions were tested ( in each time step ) for the critical distance @xmath33 calculated from this enlarged cross section : @xmath34 when this condition is fulfilled for a given pair of pions invariant mass @xmath28 and momenta of pions @xmath35 are remembered ( stored into file ) .
positions and momenta of pions fulfilling condition ( [ dce ] ) are left unchanged in the program . in this way probability for the process of pion annihilation
is enhanced to the level comparable with the elastic pion - pion interactions . at the same time process of the expansion of pion gas is not influenced by this high probability of pion - pion annihilitaion because pions are not annihilated finally .
just possibility of this process is recorded .
results obtained by the above described two independent methods of dilepton production simulation were nearly identical in all features ( invariant mass distribution , @xmath36 distribution , azimuthal and rapidity distribution ) tested so far .
invariant mass distribution of dilepton pairs produced in ultrarelativistic heavy ion collisions can be divided into separate regions .
dileptons with invariant mass @xmath37 gev are produced mainly in the process of parton - parton annihilation .
region of @xmath28 below 300 mev is dominated by dalitz decays of @xmath38 and @xmath39 mesons .
also electromagnetic bremsstrahlung contributes to this low mass region of dilelectron pairs .
dilepton pairs at region of @xmath28 close to 1gev are decay products of @xmath40,@xmath41 and @xmath42 mesons . in the region between 280
mev and 1 gev thermal production from @xmath3 annihilation contributes
. this list of dilepton production channels would be incomplete if we did not mention dileptons from decays of @xmath43 , @xmath44 and couples of @xmath45 mesons .
@xmath3 annihilation channel of dilepton production plays important role in the case of ultrarelativistic heavy ion collisions due to high multiplicities of secondary produced pions ( above 2000 in central pb+pb 160 gev / n collisions ) .
0.25 cm 1.7pt 0.15 cm in this section invariant mass spectra of dileptons originating from @xmath3 annihilation generated by the rescattering simulation are presented . in the paper @xcite
mainly influence of formation time parameter on the shape of dilepton spectrum was investigated . here
dependence of the dilepton mass spectrum on the initial state of the pion gas is studied .
initial momenta and positions of pions in events produced by cascade generator @xcite are uncorrelated .
it means that momentum of the pion does not depend on coordinates of its creation e.g. in transversal plane .
invariant mass spectrum obtained for these @xmath46 uncorrelated events is shown on fig.2a ) .
thick line on fig.2a ) is obtained for same values of @xmath47 and @xmath22 parameters of the rescattering model but for @xmath46 correlated events .
these events were obtained by the following modification of events produced by cascade generator @xcite : for each pion scalar product of its transversal momentum and position in transversal plane @xmath48 was evaluated . in the case of negative numerical value of the scalar product ,
position of the pion @xmath49 was replaced by the oposit position in transversal plane @xmath50 . in this way momentum - space correlation
was generated in the initial state of pion gas .
momenta of pions ( not changed ) are oriented predominantely out from the center of transversal plane .
resulting invariant mass spectrum of dileptons calculated for @xmath46 correlated events differs significantly from the spectrum of uncorrelated events .
low mass dilepton pairs are strongly enhanced in the case of @xmath46 correlated events . in fig.2b
influence of formation time parameter @xmath21 on invariant mass spectrum of dileptons is shown for @xmath46 correlated events .
small values of formation time parameter enhance production of dilepton pairs in @xmath40-meson region as it was found and discussed paper @xcite . what is physical origin of the low mass dilepton enhancement in the case of @xmath46 correlated events ? why non - zero values of formation time suppress dilepton pairs in @xmath40-mass region compared to the low mass region pairs ?
two independent effects originating from the finite size of pion gas can serve as explanation .
first one was already discussed in the work @xcite .
this consideration is reproduced very shortly here .
formation time @xcite excludes pions from the rescattering process and therefore also from the possibility of annihilation into dilepton pairs .
faster pions are forbidden to interact for longer time due to the lorentz dilation of the formation time implemented in the rescattering program .
due to the finite size of the pion gas faster pions can escape from the interaction region without any rescattering or annihilation .
thus invariant mass spectrum if dileptons is enhanced at low mass region by interactions of slow pions . at the same time
another mechanism also resulting from finite size of the expanding pion gas influences invariant mass spectrum of dileptons .
invariant mass quantity depends not only on the total size of momenta of pions being scattered but also on the angle between the momenta : @xmath51 for small values of scattering angle @xmath42 , it means for pions moving in parallel directions , invariant mass quantitiy ( [ mfi ] ) is small .
at the same time for pions moving in oposit directions invariant mass is large for the same values of momenta of pions .
this effect can play significant role if the mean free path of interacting particles is comparable with the initial size of the expanding gas .
this seems to be the case of the pion gas created in ultrarelativistic heavy ion collisions .
momentum - space correlation leads to the higher probability of collisions with small scattering angles and therefore it enhances production of low mass dileptons ( see fig.2a ) .
in the case of ultrarelativistic heavy ion collisions momentum - space correlation at the initial stage of pion gas can be generated also by transversal flow in prehadronic phase .
rapidity distribution of dilepton pairs produced by pion annihilation is determined by rapidity distribution of momentum sum @xmath52 of anninhilating pions ( @xmath53,@xmath54 are momenta of @xmath55 and @xmath56 ) .
this quantity is conserved in the process of @xmath3 annihilation , @xmath57 . in fig.3a ) and b ) rapidity distributions of pions before and after the rescattering process are shown .
rapidity distribution of pions is not influenced significantly by the rescattering process . however resulting rapidity distribution of dilepton pairs exhibits unexpected minimum at midrapidity surrounded by two maxima .
0.25 cm 1.7pt 0.15 cm the result shown on fig.3 is obtained for the simulation of pb+pb 160 gev / n events with values of rescattering model parameters @xmath58 fm / c , @xmath59 fm / c . in this case
number of collisions per pion is close to 1.0 and minimum in the rapidity distribution of dileptons is strongest .
for higher collision rates ( obtained for smaller vaues of @xmath60 and @xmath22 parameters ) the minimum is significantly smaller . for number of collisions per pion above 10 coll/@xmath61 the minimum disappeares at all .
such behaviour indicates that the minimum in the rapidity distribution of dileptons is generated by non - equilibrium process in the expanding pion gas .
this effect is studied in phd work @xcite . from experimental point of view data on rapidity distribution of dileptons produced in @xmath62 or @xmath63
collisions are rather rare .
it is not excluded that the only rapidity distribution of dileptons which seem to originate from @xmath3 annihilation was obtained in pioneering experiments of dls collaboration at bevalac accelerator in berkeley @xcite .
azimuthal asymmetries in transverse momentum distributions of particles have been clearly identified in ultrarelativistic @xcite non - central heavy ion collisions .
this phenomenon is well understood as a consequence of collective behaviour of nuclear matter or explained by absorption of secondary produced particles in the spectator matter @xcite .
recently also azimuthal asymmetries in transverse momentum distributions of less abundant hadrons - @xmath64 mesons and @xmath65 baryons @xcite have been identified in hic experiments .
however azimuthal asymmetries in transversal momentum distribution of dileptons have been addressed neither experimentally nor theoretically so far .
0.15 cm 1.7pt 0.15 cm from theoretical point of view main mechanisms generating azimuthal asymmetries in transversal momentum distributions of hadrons are not applicable for dileptons .
dileptons leave freely collision volume after being produced without any collective final state interactions or absorption process . in spite of this
the computer simulation described shortly in section 2 predicts significant second - order asymmetry in transversal momentum distribution of dilepton pairs .
result shown in fig.4 was obtained from the simulation of 40 non - central pb+pb 160 gev / n @xmath66 fm events . for each produced dilepton pair orientation of momentum @xmath67 in respect to the reaction plane ( impact parameter ) was determined .
histogram of azimuthal angles @xmath68 of dilepton pairs was filled and significant second order asymmetry was found .
the fit of the histogram shown in fig.4 to the function : @xmath69 \label{s2}\ ] ] gave numerical value of the asymmetry coefficient @xmath70 .
corresponding value of @xmath71 parameter ( simplified for our purposes ) used to characterize squeeze - out effect of hadrons at bevalac energies @xcite is : @xmath72 orientation of the reaction plane is parallel to @xmath73 axis in our simulation , azimuthal asymmetry of dileptons is oriented in the reaction plane . from the analysis of second order asymmetry in the transversal momentum distribution of pions
it comes out @xcite that the asymmetry is strongest for the rate of collisions per pion close to 4.0 coll/@xmath61 .
for higher collision rates strength of the asymmetry decreases though it does not tend to vanish in thermal equilibrium limit .
the same behaviour is expected for the azimuthal asymmetry of dileptons presented here for the first time .
mechanism of generation of the azimuthal asymmetry of dileptons is well understood @xcite at present . the asymmetry is closely related to the spatial distribution of the pion gas in transversal plane which is asymmetrical in non - central events .
two independent methods of the simulation of dilepton production from the expanding pion gas were found to generate nearly identical results .
invariant mass spectrum of dileptons produced via @xmath3 annihilation channel was shown to be sensitive to momentum - space correlation in initial stage of the pion gas .
generated rapidity distribution of dileptons exhibits unexpected minimum at midrapidity . the minimum seems to be a consequence of a non - equilibrium process in the expanding pion gas since it becomes weaker and disappears for higher number of collisions ( above 10 coll/@xmath74 ) in the pion gas . for non - central collisions
second order asymmetry in azimuthal distribution of dileptons is found .
this asymmetry is a consequence of asymmetrical overlapping region of the colliding nuclei in transversal plane .
azimuthal asymmetry of dileptons might be accessible experimentally if the reaction plane orientation is determined from hadronic signal on dilepton detector .
in this contribution we study expansion of interacting pion gas created in ultrarelativistic heavy ion collisions .
non - equilibrium and asymmetrical features of the expansion exhibit itself in rapidity and azimuthal distributions of particles participating or created in the rescattering process .
rescattering phenomenon is not related only to the hadronic phase of nuclear matter produced in hic .
it can equally be present also in prehadronic phase where it can lead @xcite to similar effects as those revealed in the pion gas simulation .
investigation of the non - equilibrium expansion of the parton gas possibly created also in non - central collisions might shed light on other interesting aspects @xcite of hic research . | results of the computer simulation of dilepton production from the expanding pion gas created in ultrarelativistic pb+pb 160 gev / n collisions are presented . finite size of the expanding pion gas influences invariant mass spectrum of dileptons .
sensitivity of the shape of dilepton mass spectrum to the initial stage of the pion gas is discussed .
second order asymmetry in azimuthal distribution of dileptons is predicted for non - central collisions .
non - equilibrium nature of the pion gas expansion is found to exhibit itself in rapidity distribution of produced dilepton pairs .
= 7.2pt = 7.2pt + + + + 0.75 cm _ peter filip _
+ 0.35 cm _ institute of physics , slovak academy of sciences , sk 84228 + 0.15 cm faculty of mathematics and physics , comenius university , sk 84215 + 0.15 cm bratislava , slovak republic _ + |
computer simulations have developed into a powerful tool to predict and optimize material properties . however , even given the ever increasing computational power , relevant time and length scales , in particular for biological and synthetic macromolecules @xcite , will remain prohibitively long for fully atomistic simulations , and multiscale methods are crucial to make progress .
while structure - based coarse - graining @xcite has been quite successful @xcite , challenges remain like transferability ( _ e.g. _ , from bulk to surfaces ) , and in particular the correct treatment ( or _ a posteriori _ deduction ) of the materials _ dynamical _ properties through the coarse - grained dynamics . for the latter , markov state models ( msms )
@xcite have been used successfully to tackle the evolution of large proteins towards their native state , bridging the gap from molecular dynamics on nanoseconds to folding on milliseconds @xcite .
the dynamics of msms is a discrete - time master equation with transition probabilities obey detailed balance on a network of long - lived ( metastable ) mesostates . here
we describe a systematic approach to construct msms with dynamics that break detailed balance .
we employ this method to study a model polymer in shear flow .
the rheology of dilute flexible polymers has been studied extensively due to their fundamental and practical relevance @xcite .
examples include biomolecules such as the von willebrand factor in blood plasma and dna in steady shear flow @xcite .
the shear drag can overcome the entropic forces favoring coiled or globular configurations and stretch the polymers , which might be a continuous or even discontinuous transition @xcite .
motion of dna tethered to a planar surface has been described as _
cyclic _ in experiments @xcite and computer simulations @xcite .
driving a system away from thermal equilibrium implies a non - vanishing entropy production . in a steady state , this entropy is produced exclusively in _ cycles _ since there are no sources and sinks for the probability .
this poses novel challenges for constructing coarse - grained models in non - equilibrium , which only very recently have been begun to be addressed @xcite . reducing the complexity by removing states implies a reduced entropy production , which severely influences dynamical properties and fluctuations @xcite . in ref .
, we have introduced non - equilibrium markov state modeling ( ne - msm ) performing the coarse - graining in cycle space instead of collecting configurations in metastable basins as for conventional msm .
the analog of these basins are now _ communities _ of cycles with similar properties .
the coarse - graining procedure preserves the entropy production of these communities , which makes our approach consistent with stochastic thermodynamics @xcite .
we study a single model polymer with @xmath0 beads in shear flow inspired by ref . .
we employ brownian dynamics ( bd ) simulations with @xmath1 for the bead positions @xmath2 , where @xmath3 represents the shear flow .
interactions with solvent particles are modeled by a random force with correlations @xmath4 , where upper indices label vector components .
the potential energy @xmath5 is split into the non - bonded short - ranged lennard - jones pair potentials @xmath6 $ ] and bonds @xmath7 that connect the nearest neighboring beads . here
, @xmath8 is the distance between the @xmath9-th and @xmath10-th bead , @xmath11 determines the strength of the non - bonded potential , and @xmath12 is the effective spring constant .
all quantities have been non - dimensionalized by rescaling lengths with the bead diameter @xmath13 and timescales with the characteristic monomer diffusion time @xmath14 .
numerical values for the strain rate @xmath15 thus correspond to the weissenberg number .
the polymer is driven into a non - equilibrium steady state through simple shear flow . while some scaling relations depend on hydrodynamic interactions @xcite , the qualitative behavior of the cyclic motion does not and in the following we neglect hydrodynamic interactions . as flow profile
we choose @xmath16 where @xmath15 is the strain rate and @xmath17 is the @xmath18 component of the center of mass of the polymer .
we found that this shift of the flow stabilizes the globular and stretched configurations as it increases the effective barrier for folding / unfolding and thus leads to a better separation between globular and extended states .
qualitatively , the same effect would be expected when including hydrodynamic interactions with the wall .
our conclusions do not depend on this detail .
the polymer is grafted onto a repulsive planar surface ( the @xmath19-@xmath20 plane with @xmath21 ) by fixing the position of the first bead to @xmath22 .
although simplified , this model reproduces the cyclic dynamics found in experiments @xcite .
a reasonable order parameter describing the folding and unfolding of the polymer is the relative end - to - end distance @xmath23 , where @xmath24 corresponds to a straight line of touching beads .
we perform bd simulations for multiple values of @xmath25 , see fig .
[ fig : polymer ] .
we find different behaviors of @xmath26 that we categorize into three regimes .
for @xmath27 the polymer remains collapsed , while for @xmath28 it is dominantly found in elongated conformations . for intermediate strain rates the polymer exhibits transitions between globular and elongated conformations , which was also found in similar simulations for free polymers @xcite and grafted polymers @xcite under shear .
the exemplary time series for @xmath29 in the inset of fig .
[ fig : polymer ] shows a clear separation of both states with random lifetimes and fast transitions .
the average folding time @xmath30 is much larger than the intrinsic rouse time @xmath31 of the polymer . as a function of strain rate @xmath15 . for @xmath29 ( yellow / bright circle )
the inset shows an exemplary time series .
also shown are two snapshots for the globule ( bottom ) and extended polymer ( top ) . ]
our goal is to construct a dynamically consistent ne - msm with as few discrete states as possible .
every discrete state @xmath32 is a set of @xmath33 particle positions , which we collect in the vector @xmath34 with @xmath35 entries .
these states , or _ centroids _ , represent discrete volumes of configuration space , _ i.e. _ , many configurations with slightly different positions . following ref . , the outline of the algorithm is as follows : the first step is to generate many ( specifically @xmath36 ) centroids using the bd simulation data as input employing a spatial clustering algorithm ( for technical details , see ref . and appendix [ sec : spatial ] ) . from the counting matrix of transitions in the bd data we construct an approximate markov process and identify cycles .
the crucial step is to group cycles into _ communities _ ( _ i.e. _ , clusters of cycles with similar properties defined through suitable order parameters ) and determine one cycle representative for each community . in the actual coarse - graining , only representatives and the states that they visit are retained .
the final step is then to rescale the markov dynamics so that it preserves the entropy production of every community as well as the total entropy production .
entropy production arises from non - vanishing probability currents @xmath37 with probability fluxes @xmath38 along transitions @xmath39 . in thermal equilibrium
the condition of detailed balance holds , @xmath40 , which implies zero currents .
the mean entropy production rate @xcite is @xmath41 with _
affinities _ @xmath42 .
in a steady state , kirchhoff s law implies that probability flows in cycles .
a cycle @xmath43 is defined as an ordered set of states , at the end of which the starting state is reached again and all other states are visited exactly once .
cycles that differ only in the cyclic permutation of their states are considered identical .
we extend the concept of affinities to cycles , yielding the cycle affinities @xmath44 , where the sum is over all edges along cycle @xmath45 .
traversing a full cycle , the entropy produced equals the cycle affinity .
graph theoretical results allow us to decompose the probability fluxes @xcite @xmath46 into cycle fluxes with non - negative weights @xmath47 .
the summation is over all cycles @xmath45 that include the edge @xmath48 .
the number of cycles quickly becomes very large .
here we employ the algorithm introduced in ref . and summarized in appendix [ sec : decomp ] to efficiently determine a subset of cycles with non - vanishing weights @xmath49 .
these cycles have non - negative cycle affinities ( and thus entropy production ) , which makes our approach conceptually different from the well - known schnakenberg theory @xcite . inserting eq . into eq . , we obtain @xmath50 , which clearly shows that all entropy production is encoded in cycles . at this point
we have determined all entropy - producing cycles . to make progress
, we now need to find similarities between cycles .
one approach is to consider the connectivity between cycles @xcite . for the globule - stretch transition
, we found this approach to be less suitable since it groups cycles that are located in different parts of configuration space . instead
, here we describe an alternative , general method to reduce the complexity of the high - dimensional configuration space to a few variables based on the topology of cycles . as our main tool
, we perform a principal component analysis ( pca ) @xcite , which is an orthogonal linear transformation returning the eigenvectors ( the principal components ) and eigenvalues of the covariance matrix of the centroid positions .
the principal component @xmath51 corresponding to the largest eigenvalue coincides with the direction in configuration space exhibiting the largest variance ; in the present case it already accounts for @xmath52% of the observed variance . in fig .
[ fig : cnf ] the normalized projections @xmath53 of centroids onto first and second principal component are plotted , where the normalization @xmath54 ensures that @xmath55 $ ] .
centroids with small ( negative ) values for @xmath56 correspond to globular configurations ( they contribute little to the observed variance of positions ) , large ( positive ) values correspond to stretched configurations . the second component indicates the variance within these states .
both globular and stretched configurations show larger fluctuations while the intermediate states with @xmath57 exhibit less fluctuations .
hence , the pca reproduces the expected , typical picture of two basins with intermediate transition states . ) .
scatter plot of the normalized projections of all centroids onto the two largest principal components .
colors indicate the end - to - end distance @xmath58 . ] going back to cycles , we now define two variables : ( i ) the cycle centers @xmath59 with @xmath60 the number of centroids in cycle @xmath45 and ( ii ) the cycle diameters @xmath61 the set of points @xmath62 is plotted in fig .
[ fig : cyc]a , where every point now represents a single cycle .
these points are clearly not random .
many cycles have a small diameter @xmath63 but different cycle centers .
we identify these cycles as _ local _ because the centroids @xmath64 in these cycles have similar @xmath56 and thus belong to a compact region in configuration space .
there is a second group of cycles with large diameter , which , consequently , we identify as _ global _ cycles .
basically the same structure is recovered when plotting @xmath65 shown in fig .
[ fig : cyc]b , indicating that local cycles have low affinities and global cycles have large affinities . ) .
( a ) scatter plot of cycle center vs. the cycle diameter , and ( b ) cycle affinity vs. cycle center .
the cycle diameter and centers are computed using a pca ( see text for details ) .
colors represent cycle communities while gray points indicate no community .
the filled black circles indicate the cycle representatives .
( c , d ) the same plots as for ( a , b ) but for three cycle communities . ( e ) the fuzzy partition coefficients computed for multiple communities .
the best result is obtained for 3 communities followed by a slightly lower value for 5 communities . ]
we can turn these insights into a more quantitative statement by partitioning the cycle space into @xmath9 communities . to this end
we employ an implementation of the fuzzy @xmath66-means algorithm ( an implementation is available in ref . ) , which assigns to each cycle a probability for belonging to a specific community .
first , we normalize all three features ( cycle diameters , cycle centers , and affinities ) by their variance to make them comparable .
these features are then used as an input for the fuzzy c - means clustering algorithm returning membership degrees @xmath67 which express the probability that observation @xmath32 belongs to community @xmath68 . to obtain an indicator of
how good the clustering results are we compute the fuzzy partition coefficient ( fpc ) that is defined as the frobenius norm of the membership matrix @xmath69 here @xmath9 is the number of chosen communities and @xmath70 the number of observations ( cycles in our case ) .
the closer the fpc gets to one the better the cycle space can be partitioned into the chosen number of clusters .
the advantage of fuzzy partitioning is that some cycles might not belong to any cycle community while others match well in multiple ones .
the minimal number of cycle communities to account for the collective folding and unfolding dynamics is @xmath71 ( see fig . [
fig : cyc]e ) , but to capture the full dynamics we found @xmath72 cycle communities to be more appropriate .
we identify three local communities colored in blue , green , and red , as well as two global communities colored in magenta and cyan , see fig . [
fig : cyc]a , b .
the blue community represents cycles that correspond to globular configurations while the green and red communities represent similar dynamics for intermediate and fully stretched configurations , respectively .
the global cycle communities connect two ( cyan ) or all three ( magenta ) local communities .
next , we replace each cycle community by one cycle that we will refer to as _ representative _ @xcite .
we find appropriate representatives by mapping mean first - passage times between states that belong to different local cycle communities . after selecting representatives
, we delete all states not belonging to any of the representatives and rescale the remaining transition rates with restrictions that : the total entropy production rate @xmath73 , all remaining cycle affinities @xmath74 and all remaining edge affinities @xmath75 are preserved .
for details see ref . and appendix [ sec : repr ] .
at this stage the coarse - grained msm still contains many states since a single cycle can traverse hundreds of states .
the important point is , however , that the coarse - grained model lost much of its original complexity as it now contains only a few cycles .
we can thus further reduce the number of states . to this end
we identify two dominant motifs , which we refer to as bridge and triangle states .
both motifs build on states that have exactly two neighbors . for bridge states
the neighbors are not connected to each other @xcite .
triangle states complete cycles that do not exhibit positive entropy productions and thus are not part of the decomposition eq . .
we iteratively search and remove these states ( for details see appendix [ sec : bridge ] ) until no more are found , which completes our coarse - graining scheme .
( a ) transition network of the polymer dynamics with 15 centroids ( filled circles ) and 5 cycles .
the colored states green , blue , and red correspond to the colors of the local cycle communities shown in fig .
[ fig : cyc ] . states with a black border are structurally very similar and constitute the transition ensemble . the arrows point in the direction of probability currents ( net flux ) , while the arrow widths represent the magnitude of currents . on average ,
the polymer dynamics follows the direction of the arrows .
arrow colors cyan and magenta correspond to the global cycles .
( b ) transition network in configuration space using the normalized projections onto the two largest principal components .
symbols are the configurations while lines indicate the cycles with the arrows pointing in the direction of the net flux . ]
the final ne - msm for strain rate @xmath29 is shown in fig .
[ fig : tnetwork]a . after removing the bridge and triangle
states the transition network contains 15 states ( centroids ) .
the collective rates for folding and unfolding of this coarse - grained model agree with those obtained from the bd simulations by construction .
moreover , the remaining five cycles now allow for detailed insights into the relevant pathways in non - equilibrium .
the three local cycles are composed of three states , the minimal number for a non - trivial , entropy - producing cycle .
the red and blue cycle connect stable globular and stretched configurations , respectively .
the green cycle represents a metastable intermediate of half - stretched configurations that do not unfold correctly but quickly fold back to the intermediate .
the global cyan cycle also contains half - stretched configurations ( structurally similar to the green cycle ) but here the unfolding reaches the final stretched states before returning to their half - stretched origin .
finally , the magenta cycle represents the full transition from globule to stretched configurations and back .
the five states @xmath76 describing intermediate , half - stretched configurations are very close in configuration space ( see fig . [
fig : tnetwork]b ) .
they constitute the analog of the transition ensemble through which the folding / unfolding has to proceed . in non - equilibrium , however , the folding and unfolding processes follow different paths through this narrow region of configuration space .
the globule to half - stretched transition proceeds along @xmath77 ( with state 4 belonging to the green cycle ) whereas the reverse half - stretched to globule transition proceeds along @xmath78 , with states 13 and 14 belonging to the cyan cycle .
the cycle topology thus reveals the dynamical trapping of the polymer in an intermediate , which can not be captured by structural information alone .
employing order parameters like distances in configuration space or lifetimes to identify mesostates will clearly miss this important feature of cyclic non - equilibrium dynamics .
another question that we can address is dissipation , the role of which for bio - processes has been investigated recently , _
e.g. _ , for self - replication @xcite and in the activation of signaling proteins @xcite .
the rate of dissipated heat @xmath79 created in each cycle community is proportional to their respective entropy production rate @xmath80 .
our analysis reveals that both the blue and red cycles are equally responsible for about 30% of the total dissipated heat , while the green , cyan , and magenta communities produce 5% , 15% , and 50% , respectively .
the latter is caused by the large conformational changes ( folding and unfolding process ) of the polymer .
the blue and red cycles , on the other hand , do not exhibit large conformational changes , therefore , the conformational changes must be on very short timescales , which is confirmed by the large probability currents shown in fig . [
fig : tnetwork]a .
we finally comment on using a pca , which is not necessarily the best method to reduce the dimensionality of configuration space , and more advanced methods such as the time - lagged independent correlation analysis exist @xcite .
moreover , it is common practice to first apply a dimensionality reduction to the continuous molecular dynamics ( or bd ) data and subsequently discretize the reduced configuration space into finite sets .
we did not follow this approach because ne - msms built from reduced configuration space exhibit a significantly lower entropy production rate than ne - msms built from full configuration space ( assuming the same number of centroids is used ) .
recent results @xcite suggest that the fluctuations of general currents are bounded by the entropy production , which directly links dynamic with thermodynamic consistency .
to conclude , we have presented a general method to systematically construct coarse - grained models composed of a few discrete states from molecular data of steadily driven systems .
specifically , we have studied the globule - stretch transition of a simple model polymer , but the method can in principle be applied to more realistic and complex molecules ( such as f@xmath81-atpase @xcite ) , delivering minimal and thermodynamically consistent markov state models .
while these models can be employed to bridge time scales , they also allow insights into the relevant transition pathways . here
we have shown that different directions can pass through the same region of configuration space , which we believe might be a general property of transitions in driven systems .
we thank f. schmid for helpful comments .
we acknowledge financial support by the dfg through the collaborative research center trr 146 ( project a7 ) .
the configuration space is discretized by employing the popular @xmath9-means clustering algorithm which divides the continuous configuration space into @xmath9 volumes and returns their centroids ( center of each volume ) .
next the bd trajectories are projected onto the centroids , storing its dynamical information as simple sequences of centroid indices .
we extract the dynamical information by counting the number of transitions @xmath82 for a given lag time @xmath83 ( time in between transitions ) .
the count matrix @xmath84 is used to approximate the transition matrix @xmath85 by @xmath86 which is also the maximum probability estimator for the true transition operator @xcite .
we additionally require @xmath87 to have two important properties : first , @xmath87 must be ergodic , i.e. each state can be reached from every other state in finite time or the transition network spanned by @xmath87 is connected .
second , all occurring transitions are reversible , i.e. if @xmath88 then @xmath89 . . ]
if the process described by @xmath87 is time - homogeneous its entries @xmath90 are ought to be time - independent . to test this
we perform a lag time analysis @xcite shown in fig .
[ lagtime_analysis ] . for equilibrium systems
the eigenvalues @xmath91 of the transition matrix @xmath87 can be converted into relaxation timescales of the system @xmath92 with slowest modes corresponding to the largest @xmath91 .
if the largest eigenvalues are time - independent , also the slowest modes are time - independent yielding a good estimation of the minimal lag time that can be used . in ne - msm
the eigenvalues can become complex , which leaves their physical interpretation still open for discussion .
however , for the polymer dynamics the largest eigenvalues sorted by either their real part or absolute value exhibit no imaginary part . for a continuous - time markov process with rate matrix @xmath93 , the transition probability can be expressed as @xmath94 .
while conventional markov state models have discrete - time dynamics , for our purposes we require a rate matrix , which we approximate from the computed transition matrix through @xmath95 all off - diagonal elements of @xmath93 are non - negative @xmath96 while the diagonal elements are @xmath97 . following the perron - frobenius theorem , @xmath87 and thus @xmath93
have an unique largest real eigenvalue @xmath98 with a corresponding eigenvector that has strictly positive entries .
this eigenvector represents the steady - state probability distribution with elements @xmath99 .
finally , the probability fluxes are obtained by @xmath100 .
here we describe an algorithm that decomposes a given probability flux matrix @xmath101 into cycle fluxes .
each element of @xmath101 transforms through @xmath102 with @xmath103 .
the sum adds the cycle weights of all cycles containing the edge @xmath39 . for the cycle decomposition
we start by dividing @xmath101 into a symmetric detailed - balance part @xmath104 and a non - negative current part @xmath105 so that @xmath106 here @xmath105 is obtained by @xmath107 with all negative elements set to zero .
the symmetric part follows as @xmath108 . to make progress , we identify all non - zero elements of @xmath104 yielding a set of trivial cycles , i.e. cycles with only two different states ( @xmath109 ) .
their cycle weights are identical to their corresponding non - zero entry @xmath110 .
the algorithm to decompose the current part @xmath105 is split into two parts .
first , searching for a non - trivial cycles ( cycle with more than 2 different states ) and second determining its cycle weights .
theoretically , the number of possible non - trivial cycles grows exponentially with the number of non - zero entries of @xmath105 .
however , if both steps run alternately the decomposition becomes computationally affordable even for a large number of states . to detect a non - trivial cycle
, we propose the following steps : 1 .
find the position of the largest element of @xmath105 , @xmath111 .
search for the shortest path ( smallest number of transitions ) from state @xmath68 leading back to state @xmath32 ( only following the non - zero transitions ) .
this step can be efficiently achieved by applying a breadth - fist / depth - first search @xcite .
3 . return the non - trivial cycle , i.e. ( @xmath112found path ) . to determine the corresponding cycle weight @xmath113 ,
we take all flux values along cycle @xmath114 and determine their smallest value becoming the cycle weight @xmath115 summing up both steps , the final algorithm reads 1 .
find a non - trivial cycle 2 .
compute its cycle weight @xmath113 3 . update the current matrix by subtracting @xmath113 along @xmath114 , @xmath116 and
repeat with the first step 4 .
the algorithm stops when the residuum @xmath117 has become smaller than a threshold .
general considerations @xcite show that the maximum number of needed non - trivial cycles is bounded by @xmath118 with @xmath119 being the number of non - zero elements of @xmath105 and @xmath120 its rank .
once the cycle communities are found , the idea for coarse - graining the msm is to pick one cycle for each community we refer to it as representative
, delete all states not belonging to any representative and , finally , rescale the remaining transition rates @xmath121 . to be thermodynamically consistent the newly computed transition rates have to preserve four quantities : the total entropy production rate @xmath80 , the cycle affinities of the representatives @xmath74 , all remaining edge affinities @xmath75 and the dissipated heat along the remaining edges @xmath122 .
for a detailed description of the coarse - graining algorithm we refer the reader to our previous publication @xcite . since we know
now how to coarse - grain a given set of cycle representatives , we address the question of how to select `` appropriate '' representatives .
any set of cycle representatives is thought to be appropriate if the graph spanned by their coarse - grained transition matrix is ergodic and the mean first passage times ( mfpt ) between local communities are preserved .
for example , the polymer dynamics for @xmath71 communities as illustrated in fig .
[ fig : cyc]c , d exhibits 2 local and 1 global community . to compute the mfpts we identify all states of the red community ,
say as set @xmath123 , and all states belonging to the blue one as set @xmath124 .
any appropriate set of representatives needs to preserve mfpt@xmath125 and mfpt@xmath126 .
especially , the conservation of mfpts is of particular importance as it ensures the coarse - grained msm to express the correct timescales .
so far no determinstic algorithm exists that returns a reliable set of representatives . for this reason we formulate a stochastic algorithm that picks candidates for cycle representatives randomly and checks for ergodicity and mfpts .
the algorithm is outlined as follows : 1 .
choose one representative per cycle community by drawing a random number .
2 . check if set of representatives span ergodic transition network . if yes , compute coarse - grained msm , else go back to step ( 1 ) .
3 . compute mfpts of the coarse - grained msm and compare to mfpts of full msm .
if mfpts match , return coarse - grained msm , else go back to step ( 1 ) .
are modified . the arrows point in the direction of probability currents . ] to coarse - grain bridge states ( state 1 in fig .
[ bridgestate]a ) we form a new connection between the two neighboring states ( state 0 and 2 in fig . [
bridgestate]a ) .
the new transition rates @xmath127 and @xmath128 have to preserve three characteristics : ( i ) edge affinities @xmath129 , ( ii ) probability currents @xmath130 and ( iii ) dissipated heat @xmath131 . using conditions ( i ) and ( ii ) ,
the new probability fluxes follow as @xmath132 to fulfill condition ( iii ) we demand that the ratio between any two probabilities of the full network are preserved @xmath133 , which includes @xmath134 , and thus @xmath135 condition ( iii ) is also the main difference between our approach and the one discussed in ref . . in our adaptation
the probability distribution of the complete network is changed , while altaner _
_ change it only locally ( @xmath136 and @xmath137 ) and hence absorb @xmath138 into @xmath136 and @xmath137 .
the disadvantage of the latter is that condition ( iii ) is not preserved for the full network and , when using the coarse - graining approach iteratively , accumulation of probability in single states might occur , which leads to unphysical results . for the coarse - graining of triangle states ( state 1 in fig .
[ bridgestate]b ) we consider all cycles that contain the edges @xmath139 .
we modify these cycles by replacing the edges @xmath139 with a new edge @xmath140
. to be thermodynamically consistent the modified cycles have to ( i ) preserve the cycle entropy production rate @xmath141 , ( ii ) preserve the edge affinities @xmath142 and ( iii ) dissipated heat @xmath143 . to rescale the transition rates we use the same rescaling algorithm as for the rescaling of the cycle representatives .
note that through restriction ( i ) it is not necessarily possible to coarse - grain all found triangle structures .
assume , for instance , that the modified cycle coincides with an already existing cycle , then , the rescaling is not unique anymore and entropy production is destroyed as only one out of two cycles survives .
42ifxundefined [ 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 10.1038/nsb0902 - 646 [ * * , ( ) ] link:\doibase 10.1146/annurev - physchem-032210 - 103335 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1021/ma8018624 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.116.058302 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1016/j.sbi.2010.10.006 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1038/nature02261 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1122/1.1835336 [ * * , ( ) ] link:\doibase 10.1126/science.283.5408.1724 [ * * , ( ) ] link:\doibase 10.1063/1.1681018 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.84.4769 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1063/1.3149860 [ * * , ( ) ] @noop * * , ( ) @noop _ _ , ( , , ) @noop ( ) link:\doibase 10.1088/1742 - 5468/2010/05/p05015 [ ( ) ] @noop * * , ( ) \doibase doi:10.1088/0034 - 4885/75/12/126001 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevlett.97.128301 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physreve.85.041133 [ * * , ( ) ] @noop _ _ , vol .
( , ) @noop * * , ( ) link:\doibase 10.3934/jcd.2015.2.1 [ * * , ( ) ] @noop _ _ ( , ) @noop @noop * * , ( ) link:\doibase 10.1063/1.4818538 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrevlett.114.158101 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.116.120601 [ * * , ( ) ] link:\doibase 10.1073/pnas.1524720113 [ * * , ( ) ] @noop * * , ( ) @noop _ _ ( , ) | we describe a systematic approach to construct coarse - grained markov state models from molecular dynamics data of systems driven into a non - equilibrium steady state .
we apply this method to study the globule - stretch transition of a single tethered model polymer in shear flow .
the folding / unfolding rates of the coarse - grained model agree with the original detailed model .
we demonstrate that the folding / unfolding proceeds through the same narrow region of configuration space but along different cycles . |
since the discovery of ( 1 ) ceres in 1801 @xcite the majority of observations of asteroids and other minor planets have been conducted in visible wavelengths . while visible light can provide very accurate positions , the interpretation of photometry is rendered ambiguous by the dependence of the observed flux on both the size and albedo of the asteroid .
this relationship is described by the equation : @xmath1 ( see * ? ? ? * for an overview and references therein for its derivation ) , where @xmath2 is the diameter in kilometers , @xmath3 is the visible geometric albedo , and @xmath4 is the absolute magnitude which is defined as the apparent magnitude the body would have @xmath5au from the sun , @xmath5au from the observer , and at @xmath6 phase angle .
albedos of solar system objects are observed to vary from only a few percent up to nearly @xmath7 percent for icy surfaces , causing diameters inferred from visible data alone to have nearly an order of magnitude uncertainty .
multiwavelength visible surveys have found statistical correlations between the albedos and the visible colors of asteroids ( e.g. * ? ? ?
* ) , however the accuracy of the mapping between these two properties for individual objects is currently being examined @xcite . using an independent method of measurement for either the albedo or the diameter allows for the unique solution of the other , given a visible @xmath4 value .
accurate diameters and albedos for a large number of asteroids enable a number of important areas of research into the history and formation of the solar system .
diameter measurements of asteroids based on infrared flux allow us to quantify the size - frequency distribution ( sfd ) of the bodies in a way independent of assumptions about the translation from h magnitude to diameter that are typically required .
both the initial accretion and formation process of asteroids as well as the subsequent collisional and orbital evolution affect the current main belt sfd , and a well - measured sfd will allow us to put constraints on these processes .
the albedo of an asteroid , meanwhile , is a strong function of its composition .
compositional gradients for the main belt have been shown in the past from both infrared and spectral surveys ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , but only for a limited number of objects .
a large survey conducted in mid - infrared wavelengths will allow main belt albedos and diameters to be produced with good accuracy ; this in turn will allow us to study the compositional gradient of the solar system and may ultimately allow us to set constraints on any major planetary migration that may have occurred .
recent work has attempted to understand the sfd of asteroids as it relates to the impact physics dominating the evolution of the main belt with the aid of numerical simulations .
@xcite analytically calculated the behavior of a population in a steady - state collisionally dominated regime to determine how the slope of a population s sfd behaved as a function of the mechanical strength of the material composing the body .
they find that while a strength - less regime yields a nominal power - law slope of @xmath8 , including the strength of the body can vary the result over a range of slopes depending on the specific circumstances .
expanding on this , @xcite simulated a variety of impacts with numerical hydrocodes to look for the effect of internal structure , impact angle , impact velocity , and impactor - target size ratio on the resultant sfd of the shattered products .
these authors have also attempted to match their numerical results to measured sfds for asteroid families to determine the initial size of the parent body , however the data used have depended on assumptions for the albedos of the asteroids .
for the larger main belt population , @xcite have modeled the evolution of various initial size distributions to determine what the current mba sfd can indicate about the sfd present at the formation of the asteroids .
their results show a clear peak in formation sfd , with few objects smaller than @xmath9 km forming directly from the protosolar nebula , however these results are also based on diameters determined from assumed albedos .
similarly , @xcite have investigated the sources and sinks of excited mbas and the effect orbital evolution has the sfd of the main belt .
studies of asteroid albedo distributions have in the past been limited by small data sets .
early measurements were made using thermal infrared detectors on ground - based telescopes to determine radiometric diameters and thus albedos for tens of objects ( e.g. * ? ? ? * ) .
the _ infrared astronomical satellite _ ( iras ) revolutionized studies of physical properties of the main belt by observing over @xmath10 asteroids in the thermal infrared , determining albedos and diameters for these bodies in a uniform way @xcite .
to date , iras represents the largest and most complete survey of asteroid albedos in the literature .
if the sizes and albedos of the members of an asteroid family are known , it is possible to use their orbital evolution to study their age . the yarkovsky effect ( see * ? ? ?
* for a review of the subject ) occurs when incident optical light is absorbed and re - emitted as thermal infrared photons in a different direction , usually due to the rotation of the body .
this difference in momentum causes changes in the orbit of the body over long time intervals . @xcite
use this effect to refine predictions for the age of the karin family based on backwards integration of orbital parameters . as karin
is one of the youngest families @xcite this effect becomes increasingly important for accurate age - dating of families the older they are .
the accuracy of this technique , however , depends on knowing sizes and albedos of the objects to a high degree of certainty .
the distribution of albedos in the main belt as a whole and broken down by region can provide us a window into the changing chemical and mineralogical processes active in different regions of the early solar system .
although the current understanding of the history of the main belt has the asteroids forming in or near their current locations , new theories are being proposed that the main belt may in fact be the result of a mixing of two distinct populations from different regions of the solar system .
migrations of the giant planets may have both cleared many of the objects that initially formed in the main belt region and repopulated this area with objects from beyond the `` snow line '' @xcite .
we then might expect the main belt to be composed of two overlapping populations , one having formed in a volatile - poor region of the protosolar disk and one forming in a volatile rich area , though the latter population would lose any surface volatiles over the age of the solar system . under this scenario albedo
may be able to trace dynamical evolution as well as chemical processing .
there are very few ways to measure the albedo of an asteroid directly . observations taken _ in situ _ by spacecraft can be used to measure both the absolute albedo and its variation across a body s surface ( e.g. * ? ? ?
* ) , as well as its diameter , though only a handful of objects have been visited by spacecraft .
imaging polarimetry can also be used to determine the integrated surface albedo for a number of objects @xcite , but appropriately calibrated instrumentation is uncommon .
similarly , direct measurements of asteroid diameters can come from a variety of techniques .
resolved imaging of an asteroid provides the simplest method of size determination , however the use of e.g. _ hst _ or keck adaptive optics ( ao ) only allows the largest few asteroids to be resolved ( e.g. * ? ? ?
* ; * ? ? ?
asteroid occultations of background stars also provide a robust method of diameter measurement @xcite , however the logistical constraints of occultation events make obtaining a large sample difficult , and shape measurements are only instantaneous projections .
radar measurements can provide precise distances , rotation rates , projected profiles , and with sufficient data 3d shape models @xcite , but returned fluxes fall off quickly with distance which limits the number of objects observable with this technique .
indirect measurements of asteroid diameters and albedos can provide a wealth of information both for individual objects and the population as a whole . these techniques
typically can observe a large number of objects in a relatively short period of time , provide uniform data for an entire population , and have understood biases that allow for determination of the true , underlying distributions .
careful calibration of indirect measurements is required , but once established these techniques can provide highly accurate measurements of asteroid physical properties for a large number of bodies . for objects that have known orbits , measurement of the infrared flux emitted from the surface can be used to constrain the diameter of the body ( see [ sec.models ] for a discussion of the method used here ) .
infrared imaging can be accomplished rapidly when integrated in an all - sky survey .
thermal infrared measurements of a large sample of asteroids represent the best way to determine robust diameters rapidly for many thousands of objects . in this paper
, we present preliminary results from the _ wide - field infrared survey explorer _ ( wise ) space telescope , the next - generation all - sky infrared survey , focusing here on the cryogenic observations of main belt asteroids .
launched on 2009 december 14 , wise is a thermal infrared space telescope that performed an all - sky survey from 2010 january 14 until it exhausted the telescope - cooling hydrogen ice on 2010 august 5 .
the survey continued during the warmup of this secondary tank with limited sensitivity in the longer wavelengths until the primary coolant , responsible for maintaining detector temperature , exhausted 2010 october 1 .
wise subsequently entered a post - cryogenic mission to complete the survey of the largest mbas , continue discovering new neos , and complete a second - pass survey of the inertial sky in the two shortest wavelengths . during its cryogenic mission , wise imaged the sky in four infrared wavelengths simultaneously using dichroic beamsplitters to produce co - boresight images with band centers at @xmath11 m , @xmath12 m , @xmath13 m and @xmath14 m ( w1 , w2 , w3 , and w4 , respectively ) .
the latter two bands are particularly important for solar system studies as the dominant amount of the flux received from asteroids is from thermal emission peaking at these wavelengths .
first - pass calibration of the wise data was tuned to the fully cryogenic mission and while the final calibration currently being undertaken will finalize measurements obtained during each warm - up stage , we restrict our current analysis to objects observed during this fully cryogenic stage . pre - launch descriptions of wise were given by @xcite and @xcite , while post - launch overviews , including initial calibrations and color corrections , are presented by @xcite and @xcite .
the wise survey follows a continuous scan along lines of ecliptic longitude at a solar elongation of @xmath15 as the spacecraft orbits above the terminator of the earth .
an oscillating scan mirror compensates for this motion , providing stable images with effective integration times of @xmath16seconds .
the positions of all known minor planets are propagated to the time of each observation and checked against all transient sources in the field of view , recording the appropriate calibrated magnitudes if observed . in this way ,
thermal measurements of each minor planet cataloged by the minor planet center ( mpc ) can be found .
in addition to previously known objects , the wise processing pipeline includes the capability to discover new objects via the neowise enhancement @xcite .
while comprehensive followup of new potential near - earth objects ( neos ) has been one of the priorities of the wise team @xcite , the large number of main belt discoveries ( @xmath17 ) has prohibited sufficient immediate ground - based followup .
large - scale asteroid surveys such as spacewatch and the catalina sky survey have already provided serendipitous followup of many of the new wise mbas and will continue to do so , while future surveys are expected to recover the remainder . as described in @xcite and @xcite ,
the wise cadence resulted in an average of 10 - 12 observations of each minor planet detected over @xmath18hours . for a subset of the main belt ,
predominantly in the outer regions , observations were obtained at two or more epochs depending on the relative motion of the object and earth around the sun . in this paper , we consider only those main belt asteroids detected by neowise / wmops during the cryogenic portion of the mission , shown in a top - down view of the inner solar system in fig [ fig.pacman ] .
this consists of @xmath19 unique objects .
we obtained our data used for fitting in a method identical to the one described in @xcite and @xcite , though tuned for mbas .
specifically , we queried the minor planet center observation file for all observations submitted from wise ( observatory code c51 ) .
we then used the resultant ra - dec - time values as input for a query of the wise individual exposure archive , the `` level 1b '' data , through the gator tool provided by the infrared science archive ( irsa ) . to ensure that only the observations of the moving objects were returned , we restricted our search radius to 0.3 arcsecs from the position and 2 seconds from the time obtained for the detection from the mpc . additionally , we set a constraint of jd@xmath20 to ensure that only data during the fully cryogenic mission was used for this initial survey .
this method of data acquisition has the benefit of ensuring all observations used have been vetted both internally by the wise data pipeline and again by the mpc .
all data were processed using the first pass version of the wise pipeline , which computed dark current / sky offset levels , flagged instrumental artifacts such as latent images and diffraction spikes and performed linearity compensation .
only observations with an artifact flag cc_flags@xmath21 or @xmath22 in a band were accepted : a value of @xmath23 indicates no evidence of artifact was found by the pipeline , while @xmath22 indicates the possibility of contamination by a latent image .
as discussed in @xcite we find that the pipeline was overly conservative in artifact flagging and cc_flags@xmath24 values have similar fluxes to cc_flags@xmath21 detections while increasing the number of usable observations by @xmath25 .
the ph_qual flag was required to have a value of a , b or c to again ensure only valid detections were used .
non - linearity and saturation are a particular concern for the brightest mbas , especially in bands w3 and w4 .
the wise data reduction pipeline applies a non - linearity and saturation correction for all observations brighter than the threshold of @xmath26 mag , @xmath27 mag , @xmath28 mag , @xmath29 mag .
objects with magnitudes brighter than @xmath30 and @xmath31 were assigned errors of @xmath32mag to account for the change in the point - spread function for very bright objects , and a linear correction to the magnitudes of sources with @xmath33 was applied @xcite . following those authors , we did not use objects brighter than @xmath34 and @xmath35 for thermal modeling .
each object was required to have been observed at least 3 times in one wise band with magnitude error @xmath36 to undergo thermal modeling , as a precaution against contamination by spurious sources ( e.g. background noise , cosmic rays , stars , etc . ) . for multiple - band thermal models we required other bands to have at least @xmath37 of the detection rate of the band with the largest number of detections , usually w3 for mbas . to reduce the possibility of confusion with inertially fixed sources such as stars and galaxies
, we searched each position retrieved from the level 1b catalog in the daily and atlas coadded catalogs ( also served by irsa ) within 6.5 arcsecs , equivalent to the w1 , w2 and w3 beam sizes .
these searches looked for sources that appeared at least twice and in at least @xmath38 of the images covering that location .
any sources returned from these searches were considered to be inertial , which could contaminate the observation of the asteroid at that position .
thus that asteroid detection was discarded from the thermal modeling routine . in fig [ fig.colors ] we show the mean colors of inner solar system objects as observed during the cryogenic phase of the wise mission with the mbas highlighted in black ( see * ? ? ?
* ; * ? ? ?
* for discussions on the other populations shown in this figure ) .
the bifurcation in the w1-w2 color observed for the mbas traces the two dominant albedo groupings in the main belt ( see below for further discussion ) .
the mbas span a wide range of colors and are bounded by the neo and jupiter trojan populations .
note that mbas occupying color - space typically associated with one of the other populations are candidates for objects that may have been misidentified as mbas during preliminary orbit fitting .
we show in fig [ fig.colordist ] the color for each object as a function of heliocentric distance at the time of observation .
color , especially in bands dominated by thermal emission , is a strong function of temperature of the body and thus distance from the sun .
figure [ fig.velocity ] shows the sky - plane velocity for each object compared with its w3-w4 color .
the mean sky - plane velocity for mbas is @xmath32deg / day ; objects with velocities significantly larger than this are candidates for neos misclassified as mbas , requiring further followup .
future work will combine color and sky - plane velocity cuts to distinguish mbas with poorly known orbits from neos and trojans .
in contrast to the visible flux received from an asteroid , which is reflected sunlight , the mid - ir flux beyond @xmath39 m from an object in the main belt is almost completely thermal emission from that body . for an object with an established orbit , the phase angle and distances to the asteroid from the earth and the sun
are well known , thus the observed thermal flux can be converted into a total emitted flux at the surface . by making assumptions for some physical surface properties , a diameter can be computed from a single band detection . when simultaneous thermal measurements at multiple wavelengths are available the beaming parameter of the surface material may also be fit to the data .
the beaming parameter ( @xmath40 ) represents the deviation of the thermal emission from that of a smooth non - rotating sphere due to rotation and surface roughness , and is used to consolidate the uncertainty in the values of the surface thermal properties , including emissivity .
when visible - light data are additionally available , the visible albedo ( @xmath3 ) becomes a free parameter that can now also be fit using the two data sets in conjunction .
further data such as w1 or w2 measurements that are dominated by reflected sunlight allow us to derive an independent measure of the ratio of visible to nir reflectance as well . in order to analyze the thermal infrared asteroid measurements from the iras satellite ,
@xcite developed the `` standard thermal model '' ( stm ) for asteroids , calibrated against measurements from ( 1 ) ceres and ( 2 ) pallas . in this model ,
the beaming parameter was held constant to @xmath41 based on the ground - truth occultation observations of the calibrator asteroids .
subsequent work indicated that this model might not be appropriate for smaller asteroids , thus @xcite modified stm to a form appropriate for a `` near - earth asteroid thermal model '' ( neatm ) , where @xmath40 is allowed to vary . while designed specifically to account for the breakdown of stm when considering neos , neatm can be applied quite readily to a wide range of bodies in the solar system ( e.g. * ? ? ? * ; * ? ? ? * ) .
@xcite compared neatm with a full thermophysical model of a cratered surface and found that for low phases both models produce consistent results .
@xcite present an investigation of objects with a beaming parameter at the theoretical maximum of @xmath42 which occurs for a body rotating so quickly it is latitudinally isothermal .
we have performed preliminary thermal modeling of mbas based on the wise first - pass data processing pipeline described above , covering observations taken during the cryogenic phase of the mission .
we modeled each object as a non - rotating sphere with triangular facets and variable diameter , beaming parameter , visible albedo , and nir reflectance ratio as appropriate for the data .
relative distances and phase angles were computed for each measurement to ensure changing distances do not bias the resultant fits .
although we do not expect all , or even most , asteroids to have a spherical shape , our observations covering @xmath18hours smooth out rotation effects and allow us to determine the effective diameter of a spherical body with the same physical properties .
long period rotators ( @xmath43days ) with large amplitudes , e.g. binary asteroids with mass fraction @xmath44 , will have poor fits resulting in a moderate mis - measurement of albedo and diameter .
future work will address the light curve component of our data set to determine the minimum and maximum sizes of our targets in order to estimate first - order shape models as well as estimates of the fraction of binaries in the main belt . at each instance , the temperature on every facet was computed and color corrected based on the values in @xcite .
the emitted thermal flux for each facet was computed with the neatm model and night - side facets were assumed to contribute zero flux .
a reflected light model was used to determine the reflected component in each band for all illuminated facets visible to wise at the time of observation ; this was most important for w1 and w2 .
the model reflected and emitted light was summed for all facets and converted to a model magnitude using the jansky flux of a zero magnitude source provided in @xcite and modified according to the text for red sources : @xmath45jy for w1 , @xmath46jy for w2 , @xmath47jy for w3 and @xmath48jy for w4 .
note that the modifications in the w3 and w4 zeropoints are the result of adjusting the central wavelengths of these bands to @xmath49 m and @xmath50 m to correct the discrepancy observed in the calibration tests between blue - spectrum and red - spectrum objects .
these model magnitudes were then compared with the measured magnitudes , and the model was iterated through a least - squares fitting routine until the best fit was found . for objects with two thermally dominated bands ,
the beaming parameter was allowed to vary , while for those with only one thermal band we used a fixed value of @xmath51 , based on the peak of the @xmath40 distribution of mbas that were fit with a variable beaming parameter ( see below ) . for objects where no detected band was dominated ( @xmath52 ) by reflected light
we assumed a nir reflectance ratio of @xmath53 as was found for mbas with fitted nir reflectances ( see below ) .
these objects typically were not detected in w1 and either were not detected or had both thermal emission and reflected light in w2 .
monte carlo simulations were performed in each case to determine the errors on all variable parameters . for high
s / n cases where the quoted error on the measured magnitude only represented the statistical error , we set a floor of @xmath54 , representing the absolute error on the photometry @xcite . for all objects we assumed that the emissivity @xmath55 and we assumed that the magnitude - phase slope parameter ( c.f . * ?
* ) was @xmath56 unless otherwise given by the mpc or in the lightcurve database ( lcdb , * ? ? ? * ) .
the quoted errors on the modeled parameters are equal to the weighted standard deviation of all monte carlo trial values . for objects with
fixed beaming parameter , an error of @xmath57 was assumed to allow for proper error determination of derived parameters based on the mean and standard deviation of all best - fitting beaming parameters for objects with fitted values ( see [ sec.etavar ] ) .
similarly , for objects with fixed ir reflectance ratios , we assume an error bar of @xmath58 based on the mean and standard deviation of objects with fitted ir reflectance ratios ( see [ sec.ratio ] ) .
we note that as the flux calibrations presented by @xcite set a limit on the computed diameter accuracy for sources in the wise data of @xmath59 .
this error implies a minimum fractional error for albedo of @xmath60 assuming a perfect h magnitude .
these values are in addition to any poissonian error inherent to the observations , though for most of the objects presented here the calibration errors dominate our solutions . in total @xmath19 main belt asteroids , selected from the cryogenic phase of the survey , had sufficient number and quality of detections to allow us to perform thermal modeling and determine their effective diameter . of these ,
@xmath61 objects had orbital arcs shorter than @xmath62 days ; as such their orbits have a larger uncertainty than the rest of the population , which corresponds to uncertainty in their geocentric and heliocentric distances , which will naturally increase the error on their calculated diameters .
additionally , @xmath63 objects also had available optical data allowing us to calculate albedo as well as diameter .
both of these latter two populations are changing continuously , as ground - based surveys submit serendipidous visible observations of neowise - discovered asteroids ( allowing us to fit albedos and allowing the mpc to fit better orbits ) , and as the mpc links wise observations with previous one - night stands and lost asteroids .
we provide a table of our best fits for main belt asteroids from the pass 1 processed cryogenic survey data online at : _
/ staff / bauer / neowise_pass1/_. this table contains : the mpc - packed format name of the object ; the @xmath4 and @xmath64 values used ; the diameter , albedo , beaming parameter , and infrared albedo as well as associated error bars ; the number of observations in each wise band that were used for fitting ; and the mean modified julian date of the observations . for objects observed at multiple epochs with fits consistent across all observations , each epoch
is presented as a separate row in the table . as discussed in
[ sec.unusual ] objects where multiple epochs were forced to fit to a single model because the separate fits were divergent are presented as a single row .
objects without optical data at the time of publication have `` nan '' ( `` not a number '' ) values for absolute magnitude and albedo .
similarly , objects for which an infrared albedo could not be fit nor had literature optical data that could be used with an assumed reflectance ratio to estimate an infrared albedo ( see [ sec.ratio ] ) have `` nan '' values in this entry .
using fluxes from the wise data , and a faceted neatm model , we are able to determine diameters for our observed objects . in fig [ fig.dcum ] we show the cumulative preliminary raw size - frequency distribution ( prsfd ) for the three major regions of the main belt : the inner - main belt ( imb , those objects with @xmath65au @xmath66au ) , the middle - main belt ( mmb , objects with @xmath67au @xmath68au ) , and the outer - main belt ( omb , objects with @xmath69au @xmath70au ) . in all cases ,
perihelion distance was required to be beyond the orbit of mars , @xmath71au . also plotted on the distribution histograms are 100-trial monte carlo ( mc ) simulations of the diameter distribution including the appropriate measured errors .
this mean mc distribution and associated error are shown as points ; the error bar sizes are smaller than the point size .
we see no significiant change between the mc distribution from the distribution of best - fit diameters .
we find the slope of the prsfd for smaller objects in all three subpopulations to be consistent with the @xmath72 value found by @xcite , however debiasing will be critical to determining the true value of this slope .
we observe a significant change in slope for the prsfd between @xmath73 km , consistent with the location of the `` kink '' seen in the debiased @xmath4 distribution by @xcite .
a debiasing campaign of the observed population currently underway will allow us to explore the true sfds of the populations and will be discussed in future work .
for objects with detections in at least two thermal bands , we allow the beaming parameter to vary during the thermal model fitting .
we are able to fit beaming parameters for @xmath74 mbas .
we find a wide range of best - fit beaming parameters between the theoretical limits of @xmath75 and @xmath76 , with a peak value of @xmath51 and standard deviation of @xmath57 . in fig [ fig.etascat ]
we show the beaming parameter found for all objects with fitted values as a function of a variety of orbital and physical parameters .
beaming parameter has a weak dependence on semi - major axis ( fig [ fig.etascat]a ) , an effect that is more pronounced for the relationship with phase angle ( fig [ fig.etascat]g ) .
note that due to the constraints of the pointing of wise over the course of the survey to solar elongations @xmath77 , phase angle , heliocentric distance and geocentric distance are strongly correlated , though ecliptic latitude of the observations as well small changes in the exact pointing over the survey weaken this relation .
@xcite has shown that for the phase angles we typically observe mbas at ( @xmath78 ) the differences in calculated diameter and beaming parameter between neatm and more realistic thermophysical models is minimal over a large range of observing geometries . from the raw distribution ,
the beaming parameter shows no dependence on size ( fig [ fig.etascat]b ) , eccentricity ( fig [ fig.etascat]c ) , inclination ( fig [ fig.etascat]d ) or absolute magnitude ( fig [ fig.etascat]e ) .
we find a best - fit linear relation to the running average of beaming parameter as a function of phase of : @xmath79 where @xmath40 is the beaming parameter and @xmath80 is the phase angle in degrees .
this is consistent with the results found in @xcite for the neos from wise , but differs significantly from the results of @xcite who found a best - fitting line of @xmath81 . for phase angles within the main belt ( @xmath82 ) the average beaming ranges from @xmath83
although the spread around this value is large
. thus @xmath84 for objects with only a single thermal band is a reasonable assumption for objects in the main belt .
debiasing will allow us to account for any detection - limit effects that may bias the selection of objects that have sufficient data for fitting of the beaming parameter .
note that while the running average over the beaming parameter shows a dependence on albedo ( fig [ fig.etascat]f ) and subsolar temperature ( fig [ fig.etascat]h ) , albedo determinations are very sensitive to survey biases , both from wise and optical followup , while subsolar temperature is a function of beaming parameter as per the equation from the definition of neatm in @xcite : @xmath85^{0.25 } \label{eq.tss}\ ] ] ( where @xmath86 is the subsolar temperature , @xmath87 is the bond albedo , @xmath88 the incident solar flux , @xmath40 the beaming parameter , @xmath89 the emissivity , and @xmath90 the stefan - boltzmann constant ) , and so this can not be used as an independent constraint .
object - wide bins stepped by @xmath91 objects .
the picket - fence effect in the absolute magnitude is an artifact of the @xmath92mag reported precision of @xmath4 for most objects . ]
we show in fig [ fig.etahist ] histograms of the beaming parameter distribution for the inner- , middle- , and outer - main belt populations . also shown as points are the mean distribution and error from a 100-trial monte carlo simulation of the distribution using the error bar on each fitted beaming parameter .
all populations show longer tails toward higher values of beaming parameter than toward lower , and have consistent shapes .
the shift in peak beaming parameter with phase can be seen as the change in distribution between populations with different average phase angles .
with the inclusion of visible data in our modeling , we determine albedos as well as diameters for the asteroids discussed here .
we use the published @xmath4 and @xmath64 values for all asteroids , available from the mpc . during the confirmation of the calibration of
wise for asteroids , @xcite investigated the need for an offset in h to account for systematic errors in h values , but found that no offset was required ( c.f .
* who found a @xmath32mag shift ) .
the @xmath4 magnitudes were assigned a random error of @xmath93mag .
we perform monte carlo simulations of our visible light measurements as well as of the thermal measurements to quantify the error on albedo , however in all cases the minimum error on albedo will be @xmath94 @xcite for objects with optical data and one good thermal band .
we have sufficient optical data to determine albedos for @xmath63 mbas . though our individual albedos have large error bars
, the population statistics can still provide us with a window into the state of the surface composition of the main belt . in fig [ fig.alball ]
we show the differential preliminary raw albedo distribution ( prad ) of all the inner- , middle- , and outer - main belt asteroids in our survey .
we then take the fitted albedos and their respective error bars and perform a 100-trial mc simulation of these values to find a mean distribution with errors , shown as points . in all cases , the peaks of the distributions broaden slightly in the mc simulation , which is expected . in log - albedo - space the differential distribution is well described by a bimodal gaussian distribution .
we show our best - fitting double - gaussian ( fitted to the mean distribution found through the mc simulations ) as the smooth dotted curve under each set of points .
the bimodality in albedo likely traces the difference between the two major branches of asteroid composition : the s - type asteroids with high albedos and the c - type asteroids with low albedos @xcite .
@xcite investigate the specific link between albedo and a variety of taxonomic classification systems .
as discussed above , however , the prad will naturally include the observational biases of the ground - based telescopes used to determine the optical magnitudes needed to find the visible albedo , favoring higher albedo asteroids and over - representing their contribution to the total population , particularly in the inner main belt .
debiasing , currently being undertaken , will allow us to quantify and remove this effect .
the mean value and width of the gaussian that best describes the dark peak of the prad for each population is consistent across populations , with mean albedo @xmath95 and a dispersion of @xmath96 .
note that the gaussian error bars on the @xmath97 value are in log space , and thus asymmetric in native units .
unlike the dark asteroids , the bright complex shows a distinct change in the mean value in the prad as one moves out in the main belt .
the mean albedo of the bright peak for the gaussian describing each population is : @xmath98 , @xmath99 , and @xmath100 , with widths of @xmath101 , @xmath102 , and @xmath103 .
the objects discovered by neowise that have not had optical followup will add a significant bias to the prad as these objects are most likely to be ones missed by optical surveys , i.e. lower albedo asteroids . during the cryogenic portion of the survey , neowise observed @xmath104 previously unknown main belt asteroids with data of sufficient quality for thermal modeling and with orbital arcs longer than @xmath105 day ( and
thus not considered `` one night stand '' observations by the mpc ) .
some @xmath106 additional asteroids were given temporary designations by the mpc but did not have sufficient arc length to calculate an orbit . while they do not have known orbits and thus ca nt be classified as members of the main belt or not , we can use this to set an upper limit on the number of lost mbas . as these objects
are linked to older precovery data or are serendipitously followed up our count of discovered objects with computed orbits will increase . out of these discovered objects
@xmath107 have optical photometric data as well as thermal infrared , allowing for albedo determination .
it should be noted that while many of the discovered objects have optical observations , there is a strong bias in favor of recovery of the highest albedo discoveries by subsequent ground based observations .
the visible light received from an object is directly proportional to the albedo of that object , while the thermal infrared flux is driven by the temperature of the surface , which is only weakly dependent on albedo as shown in eq [ eq.tss ] .
as such , optical surveys show a significant albedo bias towards brighter objects for both discovery and recovery observations , while the neowise infrared survey is more sensitive to detection of low albedo objects .
any survey will have inherent biases in the data set and it is necessary to account for them before the true albedo distribution can be determined for a population . ,
but for only the neowise - discovered mbas that also received optical photometric followup .
] we show in fig [ fig.discalb ] the albedo distribution of the neowise - discovered mbas with optical photometry for the inner- , middle- , and outer - belt subpopulations .
even though the optical followup will be biased toward favoring higher albedo objects , the neowise discoveries are dominated by low albedos , as these are the objects that were initially missed by the ground - based optical surveys .
all three distributions can be described by a single gaussian function with mean albedo between @xmath108 and @xmath109 , consistent with the values found for the dark complex in the whole population above .
we can use the albedos of the discovered objects and the roughly equal number of objects in each subgroup as an initial attempt to constrain the effect of these lost objects on the greater albedo distribution .
we show in fig [ fig.albwithlost ] a revised albedo distribution including this toy model for the albedos of the @xmath110 main belt asteroids without followup photometry , based on the albedo distribution of the neowise - discovered objects .
as expected the primary effect is to increase the relative abundance of dark objects in each region of the belt .
, but now including a simple model for the objects discovered by neowise that have no optical followup photometry . ]
we note that the more detailed way to properly account for the uncertainties introduced by objects without follow - up and objects for which reliable orbits could not be determined is through careful modeling of survey biases .
a debiased study of the main belt asteroids will be subject of future work , and will allow us to determine the true size and albedo distributions of these objects .
we also investigate the distribution of albedo as a function of orbital parameters , in particular semimajor axis ( @xmath111 ) , eccentricity ( @xmath112 ) and inclination ( @xmath113 ) .
figure [ fig.apcolor ] shows the distribution of albedos as a function of semimajor axis .
we have color - coded the points by albedo using a `` weather - map '' palette divided evenly in @xmath114 space , and use this same color code for all subsequent plots .
dark colors ( black , grey , dark blue , blue ) indicate objects in the low albedo complex , while brighter colors ( magenta , red , orange , yellow ) indicate members of the high albedo complex .
objects colored yellow have very high albedos and are concentrated in the hungaria region and inner - main belt near the orbit of ( 4 ) vesta .
these largest albedos found are likely artifacts of using @xmath56 to calculate the @xmath4 value : while the thermal models for the diameters of these objects show no errors , using the literature @xmath4 value forces anomalously high albedos .
we investigate the use of different @xmath4 and @xmath64 values for these highest albedo objects in a future publication .
space and are used to denote albedo in subsequent plots . ]
we show in figures [ fig.aip ] and [ fig.aep ] the plot of semimajor axis vs. inclination and eccentricity , respectively , using the same colors denoted above .
asteroid albedos are not homogeneously distributed throughout the main belt , but rather are clumped in a - e - i space , correlating with the positions of known asteroid families @xcite .
mbas have been previously shown to cluster in color - space @xcite and show similar reflectance spectra ( e.g. * ? ? ?
* ; * ? ? ?
indicative of the common origin of members of families , which are the result of a catastrophic breakup of a single parent body @xcite .
the clustering of albedos is a further confirmation of this origin .
a future paper in this series will investigate the use of albedo as an added criterion to the orbital parameters typically used to determine family membership . .
the reflectance spectrum for most asteroids ( but not all , e.g. b - types ) show an increasing value as the wavelength moves from the visible to near - infrared ( nir ) region of the spectrum ( e.g. * ? ? ?
* etc . ) .
if these trends continue into the w1 and w2 bandpasses we would expect the reflectance observed there to be higher than observed in the optical .
we assume in our thermal model that the reflectance in w1 and w2 is the same ; depending on the location and depth of absorption bands this may not be universally true , but this assumption provides a generic constraint from which we can identify interesting objects that do not follow this assumption .
reflectance depends on both the albedo and the @xmath64 slope parameter , both of which can not be assumed to be wavelength independent .
we do not have sufficient phase coverage to fit @xmath115 and thus disentangle its effect from that of @xmath116 , so we present only the nir reflectance ratio for objects in the main belt with sufficient signal in w1 and/or w2 to be able to fit this value , a total of @xmath117 objects .
we show in fig [ fig.ratioscat ] plots of the nir reflectance ratio compared to a range of physical and orbital parameters , as well as running averages for those distributions .
all objects were detected in thermal emission , however only objects with sufficient reflected light were able to provide fitted nir reflectance ratios .
as such objects with higher ir albedos will be more likely to have a fitted reflectance ratio ( this is comparable to the biases inherent in optical surveys ) .
evidence of this is apparent in fig [ fig.ratioscat]e , where the running average of the ratio climbs for objects with fainter @xmath4 values .
we note that as the visible albedo @xmath3 is intricately linked with the nir / vis ratio it can not be considered an independent variable . the structure seen in the running average in fig [
fig.ratioscat]f is expected to be heavily influenced by the debiasing of the main belt population currently being undertaken .
future work @xcite will explore the connection between taxonomic types derived from spectroscopy and photometry and the ir reflectance ratio found in the wise data .
object - wide bins stepped by @xmath118 objects .
the picket - fence effect in the absolute magnitude is an artifact of the @xmath92mag accuracy of @xmath4 for most objects . ] in fig [ fig.ratiodist ] we show the raw differential distribution of nir reflectance ratios for the imb , mmb , and omb asteroids with fitted ratios .
also shown are the mean distribution and associated errors derived from a 100-trial monte carlo simulation of all measured reflectance ratios .
all three populations show a peak between @xmath119ratio@xmath120 while the running average vs. semi - major axis varies across the main belt from @xmath121ratio@xmath122 .
this is consistent with the values found for the neos by @xcite . for objects without fitted nir reflectance ratios
, we use the mean of all fitted values of @xmath53 and an error bar based on the associated standard deviation of @xmath58 for modeling purposes .
asteroid families were first identified as groups of objects that clustered tightly in orbital element - space by @xcite nearly a century ago .
subsequent work has confirmed that families originate from the catastrophic breakup of a single parent asteroid after an impact ( see * ? ? ? * for a recent review of the current state of the field ) .
this single mineralogical origin causes families to cluster tightly not only when comparing orbital elements but also when investigating colors @xcite and reflectance spectra ( e.g. * ? ? ?
* ; * ? ? ?
* etc . ) .
the sfd of asteroid family members can also act as a tracer of the physical properties of the original parent body and can even be used to constrain the impact velocity and angle @xcite .
however , a major deficiency in models to date has been the lack of measured diameters for the family members , forcing these values to be assumed based on the apparent magnitude of the object .
albedo measurements of the largest bodies in a family are often available from the iras data set @xcite and can be used to assume an albedo for all family members , but this can add a significant and systematic error to the diameters used , especially in the cases where families may be mixed or where the largest body in a family may not be associated with the other members ( e.g. * ? ? ?
there are a number of methods that can be used to determine which asteroids are members of a given family .
the hierarchical clustering method ( hcm , * ? ? ?
* ) , a commonly used technique , takes the differences in velocities between the proper orbital elements of objects to reveal dynamical associations .
@xcite used this method to identify @xmath123 families out of @xmath124 mbas with low inclinations .
we use these @xmath123 families as the baseline for our analysis , selecting those objects that appear both in that list and in the wise observations . as all these objects were discovered by optical surveys , there will be an inherent bias in the albedos favoring brighter family members . future work will address this bias and explore the use of albedo in conjunction with dynamical orbital properties to identify members of asteroid families and to reject interloper objects . of the @xmath123 families identified by @xcite
we find that @xmath125 have more than @xmath126 members observed during the cryogenic wise mission . due to the limitations of proper orbital element calculations , high inclination objects
are not included in the astdys list of proper orbital elements @xcite .
as such , high inclination families are likewise not represented in the list of family members .
we have , however , included in our analysis the pallas family , identified by nesvorny ( private communication ) through the same methods as the @xmath123 published families .
we also include the asteroids located in the hungaria region ; while not canonically included in the list of dynamical families , recent work by @xcite and @xcite support the classification of objects near hungaria in orbital space as a true dynamical family .
these two added groups bring our total considered population up to @xmath127 families .
we show in fig [ fig.aipfam ] the proper inclination against the proper semimajor axis for all objects identified as a member of one of the families considered here .
note that for the hungaria and pallas families proper orbital elements were not available and so osculating elements were used .
similarly , we show in fig [ fig.aepfam ] the proper eccentricity against the proper semimajor axis . in both plots
, we use the same color scheme as shown in fig [ fig.apcolor ] .
it is quite apparent in these plots that families characteristically have uniform albedos , however there are notable exceptions . . ] . ]
we show in fig [ fig.dcumfam ] the cumulative prsfd for each of the families observed during the wise survey , as well as monte carlo simulations of all distributions with appropriate error bars . without debiasing ,
the prsfd can not be assumed to represent the true size distribution of the entire family population , as neither the neowise - inherent biases nor the biases in family selection have been accounted for .
approximately @xmath128 of these families show kinks at the large end of the distribution inconsistent with a simple power - law . as
small number statistics dominate families especially at the largest sizes , and because wise did not survey the entire main belt before the exhaustion of cryogen , precise debiasing is required to confidently measure the shape of the true sfd , especially at the largest sizes for each family . in fig [ fig.albfam ]
we show the normalized prad for each family , along with the monte carlo results for each distribution . as was evident in fig [ fig.aipfam ] and fig [ fig.aepfam ] most families
show a single - peaked albedo distribution , however as mentioned above debiasing will be critical to proper interpretation of these distributions . about @xmath129 of families
include a small population of objects with non - characteristic albedos ; these may be the result of an improper association of background objects into the family .
however , another @xmath0 of families show significant mixing between two albedo types that can not be solely due to the intrusion of a few background objects .
in particular the nysa - polana and tirela families show nearly parity between the low and high albedo objects in the preliminary raw distributions . among the families with sufficient data for study were the karin and koronis families .
the karin family is believed to be a very young family , with an age of @xmath130myr , that formed from the breakup of a member of the much older ( @xmath131gyr ) koronis family @xcite .
we fit a gaussian to the observed family albedo distributions .
we find that the karin family has a lower mean albedo from this fit ( @xmath132 ) than the koronis family ( @xmath133 ) where the error bars represent the width of the best - fit gaussian .
@xcite gives an overview of space weathering effects , a process that is generally thought to darken and redden surfaces of atmosphereless bodies in the solar system .
our result is in apparent contradiction with this hypothesis under the assumption that the compositions of both families are identical .
it is possible that variations in the composition , the presence of interlopers , or differentiation of the proto - koronis parent body could result in this observation .
this analysis will be strengthened by the identification of more neowise - observed family members ( there were @xmath134 identified members of the karin family and @xmath135 members of the koronis family in this data set ) and the acquisition of additional compositional information .
we also observed @xmath136 asteroids that were identified as members of the baptistina family . @xcite postulate that a fragment from the breakup of the baptistina family was the impactor responsible for the k / t mass extinction event .
however , these authors assumed an albedo for the family members of @xmath137 .
we find that the best fitting gaussian to the baptistina family member albedos has a mean of @xmath138 where the error bar indicates the width of the gaussian distribution .
the method of age determination used by @xcite depends on the albedo assumed , and the calculated age @xmath139 is proportional to albedo following @xmath140 . thus our measured albedo results in an age for the breakup approximately @xmath141 times younger than that found with the lower assumed albedo ( from @xmath142myr to @xmath143myr ) , reducing the likelihood that the baptistina breakup generated the k / t impactor ( c.f .
* ) . also of interest
are the families that show characteristic albedos distinct from the distribution observed in the main belt in fig [ fig.alball ] .
for example , the eos and aeolia families have characteristic albedos that fall in between the two peaks of the mba albedos .
standing out as an outlier is the hungaria family : while showing a fairly strong coherence in albedo within the family , that characteristic albedo is incredibly high ( see below ) .
future work will investigate whether variations in the @xmath64 slope parameter used to compute the @xmath4 absolute magnitude ( and thus the reflected - light albedo ) from the assumed value of @xmath56 typically used could account for the very large albedos calculated in these preliminary results .
but for the prad .
monte carlo simulations of the distributions are plotted as grey points with error bars .
the distributions have been normalized to unit area for easier comparison . ]
a small fraction of our objects had fits that did not conform to the general trends seen for the population as a whole .
these highly unusual fits may be caused by strange physical parameters , incorrect associated properties ( e.g. @xmath4 mag , orbital elements , etc ) , or a break - down of the neatm model . in any case , these objects warrant further inspection .
the most obvious candidates for this category of unusual fits are the asteroids with very high visible geometric albedos ( @xmath144 ) .
we find @xmath145 objects in our survey have computed albedos that fall into this range , mostly contained within the region of orbital element space occupied by the hungaria and vesta families ( out of @xmath146 objects found in the hungaria region and @xmath147 in the vesta family ) , implying a possible mineralogical origin . the asteroid ( 434 ) hungaria , the lowest numbered member of its namesake region , has a relatively large albedo of @xmath148 which can be explained by a composition dominated by the iron - poor mineral enstatite however some of the observed spectral features may require contamination from a darker , external source @xcite .
this may indicate that albedos can range larger than @xmath149 , but albedos significantly higher than this are likely suspect .
visual inspection of the thermal model fits of these objects shows that these high albedos are not due to a failure of the thermal model . for those objects where ir reflectance factor could also be fitted , we find values typically with reflectance ratio@xmath150 , while the distribution of beaming parameters for these objects is similar to that of the general main belt population .
if the @xmath4 or @xmath64 values for these objects were very far from the true values this could result in the unusual calculated albedos . by increasing the error bar we assume for the @xmath4 value and setting the ir reflectance ratio to a constant value of ratio@xmath151 ( the average for mbas ) we are able to find good fits of objects with reflected light in w1 and/or w2 .
under this assumption , these objects all return fits with albedos comparable to that of ( 434 ) hungaria .
however all these cases require the @xmath4 magnitudes to be @xmath152mag fainter than the values given by the minor planet center .
a mis - identification of @xmath4 could be a symptom of an improper assumption for the @xmath64 slope parameter .
in particular if @xmath64 should be much larger ( e.g. equal to or greater than the value for ( 44 ) nysa of @xmath153 this would result in an @xmath4 value that was too bright by @xmath154mag or more , which would thus give an albedo that was a factor of @xmath155 too large .
this can account for some , but not all , of our improbably high albedos .
large light curve amplitudes may also contribute to an absolute magnitude that is too bright .
future work will address these objects in detail .
in addition to the high albedo objects , we find a very small number of objects with beaming parameters at or close to the theoretical limit of @xmath42 .
while some of these fits can be rejected upon visual inspection e.g. because of a single bad point in w4 dragging the fit to higher beaming parameter , at least @xmath156 objects appear to be legitimate fits with beaming parameters indicative of latitudinally - isothermal surfaces , though none of the fits show any significant lightcurve variations
. further investigation will be critical to determine if they show rapid rotation or very high thermal inertia needed to explain this beaming parameter .
approximately @xmath157 mbas were observed by wise at two different epochs .
we initially treated each epoch separately for fitting purposes , and in the majority of cases the fits were within the expected error of each other .
for those that were nt , we recomputed the best fitting model using both epochs together and forcing the physical parameters to be identical . we find @xmath158 objects for which the two - epoch fits could not produce a good fit at one or both epochs .
while some of these objects may have very long rotation periods(@xmath159days ) and thus different projected diameters between epochs , others do not appear to show any light curve variation during our observations and may be cases of objects showing a significant difference between the temperatures of the morning and afternoon hemispheres of the body .
we also observe @xmath160 objects with peak - to - trough light curve variations larger than @xmath161mag in w3 and average magnitude measurement error smaller than @xmath162 after removing spurious measurements . as these are not fitted amplitudes
they represent a minimum for the light curve amplitude for the body observed .
an example of one of these objects , ( 61469 ) , is shown in fig [ fig.lc ] .
there are also many objects with amplitudes smaller than this cutoff with readily apparent rotational effects , too
. a future work will investigate specific light curves to determine the period and amplitude of the objects . ,
@xmath163 and @xmath164 shown in red , green and cyan ( top to bottom ) respectively .
the dashed line shows the modeled magnitudes for the best - fitting sphere .
the period of this asteroid is approximately @xmath165hr assuming a double - peaked lightcurve . ]
we have presented an initial analysis of main belt asteroids detected by neowise during the cryogenic portion of the wise mission . with infrared fluxes of sufficient quality to determine diameters and albedos for @xmath19 mbas , we show the power and great potential contained in this dataset .
these data allow us to probe the composition , structure , and history of the main belt in ways that were previously impossible . for objects with thermal emission detected in two or more bands we allowed the beaming parameter to vary .
we find a mean beaming parameter of @xmath166 however we do see evidence of a phase dependence for the beaming parameter , ranging from @xmath167 for low phases to @xmath168 for higher phases ( within the main belt ) .
the best fit linear relation between beaming and phase is @xmath169 which is a much shallower relation than seen previously in @xcite , but consistent with @xcite which included the near earth objects with fitted beaming parameters as well . as was observed in the iras data ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) the albedos of main belt asteroids are strongly bimodal : a bright complex ( @xmath170 ) and the dark complex ( @xmath171 ) .
we find both peaks to be well described by gaussian distributions in log - albedo space .
we find that the reflectance of asteroids in the w1 and w2 bandpasses is typically larger than the albedo found in visible light .
the best fit ratio of reflectance ranges from @xmath172 in the imb to @xmath173 in the omb however the spread of values is large , and the final distribution will depend strongly on the debiasing .
we identify albedo clusters in @xmath174 space corresponding to the locations of asteroid families .
albedo is another coherent property of dynamical families in addition to orbit @xcite , color @xcite and reflectance spectrum ( e.g. * ? ? ?
* ; * ? ? ?
* etc . ) .
albedo can also be used to trace the halos of similar objects that surround some families ( e.g. vesta , eos , etc .
* ) that may be evidence of a collisional breakup very early in the age of the solar system . using asteroids previously identified though hcm techniques to be members of collisional families
we show that most , but not all , families have a characteristic albedo .
critical to any interpretation of the observations presented here is a careful accounting of the biases in both the wise survey data as well as the optical data used to derive albedos .
we are currently undertaking an extensive debiasing campaign with the goal of producing unbiased size and albedo distributions for the main belt .
this will be the subject of the next paper in this series .
j.r.m . was supported by an appointment to the nasa postdoctoral program at jpl , administered by oak ridge associated universities through a contract with nasa .
thanks m. delb and m. mueller for providing access to their thermal modeling code which was helpful in early test cases . this publication makes use of data products from the wide - field infrared survey explorer , which is a joint project of the university of california , los angeles , and the jet propulsion laboratory / california institute of technology , funded by the national aeronautics and space administration .
this publication also makes use of data products from neowise , which is a project of the jet propulsion laboratory / california institute of technology , funded by the planetary science division of the national aeronautics and space administration .
this research has made use of the nasa / ipac infrared science archive , which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration .
we thank the worldwide community of dedicated amateur and professional astronomers devoted to minor planet follow - up observations .
we are deeply grateful for the outstanding contributions of all members of the wise and neowise teams .
cutri , r.m . ,
wright , e.l . , conrow , t. , bauer , j. , benford , d. , brandenburg , h , dailey , j. , eisenhardt , p.r.m .
et al . _ , 2011 , _ http://wise2.ipac.caltech.edu/docs/release/prelim/expsup/wise_prelrel_toc.html _
juri , m. , ivezi , z. , lupton , r.h . , quinn , t. , tabachnik , s. , fan , x. , gunn , j.e . ,
hennessy , g.s . ,
knapp , g.r .
, munn , j.a . , pier , j.r . ,
rockosi , c.m . ,
schneider , d.p .
, brinkmann , j. , csabai , i. & fukugita , m. , 2002 , aj , 124 , 1776 .
nesvorn , d. , bottke , w.f . , vokrouhlick , d. , morbidelli , a. & jedicke , r. , 2006a , asteroids , comets , meteors , proc . of the 229th symp . of the iau , daniela , l. , sylvio ferraz , m. , and angel , f.j .
( eds ) , cambridge university press , 289 .
ostro , s.j .
, hudson , r.s . , benner , l.a.m .
, giorgini , j.d . ,
magri , c. , margot , j.l .
& nolan , m.c . , 2002 , asteroids iii
, w. f. bottke jr .
, a. cellino , p. paolicchi , and r. p. binzel ( eds ) , university of arizona press , 151 .
wright , e.l . , eisenhardt , p. , mainzer , a.k . ,
ressler , m.e .
, cutri , r.m . ,
jarrett , t. , kirkpatrick , j.d .
, padgett , d. , _ et al .
_ , 2010 , aj , 140 , 1868 .
zappal , v. , cellino , a. , farinella , p. & knezevic , z. , 1990 , aj , 100 , 2030 . | we present initial results from the wide - field infrared survey explorer ( wise ) , a four - band all - sky thermal infrared survey that produces data well suited to measuring the physical properties of asteroids , and the neowise enhancement to the wise mission allowing for detailed study of solar system objects . using a neatm thermal model fitting routine we compute diameters for over 100,000 main belt asteroids from their ir thermal flux , with errors better than @xmath0 .
we then incorporate literature values of visible measurements ( in the form of the h absolute magnitude ) to determine albedos . using these data
we investigate the albedo and diameter distributions of the main belt .
as observed previously , we find a change in the average albedo when comparing the inner , middle , and outer portions of the main belt .
we also confirm that the albedo distribution of each region is strongly bimodal .
we observe groupings of objects with similar albedos in regions of the main belt associated with dynamical breakup families .
asteroid families typically show a characteristic albedo for all members , but there are notable exceptions to this .
this paper is the first look at the main belt asteroids in the wise data , and only represents the preliminary , observed raw size and albedo distributions for the populations considered .
these distributions are subject to survey biases inherent to the neowise dataset and can not yet be interpreted as describing the true populations ; the debiased size and albedo distributions will be the subject of the next paper in this series . |
together with gravity neutrinos drive the evolution of several stellar phenomena such as supernovae , binary mergers , and black hole accretion disks .
they are also a key ingredient in the synthesis of heavy elements @xcite , the production of gamma ray bursts @xcite and kilonova @xcite . despite the numerous efforts to fully understand these phenomena there are still puzzles to address .
unrevealing the missing pieces requires complex and computationally expensive simulations of the system under study that bring together details of nuclear physics and gravity .
furthermore , due to the complexity of the problem , it is necessary to post - process the results to shed light on other derived aspects such as the element abundances and neutrino detection . the strong gravitational field generated by a binary neutron star merger changes the spectral properties of the emerging neutrinos .
this will affect any physical quantity related to them .
for example , the emitted neutrinos can interact with matter outflowing the merger . via weak interactions neutrinos will change the neutron to proton ratio ( electron fraction ) of the outflow setting the path that a nuclear reaction chain follows .
however , strong gravity effects , such as redshifts and bending of neutrino trajectories , affect the neutrino fluxes altering the electron fraction . as a consequence the final abundances differ from those obtained in the absence of the gravitational field @xcite . on another hand ,
if the merger occurs in the milky way or in the local galaxy group chances are that we will able to detect those neutrinos with current and future facilities @xcite .
it is therefore interesting to study the behavior of neutrinos emitted from binary systems . in this proceedings we present post - processed neutrino emission results from state of the art 3d fully relativistic binary neutron star simulations with magnetic fields , neutrino cooling and different equations of state .
the original simulation results of the coalescence have been presented in reference @xcite . there , the evolution of the matter in connection to the equation of state ( eos ) is discussed , as well as the gravitational wave emission and neutrino detection rates . here ,
we expand our discussion of neutrino emission focusing on the evolution of the neutrino surface and its relation to the eos .
the eos connects the macroscopic thermodynamical variables to the microphysics of the neutron star .
observables such as maximum masses and radii carry important information about the behavior of nuclear matter at high densities . similarly , our theoretical models of nuclear interactions should correctly account for these observables . among other macroscopic variables ,
the eos predicts the temperature , density and matter composition .
the neutrino absorption and emission rates are correlated to these same quantities .
a microphysical eos is therefore fundamental to track the merger dynamics and the neutrino emission .
we consider three different eos ( tables publically available at www.stellarcollapse.org ) , nl3 , dd2 , sfho , all of them in the frame of the statistical model of hempel and schaffner - bielich @xcite .
they differ in the relativistic mean field model used to describe the nuclear interaction .
the nl3 eos is based on the interaction model of reference @xcite , the dd2 on @xcite and the sfho on @xcite .
these eos predict different radii and maximum masses .
the eos is said to be _ soft _ if results in a small radius for a given mass and is _ stiff _ if for the same mass the predicted radius is larger .
our simulations consider the coalescence of two neutron stars each with a mass of @xmath0 .
the sfho predicts for that mass a neutron star radius of @xmath1 12 km .
the dd2 predicts a close to @xmath2 km radius while for the nl3 the radius is @xmath3 km . in that order of ideas
the sfho is the softest of the eos considered , the dd2 is an intermediate eos and the nl3 is the stiffest one .
post - merger neutrino luminosities can be as large as @xmath4 erg / s .
this huge amount of neutrinos can reach detectors on earth and tell us valuable information about the merger and its evolution . at high matter densities ( @xmath5 g/@xmath6 )
neutrinos are trapped : the neutrino mean free path @xmath7 is shorter than the physical dimensions of the merger , i. e. matter is neutrino opaque .
once the density decreases the neutrino mean free path increases and neutrinos escape carrying energy away and contributing to the cooling of the system . the point of last scattering , the place where neutrinos are free , is known as the neutrino surface . a common approach to determine this surface is to find the places where the optical depth @xmath8 ( see e.g. @xcite ) . in terms of the opacity @xmath9 , and
the mean free path @xmath10 , the optical depth @xmath11 is given by @xmath12 along the propagation direction @xmath13 . in this post - processing we chose to study the neutrino emission as seen from an observer located at infinity in the @xmath14-axis .
then , the neutrino surfaces are defined by the two - dimensional hypersurfaces where the optical depth of the merger is perpendicular to the equatorial plane .
this turns the optical depth into @xmath15 we integrate equation [ opticaldepthz ] changing the lower limit @xmath16 until the value @xmath17 is reached .
the themodynamical conditions of the neutrino surface are therefore the conditions of the matter at @xmath16 . in equation [ opticaldepthz ]
the mean free path depends of the cross sections @xmath18 , @xmath19 the sum above goes over all the relevant scattering and absorption processes @xmath20 that neutrinos undergo as they diffuse through matter , with @xmath21 the number density of the target .
@xmath22 is the thermally averaged ( `` weighted '' ) cross section , @xmath23 where @xmath24 is the neutrinos fermi - dirac flux @xmath25 and @xmath26 is the neutrino occupation number .
we assume the temperature is equal to the local temperature and zero chemical potential .
this procedure removes the energy dependence of the neutrino surface .
note , that the scattering neutrino surfaces introduced above differ from the effective neutrino surfaces ( see e.g. @xcite ) .
the first ones correspond to the last scattering surfaces , where absorption and scattering neutrino processes are treated equally while in the calculation of the effective neutrino surfaces more emphasis is given to absorption processes .
the effective surfaces would be however diffuse in area but will give a better estimate of the neutrino temperature , while the scattering surfaces will give a better estimate of the effective area of decoupling .
as our calculation for average energies ( explained below ) is thermally weighted over the fermi - dirac distribution we expect that they will be good estimates . in this analysis
we consider neutrino scattering from protons , neutrons and electrons , as is a good approximation to assume that matter is dissociated at the typical values of @xmath27=10 mev ( and above ) found in simulations . in this way we have the charged current processes @xmath28 and
@xmath29 we also consider neutral current processes , elastic scattering from electrons and neutrino - antineutrino annihilation , which affect on equal footing all neutrino flavors .
we find proton and neutron number densities assuming charge neutrality @xmath30 , and the electron number density assuming equilibrium of thermal electrons and positrons with radiation .
details on the cross sections of the above reactions can be found in ref .
@xcite . as the heavy lepton neutrinos , tau and muon ,
do not undergo the process of equations [ nuabsorption ] and [ anuabsorption ] they are the first to decouple form matter . also as the medium is neutron - rich , due to the original neutron richness of the neutron stars , the last ones to decouple are electron neutrinos given the predominance of the reaction of equation [ nuabsorption ] .
using our results for the neutrino surfaces we can make estimates of the number of neutrinos leaving the merger per second @xmath31 and average energies @xmath32 .
for the first one we have @xmath33 and for the energy rate ( energy per sec ) @xmath34 in the equations above the integral over @xmath35 corresponds to an integral over the neutrino surface .
our estimate for the average neutrino energy is then given by @xmath36 in order to transform the average energy to another reference frame we make use of the fact that the quantity @xmath37 , with @xmath38 the specific intensity is an invariant , @xmath39 noting the quantities measured by an observer with `` tilde '' , so the observed energy is @xmath40 , and @xmath41 is the observed area differential , and without tilde analogous emitted quantities @xmath42 and @xmath35 , then we have @xmath43 which leads to @xmath44 we can rewrite this in terms of the emitted quantities given that @xmath45 , with @xmath46 the redshift ( in the schwarzchild metric for simplicity ) , assuming that the emission is isotropical @xmath47 , and that distances are stretched by a factor of @xmath48 , @xmath49 we then get @xmath50 , @xmath51 and the observed average neutrino energy is just the ratio of the above expressions , @xmath52 .
following the methodology of the previous section we find the neutrino surfaces for the three different neutrino flavors .
the electron antineutrino surface at @xmath53 ms after the merger for the dd2 eos is presented in figure [ eanudd2 ] . the height represents the distance , measured from the equatorial plane , at which neutrinos decouple from matter , whereas the color scale shows the matter temperature at these last points of neutrino scattering . with these temperatures we determine the average neutrino energies as described above .
figure [ eanudd2 ] shows the time evolution of the electron antineutrino surface for the dd2 eos as seen from the @xmath14-axis .
the time after the merger is as indicated in the figure .
although , electron antineutrinos achieve larger temperatures , around @xmath54 mev , at earlier times near the center for t=2.5 ms after the merger , larger areas are warmed up at latter times . the highest temperature for electron
antineutrino decreases @xmath55 12 mev at t= 7.4 ms , but matter is overall warmer leading to the larger neutrino luminosities as reported in @xcite .
for even latter times , t=12.3 and 17.2 ms , the flux of neutrinos has contributed to the matter cooling and the neutrinos temperatures decrease .
table [ energytable ] shows the deduced average neutrino energies for electron antineutrinos and electron neutrinos at the corresponding times after merger . in parenthesis
are indicated the energies at the emission point while out of the parenthesis are the observed quantities at infinity .
.observed(emitted ) neutrino average energies for electron neutrinos and electron antineutrinos for the dd2 eos . [ cols="<,^,^,^,^ " , ] [ eosenergytable ]
binary neutron star mergers represent a wonderful laboratory to contrast our theoretical predictions with observations . from one side
the occurrence of an event will produce a gravitational wave signal sufficiently strong to be detected at existent facilities such as advanced ligo . on the other hand
, we will be also able to detect neutrinos emitted during the merger with current and future water cherenkov detectors ( such as super kamiokande ) .
furthermore , we expect that from these mentioned signals we will gain information from the eos and consequently about the nuclear force . in this proceedings
we have discussed the influence that the eos has on the neutrino emission from binary neutron star mergers .
we have followed the evolution of the neutrino surface for the dd2 eos , and found that for earlier times the neutrino production is low due to the higher densities of the system . at later times ( about 2.5
ms after merger ) the neutrino production peaks as the matter distribution is more even and the temperatures increase . at this stage
the neutrino surface temperatures are high and neutrinos have an average energy at infinity around 18 mev .
finally , the emission decreases as the system losses energy via neutrinos .
this quasi - steady state last for some milliseconds .
we showed that the stiffness of the eos changes the neutrino energies .
a soft eos like the sfho will produce hotter and higher energy neutrinos compared to the stiffer nl3 eos .
for all cases the neutrino emission will last tens of milliseconds .
this together with the higher neutrino energies of the merger will allow us to distinguish this signal from a supernovae one which is estimated to be around 10 seconds .
fully relativistic hydrodynamical simulations with realistic eos represent a valuable method to understand the evolution of neutron star mergers .
together with an accurate description of neutrino fluxes and nuclear reaction networks that determine the elements produced during these events will shed light on our understanding of the production of heavy elements .
future efforts will be focused on this problem .
we thank our collaborators l. lehner , c. palenzuela , s. l. liebling , d. neilsen , e. oconnor , and m. anderson .
o. l. caballero , g. c. mclaughlin and r. surman apj , * 745 * , 170 ( 2012 ) shigehiro nagataki and kazunori kohri , prog .
phys . , * 108 * ( 2002)789 .
g. c. mclaughlin and r. surman , prd , * 75 * ( 2007 ) 023005 .
o. l. caballero , g. c. mclaughlin , & r. surman , phys .
d , * 80 * 123004 ( 2009 ) m. hempel and j. schaffner - bielich , nucl . phys .
a * 837 * , 210 ( 2010 ) doi:10.1016/j.nuclphysa.2010.02.010 [ arxiv:0911.4073 [ nucl - th ] ] .
g. a. lalazissis , j. konig and p. ring , phys .
c * 55 * , 540 ( 1997 ) doi:10.1103/physrevc.55.540 [ nucl - th/9607039 ] .
s. l. shapiro and s. a teukolsky , _ black holes , white dwarfs and neutron stars _ , wiley , new york , 1983 .
a. perego , s. rosswog , r. m. cabezn , o. korobkin , r. kppeli , a. arcones and m. liebendrfer , mon . not .
* 443 * , no . 4 , 3134 ( 2014 ) [ arxiv:1405.6730 [ astro-ph.he ] ] | neutron - star mergers are interesting for several reasons : they are proposed as the progenitors of short gamma - ray bursts , they have been speculated to be a site for the synthesis of heavy elements , and they emit gravitational waves possibly detectable at terrestrial facilities .
the understanding of the merger process , from the pre - merger stage to the final compact object - accreting system involves detailed knowledge of numerical relativity and nuclear physics . in particular , key ingredients for the evolution of the merger are neutrino physics and the matter equation of state .
we present some aspects of neutrino emission from binary neutron star mergers showing the impact that the equation of state has on neutrinos and discuss some spectral quantities relevant to their detection such as energies and luminosities far from the source . |
sparse matrices are particularly useful in applied and computational mathematics because of their low storage complexity and fast implementation as compared to dense matrices . of late
, significant progress has been made to incorporate sparse matrices in compressed sensing , with @xcite giving both theoretical performance guarantees and also exhibiting numerical results that shows sparse matrices coming from expander graphs can be as good sensing matrices as their dense counterparts .
in fact , blanchard and tanner @xcite recently demonstrated in a gpu implementation how well these type of matrices do compared to dense gaussian and discrete cosine transform matrices even with very small fixed number of nonzeros per column ( as considered here ) . in this manuscript
we consider random sparse matrices that are adjacency matrices of lossless expander graphs .
expander graphs are highly connected graphs with very sparse adjacency matrices , a precise definition of a lossless expander graph is given in definition [ def : expander ] .
[ def : expander ] @xmath0 is a lossless @xmath1-expander if it is a bipartite graph with @xmath2 left vertices , @xmath3 right vertices and has a regular left degree @xmath4 , such that any @xmath5 with @xmath6 has a set of neighbors @xmath7 with @xmath8 neighbors .
note that these graphs are _ lossless _
because @xmath9 , they are also referred to as _ unbalanced expanders _ in the literature because @xmath10 and a @xmath11-lossless expander graph has an _ expansion _ of @xmath12 .
such graphs have been well studied in theoretical computer science and mathematics and have many applications .
probabilistic constructions of such graphs using random left - regular bipartite graphs with optimal parameters exist but deterministic constructions only achieve sub - optimal parameters , see @xcite or @xcite for a more detailed survey . using a novel technique of _ dyadic splitting of sets _ , this work derives quantitative guarantees on the probabilistic construction of these graphs in the form of a bound on the tail probability of the size of the _ set of neighbors _ , @xmath13 for a given @xmath5 , of a randomly generated left - degree bipartite graph .
moreover , this tail bound proves a bound on the tail probability of the _ expansion _ of the graph , @xmath14 . in addition , we derive the first phase transitions showing regions in parameter space that depicting when a left - regular bipartite graph with a given set of parameters is guaranteed to be a lossless expander with high probability .
similar results in terms of the adjacency matrices of these graphs is also presented .
another contribution of this work is the derivation of sampling theorems comparing performance guarantees for some of the algorithms proposed for compressed sensing using such sparse matrices as well as the more traditional @xmath15 minimization compressed sensing formulation .
it also provides phase transitions of @xmath15 minimization performance guarantees for such sparse matrices compared to what @xmath16 restricted isometry constants ( @xmath17 ) analysis yields for gaussian matrices .
our main result is the presentation of formulae for the expected cardinality of the _ set of neighbors _ of @xmath11-lossless expander graphs and the sparse non - mean zero matrices from these graphs .
based on this , we present a tail bound on the probability that this cardinality will be less than the expected value .
we start by defining the class of matrices we consider and a key concept of a _ set of neighbors _ used in our derivation .
[ def:1set_expansion ] let @xmath18 be an @xmath19 matrix with @xmath4 nonzeros in each column .
we refer to @xmath18 as a random a ) sparse expander ( se ) if every nonzero has value @xmath20 and b ) sparse signed expander ( sse ) if every nonzero has value from @xmath21 .
the support set of the @xmath4 nonzeros per column of these matrices are drawn uniformly at random and independently for each column .
an se matrix is an adjacency matrix of @xmath11-lossless expander graph while an sse matrix have random sign patterns in the nonzeros of an adjacency matrix of a @xmath11-lossless expander graph .
if @xmath18 is either an se or sse it will have only @xmath4 nonzeros per column and since we fix @xmath22 , @xmath18 is therefore extremely sparse .
we formally define the _ set of neighbors _ in both graph theory and linear algebra notation to aid translation between the terminology of the two communities .
denote @xmath23 as a submatrix of @xmath18 composed of columns of @xmath18 indexed by the set @xmath24 with @xmath25 .
[ def : neighbours ] consider a bipartite graph @xmath26 where @xmath27 is the set of edges and @xmath28 is the edge that connects vertex @xmath29 to vertex @xmath30 .
for a given subset of left vertices @xmath31 its set of neighbors @xmath32 is defined as @xmath33 . in terms of the adjacency matrix , @xmath18 , of @xmath26
the set of neighbors of @xmath23 denoted by @xmath34 , is the set of rows with at least one nonzero .
henceforth , we will only use the linear algebra notation @xmath34 which is equivalent to @xmath35 . note that @xmath36 is a random variable depending on the draw of the set of columns , @xmath24 , for each fixed @xmath18 .
therefore , we can ask what is the probability that @xmath36 is not greater than @xmath37 , in particular where @xmath37 is smaller than the expected value of @xmath36 .
this is the question that theorem [ thm : prob_bound_1set_expansion ] attempts to answers .
[ thm : prob_bound_1set_expansion ] for fixed @xmath38 and @xmath39 , let an @xmath19 matrix , @xmath18 be drawn from either of the classes of matrices defined in definition [ def:1set_expansion ] , then @xmath40}\ ] ] where @xmath41 , and for random variables @xmath42 and @xmath43 , @xmath44 is given by @xmath45},\end{gathered}\ ] ] where @xmath46 is the shannon entropy function of base @xmath47 logarithm .
consequently : 1 .
if no restriction is imposed on @xmath37 then the @xmath48 for @xmath49 take on the expected values of @xmath36 , which are given by @xmath50 for @xmath51 ; 2 . else if @xmath52 is restricted to be less than @xmath53 , then the @xmath48 for @xmath49 are the unique solutions to the following polynomial system @xmath54 for @xmath55 with @xmath56 for each @xmath57 .
theorem [ thm : prob_bound_1set_expansion ] gives a bound on the probability that the cardinality of a union of @xmath58 sets each with @xmath4 elements is less than @xmath59 .
figure [ fig : cardinalities1 ] shows plots of values of @xmath60 ( size of set of neighbors ) for different @xmath58 taken over 500 realizations ( in blue ) , superimposed on these plots is the empirical mean values of @xmath60 over the 500 runs ( in red ) and the @xmath61 in green . and @xmath62 , over @xmath63 realizations , plots ( in blue ) of the cardinalities of the index sets of nonzeros in a given number of set sizes , @xmath58 .
the dotted red curve is mean of the simulations and the green squares are the @xmath61.,scaledwidth=40.0% ] furthermore , simulations illustrate that the @xmath61 are the expected values of the cardinalities of the union of @xmath58 sets , @xmath60 , as shown in figure [ fig : cardinalities2 ] , where we show the relative error between @xmath61 and the empirical mean values of the @xmath60 , denoted by @xmath64 , realized over @xmath63 runs , to be less than @xmath65 . and
@xmath62 , over @xmath63 realizations , plots of the relative error between the mean values of @xmath59 ( referred to as @xmath64 ) and the @xmath61.,scaledwidth=40.0% ]
we now use theorem [ thm : prob_bound_1set_expansion ] with the @xmath15-norm restricted isometry property ( rip-1 ) , introduced by berinde et .
al . in @xcite
, to deduce the corollaries that follow which are about the probabilistic construction of expander graphs , the matrices we consider , and sampling theorems of some selected compressed sensing algorithms .
firstly , using only the expansion property of these graphs we can draw the following corollary from theorem [ thm : prob_bound_1set_expansion ] . [ cor : prob_bound_1set_expander ] for fixed @xmath38 and @xmath66 ,
let an @xmath19 matrix , @xmath18 be drawn from the class of matrices defined in definition [ def:1set_expansion ] , then @xmath67},\ ] ] where @xmath68 with @xmath69 .
theorem [ thm : prob_bound_1set_expansion ] and corollary [ cor : prob_bound_1set_expander ] allow us to calculate @xmath70 where the probability of the probabilistic constructions in definition [ def:1set_expansion ] not being a @xmath71-lossless expander is exponentially small . using corollary [ cor : prob_bound_1set_expander ] and the rip-1 results in @xcite we derived a bound for the probability that a random draw of a matrix with @xmath72s or @xmath73s in each column fails to satisfy the lower bound of the rip-1 constant ( @xmath74 ) and hence fails to come from the class of matrices given in definition [ def:1set_expansion ] , for details see @xcite . from this bound we deduce the following corollary which is a sampling theorem on the existence of lossless expander graphs .
[ cor : prob_expander_existence2 ] consider @xmath66 and @xmath4 fixed . if @xmath18 is drawn from the class of matrices in definition [ def:1set_expansion ] and any @xmath58-sparse @xmath75 with @xmath76 while
@xmath77 and @xmath78 then for @xmath79 and @xmath80 @xmath81 exponentially in @xmath82 , where @xmath83 is the largest limiting value of @xmath84 for which @xmath85 for each fixed @xmath66 and each fixed @xmath4 , @xmath83 in corollary [ cor : prob_expander_existence2 ] is a function of @xmath86 and a phase transition function in the @xmath87 plane . below the curve of @xmath83
the probability in goes to one exponentially in @xmath82 as the problem size grows .
that is if @xmath18 is drawn at random with @xmath72s or @xmath88s in each column and having parameters @xmath89 that fall below the curve of @xmath83 then we say it is from the class of matrices in definition [ def:1set_expansion ] with probability approaching one exponentially in @xmath82 . in terms of @xmath90 for @xmath91 and @xmath92 , corollary [ cor : prob_expander_existence2 ]
say that the probability @xmath93 goes to one exponentially in @xmath82 if the parameters of our graph lies in the region below @xmath83 .
this implies that if we draw a random bipartite graphs that has parameters in the region below the curve of @xmath83 then with probability approaching one exponentially in @xmath82 that graph is a @xmath11-lossless expander . for fixed @xmath94 and @xmath95 with @xmath4 varied.,scaledwidth=40.0% ] figure [ fig : phase_transition1 ] shows a plot of what @xmath83 converge to for different values of @xmath4 with @xmath96 and @xmath82 fixed .
it is interesting to note how increasing @xmath4 increases the phase transition up to a point then it decreases the phase transition . essentially beyond @xmath97
there is inconsequential gain in increasing @xmath4 .
this vindicates the use of small @xmath4 in most of the numerical simulations involving the class of matrices considered here .
note the vanishing sparsity as the problem size @xmath89 grows while @xmath4 is fixed to a small value of @xmath98 .
corollary [ cor : prob_expander_existence2 ] can also be arrived at based on similar probabilistic constructions of expander graphs first proven by pinsker in @xcite with more recent proofs in @xcite .
to put our results in perspective , we compare them to the phase transitions derived from the constants from the construction in @xcite , shown in figure [ fig : phase_transition2 ] . in corollary [ cor : prob_expander_existence2 ] to @xmath99 derived from the alternative construction proven in @xcite.,scaledwidth=40.0% ]
furthermore , for moderate values of @xmath96 this allows us to make quantitative sampling theorems for some compressed sensing reconstruction algorithms . as usual in compressed sensing , in addition to @xmath15-minimization quite a few _ combinatorial _
greedy algorithms have been proposed for these sparse non - mean zero matrices .
these algorithms iteratively locates and eliminate large ( in magnitude ) components of the vector , @xcite .
they include sequential sparse matching pursuit ( ssmp ) , see @xcite ; and expander recovery ( er ) , see @xcite .
besides , theoretical guarantees have been given for @xmath15 recovery and some of the greedy algorithms including ssmp and er .
base on these theoretical guarantees , we derived sampling theorems and present here phase transition curves which are plots of phase transition functions @xmath100 of algorithms such that for @xmath101 , a given algorithm is guaranteed to recovery all @xmath58-sparse signals with overwhelming probability approaching one exponentially in @xmath82 .
computed over finite values of @xmath102 with @xmath4 fixed and the different @xmath103 values for each algorithm - 1/4 , 1/6 and 1/16 for er , @xmath15 and ssmp respectively.,scaledwidth=40.0% ] figure [ fig : pt_alg_compare1 ] compares the phase transition of thee above mentioned algorithms .
remarkably , for er recovery is guaranteed for a larger portion of the @xmath104 plane than is guaranteed by the theory for @xmath15-minimization using sparse matrices ; however , @xmath15-minimization has a larger recovery region than does ssmp .
figure [ fig : pt_alg_compare2 ] shows a comparison of the phase transition of @xmath15-minimization as presented by blanchard et .
al . in @xcite
for dense gaussian matrices based on @xmath17 analysis and the phase transition we derived here for the sparse binary matrices coming from lossless expander based on @xmath74 analysis .
this shows a significant difference between the two with sparse matrices having better performance guarantees .
however , these improved recovery guarantees are likely more due to the closer match of the method of analysis than to the efficacy of sparse matrices over dense matrices . , @xmath105 , for gaussian matrices derived using @xmath17 and @xmath106 for adjacency matrices of expander graphs with @xmath107 , @xmath108 , and @xmath94.,scaledwidth=40.0% ]
due to space constraints the details of the proofs are skipped and the interested reader is referred to @xcite .
it is however important to briefly describe the key innovations in the derivation of the main result , theorem [ thm : prob_bound_1set_expansion ]
. for one fixed set of columns of @xmath18 , denoted @xmath23 , the probability in can be understood as the cardinality of the unions of nonzeros in the columns .
our analysis of this probability follows from a nested unions of subsets using a _
dyadic splitting _ technique . given a starting set of columns we recursively split the number of columns from this set and the resulting sets into two sets of cardinality of the ceiling and floor of the cardinality of their union until a level when the cardinalities are at most two . resulting from this type of splitting is a regular binary tree where the size of each child is either the ceiling or the floor of the size of it s parent set .
the probability of interest becomes a product of the probabilities involving all the children from the dyadic splitting of @xmath34 .
the proof therefore reduces to upper bounding this product .
furthermore , in the binary tree resulting from our dyadic splitting scheme the number of columns in the two children of a parent node is the ceiling and the floor of half of the number of columns of the parent node . at each level of the split
the number of columns of the children of that level differ by one .
the enumeration of these two quantities at each level of the splitting process is necessary in the computation of the bound in .
this led to another novel technical result in our derivation , i.e. _ dyadic splitting lemma _
( lemma 2.5 in @xcite ) .
this work derived bounds on the tail probability of the cardinality of the _ set of neighbours _ of expander graphs resulting into better order constants than the standard probabilistic construction . using this bound and @xmath74 analysis
, we deduce sampling theorems for the existence of expander graphs and their adjacency matrices .
the derivation of the tail bound used a novel technique of _ dyadic set splitting_. we also compared quantitatively , performance guarantees of compressed sensing algorithms which show greater phase transitions for er than @xmath15-minimization which in turn is greater than ssmp .
a comparison of @xmath15-minimization for dense and sparse matrices shows a higher phase transition for sparse matrices .
r. berinde , a. gilbert , p. indyk , h. karloff , and m. strauss , `` combining geometry and combinatorics : a unified approach to sparse signal recovery , '' in _ communication , control , and computing , 2008 46th annual allerton conference on_.1em plus 0.5em minus 0.4emieee , 2008 , pp .
798805 .
s. jafarpour , w. xu , b. hassibi , and r. calderbank , `` efficient and robust compressed sensing using optimized expander graphs , '' _ information theory , ieee transactions on _ , vol .
55 , no . 9 , pp . 42994308 , 2009 .
w. xu and b. hassibi , `` further results on performance analysis for compressive sensing using expander graphs , '' in _ signals , systems and computers , 2007 .
acssc 2007 .
conference record of the forty - first asilomar conference on_.1em plus 0.5em minus 0.4emieee , 2007 , pp .
621625 .
m. capalbo , o. reingold , s. vadhan , and a. wigderson , `` randomness conductors and constant - degree lossless expanders , '' in _ proceedings of the thiry - fourth annual acm symposium on theory of computing_.1em plus 0.5em minus 0.4emacm , 2002 , pp .
659668 .
r. berinde and p. indyk , `` sequential sparse matching pursuit , '' in _ communication , control , and computing , 2009 .
allerton 2009 .
47th annual allerton conference on_.1em plus 0.5em minus 0.4emieee , 2009 , pp .
j. blanchard , c. cartis , j. tanner , and a. thompson , `` phase transitions for greedy sparse approximation algorithms , '' _ applied and computational harmonic analysis _
, vol . 30 , no . 2 , pp .
188203 , 2011 . | we revisit the probabilistic construction of sparse random matrices where each column has a fixed number of nonzeros whose row indices are drawn uniformly at random .
these matrices have a one - to - one correspondence with the adjacency matrices of lossless expander graphs .
we present tail bounds on the probability that the cardinality of the _ set of neighbors _ for these graphs will be less than the expected value .
the bounds are derived through the analysis of collisions in unions of sets using a _
dyadic splitting _ technique .
this analysis led to the derivation of better constants that allow for quantitative theorems on existence of lossless expander graphs and hence the sparse random matrices we consider and also quantitative compressed sensing sampling theorems when using sparse non mean - zero measurement matrices .
-5 mm |
geometrically frustrated spin systems@xcite have been extensively investigated because of the fascinating fundamental physics they display .
more recently , they have attracted interest due to their frequently possessing ground states with spiral magnetic structures.@xcite recent interest in such structures has been engendered by the rediscovery of systems where the magnetic ordering does not have a center of symmetry , and there is perforce a loss of a center of symmetry in the crystal structure as well .
this helps to couple lattice and spin degrees of freedom and gives rise to a plethora of magnetoelectric phenomena.@xcite the spinel structure type ( displayed in figure[fig : struc ] ) with the general formula ab@xmath0x@xmath1 , has been extensively studied in this light because of its ability to host magnetic cations on both its tetrahedrally coordinated a sublattice as well as its octahedrally coordinated b sublattice .
additionally , both sublattices are geometrically frustrated with octahedral b sites forming a pyrochlore network , and the tetrahedral sites forming a diamond lattice .
the former intrinsically displays geometric magnetic frustration associated with the difficulty of decorating the vertices of a tetrahedra with spins that are antialigned .
the latter is frustrated because of competing near- and next - near neighbor interactions.@xcite o@xmath1 structure showing edge - shared laths of bo@xmath6 octahedra ( blue - grey ) with b b magnetic coupling across the edges .
tetrahedrally coordinated a atoms ( dark grey ) connect the octahedral laths , and each a atoms has four b near - neighbors .
oxygen are orange .
[ fig : struc],width=302 ] conventional wisdom , however , is that the dominant magnetic interactions are between the a and the b sublattices provided there are magnetic ions on both sites . adding to the complex nature of magnetic interactions in the spinel structure is that superexchange interactions between a - sites are mediated by atoms on the b - site ( even when the b site is non - magnetic ) , and in the same vein , b - b next - nearest - neighbor interactions are mediated by the a site ion.@xcite in the cubic spinel , the number of near neighbors ( nn ) and next - near - neighbors ( nnn ) for a are 4a(nn ) , 12b(nn ) , 12a(nnn ) , and 16b(nnn ) . for b ,
the distribution is 6a(nn ) , 6b(nn ) , 8a(nnn ) , and 12b(nnn ) .
in addition , most of the coupling pathways are multiply degenerate .
these result in next - near neighbors contributing in an important manner to the overall magnetic ordering behavior.@xcite the high connectivity of the spinel lattice has been shown to give rise to a vast array of interesting magnetic phenomenon .
@xcite chromium spinels , where the b sublattice is fully occupied by cr@xmath7 , present a unique opportunity to examine these complex magnetic interactions due to the fact that cr@xmath7 will not invert to the tetrahedral a sites unlike many other transition metals , as a result of the very strong crystal field stabilization of octahedral @xmath3 cation ( @xmath8 crystal field ) . here
we prepare and study clean samples of solid solutions of zncr@xmath0o@xmath1 and cocr@xmath0o@xmath1 , two well studied chromium spinels with very different magnetic ground states .
zncr@xmath0o@xmath1 , with a non - magnetic cation occupying the a site , undergoes a spin - driven jahn - teller - like distortion at low temperatures which results in a nel - type antiferromagnetic order.@xcite cocr@xmath0o@xmath1 , with magnetic cations on both sublattices , displays collinear ferrimagnetic as well as non - collinear spiral magnetic ordering.@xcite by replacing the zn with co our goal is to introduce magnetic interactions into an otherwise non - magnetic lattice and observe the effect this would have on the magnetic frustration of the system .
this series is particularly interesting because it provides the opportunity to study the effects of dilute magnetic interactions in the absence of changes to the superexchange angle ( co@xmath9 and zn@xmath9 have very similar ionic radii ) or site disorder ( because of the strong octahedral site preference of cr@xmath7 ) .
we find for low concentrations of co , a seeming increase in frustration resulting from spin disorder which results in a glassy magnetic state . at higher co concentrations
there is evidence that the spins form non - collinear structures reminiscent of the cocr@xmath0o@xmath1 end member .
monte carlo simulations of heisenberg spin systems have been performed which allow comparisons with the experimental magnetic susceptibility , and yield the appropriate range of @xmath10 couplings between the different sites to be estimated .
they also suggest the minimal models required to describe the gross aspects of the magnetic susceptibility in these systems .
we also demonstrate here that appropriate scaling of curie - weiss plots of the inverse magnetic susceptibility as a function of temperature allows for systems with purely antiferromagnetic interactions to be distinguished by systems with more complex interactions .
polycrystalline samples in the series were prepared by solid state routes .
appropriate stoichiometric amounts of cobalt oxalate ( coc@xmath0o@xmath1@xmath112h@xmath0o ) , zno , and cr@xmath0o@xmath12 were mixed and ground with ethanol in an agate mortar .
the powders were then pressed into 13 mm pellets and calcined in alumina crucibles at 800@xmath13c for 12h .
these pellets were then reground , pressed back into pellets , and fired at 1150@xmath13c for 12h .
the pellets were then briefly annealed at 800@xmath13c for 24h . during all heatings ,
the pellets were placed on a bed of powder with the same stoichiometry to minimize reaction with the crucible .
x - ray diffraction patterns were obtained using cu - k@xmath14 radiation on a philips xpert mpd diffractometer operated at 45kv and 40ma .
phase purity was subsequently determined by refining the patterns using the rietveld method as implemented in the xnd rietveld code.@xcite dc magnetization measurements were carried out using a quantum design squid magnetometer .
classical monte carlo simulations of the magnetic behavior of the system were performed using the alps project s spinmc application @xcite .
we generated a lattice with the appropriate fraction of a sites occupied for each simulation run in order to perform disorder averaging .
the magnetic interactions were modeled by a nearest - neighbor heisenberg hamiltonian with antiferromagnetic couplings @xmath15 , @xmath16 and @xmath17 where the subscripts specify the sublattice types of the two spins connected by a given interaction . since the number of @xmath15 interactions in the system does not vary with @xmath5 , the other two couplings are varied relative to @xmath15 , which is set to unity ( in monte - carlo units ) throughout .
the inverse magnetic susceptibility of each simulated system was fit to the experimental data by scaling both sets of data such that they satisfy the curie - weiss relationship at at high temperature .
this is the same method used to plot the experimental data described below
see equation ( [ eqn : cw2 ] ) .
it should be noted , however , that in these simulations , `` high temperature '' means high relative to the temperature at which the inverse susceptibility strongly deviates from linear behavior . as discussed in what follows , the experiments could not actually be performed above @xmath18 where the curie - weiss law is strictly applicable .
numerical simulations could indeed verify the following analytic expression for the curie - weiss temperature of the nearest - neighbor model @xmath19 after a reasonable fit to the susceptibility was found for each experimental dataset , corresponding to the different @xmath5 values , the simulations were run again using a newly generated random lattice as many as 3 times to determine the effect of disorder .
a disorder average was then performed by averaging the resulting inverse susceptibilities and the disorder error bars were taken to be the standard deviation of this average .
finally , it should be noted that since the simulations were run until the monte carlo error was negligible ( relative errors of about @xmath20 ) , all numerical error bars shown below represent only the disorder error bars with the exception of the disorder - free @xmath21 end member .
co@xmath22cr@xmath0o@xmath1 . ( a ) points are data and solid grey lines represent the rietveld least - squares analysis .
( b ) evolution of the cubic cell parameter with co content @xmath5 .
the line connecting end members is drawn as a guide to the eye to illustrate the vgard law .
( c ) shows the steady decrease with @xmath5 of the internal positional parameter @xmath23 of oxygen .
[ fig : xrd],width=302 ] a portion of the x - ray diffraction patterns of the solid solution series are presented in figure[fig : xrd ] .
all samples were found to be single phase with no extraneous peaks .
the patterns were fit using the rietveld method to the normal cubic spinel crystal structure with appropriate mixed occupancies on the a site .
the cubic cell parameter and the position of the oxygen atom @xmath24 in the unit cell were obtained from the refinement . because co and zn are relatively similar in terms of their x - ray scattering powers , no attempt was made to extract the relative concentrations of these ions in the lattice .
the cell parameter @xmath25 and oxygen position @xmath23 extracted from the rietveld refinements are presented in figure[fig : xrd](b ) and ( c ) .
co@xmath9 and zn@xmath9 in tetrahedral coordination have very similar ionic radii , 0.58 and 0.60 respectively,@xcite with co@xmath9 the slightly smaller ion . increasing co in the compounds should therefore lead to a contraction of the cell parameter .
we find , however , a small expansion of the unit cell volume .
this is consistent with previously reported structures of the end members.@xcite a possible reason for the expansion is that tetrahedral co@xmath9 is more ionic than zn@xmath9 and the substitution of zn by co introduces a b cation - cation repulsions which expand the lattice .
expansion in the cell parameters in order to minimize repulsion between neighboring cations has been suggested in other spinel systems.@xcite changes in the @xmath23 parameter associated with o atoms in the unit cell occur as the structure tries to accommodate cations of different sizes .
when @xmath26 , the anions are in an ideal cubic - close - packed arrangement , with perfect cro@xmath6 octahedra , while values of @xmath27 indicate an increase in size of the tetrahedron with a corresponding shrinkage and trigonal compression of the octahedron.@xcite as seen from figure[fig : xrd](c ) , there is a systematic decrease in @xmath23 with @xmath5 which brings it closer to the ideal value .
this decrease implies that the tetrahedral sites shrink across the series , in keeping with the smaller size of co@xmath9 , but in contradiction with the unit cell expansion .
.results from fitting the inverse magnetic susceptibility data to the curie - weiss equation .
@xmath28 is taken as the @xmath5-intercept of the inverse susceptibility curve while @xmath29 is chosen as the first point of inflection in the first derivative of the susceptibility with respect to temperature .
@xmath30 is defined as the absolute ratio of @xmath28 to @xmath29 .
the experimental effective moment is calculated using the relation @xmath31 where @xmath32 is the slope of the inverse susceptibility versus temperature . the estimated moments (
spin - only , and unquenched ) are obtain using @xmath33 .
[ table : magnetism ] [ cols="<,^,^,^,^,^,^ " , ] it turns out that a simple quantitative understanding of the susceptibility data in figure [ fig : cw ] is possible as well .
we performed classical monte carlo simulations of the nearest - neighbor heisenberg model on a b site spinel with a site doping and found a series of heisenberg coupling parameters @xmath34 and @xmath35 that bring the simulations into rather close agreement with the experimental susceptibility measurements .
the best fits obtained are shown in figure [ fig : simul1 ] and the resulting coupling constant estimates are summarized in table [ table : best_fits ] .
the fits were first found using systems consisting of @xmath36 conventional cubic unit cells ( 1024 b sites ) with periodic boundary conditions .
we then checked the finite - size scaling on systems of @xmath37 unit cells ( 2000 b sites ) .
it should also be noted that these best fits were determined by searching the parameter space , not by an optimization algorithm , so that quite likely even better agreement is possible using the same model . and the following coupling parameters : @xmath34 = 0.45 and @xmath35 = 0.90 . [ fig : simul2],width=340 ] we were encouraged to see that all of the best @xmath34 and @xmath35 parameters do not differ much from each other ( the @xmath38 data had too weak a dependence on @xmath17 for us to determine it effectively ) .
evidently , the bulk magnetic behavior of the entire series of materials is well described by taking @xmath16 to be comparable to @xmath15 and taking @xmath17 to be about half of @xmath15 .
accordingly , it was considered useful to actually fix all the couplings to such values and only vary @xmath5 the resulting scan ( see figure [ fig : simul2 ] ) shows that it is indeed the value of @xmath5 that most strongly determines the bulk magnetic properties of the system .
the suprise is that for the range of parameters studied by which seem to be close to the experimentally measured susceptibility , @xmath16 and @xmath15 are effectively equal , rather than @xmath16 being significantly larger than @xmath15 ; the frequent expectation .
lyons , kaplan , dwight , and menyuk ( lkdm ) @xcite have proposed the following inequality : @xmath39 implies collinear nel ordering of moments in spinels such as cocr@xmath0o@xmath1 . for @xmath40 non - collinear ordering is suggested . since in cocr@xmath0o@xmath1 , @xmath41 = @xmath42 , we expect that if @xmath43 , the ground state will be non - collinear , which is certainly true for cocr@xmath0o@xmath1 .
so the monte carlo estimates of @xmath15 and @xmath16 are consistent with the experimental realization of non - collinear ground states.@xcite recently ederer and komeilj@xcite have performed detailed density functional calculations on the @xmath5 = 1 end - member , cocr@xmath0o@xmath1 .
they suggest that @xmath17 is important ( ignored by lkdm ) as determined here as well .
while @xmath17 in their calculations is determined to be typically less than @xmath15 , they estimate that @xmath16 can be even twice as large as @xmath15 which is not the result obtianed in the preliminary analysis presented here . ,
width=340 ] the results of the field - cooled susceptibility traces was used to construct a tentative magnetic phase diagram for the solid solutions as depicted in figure[fig : phase ] .
we use changes in the slope of the field - cooled susceptibility as a function of temperature to obtain the boundaries between different magnetic phases .
regardless of composition , all samples demonstrate paramagnetic behavior at high temperatures and begin to show transitions to long range order starting near 100k .
the end member zncr@xmath0o@xmath1 is the only compound which exhibits long range antiferromagnetic order because of its structural distortion .
we make the reasonable assumption that even small amounts of co result in glassy behavior as indicated by the line separating frustrated antiferromagnetism from glassy states . for concentrations of @xmath44 0.6
the magnetic behavior begins to look more like that of cocr@xmath0o@xmath1 with a collinear ferrimagnetic transition between 50k and 100k followed by a transition into a non - collinear spiral at lower temperatures .
the behavior near @xmath5 = 0.6 is rather complex and more experiments will be particularly important for this region of the phase diagram .
the transition temperatures for the spiral appear to be slightly suppressed by zn substitution with respect to the end member as illustrated by the positive slope of the phase boundary as the cocr@xmath0o@xmath1 end member is approached . for the co - rich compounds at the lowest temperatures ,
the structures begin to develop some spiral short range order ( sro).@xcite in summary , we have prepared and studied the magnetic properties of a solid solution of zncr@xmath0o@xmath1 and cocr@xmath0o@xmath1 .
we have found that replacing non - magnetic cations with magnetic cations in the tetrahedral sites of a well ordered spinel results in a transition from glassy behavior at low concentrations of magnetic cations into a n@xmath45el - like ferrimagnetic order as well as commensurate and incommensurate magnetic spirals .
these results are significant because they demonstrate that the presence of magnetic cations on the a - site can have the effect of relieving magnetic frustration on the b - site by allowing the system to find a lower energy state in the form of non - collinear magnetic spirals .
monte carlo simulations suggest the minimum physiccs required to describe trends in the magnetic behavior , and suggest the range of strengths of the different magnetic couplings .
we also present a useful form for plotting susceptibility data as a function of temperature which allows us to easily recognize the existence of uncompensated magnetic interactions in otherwise antiferromagnetic systems .
ems would like to acknowledge encouraging discussions with leon balents .
the national science foundation for support through a career award ( nsf - dmr0449354 ) , and for the use of mrsec facilities ( award nsf - dmr0520415 ) .
jed was supported by the rise program at the ucsb mrl . | we present the magnetic properties of complete solid solutions of zncr@xmath0o@xmath1 and cocr@xmath0o@xmath1 : two well - studied oxide spinels with very different magnetic ground states .
zncr@xmath0o@xmath1 , with non - magnetic @xmath2 cations occupying the a site and magnetic @xmath3 cations on the b site , is a highly frustrated antiferromagnet .
cocr@xmath0o@xmath1 , with magnetic @xmath4 cations ( three unpaired electrons ) on the a site as well , exhibits both nel ferrimagnetism as well as commensurate and incommensurate non - collinear magnetic order .
more recently , cocr@xmath0o@xmath1 has been studied extensively for its polar behavior which arises from conical magnetic ordering . gradually introducing magnetism on the a site of zncr@xmath0o@xmath1 results in a transition from frustrated antiferromagnetism to glassy magnetism at low concentrations of co , and eventually to ferrimagnetic and conical ground states at higher concentrations .
real - space monte - carlo simulations of the magnetic susceptibility suggest that the first magnetic ordering transition and features of the susceptibility across @xmath5 are captured by near - neighbor self- and cross - couplings between the magnetic a and b atoms .
we present as a part of this study , a method for displaying the temperature dependence of magnetic susceptibility in a manner which helps distinguish between compounds possessing purely antiferromagnetic interactions from compounds where other kinds of ordering are present . |
in the papers @xcite @xcite we started the systematic study of the special class of effective theories of strong interactions , which we call _ localizable_. roughly speaking , these are theories with the @xmath0-matrix which can be obtained in a _ perturbative way _ in the frame of an _ effective field theory _ that contains auxiliary resonance fields along with the fields of true asymptotic states ( those stable with respect to the strong interaction ) .
our goal is to construct an efficient method for calculating the amplitudes of physical processes .
this means that we need to develop the systematic scheme of perturbative calculations in the framework of infinite component effective field theory .
there are few obstacles that usually prevents the effective theory concept to become a useful computational tool .
the main one is the presence of an infinite number of coupling constants , which requires introducing an infinite number of renormalization prescriptions .
it is clear that we hardly get some predictive power for our theory unless find some _ regularity _ in the system of those prescriptions .
further , if one admits an unlimited number of the resonance fields in a theory ( as we do ) , then even the amplitude of a given loop order ( say , tree level ) acquires contributions from the infinite number of graphs .
the problem of convergence of the latter _ functional _ series is the problem of correct definition of the loop expansion .
our approach suggests a solution .
first , we systematize the set of renormalization prescriptions . in @xcite
we demonstrated that only the combinations of coupling constants that we call as _ resultant _ parameters require fixing to obtain renormalized @xmath0-matrix at any given loop order .
next , we show that even these renormalization prescriptions can not be taken arbitrary it is the convergence requirements giving a hand . here
is the basic idea .
any requirement of convergence of functional series always can be thought as a _ restriction for the parameters _ appearing in those series . in other words , _ the couplings , and so the renormalization prescriptions , are unavoidably restricted for each step of perturbation expansion to make sense_. it is this circumstance that eventually gives rise to the system of bootstrap equations for physical parameters ( see @xcite ) and allows to make numerical predictions .
briefly speaking , we develop dyson s perturbation scheme for the infinite component effective theories taking seriously the problems of mathematical correctness and self - consistency . in this talk we try to make a short overview of our strategy and explain the main postulates .
the first ( and , probably , the main ) step toward a classification of the renormalization prescriptions is transition to the _ minimal parametrization _ discussed in @xcite .
this allows one to rewrite every given graph in terms of the minimal propagators and minimal effective vertices .
the numerator of the minimal propagator is just a covariant spin sum considered as a function of four independent components of momentum . at tree level
the minimal vertices are just the hamiltonian on - shell vertices with the wave functions crossed out . to be precise
, one shall consider an @xmath0-matrix element of the formal infinite sum of all the hamiltonian items constructed from a given set of , say , @xmath1 ( normally ordered ) field operators .
the members of this sum differ from each other by the hamiltonian coupling constants , by number of differentiation operators and/or , possibly , by their matrix structure .
the matrix element under consideration should be calculated on the mass shell , presented in a lorentz - covariant form and considered as a function of @xmath2 independent components of the particle momenta .
the wave functions should be crossed out .
the resulting structure we call as the @xmath1-leg _ minimal effective vertex of the 0-the level_. every minimal effective vertex presents a finite sum of scalar formfactors dotted by the corresponding tensor structures , each formfactor being a formal power series , or even just a polynomial in relevant scalar kinematical variables .
l - th level _ effective vertex only differs from that described above by the presence of bubbles ( @xmath3 loops in total ) attached to the same point as the external legs
. the numerical coefficients ( eventually supplied with the index @xmath3 ) in the formal power series that describe the corresponding formfactors are called as the _ @xmath3-th level minimal parameters_. the above - mentioned resultant parameters are certain sums of the minimal ones .
the resultant parameters are all independent , as far as one considers as independent all the hamiltonian couplings .
besides , as shown in @xcite , every _ amplitude graph _
depends only on minimal parameters . in turn , the full sum of @xmath4-th loop order graphs describing certain scattering process can be expressed solely in terms of the resultant parameters with level index @xmath5 .
at least a few words should be said here about the renormalization prescriptions . in @xcite
it is shown that , if the renormalization point is taken on shell , the resultant parameters are the only quantities that require formulating renormalization prescriptions .
actually , our use of the minimal parametrization implies that we rely upon the scheme of _ renormalized perturbation theory_. in this scheme one starts from the physical action written in terms of _ physical _ parameters and adjusts the counter terms in such a way that the numerical values of those parameters remain unchanged after renormalization .
it is this fact that allows us to obtain the bootstrap restrictions for the physical ( in principle _ measurable _ ) parameters later on .
one of the most important requirements which we make use of when constructing the meaningful items of dyson s perturbation series is that of polynomial boundedness .
namely , the full sum of @xmath0-matrix graphs with a given set of external lines and fixed number @xmath4 of loops must be polynomially bounded in every pair energy at fixed values of the other kinematical variables .
there are two basic reasons for imposing this limitation .
first , from the general postulates of quantum field theory ( see , e.g. , @xcite ) it follows that the full ( non - perturbative ) amplitude must be a polynomially bounded function of its variables .
second , from experiment it follows that this is quite a reasonable requirement .
since we never fit data with non - perturbative expressions for the amplitude , it is natural to impose the polynomial boundedness requirement on a sum of terms up to any fixed loop order and , hence , on the sum of terms of each order .
similar argumentation also works with respect to the bounding polynomial degrees . to avoid unnecessary mutual contractions between different terms of the loop series , it makes sense to attract the following _ asymptotic uniformity _ requirement . _ the degrees of the bounding polynomials which specify the asymptotics of the amplitude of a given loop order must be equal to those specifying the asymptotics of the full ( non - perturbative ) amplitude of the process under consideration .
_ surely , this latter degree may depend on the type of the process as well as on the values of the variables kept fixed .
the condition of asymptotic uniformity ( or , simply , uniformity ) is concerned with the asymptotic behavior of the items corresponding to different loop orders .
it tells us not too much about the rules needed to convert the disordered sum of graphs with the same number of loops ( and , of course , describing the same process ) into the well - defined ( _ summable _ ) functional series . to solve the latter problem we rely upon another general principle which we call as
_ summability requirement_. it is formulated as follows . _ in every sufficiently small domain of the complex space of kinematical variables there must exist an appropriate order of summation of the formal sum of contributions coming from the graphs with a given number of loops , such that the reorganized series converges .
altogether , these series must define a unique analytic function with only those singularities which are presented in contributions of individual graphs .
_ at first glance , the summability ( analyticity ) requirement may seem somewhat artificial .
this is not true .
there are certain mathematical and field - theoretical reasons for taking it as the guiding principle that provides a possibility to manage infinite formal sums of graphs in a way allowing to avoid inconsistencies .
it is , actually , _ both _ the summability and uniformity principles that allow us to use the cauchy formula and obtain well defined expression for the amplitude of a given loop order @xcite .
we would like to stress that the requirements of uniformity and summability are nothing but independent subsidiary conditions fixing the type of perturbation scheme which we only work with .
surely , there is no guarantee that on this way one can construct the most general expressions for the s - matrix elements in the case of effective theory .
nevertheless , there is a hope to construct at least a meaningful ones presented by the dyson s type perturbation series only containing the well - defined items .
the explicit calculations are better illustrated by the concrete examples @xcite , here we just briefly sketch the strategy .
one needs to consider any process and classify all the resultant parameters contributing to it at a given loop order .
then , using the stated above principles of uniformity and summability , one applies the well known cauchy formula to the amplitude of the given order at various regions in the space of kinematical variables .
several things happen during this process .
first , the summability principle helps to identify the parameters of the amplitude singularities as a combination of resultant parameters which are already fixed by renormalization conditions . with this in hands , and with the degree of bounding polynomial given by uniformity principle , the cauchy formula allows one to write down a _ well defined expression for the amplitude in a given domain ( layer ) of the space of kinematical variables . _
next , one obtains _ different expressions for the same amplitude in different layers _ , different couplings and masses contributing to each of them .
the layers intersect , so one should equate the expressions in the common domains of validity to ensure self - consistency ( usually , it appears to be equivalent to requirement of crossing symmetry of the given loop order amplitude ) .
the latter step gives an infinite number of numerical relations for the resultant parameters and thus , as explained above , for the renormalization prescriptions or , the same , for the physical parameters of a theory .
we call these relations as the _ bootstrap equations _ of a given loop order . it should be noted , that , although the bootstrap for each loop order amplitude gives restrictions for _ physical _ parameters ( the latter are independent of the loop order ) , the structure of bootstrap equations themselves varies from order to order .
the search for the _ solution _ of all these equations is still beyond our scope , what we can only do so far is to test them with experimental data .
the results are presented in two other talks , see @xcite .
we are grateful to v. cheianov , h. nielsen , s. paston , j. schechter , a. vasiliev and m. vyazovski for stimulating discussions .
100 a. vereshagin , _ `` explicit equations for renormalization prescriptions in the case of pion - nucleon scattering '' _ , this conference ; k. semenov - tian - shansky , _
`` renormalization in effective theories : prescriptions for kaon - nucleon resonance parameters '' _ , this conference .
a. vereshagin and v. vereshagin , phys .
d * 59 * ( 1999 ) 016002 .
a. vereshagin , v. vereshagin , and k. semenov - tian - shansky , zap .
nauchn . sem .
pomi * 291 * , part 17 , ( 2002 ) 78 ( in russian ) .
hep - th/0303242 ( the english version will appear in `` j. math .
( ny ) * 125 * , iss . 2 ( 2005 ) 114 )
. a. vereshagin and v. vereshagin , phys . rev .
d * 69 * ( 2004 ) 025002 .
s. weinberg , physica a * 96 * ( 1979 ) 327 .
s. weinberg , _ the quantum theory of fields _ , vols . 1 - 3 ( cambridge university press , cambridge , england , 2000 ) .
n. n. bogoliubov , a. a. logunov and i. t. todorov .
_ foundations of the axiomatic method in quantum field theory _ ( nauka , moskow , 1969 ; in russian )
. j. f. chew . _ the analytic s - matrix _
( w. a. benjamin , inc . new york , 1966 ) . | we summarize our latest developments in perturbative treating the effective theories of strong interactions .
we discuss the principles of constructing the mathematically correct expressions for the s - matrix elements at a given loop order and briefly review the renormalization procedure .
this talk shall provide the philosophical basement as well as serve as an introduction for the material presented at this conference by a. vereshagin and k. semenov - tian - shansky @xcite . |
mass loss dominates stellar evolution on the asymptotic giant branch ( agb ) .
the mass loss probably begins on the early agb , or even the first red giant branch @xcite , but culminates during the last 10% of the agb , the thermal - pulsing agb ( tp - agb ) .
observed mass - loss rates range from @xmath6 for semiregulars and miras , up to a few times @xmath7 for oh / ir stars and obscured carbon stars .
typically , the mira phase lasts between @xmath8 and a few times @xmath9 yr ( e.g. @xcite ) . the oh / ir star or the obscured carbon star phase may last @xmath10yr .
because the mass - loss rates increase during the agb evolution , circumstellar shells may be expected to have a centrally - condensed , compact appearance . but
mass - loss fluctuations during the thermal - pulse cycle can lead to detached circumstellar shells @xcite , and observations of detached shells have been interpreted as such @xcite .
evidence for detached shells came from the iras point source data , when @xcite found carbon stars with excess flux at 60@xmath11 m .
@xcite , in the same way , found oxygen - rich stars ( m stars ) with detached shells ; they interpreted these in terms of mass - loss variations during the thermal - pulse cycle . from iras
spectroscopy @xcite found evidence for a detached shell with a small inner radius around the o - rich star r hya ( see also @xcite ) .
photometric evidence may be affected by cirrus @xcite , and direct imaging can give more stringent evidence for detached shells .
izumiura et al .
( 1996 , 1997 ) imaged detached shells around the carbon stars y cvn and u ant , with u ant even showing a _
double _ detached shell .
a spectacular thin co ring was found around the carbon star tt cyg @xcite .
@xcite found a detached shell around the carbon star u cam .
@xcite reported parsec - sized thin dust shells around two carbon - rich post - agb stars , afgl 2688 and afgl 618 , indicating episodic mass - loss variations with timescales of a few times 10@xmath12 years .
a thermal - pulse explanation is likely for all these cases .
circumstellar shells could grow to very large dimensions during the agb evolution . in principle , a gradual and gentle slow - down of the earliest ejected shell(s ) by the ism could allow subsequently expelled matter to catch up , and yield very large , detached - appearing shells independent of the thermal - pulse cycle .
@xcite , using iras survey data for 512 agb stars and young pne ( mostly within 1 kpc ) , found a total of 76 stars with circumstellar shells greater than 2@xmath0 in the 60@xmath11 m data .
of these , four shells are larger than 4 pc in diameter ; all four appear detached , having inner diameters of between 0.5 and 1.4 pc .
( the detection of these shells is complicated by the often very bright central sources , requiring deconvolution techniques to separate the shell from the psf of the stellar source ) .
such fossil shells carry the imprint of the mass - loss history and the structure and density of the ism .
they could help to explain the presence of halos around planetary nebulae ( but see @xcite ) and the interaction of the most evolved of them with interstellar matter .
clearly , a search for extended shells - the larger , the better - is of interest . here
we report on an extreme dust shell around the source iras 02091 + 6333 , with a diameter of 4 pc .
we interpret the shell not in terms of mass - loss variations , but as a wall of swept - up ism .
we suggest that rings and walls around evolved stars may not have a single explanation .
the dust ring was discovered by one of us ( rw ) during a search for extremely extended structures , such as holes and emission regions around planetary nebulae , on the iras skyview maps provided by nasa .
it is visible at 60@xmath11 m and - slightly better - at 100@xmath11 m , and attracted our attention because of its well - defined shape and the presence of a stellar , almost centrally located object , prominent at 12@xmath11 m but fainter at 25@xmath11 m and absent at longer wavelengths .
the shape of the dust emission suggests that the central source is responsible for the ring , which has the appearance of a detached shell .
the shell is brightest at 100@xmath11 m indicative of cold dust .
interestingly , the shell , particularly its brightest part towards the south , is also optically identifiable as an absorbing cloud on the possi blue - sensitive ( o ) plates .
the stellar field is very dense ( consistent with low foreground extinction for a line of sight near the galactic plane ) and the lack of blue stars in the small patch is highly noticeable .
we estimate from the appearance of the patch that @xmath13mag .
the processed iras images , retrieved from the ipac skyview images , and also remade from the original ( recalibrated ) scans using the iras software telescope @xcite , are shown in fig .
the emission at 100@xmath11 m has a slightly elliptical shape , with the long axis in the n - s direction .
it consists of a ring of emission which is thicker ( and brightest ) in the south and least thick in the west .
there is a distinct , approximately round , central cavity .
the outer dimensions are 19@xmath0 @xmath1 16@xmath0 , corresponding to 3.9 @xmath1 3.2 pc at @xmath3pc ( see below ) .
the ring has a mean apparent thickness of @xmath41.0 pc , and the cavity is @xmath41.6 pc in diameter .
the resolution of iras at 100-@xmath11 m is sufficient to resolve the cavity , but the thickness of the ring is likely affected by the iras beam .
the emission spur towards the south - west is probably unrelated ( it has a different colour temperature ) .
the total flux of the ring is about 20 jy at 60@xmath11 m and 70 jy at 100@xmath11 m : both values are uncertain due to the need to subtract a sloping background . [ iras.images ] the concave thickening of the ring in the south is suggestive of an interaction with the ism of the galactic plane which runs approximately e - w , 2.3@xmath14 below the object . but this interaction has had no marked effect on the outer shape of the ring .
a subtraction of the 60-@xmath11 m from the 100-@xmath11 m map , carried out using the unprocessed images , showed that there is no obvious radial dependence within the ring .
the black - body dust temperature is about 35 k. the large distance from the central star indicates the shell is heated externally ( e.g. , @xcite ) .
the temperature should be compared to typical values for galactic cirrus emission of 1520 k @xcite .
the flux densities of the central source ( table 1 ) are consistent with photospheric emission .
optical images from the palomar sky surveys show a star coincident with the central source .
although rather bright , very red , and located not far from the galactic plane ( @xmath15 ) in one of the most studied regions of the milky way , simbad does not list individual observers .
it is present in the guide star catalog ( gsc - id 0404101743 ) , in the ac2000 , and was covered by the tycho experiment onboard of hipparcos .
data for the star are listed in table 1 .
one long - slit spectrum ( 300 sec ) of the star was taken with the instrument afosc attached to the 1.82 m telescope of the asiago observatory .
it was obtained on dec 1@xmath16 1999 , at an air mass of @xmath41.4 .
the spectral range covered was 3940 7860 and the dispersion was 4.9 / pixel .
another spectrum was kindly provided by maurice gavin , taken on dec 21st , 1999 at the worcester park observatory ( uk ) at a dispersion of 40 / pixel .
both spectra were reduced in the usual way . the spectra show only the star itself , i.e. we found no traces of any nebulosity in its immediate neighborhood .
the tio bands are very well visible and establish the spectral type as m3 or m4 .
we could however not determine the luminosity class .
the discussion below shows that a giant is most probable and we will assume the object to be an m3iii star .
several photometric runs in bvri were kindly done for us by roger pickard , who used his 40 cm newtonian reflector .
he reported @xmath17 = 2.06 .
.[star.data ] parameters for the central star [ cols= " < , < , < , < , < , < " , ] the central star appears to have a slight offset of @xmath18 from the center of the shell .
this apparent offset is in part caused by the difference in brightness between the n and s part of the shell .
compared to the peak brightness positions , the star is very well centred , 7@xmath19 from the two components .
although the star is close to the centre of the ring , in such a dense stellar field the chance of an accidental superposition of the shell and the central source has to be addressed .
the iras catalogue lists 49 sources located within @xmath20 radius of the present position .
of these , 36 are detected at 12@xmath11 m and 20 are detected at both 12 and 25@xmath11 m .
one of these has the colours of an hii region and can be removed .
of the 36 candidates , 14 appear to be associated with a gsc star of @xmath21 , within 5 arcsec .
the v[12 ] colours indicate that 6 of these are early - type stars .
the remaining 8 stars ( including iras 02091 + 6333 ) have the colours of rgb or agb stars : the density of such stars in this field is 1 per 1400 square arcminutes ( we note that the association of iras 02091 + 6333 is , at 1.6 arcsec , the closest of the 14 sources . ) using the apparent offset of 1.5@xmath19 , the chance that the central star and ring are unrelated is 0.5% : the association is significant at the 3@xmath22 level . even without the presence of an optical star , the significance of an association between the ring and the central iras source would be close to 3@xmath22 .
a relation between the ring and its bright central agb or rgb star is therefore plausible .
one should however also take into account how common such rings are in the iras images .
this close to the galactic plane , the background emission is extensive and structured . but within a field of @xmath23 , there is no other ring - like structure : the background emission shows elongated ridge - like structures instead .
the difference between these structures and the ring is shown by the fact that most of the iras psc sources in this region correspond to parts of the ridges , tracing brightness fluctuations which are common features of these structures .
in contrast , the ring is quite smooth and was not picked up as one or more point sources .
although confusion can not be fully ruled out , a physical association between star and ring appears plausible .
for an m4iii or m3iii star , @xmath24 or @xmath25 respectively @xcite .
the observed @xmath26 , gives an extinction @xmath27 or @xmath28 , for an m3iii star .
the dereddened v@xmath29 $ ] colour of 4.9 indicates an m3 rather than m4 star , with @xmath30 ( e.g @xcite ) for which we find a distance of @xmath31pc .
this extinction and distance are not unreasonable if compared to nearby stars of known spectral type , luminosity class and with reliable photometry : within a radius of 40@xmath0 four such stars were found in simbad : two are about @xmath32pc distant and have an @xmath33 of 0.15 and 0.30 , respectively , and the other two are about 1 kpc away and have values of 0.59 and 0.43 , respectively
. some of the extinction towards the central star may be circumstellar .
we note that the w35 complex , located only 2 degrees away in the perseus arm , is at a distance of 2.35 kpc ( e.g. @xcite ) .
if an extinction sheet is associated with this spiral arm , iras 02091 + 6333 is likely located in front of it .
the ob association cam ob1 is found a few degrees away at 900 pc @xcite .
the likely distance suggests that the m3 star is of luminosity class iii rather than i. furthermore , its circumstellar obscuration can not be very large .
the iras psc flux densities and the v[12 ] colour are consistent with photospheric emission from an m3 star : there is no evidence for hot circumstellar dust , and the present mass - loss rate must be small .
the msx flux densities ( table 1 ) agree with iras but the 14 micron point seems a little higher , a difference that may be caused by the larger beam width of msx , which was a smaller telescope .
evidence for mass loss ( the detached shell ) , generally indicates a thermal - pulsing , variable agb star .
@xcite shows that some rgb stars also show dust emission , but only close to the star .
but there is no strong evidence for variability of the m3iii star .
the 228 tycho measurements include some fainter ones , but the magnitude is close to the limiting magnitude of the instrument and the catalog does not flag the star as variable .
a comparison of its brightness in a possi overlapping region ( plates e / o 597 , taken on 1952 - 09 - 15/16 , and e / o 878 , taken on 1953 - 10 - 30/31 ) also does not show any obvious variability .
monitoring by roger pickard between january 2000 and january 2001 indicates a possible brightening at v from 10.5 to 10.1 , and from 12.5 to 11.7 at b , but this requires further confirmation .
the bolometric magnitude is estimated as @xmath34 , based on the bc(v ) of @xcite .
this is very uncertain as the bolometric correction is a steep function of colour .
the corresponding luminosity @xmath35 is , given the uncertainty , consistent with a tp - agb star ( e.g. @xcite ) .
low - level variability would be consistent with a pre - mira evolutionary stage .
but the data does not rule out a location near the tip of the rgb ; neither can a tp - agb star during the helium - burning phase of the tp cycle be ruled out , when luminosity and the mass - loss rate are reduced .
the fact that part of the shell is seen in absorption gives an indication of its mass .
we use the conversion ratio @xmath36 .
the absorption region is roughly @xmath37 .
this gives a total mass of @xmath38 for @xmath39mag .
this estimate is by necessity rough , since the extinction is estimated and the size of the region has significant uncertainties .
the iras flux ( 70 jy at 100 @xmath11 m ) gives a second determination of the mass of the shell . using a black - body dust temperature of 35 k
, we derive a total dust mass of @xmath40 , where @xmath41 is the distance in kpc @xcite . for a distance of @xmath42pc and a gas - to - dust mass ratio of 200 , the total mass in the ring
is @xmath43 m@xmath44 .
most of this mass will be in the brighter southern region .
the agreement between the two estimates is deceptive , since the uncertainties in these calculates are easily a factor of 2 .
however , it indicates a relation between the absorption region and the ring . in general , distant shells around agb stars
have been interpreted in terms of episodically enhanced mass loss , related to the thermal - pulse cycle @xcite .
assuming an expansion velocity of 10 km / s , the outermost gas in the shell would have been ejected @xmath45yr ago , and the gas closest to the star @xmath46yr ago .
the duration of the mass - loss event would not be inconsistent with the duration of the thermal - pulse cycle @xcite .
however , this would require an equally long time of quiescence to explain the cavity , whilst the duration of the low - luminosity part of the cycle , during helium burning , is only 1015 per cent of the cycle .
the relatively large mass of the shell suggests that much of the the gas does not originate from the star .
given the uncertainties in distance , integrated iras flux density , and temperature , the mass could be as low as 1 m@xmath44 in which case a significant contribution from stellar mass loss could not be excluded .
in view of these two points , we suggest that the shell reflects ism sweep - up . in this model , the agb wind , moving out at @xmath47 into the ism , is creating a a denser shell consisting of both components .
models for the early phases of such sweep - up have been calculated by @xcite .
the typical cooling time for the post - shock region is @xmath48 , with density increasing to @xmath49 , reducing as the front slows down @xcite .
the interaction region can thus be considered momentum - driven , in agreement with @xcite .
the velocity of the interface declines as @xmath50 or @xmath51 , until its velocity drops below the sound speed in the ism outside the shell .
@xcite do not consider cases where the amount of matter swept up from the ism exceeds the mass contributed by the stellar wind .
the large mass of the present shell suggests the present shell has evolved past this phase .
[ momentum.flow ] , wind velocity @xmath52 and ism density @xmath53 .
this phase ends when the velocity drops below the sound speed of the ambient medium.,width=321 ] using the measured diameter of the shell , and assuming its mass reflects the ism , suggests a local ism density ( outside of the shell ) of a few particles @xmath54 .
[ momentum.flow ] shows the early development of the wall . in this case
, the velocity of the interface ( the wall ) drops to 1kms@xmath55 after about @xmath56 .
the early phase ends once the thermal pressure @xmath57 of the ambient ism becomes non - negligible , causing the interface to stall .
this happens once @xmath58 for the ambient pressure , we use a typical value of @xmath59 . assuming that the interface has stalled at a distance of 2pc from the star ( the present radius of the shell )
, the equation above yields a mass - loss rate @xmath60 .
this value is typical for semi - regular variables , which have not yet evolved into mira variables , and is sufficiently low that the 12-@xmath11 m emission would still be dominated by the photosphere .
much of the stellar envelope mass on the agb may be lost in such a moderate wind before the onset of the mira phase ( the narrow @xmath61 relation for short - period miras @xcite suggests that their masses are already reduced to close to the final white dwarf mass before the mira phase begins ) .
iras 02091 + 6333 could have built a hydrogen wall during its pre - mira evolutionary phase , before reaching the mass - loss rates associated with the tip of the agb .
however , the resulting shell will not be stable : once the interface becomes trapped , the swept - up shell begins to expand into the ambient ism at its sound speed . using a typical value of the sound speed of @xmath62 , the ring reaches a thickness of 1pc in @xmath63 . as a rough estimate , we may assume that the shell ceases to be observable once its density has dropped to twice that of te surrounding ( presumably uniform ) medium .
for a thin shell : @xmath64 which gives @xmath65 the observed thickness of the ring is roughly 1 pc , but this is affected by the iras 10-@xmath11 m resolution and the actual thickness is probably rather less . within the limited accuracy , we find agreement between the possible parameters of such a wall and the observed ring of iras 02091 + 6333 .
this suggests that it is possible to create such a wall even before the mira phase on the agb , and that the wall could survive for an appreciable fraction of the thermal - pulsing agb ( @xmath66 ) . at a constant mass - loss rate ,
the interface will stall in less than @xmath67 after the onset of the mass loss .
however , the mass loss likely slowly increases over time , effectively resetting the clock and re - creating the shell .
the resulting shell may be seen for up to @xmath63 following an increase in mass - loss rate .
the precise value depends on ambient density and mass - loss rate .
the limited life time of the ring indicates it can not pre - date the present evolutionary phase of the star .
we therefore suggest that iras 02091 + 6333 presents a case where an agb star , having undergone mass loss , perhaps intermittently over some fraction of its agb life time , has surrounded itself with a swept - up wall of dust and hydrogen more than 1pc away .
two aspects of the wall need further consideration : first , it is roughly ( but not perfectly ) spherical ; second , its intensity is strongest towards and away from the galactic plane .
the morphology , in the model of the previous subsection , will be determined by the intrinsic distribution of the stellar wind and by the structure of the ism .
the ism shows structure on many size scales , however the shell is small compared to its large - scale structure . if the shell is running into a region of higher density , this is most likely to have occured near its outer edge as this is the region contributing most to the volume .
this would affect the intensity distribution , but would be less likely to cause large deviations from spherical symmetry .
stellar winds on the agb tend to be symmetric , until the star approaches the tip of the agb ( e.g. @xcite ) .
however , the effects of the peculiar motion of the star with respect to the ism should be considered .
the proper motion indicates a velocity of 45kms@xmath55 .
about half of this is due to the effect of the solar motion and galactic rotation ; the remaining velocity vector has a large uncertainty .
the proper motion of the ism is also not easily measured , but considerable deviations from circular rotation may occur in both the young stellar population and the ism ( e.g. @xcite ) .
a recent study of proper - motion vectors of stars in the cam ob1 association @xcite , at a distance of 900 pc and at @xmath68 , shows a number of distant ob - type stars with similar proper motion vectors to iras02091 + 6333 .
the relative velocity between star and ism is therefore not well determined .
to first order , the effect of stellar motion is to change the wind velocity with respect to the ism .
the stalling radius depends on the square root of this velocity , and thus the shape of the shell is relative insensitive to small peculiar velocities .
but at velocities much larger than 10kms@xmath55 , the effects become large and the star may position itself off - centre over the life time of the shell . supernova remnants often show a bilateral structure very similar to that of iras 02091 + 633 and these structures are generally oriented in the same way with respect to the galactic plane @xcite .
the explanation is believed to be shaping by the galactic magnetic field , which tends to run along the plane .
the shaping can either occur directly by the magnetic field , or indirectly if the ism is stratified by the field .
however , for nebulae as small as our ring these effects are expected to be negligible .
the similarity is likely accidental .
planetary nebulae do not show a preferential alignement with respect to the galactic plane @xcite . assuming that the wall around iras 02091 + 6333 is due to agb
ism interaction , the question arises how this structure will evolve .
the interface between the wind and the ism will have stalled and the inner edge of the wall can be considered stationary .
the outer edge is expanding at the velocity of sound , @xmath69kms@xmath55 .
the thermal - pulsing agb lasts several @xmath70yr , allowing this edge to travel several parsec , beyond the phase of observability .
once the star enters the post - agb phase , stellar evolution speeds up dramatically and the wall will evolve much slower than the inner nebula and its star .
after the agb , the circumstellar wind quickly becomes ionized as the star heats up . at the same time
a hot , much faster stellar wind sweeps up the inner agb wind , with the swept - up region forming ( part of ) a planetary nebula , with a typical radius of 0.10.5pc @xcite .
the planetary nebula phase may last @xmath71yr @xcite . in our model , this swept - up shell is itself embedded in the much larger wall .
outer rims and haloes are seen in some planetary nebulae but they do not have a unique explanation @xcite and their presence can not be taken as a vindication of this scenario . in iras 02091 + 6333 the wall is visible as an iras shell .
once the expansion dilutes the wall , this shell becomes less and less distinct .
however , inside the wall is still a region where the density is much lower , in which the agb wind keeps the surrounding medium at bay .
thus , instead of a shell , one can expect the appearance of an infrared cavity in which the planetary nebula may be located .
such a cavity has been found around the old nova ck vul @xcite .
these authors classify ck vul as a late - thermal - pulse nova , similar to nova sakurai . in this case
, the nova is expected to be located inside an old planetary nebula .
( a hydrogen nebulosity has been found around ck vul but has ( sofar ) not been indentified as an old planetary nebula ) .
the cavity has a radius of approximately 1pc .
we suggest that this object may be an example of an old wall
. the presence of other , very large cavities around planetary nebulae will be the topic of a future paper .
we present the discovery of a large dust shell with a detached appearance , surrounding a previously unknown but bright agb star . an association between the star and the dust shell is likely , given the statistics of confusion in this region .
the spectral type of the star , magnitude and colour , is used to derive the extinction , and a distance of @xmath72 .
the gas mass of the shell is estimated as @xmath73 .
the large mass and size of the shell make an origin in a thermal pulse , as normally assumed for detached shells around agb stars , unlikely .
instead we propose a model in which an agb wind of low mass - loss rate has swept out a cavity in the ism ; the swept - up material forms the present shell .
we show a typical calculation in which the snowplough phase lasts for @xmath74rm . at the end of this phase ,
the shell is subsonic , stalls and begins to expand into ism at its sound speed . in our model ,
the shell remains visible for a further @xmath75yr .
after this , the shell would be visible as an apparent hole in the ism surrounding the central agb star , or the later planetary nebula wich follows the agb evolutionary phase .
a star with a similar size dust shell is r crb @xcite , where a huge nearly symmetrical dust shell has been discovered .
this shell may have a similar origin .
hydrogen walls are also known around the sun and nearby stars , from the interaction between the solar wind and the ism ( e.g. @xcite ) : they are a common feature of star
ism interactions .
the interaction between an agb wind and its surrounding ism has largely been ignored in the analysis of circumstellar environments , with the exception of @xcite .
the shell of iras 02091 + 6333 suggests that the effects of such an interaction should be further explored .
we are very grateful to roger pickard who provided us with the results of his photometric runs of the central star of the dust shell and to maurice gravin for spectroscopy .
special thanks also go to sonia temporin , gernot grmer , ernst dorfi and stefan kimeswenger .
fernando comeron took time to explain to us the various aspects of low - velocity hydrodynamics .
this research has made use of the vizier and simbad databases , maintained by the cds , of the ipac on - line iras and msx databases , and of the iras software telescope system maintained by sron , groningen . | we present the discovery of a huge ( 19@xmath0 @xmath1 16@xmath0 ) dust ring surrounding a bright ( @xmath2 = 10.60 ) red star .
the dust ring has , at @xmath3pc , a diameter of 4 pc , and a central hole of @xmath41.5 pc across .
part of the shell is also seen as an absorption nebulosity .
the star is classified as a m3iii agb star . among agb stars
its detached shell is of unrivalled size .
detached shells around agb stars are normally interpreted in terms of thermal pulses . however , in this case a significant fraction of the shell may consist of swept - up ism ; the detached appearance can be explained with wind
ism interaction .
we present a model where the agb wind has been stopped by the surrounding ism , and the swept - up shell is now expanding at the sound speed .
the model predicts that the ring will disperse over a few times @xmath5yr , and eventually will leave a large hole in the ism surrounding the agb star or its future planetary nebula . |
the apparent interferometric sizes of variable red giants vary dramatically - up to a factor of 3 - with wavelength and pulsation phase @xcite + @xcite .
these large variations challenge current hydrodynamical and line opacity models of these stars , indicating that their extended atmospheres are extremely complex .
simultaneous spectroscopic and high angular resolution observations over a wide wavelength range are key to understanding the various phenomena these stars exhibit , including photometric pulsation , high mass - loss rates , dust nucleation , and formation of molecular layers and masers .
high angular resolution thermal infrared observations of bright red giants in _ narrow bandpasses _ ( @xmath3 ) have been available for some time @xcite .
the keck interferometer nuller ( kin ) @xcite and the midi instrument on the vlti @xcite offer much greater spectral coverage , typically over the whole 813 @xmath0 m region , with a sensitivity goal of about 1 jy . the keck interferometer is nearly ideal for the study of such stars .
its 85 m baseline produces a fringe spacing of 27 mas at 11 @xmath0 m , commensurate with the expected angular scales of most variable red giants located within 1kpc ( photosphere diameter of a few mas , with thermal emission from dust peaking within several stellar radii ) .
the detection is carried out by means of a low - resolution spectrometer ( @xmath4 35 ) , sensitive across a bandpass which spans the broad opacity peaks characteristic of various types of o - rich and c - rich dusts .
rs crb was the first star successfully observed with the keck interferometer s mid - infrared ( mir ) beam combiner .
it is a bright semi - regular variable of type _ a _ , quite similar to a mira but with smaller photometric variations ( @xmath5 @xmath6 1 - 2 ) .
it has a mean pulsation period of 332 days @xcite and is classified as a red giant of spectral type m7 , with a 12 @xmath0 m flux of 53jy @xcite .
it has an oxygen - rich dust shell with strong silicate emission features detected by the iras low resolution spectrometer ( lrs ) ( class se6 t , @xcite ) .
ita et al .
@xcite detected sio maser emission around rs crb , which can be regarded as a tracer of dense molecular gas in the inner region ( @xmath6 2 stellar radii ) of the circumstellar envelope ( e.g. vlba images of tx cam @xcite and other miras @xcite ) .
we present a measurement of the wavelength - dependent fringe visibility of rs crb from 812 @xmath0 m , a spectral region where both the central star and its surrounding dust shell are visible . in combination with the intensity spectrum measured across the same band with the individual keck telescopes , it is possible to determine the relative brightness of the two components , the spatial extent of the shell , and its opacity spectrum .
the observations were made on august 8 2004 , using a subset of the kin optics @xcite as a two - input beam combiner , without most of the special features needed to produce the high fringe contrast ( @xmath7 0.99 ) required for nulling .
the telescopes each fed one of the two inputs of the primary modified mach- zehnder ( mmz ) symmetric beam combiner @xcite .
one of the two resulting outputs was sent to kali , a low - resolution mir spectrometer @xcite .
kali uses a boeing si : as bib detector , cooled to 4k , with the optics and radiation shield cooled to @xmath8 k using ln@xmath9 .
a cold internal pinhole ( equivalent diameter 2@xmath10/d ) at an intermediate focus provides spatial filtering .
direct - view prisms disperse the n - band light into 14 spectral channels .
the central wavelengths for the channels were measured using fourier transform spectroscopy ( fts ) scans of a hot filament .
they ranged from 8.1612.01 @xmath0 m , with the channel width varying from @xmath6 0.2 @xmath0 m on the red end to @xmath6 0.4 @xmath0 m on the blue end .
rs crb was observed at a single , quasi constant projected baseline ( 80.6 80.9 m ) , along with a set of calibrator stars ( early k giants ) with known near - infrared ( nir ) diameters . to allow for the differential limb darkening between the nir and mir wavelengths
, we adopted conservative error bars on the estimated calibrators diameters in the mir .
a list of the stars and their properties is given in table [ starlist ] .
the observation sequences were similar to those used at nir wavelengths for the keck interferometer @xcite .
the fringe amplitude was measured accurately through a fast ( 40hz ) optical path difference modulation .
each star was visited between 3 and 4 times , with typically one or two minutes spent tracking the fringe during a visit .
the most important source of uncertainty in these visibility measurements was the removal of the strong and variable thermal background .
for each of the two beams , the absolute stellar signal was determined through sky chopping .
a spatial chop was implemented using the fast tip - tilt mirrors of the keck adaptive optics systems @xcite at a frequency of 5hz .
because of beamwalk on the optics upstream of the tip - tilt mirror , a fraction of the thermal background is also modulated . to correct for this bias
, the demodulated background signal was determined through a similar chopping measurement on blank sky .
after this calibration , stellar signals derived from the various single aperture observations still exhibited fluctuations at the @xmath6 5 - 10% level , which we hope to overcome in the future with a symmetric three position chop .
this is currently the main contributor to the visibility error bars , both on rs crb and the calibrators .
the data reduction was otherwise similar to that used for nir visibility measurements at pti and ki @xcite .
the calibrator observations were used to measure the `` system visibility '' ( the response of the interferometer to a point - source ) , which did not show any significant time dependence .
its value during the rs crb observations was estimated by combining the system visibilities measured on all three calibrators ( table 1 ) .
the resulting calibrated visibility curve for rs crb is plotted in fig [ v2 ] , showing that the source is clearly resolved .
since all measurements presented here were obtained with a single baseline orientation , only models with spherical symmetry are considered .
there is also astrophysical justification for this choice , both on a theoretical and observational basis .
the envelope surrounding an agb star can be described to first order by a spherically symmetric outflow at a constant velocity of @xmath11 km / sec , in which the material is heated by friction between gas and dust particles and cooled primarily by the line radiation of molecules such as h@xmath9o @xcite . also , maser observations of _ single _ evolved stars generally show ring - like structures which are much more compatible with spherical geometries than disk - like ones @xcite . finally , recent nir interferometric phase closure measurements failed to detect asymmetric structures in a sample of 10 m - class semi - regular variables ( s. ragland , priv .
comm . ) .
we do not expect single component brightness distributions to accurately describe dusty red giants in the mir . yet
, a uniform disk fit allows a simple representation of the visibility data at each wavelength , and we use it here as a convenient starting point .
the resulting apparent diameter , plotted in fig .
[ ud_vs_wavelength ] , increases rapidly between 8.2 and 10 @xmath0 m and reaches a plateau at longer wavelengths .
the apparent size and its chromatic variations are large , pointing to a variable opacity effect in the upper atmosphere , rather than to true changes in the physical dimension of a single photosphere . adopting now a two component representation , consisting of a central photosphere surrounded by a dust shell with a variable opacity ,
the visibility is sensitive to the size of the shell and its brightness relative to the central star , and depends only weakly on the size of the stellar photosphere . a straightforward interpretation to the apparent size variations in fig .
[ ud_vs_wavelength ] is that the relative brightness of the shell increases with wavelength across the measured band , becoming dominant at the long wavelengths .
this behavior correlates nicely with the se6 classification of rs crb in the silicate dust sequence @xcite , which shows a similar variation of the infrared excess versus wavelength .
we now model the shell as infinitely thin , with angular diameter @xmath12 , optical depth @xmath13 and temperature @xmath14 , surrounding a blackbody photosphere of angular diameter @xmath15 and temperature @xmath16 .
this model is similar to the one previously applied to molecular layers around red giants @xcite .
it predicts both the visibility spectrum and the intensity spectrum across the observed bandpass .
we explored a grid of model parameters \{@xmath16 ,
@xmath15 , @xmath17 , @xmath12 } , computing @xmath18 per spectral channel for the visibility at each point in the grid .
the spectrum of rs crb derived from our single keck telescope observations ( fig .
[ photometry_plot ] ) is used to determine @xmath13 at each point .
we constrain the stellar temperature @xmath16 to be 3100 @xmath19 150k , corresponding to the range of effective temperatures proposed for m7 giants @xcite .
the value of @xmath15 can be estimated from a blackbody fit to the nir flux @xcite .
the results range from 3.74.2 mas between the j and m bands .
this estimate is likely crude , however , as it neglects the effect of the circumstellar envelope . in light of this uncertainty , we allowed @xmath15 to range from 2.55 mas .
the shell diameter was allowed to vary between 1 and 50 times @xmath15 , while the maximum shell temperature was set to 1500k , a reasonable upper limit for the grain sublimation temperature .
the error bar on each parameter is computed as the deviation that increases the @xmath20 by one after reoptimizing all the other parameters .
the resulting values are : @xmath15 = @xmath21mas , @xmath12 = @xmath22mas , @xmath14 = @xmath23k , and @xmath24 = 0.04 - 0.3 .
the agreement between the model and the observed @xmath25 is excellent ( @xmath20 per point = 0.38 , fig .
[ v2 ] , plain curve ) showing that the visibility data at all wavelengths can be reproduced by a 2-component model with a fixed , _
wavelength independent _ geometry , and an opacity profile fixed by the observed intensity spectrum .
the shell diameter @xmath12 is the most accurately constrained parameter .
it indicates that most of the flux at long wavelengths - probing the dust emission zone- originates from a region located around 7 - 8 stellar radii .
interestingly , this is comparable to the inner dust shell size derived by the midi instrument for the mira star rr sco @xcite . for wavelengths shorter than 10 @xmath0 m
, the contribution from the central star is visible and well constrained by the observed @xmath25 . for the stellar temperature range adopted , the fit yields a central object diameter consistent with the nir photometry .
we expect the photospheric diameter as measured at shorter nir wavelengths to be smaller , due to the effect of molecular layers around this type of star @xcite which is more prominent at long infrared wavelengths @xcite .
this has now been confirmed by iota interferometric observations at 1.65@xmath0 m - a region free of significant molecular absorption - , which yield an uniform disk diameter of 3.15 @xmath19 0.12 mas for rscrb ( john monnier , priv .
june 2005 ) .
the model fit depends on the overall flux from the shell rather than on @xmath14 and @xmath13 individually , so those parameters are not separately well constrained by the fit .
@xmath14 must be at least @xmath26k to account for the observed shell flux at the shortest wavelengths , and is presumably below @xmath27k , as that is the maximum radiative equilibrium temperature allowing silicate grains to survive in a stationary flow around a typical mass - losing agb star @xcite .
within this temperature range , the corresponding @xmath13 varies over almost an order of magnitude .
although the overall dust opacity is not well constrained , its _ shape _ is ( fig .
[ tau ] ) .
the sharp opacity rise observed between 8 and 9.7 @xmath0 m is characteristic of amorphous silicates .
the shoulder seen near 11 @xmath0 m is often seen in low mass - loss rate agb stars @xcite and may indicate the presence of another solid , such as amorphous alumina .
it could for instance be in the form of mg - fe aluminosilicate compounds @xcite .
alternatively , this 11 @xmath0 m feature could reflect the presence of mg - rich crystalline forms of olivine such as forsterite ( @xmath28 ) , which exhibits an absorption peak around 11.2@xmath0 m @xcite .
our spectral resolution does not allow to choose between these two different scenarios at present .
using broad - band thermal interferometry with the twin keck telescopes , we have measured the size and opacity profile of the dust shell around rs crb , an agb star with strong silicate emission features .
the dust shell `` radius '' derived in our model can be understood as a characteristic size of the weighted brightness distribution of the shell .
strikingly , a single shell radius value of 7 - 8 @xmath29 allows to fit the data over the whole 8 - 12 microns range , suggesting that most of the detected emission arises from dust in this region . the opacity profile which fits both the intensity and
visibility spectra indicates the presence of amorphous silicates , and possibly other dust species like amorphous alumina or mg - rich olivine in this region .
the @xmath25 measurements obtained at the blue end of our bandpass , where the dust opacity is low , constrain the size and fractional amount of flux coming from the `` central '' star . these first results obtained with the keck interferometer
already demonstrate the value of wide band long baseline thermal infrared interferometry .
interferometric observations allow non ambiguous determination of the spatial brightness distribution very close to the central source , information that can not be derived from spectroscopy alone .
finally , the wide spectral coverage provides access to different physical regimes and can be used to simultaneously constrain multiple components of an astronomical source .
the keck interferometer is funded by the national aeronautics and space administration ( nasa ) .
part of this work was performed at the jet propulsion laboratory , california institute of technology , and at the michelson science center ( msc ) , under contract with nasa .
the keck observatory was made possible through the generous financial support of the w.m .
keck foundation .
finally , the authors wish to recognize and acknowledge the very significant cultural role and reverence that the summit of mauna kea has always had within the indigenous hawaiian community .
we are most fortunate to have the ability to conduct observations from this mountain .
beichman c.a . ,
woolf , n.j . and lindensmith c.a . , 1999 , the terrestrial planet finder : a nasa origins program to search for habitable planets , jpl publication 99 - 3 .
colavita , m.m .
et al . 1999 ,
apj 510 , 505 colavita m.m .
2003 , apj , 592 , 83 cotton w. et al .
2004 , a&a , 414 , 275 creech - eakman m.j .
2003 , in spie conf .
proceedings 4841 , 330 danchi w. et al .
1994 , aj , 107 , 1469 diamond p.j . &
kemball a.j .
1999 , iau symposium 191 , edited by t. le bertre , a. lebre and c. waelkens , 195 dyck h.m .
1996 , 111 , 1705 fogel m.e . & leung c.m .
1998 , apj , 501 , 175 .
habing h.j .
1996 , a&a rev . 7 , 2 , 97 hinkle k.h . &
barnes t.g .
1979 , apj , 227 , 923 ita y. et al .
2001 , a&a , 376 , 112 kerschbaum f. & hron j. 1994 , a&as , 106 , 397 kholopov et al .
1992 , bulletin dinformation centre donnees stellaires , vol 40 , 15 koike c. et al .
2003 , a&a 399 , 1101 koresko c.d .
, mennesson b.p . ,
serabyn e. , colavita m. , akeson r. , swain m. 2003 , spie 4838 , 625 leinert c. et al .
2003 , astrophysics & space science , 286 , 1 , 73 mennesson b. et al .
2002 , , 579 , 446 mennesson b. et al . 2003 , in conf . towards other earths : darwin / tpf and the search for extrasolar terrestrial planets , ed .
m. fridlund & t. henning , ( noordwijk : esa - sp-539 ) , 525 . monnier j. et al .
2004 , , 605 , 436 mutschke h. et al .
1998 , a&a 333 , 188 ohnaka k. et al 2005 , a&a , 429 , 1057 perrin g. et al .
1998 , a&a , 331 , 619 perrin g. et al .
2004 , a&a , 426 , 279 scholz m. 2001 , , 321 , 347 serabyn e. and colavita m. m. 2001 , applied optics , 40 , 1668 serabyn e. et al .
2004 , spie conf .
5491 , 806 , ed . w.a .
traub sloan g.c . & price s.d .
1998 , , 119 , 141 thompson r.r . , creech - eakman m.j . &
van belle g.t .
2002 , apj , 124 , 1706 tsuji t. 1988 , a&a , 197 , 185 tuthill , p.g . ,
haniff , c.a . & baldwin , j.e . , 1995 , , 277 , 1541 tuthill , p.g . ,
monnier , j.d . & danchi , w.c .
, 2000 , spie conf .
4006 , 491 , , ed .
lna & a. quirrenbach waters r. et al .
1998 , a&a , 331 , l61 weiner j. et al .
2000 , , 544 , 1097 weiner j. 2004 , , 611 , 37 willson l.a .
2000 , ara&a , 38 , 573 wizinowich p. et al .
2003 , spie proc . 4839 , 9 yamamura , i. , de jong t. , & cami j. , 1999 , a&a , 348 , l55 ccccc rs crb & m7 sra & @xmath14 & 411 & + @xmath30 crb & k2iii & 2.81 @xmath19 0.1 @xmath31 & 676 & @xmath32 + 39 cyg & k3iii & 2.82 @xmath19 0.1 @xmath31 & 604 & @xmath33 + @xmath30 cyg & k0iii & 4.47 @xmath19 0.2 @xmath34 & 598 & @xmath35 + + | we report interferometric observations of the semi - regular variable star rs crb , a red giant with strong silicate emission features .
the data were among the first long baseline mid - infrared stellar fringes obtained between the keck telescopes , using parts of the new nulling beam combiner .
the light was dispersed by a low - resolution spectrometer , allowing simultaneous measurement of the source visibility and intensity spectra from 812 @xmath0 m .
the interferometric observations allow a non - ambiguous determination of the dust shell spatial scale and relative flux contribution . using a simple spherically - symmetric model , in which a geometrically thin shell surrounds the stellar photosphere
, we find that @xmath1 30% to @xmath1 70% of the overall mid - infrared flux - depending on the wavelength - originates from 7 - 8 stellar radii .
the derived shell opacity profile shows a broad peak around 11 microns ( @xmath2 0.06 ) , characteristic of mg - rich silicate dust particles . |
one of the most promising applications of carbon - based new materials like carbon nanotubes or graphene is the area of transparent conducting layers.@xcite carbon - based materials have many advantages over widely used oxides like indium tin oxide ( ito ) in terms of better flexibility and no toxicity ; however , their basic optical and electrical properties have not reached those of conventional transparent conductors so far .
the field has been broadened recently by the possibility of separating nanotubes by electronic type.@xcite by this method , not only highly enriched semiconducting or metallic networks can be prepared , but also extremely purified mixed samples.@xcite these samples offer a unique opportunity to study the role of intrinsic conductivity vs. intertube contacts in nanotube networks.@xcite intertube connections have been studied previously on junctions built from individual nanotubes@xcite with similar or dissimilar electronic character ( ss , mm or sm , respectively , where m stands for metallic and s for semiconducting tube ) . in macroscopic networks containing predominantly one type of nanotube , and by measuring both the transport and frequency - dependent conductivity , these roles can be even more precisely determined .
impedance studies by garrett et al.@xcite showed that above a characteristic frequency of the order of khz , the values reflect the intrabundle conductance instead of the combined values of bundles and junctions that is measured by the dc method .
thus infrared spectroscopy , which starts at a few thz , clearly yields the intrinsic conductivity of the bundled nanotubes . in this paper
, we report the transparent and conducting properties of separated metallic and semiconducting single - walled nanotube ( swnt ) films and compare them to those of an ultrahigh purity reference sample .
the contactless measurement of the frequency - dependent conductivity and the four - point dc transport results on the same material can be directly related and the role of intrinsic conductivity vs. that of intertube contacts established .
we also study the effect of doping on the transport and optical properties . in order to improve the conductivity and avoid chemical reactions at defects , we use mild p - doping by nitric acid vapor at room temperature.@xcite as this procedure
is shown to result in a doped state unstable over time , it is only regarded as a proof - of - concept experiment .
nevertheless , two important practical questions can be addressed : 1 . if we find a stable doping method , which kind of tubes should be used to obtain the best transparent conducting properties ; 2 .
if stability is required over optimal conducting properties , which kind is the most stable against accidental doping ?
we find the answer by comparing optical transmittance and dc conductivity measurements on all three kinds of samples .
we used very high purity swnt samples commercialized by nanointegris.@xcite starting material was arc - discharge p2 by carbonsolutions.@xcite separation of the nanotubes was performed by density gradient ultracentrifugation ( dgu),@xcite resulting in separated metallic and semiconducting samples with 95% nominal purity , and a mixed ( reference ) sample with 99% swnt content . for the latter , we assume a composition of 1/3 metallic and 2/3 semiconducting tubes .
( the fact that all three samples underwent identical treatment starting from the original p2 material ensures that extrinsic factors due to sonication and other steps do not influence the comparison .
these effects were extensively discussed in ref . . )
the mean diameter of the nanotubes is 1.4 nm and their length varies from 100 nm to 4 @xmath1 m . from the aqueous suspensions of surfactant - covered nanotubes we have prepared samples of different thickness using vacuum filtration@xcite through an acetone soluble filter .
the thickness of the films was controlled by the applied amount of solution .
these layers ( three of each sample , differing in thickness ) were subsequently relocated over a 1 cm x 1 cm x 1 mm quartz ( suprasil ) substrate . to remove any remnant of the solvent and traces of accidental atmospheric doping
, the samples were annealed at 200@xmath2c for 13h .
self - supporting thin films were prepared by stretching the nanotube layer over a hole created in a graphite disk .
using this kind of samples enables us to measure transmission without the perturbation caused by substrates and to calculate easily the optical functions from transmission.@xcite doping of the films was performed by subjecting them to nitric acid vapor overnight at room temperature .
this mild treatment causes hole doping in the @xmath3-electron system,@xcite without converting sp@xmath4 carbon atoms into sp@xmath5 by carboxylic group addition.@xcite scattering type near - field infrared microscopy ( s - snom ) were taken with a neasnom nano - ftir instrument ( neaspec gmbh ) using a quantum cascade laser with 10.5 @xmath1 m central wavelength .
the s - snom technique is described in detail elsewhere.@xcite briefly , the s - snom uses a metal coated afm tip with the radius of curvature of 20 nm enabling the operation at ultrahigh spatial resolution and near - field interaction between the illuminated tip and the sample.@xcite besides the usual afm topography data the interferometric detection of the optical signal reveals local optical information including the absolute scattering efficiency ( amplitude ) as well as phase of the scattering.@xcite the observed optical contrasts are strongly related to the complex dielectric function of the studied material.@xcite wide range ( far - infrared through ultraviolet ) optical measurements were performed on the self - supporting nanotube networks using the following spectrometers : a bruker ifs 66v / s fourier - transform ( ftir ) interferometer for the far - infrared ( fir ) and mid - infrared ( mir ) range , a bruker tensor 37 ftir in the near infrared ( nir ) , and an ocean optics qe65000 instrument for the ultraviolet - visible ( uv - vis ) region . in the case of samples on quartz substrate , only data from the uv - vis range were collected to obtain the transmission value at 550 nm ( 18180 @xmath6 ) . for each nanotube network on the quartz substrate ,
we measured the four - point resistivity using a keithley 192 digital multimeter , and calculated their sheet resistivity ( @xmath7 ) applying the van der pauw formula.@xcite small dots of colloidal silver were used for contacting the samples .
all measurements were performed at room temperature .
thickness of the self - supporting films was measured by atomic force microscopy following the procedure described in ref . .
from the thickness and the spectra of each type of network , we determined the absorption coefficient at 550 nm and used these values to estimate the thickness of the samples used for resistivity , in order to determine the conductivity values .
the absorption coefficients measured on the undoped samples are 3.39@xmath8 10@xmath9 @xmath6 ( r ) , 3.18@xmath8 10@xmath9 @xmath6 ( s ) , and 2.36@xmath8 10@xmath9 @xmath6 ( m ) , respectively .
these values compare very well with those measured earlier on laser - deposited films prepared by the same procedure.@xcite
figure [ fig : afm ] shows the local structure of the films of all three types .
all films consist of bundles with about 50 nm average thickness as seen in both the topographic and optical images .
the s - snom technique results in both amplitude and phase values of the scattered infrared light .
the frequency range of the infrared laser is 955 - 1030 @xmath6 , where the exciting radiation interacts with the free ( drude ) carriers from the metallic nanotubes .
both amplitude and phase show a remarkable uniformity , proving the high purity of our samples .
additionally , the scattering amplitude from the metallic samples was found to be consistently higher than that from the semiconducting ones , at four distinct wavelengths , 970 , 985 , 1000 and 1015 @xmath6 , indicating higher conductivity .
a full account of s - snom measurements in a wider frequency range and using more sophisticated evaluation methods to obtain optical functions@xcite will be published elsewhere .
frequency - dependent transmission spectra of self - supporting thin films in the fir / uv range for metallic ( m ) , semiconducting ( s ) and mixed reference ( r ) samples before ( dark color ) and after p - doping ( light color ) .
solid black bars mark the first two semiconducting ( s@xmath10 and s@xmath11 ) and the first metallic ( m@xmath10 ) transition ; red dashed bar indicates the frequency corresponding to 550 nm wavelength.,width=321 ] transmission spectra of self - supporting thin films of the mixed and separated nanotubes before and after doping are shown in fig .
[ fig : trans ] .
these wide - range spectra prove that separation is effective : in the semiconducting sample the s@xmath10 and s@xmath11 transitions are very intense , while the m@xmath10 transition can hardly be seen ; in the case of the metallic sample the intensity of the m@xmath10 transition is high and the peaks representing the transitions of semiconducting nanotubes are weak .
the intrinsic frequency - dependent conductivity can be calculated from the wide - range spectra and the thickness of the films by kramers - kronig transformation of the transmittance@xcite and contributions from individual transitions can be determined by the fitting procedure given in ref . .
we show in fig .
[ fig : sep ] the optical conductivity curves corresponding to the s@xmath10 and m@xmath10 transitions , respectively ; these curves represent the envelopes of the transitions of semiconducting and metallic tubes of different diameter . for comparison , we include in fig . [
fig : sep ] similar curves obtained for a commercial p2 sample.@xcite the ratio of the areas a(s@xmath10)/a(m@xmath10 ) , as expected , scales with the semiconductor / metal ratio : 1.03 and 0.98 for p2 and r , respectively , 9.67 for s and 0.14 for m. additional information included in these numbers is that upon dgu treatment , the composition of the reference sample does not change considerably .
a slight blueshift of @xmath12 100 @xmath6 occurs from p2 to r , and no further shift upon separation .
the shift indicates that the purified mixed sample contains slightly more small - diameter nanotubes than the original .
the optical conductivity increases about 18 per cent upon the purification step for both the semiconducting and metallic transition . optical conductivity in the s@xmath10 and m@xmath10 transition region for the three samples , indicating the relative semiconducting to metallic nanotube content . in the bottom panel ,
the same transitions for a commercial p2 sample are shown ( ref . ) .,width=302 ] doping has the largest noticeable effect on the reference and semiconducting samples ( fig .
[ fig : trans ] ) , showing an increase of transmittance in the near infrared .
the disappearance of the first and second interband transitions proves the high p - doping efficiency .
( the appearance of the new transmission minimum between s@xmath10 and s@xmath11 is most probably originating in excitonic effects@xcite and will be discussed elsewhere . ) in addition , in all three samples the far - infrared transmission is decreasing , due to the free carriers introduced by doping.@xcite frequency - dependent optical conductivity of doped and undoped nanotube films .
the color code is the same as in fig .
[ fig : trans ] .
note the logarithmic frequency scale .
squares with error bars represent measured dc conductivity values.,width=321 ] figure [ fig : cond ] shows the wide - range optical conductivity curves of all doped and undoped samples . the most striking difference ( apart from the interband transitions ) appears below 2000 @xmath6 : all doped samples have a strong drude contribution to the optical conductivity and therefore can be considered metals . for comparison
we also show the dc conductivity of the samples from fig .
[ fig : cond1 ] obtained by averaging the respective data shown in table [ tab : cond ] .
the low - frequency conductivity scales qualitatively with that obtained from the transport measurements , the latter being consistently lower , as expected for a heterogeneous structure involving contacts.@xcite we regard this behavior as compelling evidence for the electronic structure of the nanotubes being mostly responsible for the conductivity enhancement upon doping , with contact effects playing a secondary role .
a similar conclusion was drawn by miyata et al.@xcite who explained the selective conductivity enhancement by the differences in the electronic density of states of metallic and semiconducting tubes and their change upon doping .
hole doping results in an increased carrier ( hole ) density and partially filled valence bands , leading to enhanced intrinsic conductivity of the nanotubes .
the carrier density changes show up in the optical spectra,@xcite with the free carrier absorption increasing and the interband transition intensities decreasing because of a decrease in the density of initial states for these transitions .
.[tab : cond]conductivity data for metallic , semiconducting and reference films of different thicknesses . [ cols="<,^,^,^,^,^ " , ] conductivity data are summarized in table [ tab : cond ] .
these data show that the highest relative increase ( 12.8 - 17.5 ) is observed for the semiconducting sample , followed by the reference ( 7 - 10.9 ) while the lowest increase is shown by the metallic tubes ( 1.5 - 2.2 ) .
( the thickness dependence of the conductivity enhancement is related to the percolation nature of conductance in the films.@xcite ) in absolute values , the doped reference sample gives the highest sheet conductivity .
miyata et al.@xcite performed sulfuric acid doping on laser ablated metallic and reference nanotube samples .
they obtain a sheet conductance enhancement of 1.58 for their metallic and 19.8 for the reference sample . comparing those with our numbers above and taking into account the difference in both starting material and doping agent
, we find the agreement remarkable .
studies on individual nanotube networks cited above@xcite resulted in the observation that the resistivity of junctions between semiconducting and metallic nanotubes - which create schottky barriers - is two orders of magnitude higher than the resistivity between tubes of the same electronic type . applying this model to our measurements ,
two observations should follow if intertube contacts played the key role in network conductivity : 1 . in the undoped samples , both s and m type networks should have a higher dc conductivity than r , while at higher frequencies , the conductivity should scale with the metallic content ; and 2 .
doping of a system where both metallic and semiconducting tubes occur , will result in dramatic decrease of intertube resistivity and consequent increase in dc conductivity . if intertube connections dominated during doping,@xcite the effect for the reference sample would be much larger in the transport than in the optical data . since our observations point to rather the opposite , we conclude that although intertube connections are important , they do not dominate transport properties of high - quality nanotube networks as these .
a pure mechanical explanation for the conductivity enhancement upon acid treatment has also been suggested:@xcite according to this model , the only role the acidic dopant plays is to `` clean '' the intertube contacts from surfactant molecules present in the undoped sample .
the presence of surfactant in the material has been , unfortunately , never proven , nor has any reaction been proposed between surfactant and acid . in our wide - range spectra , the intense infrared absorption of these organic molecules should be observed if they were present in significant concentration .
also , our films are washed copiously with water after filtering and we believe that this treatment removes surfactants much more effectively and with much less residue than any unspecified reaction with an acid .
dc sheet conductance s@xmath13 vs optical density ( -log t ) at 550 nm for the metallic , semiconducting and mixed reference swnt thin films after annealing and doping , respectively .
the shaded area represents the application region of ito.,width=302 ] for comparison of the transparent nanotube networks , we applied a recently introduced figure of merit,@xcite which is the inverse of that given by jain and kulshreshtha.@xcite this value is analogous to that defined by gordon@xcite but contains the transmission at a single wavelength instead of the integrated visible transmission , and similar to @xmath14 in ref . . fig .
[ fig : cond1 ] shows the dc sheet conductance s@xmath13 as a function of optical density ( _ -log t _ ) at 550 nm , the wavelength of choice for solar cell applications . due to the fact that both the optical density and sheet conductivity are proportional to the thickness of the film , for each sample the measured values
can be fitted with a linear function and the determination of the thickness is not necessary .
higher slope of this line indicates better quality of the film as a transparent conductor .
the shaded area represents the region of ito layers already used in technological applications ( sheet resistance r@xmath15 140 @xmath16 , @xmath17 0.7).@xcite extrapolating the values to lower thickness ( dashed lines in fig . [
fig : cond1 ] ) indicates that both the doped reference sample and the doped semiconducting sample reach the minimum parameters of the ito region , while the doped metallic one barely misses it .
figure [ fig : cond1 ] shows that doping causes the conductivity to increase in all samples , but the relative increase is different depending on electronic structure . at the chosen wavelength of 550 nm , the increase in slope of the lines in fig .
[ fig : cond1 ] , therefore the increased performance as transparent conductor , is determined by the change in conductivity rather than absorbance . from the optical density values in fig . [
fig : cond1 ] , a change within 15 per cent in absorbance can be deduced . in the near infrared
, however , the transmittivity of the samples with substantial semiconductor content rises dramatically ( fig . [
fig : trans ] ) .
this means that for near - infrared applications , carbon nanotubes would be even more hopeful than for visible ones .
time dependence of the sheet conductivity of metallic , semiconducting and reference samples after doping with nitric acid .
data have been scaled to the sheet conductivity value at 60 minutes after removal from nitric acid vapor .
the dashed lines indicate the sheet conductivity of the undoped films .
the samples numbered a , b and c , respectively , increase in thickness in that order.,width=302 ] the most common practical problem with gas - phase doping is its reversibility , i.e. the dedoping process which usually starts as soon as the doping agent is removed .
indirectly , this effect is seen in fig .
[ fig : cond ] where the error bars in the dc conductivity of the doped samples are significantly larger than those for the undoped samples .
we have followed the dedoping process by measuring the sheet conductance for 24 hours after removing the samples from the nitric acid vapor .
figure [ fig : tdep ] shows these curves for a set of nine samples similar to those in fig . [ fig : cond1 ] .
the values before doping ( relative to the as - doped state ) are shown as dashed lines on each plot .
all three types of samples show a behavior tending towards saturation which can be described by the sum of two exponential functions of time , but the parameters do not seem to have real physical significance and therefore we do not discuss them here .
the most striking feature of fig . [ fig : tdep ] is that contrary to the semiconducting and mixed samples , which approach a much higher saturation conductivity than before doping , the metallic samples seem to return to their original conductance in about a day .
doping is thus not very effective for these materials , but on the other hand , they seem to be the most stable against accidental doping .
conclusions from our experiments can be drawn on two levels .
first , frequency - dependent conductivity data indicate that by acid doping , the intrinsic conductivity increases considerably , especially for semiconducting nanotubes .
although contacts play a crucial part in increasing conductivity of the networks , they are not the exclusive reason for improved electrical properties ; moreover , doping does not seem to improve contact resistance to an extent to be considered solely responsible for this effect . the other type of conclusion concerns the application possibilities of separated and doped nanotube networks , respectively .
if one chooses charge doping to increase the dc conductivity of carbon nanotubes , separation by type does not improve the results .
doped non - separated nanotubes can be best applied to substitute ito for technological applications using visible light .
however , undoped nanotubes of metallic type show other advantages : they have better conductance properties than either semiconducting or mixed ones , moreover , when doped , they recover their initial conductivity within less than 24 hours and are therefore much more stable against incidental doping , as already stated in ref . .
33ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop @noop @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) in @noop _
_ , ( , ) p. @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) | we report wide - range optical investigations on transparent conducting networks made from separated ( semiconducting , metallic ) and reference ( mixed ) single - walled carbon nanotubes , complemented by transport measurements . comparing the intrinsic frequency - dependent conductivity of the nanotubes with that of the networks ,
we conclude that higher intrinsic conductivity results in better transport properties , indicating that the properties of the nanotubes are at least as much important as the contacts .
we find that hno@xmath0 doping offers a larger improvement in transparent conductive quality than separation .
spontaneous dedoping occurs in all samples but is most effective in films made of doped metallic tubes , where the sheet conductance returns close to its original value within 24 hours . |
a standard method for investigating the evolution of horizontal branch stars is to analyze the period changes of rr lyrae variables in globular clusters .
period increases should indicate redward evolution , while decreases indicate blueward evolution , and the rates of change give information about the time scales .
the models of sweigert & renzini ( 1979 ) and lee _ et al .
( 1990 ) predict that the rr lyrae variables in the oosterhoff type i clusters cross the instability strip during their zahb phase and are therefore expected to have period decreases , which are followed later by increases , when the stars evolve away from the zahb . for the rr lyrae variables in the more metal - poor oosterhoff type ii clusters , the models predict only increasing periods , because these stars cross the instability strip after their zahb phase . since there are a few globular clusters that have been observed over an interval of 100 years , it should be possible to use these observations to test the models .
unfortunately , the observational data do not give definitive results ( cf .
smith 1995 , rathbun & smith 1997 ) .
the evolutionary period changes seem to be masked by a period change @xmath0 of irregular character .
sweigert & renzini ( 1979 ) have demonstrated that this @xmath0 could be caused by mixing events that alter the hydrostatic structure of the core and thus affect the star s pulsation period . because of this , the observed rates of change are sometimes an order of magnitude too large and of the wrong sign , when compared with the rates predicted by evolutionary theory . however , if one assumes that , for a particular cluster , the @xmath1 rate of period change is a measure of the evolutionary change , then it may be possible to compare the observations with theory .
lee ( 1991 ) took this approach and found that the observed period changes of the rr lyrae stars in five well observed clusters could be attributed to evolutionary effects , provided the @xmath0 is random and of the order of @xmath2 days per million years . _ another way to study the evolution of hb stars is to examine the distribution of periods and modes of pulsation of rr lyrae variables in globular clusters of the two oosterhoff types .
some years ago , van albada & baker ( 1973 ) postulated that the difference in the period distributions for the oosterhoff type i and ii clusters is due in part to hysteresis in pulsation . according to their theory
, the rrab variables in the type i clusters enter the instability strip as fundamental mode pulsators ( rrab stars ) and evolve from red to blue , while the stars in type ii clusters enter as first overtone pulsators ( rrc stars ) and evolve from blue to red . because of hysteresis
, the transition period at which a star switches between the fundamental and first overtone modes will be different for the two cluster types , and consequently , the mean periods of both the rrc and rrab stars will be affected as well .
the rr lyrae variables in type ii clusters should switch modes at longer periods and as a result , their mean periods will be longer than those in the type i clusters . also , the oosterhoff type ii clusters should have a higher proportion of rrc stars .
these are the observed characteristics of the two oosterhoff groups .
if the van albada & baker scenario is correct , then the transition between the @xmath3 and @xmath4type rr lyrae variables should occur near the blue edge for fundamental mode pulsation in oosterhoff type i clusters and near the red edge for the first overtone in oosterhoff type ii clusters .
using up - to - date convective pulsation models to locate the instability strip on the hr diagram , bono _
et al . ( 1994 ) and cox ( 1995 ) have presented evidence to indicate that this is , in fact , the case .
_ in the present investigation , we suggest another method for studying hb evolution : monitoring long term changes in the pulsation characteristics of double - mode rr lyrae ( rrd ) stars .
if an rrd star is evolving blueward , then over a period of time , the amplitude of the first overtone should gain in strength relative to fundamental , but if it evolves redward , the strength of fundamental mode oscillations should increase .
our study is motivated by the recent discovery of kaluzny _ et al .
( 1997 , hereafter khcr ) that v79 in m3 , previously classified as rrab , is now an rrd star with the first overtone mode dominating .
the khcr finding was based on observations made in 1996 . _
the 1996 observations of m3 were obtained on nine nights in the interval march 19 to april 2 by one of us ( rwh ) with the 1 meter jacobus kapteyn telescope at the observatorio del roque de los muchachos , la palma .
the observations , which include @xmath5 photometry for 42 rr lyrae variables on 176 frames , and the reduction procedures , have been discussed by khcr .
one of the two previously known rrd variables , v68 , was among the 42 stars and so we have included it in our investigation .
( v87 , the other rrd star was not in our field of view . )
a finding chart for the variables in m3 was published by bailey ( 1913 ) .
we determined the primary periods for v68 and v79 with a computer program that utilized stellingwerf s ( 1978 ) phase dispersion minimization ( pdm ) technique with a ( 5,2 ) bin structure . to search for the secondary period , we derived a mean light curve by fitting a cubic spline interpolating function to the bin means , then measured the residuals from this curve and applied the pdm technique to the residuals .
next , we corrected the magnitudes by subtracting the mean curve for the secondary period from the raw magnitudes and then , again applied the pdm technique to obtain a final value for the primary period .
light curves for v68 and v79 are shown in figures 1 and 2 .
the top panel of each figure shows the ` raw ' magnitudes plotted with the primary period , the first overtone .
the curves in the middle and bottom panels show respectively , the corrected magnitudes plotted with the first overtone period and the residuals plotted with the secondary ( fundamental ) period , both with the interaction frequencies ( @xmath6 ) also removed .
to assess the long term behavior of these stars , we compared our results with those of an earlier study of the rrd stars in m3 by nemec & clement ( 1989 , hereafter nc ) .
the nc study was based mainly on data from a combination of three sets of observations made in the interval 1920 to 1926 and published by larink ( 1922 ) , muller ( 1933 ) and greenstein ( 1935 ) .
nc found that v68 was an rrd star with the fundamental mode dominant , but they considered v79 to be an rrab star .
the light curve of v79 ( see figure 2 of nc ) showed night - to - night scatter , but a pdm period search of the residuals did not reveal any periodicity in the range expected for first overtone oscillations ( see figure 3 of nc ) .
they estimated that if there were any first overtone oscillations , they would have an amplitude less than @xmath7 mag .
we summarize the pulsation characteristics of v68 and v79 in table 1 . in columns 2 and 3 ,
we list their co - ordinates in arcseconds relative to the cluster center according to sawyer hogg s ( 1973 ) catalogue , in columns 4 to 8 , we list the first overtone and fundamental periods ( @xmath8 and @xmath9 ) , their corresponding amplitudes ( @xmath10 and @xmath11 ) and amplitude ratios derived from the 1996 observations and in the final column , we list the amplitude ratios for 19201926 based on nc s study .
the data of table 1 illustrate that for both stars , the first overtone oscillations have grown in strength since the 1920s , but for v79 , the change is more striking . in order to find out what happened to v79 in the intervening years , we used the pdm technique to analyze other published observations that were suitable for period searches . these included observations obtained in the intervals 1938 to 1962 ( szeidl 1965 ) and 1946 to 1948 ( belserene 1952 ) . in the pdm technique , a @xmath12 statistic which is a measure of the scatter on the light curve
is evaluated for a series of periods and the period for which @xmath12 is a minimum is considered to be the best period . in figure 3
we show plots of @xmath12 versus period for the raw data over the range 034 to 050 , for the four epochs .
the plot for the 1996 data in the bottom panel of the figure indicates that the best period is 0358 .
however , the plots for the first three epochs all show a clear minimum at a period of approximately 04833 and period searches of the residuals measured from their corresponding primary light curves did not reveal any oscillations in the overtone mode even though there were variations in the amplitude .
our analysis also indicates that the fundamental period of v79 has decreased by a substantial amount in the last 35 years .
the @xmath13 day period derived from the previous studies is significantly longer than the fundamental period @xmath140480 that we determined for the 1996 observations and listed in table 1 .
if we plot the 1996 data with the longer period , it introduces a phase shift that increases the scatter on the light curve in the bottom panel of figure 2 .
we therefore conclude that two things have happened to v79 between 1962 and 1996 .
first overtone oscillations have either commenced or increased significantly and there has been an abrupt decrease in the fundamental period . in his study of period changes of the rr lyrae variables in m3 ,
szeidl ( 1965 ) found that the @xmath15 diagram for v79 had a discontinuity between 1926 and 1938 and consequently , he was unable to detect any systematic period change for the star .
rathbun & smith ( 1997 ) cited v79 as an example of a star that has `` period changes so erratic as to be impossible to even approximately describe with a single number . ''
if the large abrupt decrease we have detected in @xmath9 the fundamental period of v79 was caused by evolution on the horizontal branch , then an @xmath16 diagram is not a useful tool for studying evolutionary period changes of rr lyrae variables , at least not for stars like v79 in m3 .
since , prior to 1962 , v79 was an rrab star with an irregular light curve : a ` blazhko ' variable , our investigation provides some insight into the actual cause of the blazhko effect in rr lyrae variables . according to smith ( 1995 ) ,
most speculation has centered on two possibilities : ( 1 ) that the effect is a `` consequence of some type of mixing of pulsational modes '' and ( 2 ) that the effect is `` related to magnetic cycles in the stars , perhaps coupled with rotation '' . since we now know that at least one ` blazhko ' variable , v79 in m3 , has exhibited mixed mode pulsations , we consider the first possibility to be more feasible .
it appears that regular monitoring of stars like v79 in m3 can provide important information about stellar evolution .
if there are any available unpublished observations of this star between 1962 and 1996 , it would be very informative to analyze them
. it would be particularly interesting to know if there has been any change in luminosity or color because this may put constraints on the boundaries of the different pulsational modes in the hr diagram .
we should also point out that there are other interesting stars in m3 that merit further investigation .
for example , nc noted that the rrab star v28 was a promising rrd candidate .
unfortunately , it was not in the field of view for our present study .
perhaps v28 has also started to switch modes or will do so in the near future .
it would be interesting to find out .
we have presented evidence that indicates that two rrd stars in m3 are evolving blueward .
are the rrd stars in oosterhoff type ii systems evolving redward ?
this question was addressed by purdue _
1995 ) in a study of the long term behavior of rrd stars in the oosterhoff type ii cluster m15 .
they noted that the fundamental mode oscillations in v30 increased in strength relative to the first overtone between 1941 and 1991 .
in fact , their analysis suggests that the change happened rather abruptly in the 1950s .
a similar situation occurs for the field rrd star , aq leo , which has periods and a period ratio similar to those of rrd stars in oosterhoff type ii systems .
et al . ( 1982 ) compared the amplitude differences between a series of observations made in 1960 - 1961 and another in 1973 - 1974 and found that the first overtone amplitude decreased by an amount @xmath17 while the fundamental mode amplitude increased by @xmath18 .
they also found that an abrupt increase in the first overtone period occurred in the early 1970s .
they stated that their findings did not give direct evidence for mode switching , but that it is a possibility .
the increase in strength of the fundamental mode oscillations and the period increase are events that are expected to occur if the star is evolving redward . _
_ our investigation has shown that high quality photometry of rr lyrae stars , particularly rrd stars and stars that exhibit the blazhko effect can provide useful information about the evolution of horizontal branch stars .
we expect that future studies will confirm this .
this work has been supported by the natural sciences and engineering research council of canada through operating grants to cmc and smr .
rwh thanks the u. k. particle physics and astronomy research council for the award of a research grant .
jk was supported by polish kbn grant 2p03d011.12 and by the nsf grant ast 9528096 to bohdan paczyski .
bailey , s. i. 1913 , harv .
78 belserene , e. p. 1952
, , 57 , 237 bono , g. , caputo , f. & stellingwerf , r. f. 1994 , , 423 , 294 cox , a. n. 1995 , in astrophysical applications of powerful new databases , asp conference series 78 , ed . s. j. adelman & w. l. wiese ( asp ) , 243 greenstein , j. l. 1935 , astron .
257 , 301 jerzykiewicz , m. , schult , r. h. & wenzel , w. 1982 , acta a , 32 , 357 kaluzny , j. , hilditch , r. , clement , c. & rucinski , s. m. 1997 , submitted to ( khcr)= astro - ph/9709002 larink , j. 1922 , astron .
bergedorf 2 , no
. 6 lee , y .- w .
1991 , , 367 , 524 lee , y .- w . , demarque , p. & zinn , r. 1990 , , 350 , 155 muller , th .
1933 , veroff .
berlin - babels .
11 , 1 nemec , j. m. and clement , c. m. 1989 , , 98 , 860 ( nc ) purdue , p. , silbermann , n. a. , gay , p. & smith , h. a. 1995 , , 110 , 1712 rathbun , p. g. & smith . h. a. 1997 , , ( october issue ) sawyer hogg , h. s. 1973 , publ .
ddo 6 , no .
smith , h. a. 1995 , rr lyrae stars ( cambridge university press ) stellingwerf , r. f. , 224 , 953 sweigert , a. v. & renzini , a. 1979 , , 71 , 66 szeidl , b. 1965 , mitt .
budapest - szabadsaghegy no.58 van albada , t. s. & baker , n. h. 1973 , , 185 , 477 | new observations of variable stars in the globular cluster m3 reveal that the rr lyrae variable v79 is a double - mode ( rrd ) variable with the first overtone mode dominating . in all previous studies , v79 was found to be a fundamental mode ( rrab ) pulsator with an irregular light curve .
this is the first observed mode switch for an rr lyrae variable and it is direct observational evidence for blueward evolution of horizontal branch stars in the oosterhoff type i cluster m3 .
it also demonstrates that there is a connection between the blazhko effect and pulsational mode mixing in rr lyrae variables .
these new observations also show that the strength of the overtone oscillations in the rrd star v68 in m3 may have increased in the last 70 years , thus indicating blueward evolution for v68 as well .
a survey of previously published investigations of rrd stars in oosterhoff type ii systems indicates that there is marginal evidence for an increase in the strength of fundamental mode oscillations in two stars : v30 in m15 and aq leo .
if these increases are confirmed by future observations , it will indicate redward evolution for rrd stars in type ii systems . |
@xmath0 it has been commonly held that black holes are thermodynamic systems @xcite .
black holes obey four laws of black hole mechanics analogous to the four laws of ordinary thermodynamics , posing a bekenstein - hawking entropy and a characteristic hawking temperature related to the surface gravity of event horizon @xcite .
finding the underlying microscopic description of this entropy is one of the most challenging subjects in theoretical physics , but it still remains obscure , best left for future development of quantum theory of gravity @xcite .
thermodynamic fluctuation theory , whose basic goal is to express the time independent probability distribution for the state of a fluctuating system in terms of thermodynamic quantities , is usually attributed to einstein who applied it to the problem of blackbody radiation @xcite .
however , despite a wide range of applications , the classical fluctuation theory fails near critical points and for volumes in the order of the correlation volume or less . in 1979
, ruppeiner @xcite introduced a riemannian metric structure representing thermodynamic fluctuation theory , and related it to the second derivatives of the entropy .
his theory offered a good meaning for the distance between thermodynamic states .
he showed that the breakdown of the classical theory was due to its failure to take account of local correlations @xcite .
one of the most significant topics of this theory is the introduction of the riemannian thermodynamic curvature as a qualitatively new tool for the study of fluctuation phenomena .
this curvature has a possible relationship with the interactions of the underlying statistical system as proposed by ruppeiner @xcite in contrast with single - component ideal gas which is a non - interacting system with zero ruppeiner curvature .
earlier , in 1975 , weinhold @xcite had proposed an approach which was based on a sort of riemanian metric defined as the hessian of the internal energy of a given system , @xmath1 , where derivatives are taken with respect to the extensive thermodynamic variables and entropy . in 1984 , mrugaa @xcite and salamon @xcite proved that these two metrics are conformally equivalent with the inverse of the temperature , @xmath2 , as the conformal factor @xmath3 since then , geometrical approaches have been intensively used to study the chemical and physical properties of various thermodynamic systems @xcite , first applied to black holes by ferrara _
@xcite to discuss the critical behavior in moduli space .
later , some authors used ruppinier geometry to study phase space , critical behavior , and stability of various types of black hole families @xcite . recently , in a series of papers @xcite , aman _
et al . _ derived ruppeiner and weinhold scalars for reissner - nordstrm ( rn ) , kerr , and btz black holes in arbitrary four dimensional space times .
one of their results is that the ruppeiner curvature is zero for rn black holes of any dimensions ; however , for rn - ads and kerr black holes thermodynamic spaces are non - trivial and curvatures diverge at extremal limits @xcite . on the other hand , it seems that a non - interacting background can not produce the thermodynamic behavior of the rn black holes , so efforts were made to redefine ruppeiner geometry using new massiue functions @xcite .
furthermore , if one uses the ruppeiner curvature , as proposed in @xcite , to find a lower volume as the limit of the applicability of classical thermodynamics , zero curvature will tell us that there is no such limit .
this seems unacceptable because we expect classical description to fail at least at the planck scale . in this paper
, we propose that a better measure of microscopic interactions will be obtained if one uses the complete phase space of extensive variables .
according to this view , in calculating ruppeiner curvature for rn black holes , one should set @xmath4 limits in the scalar curvature of the kerr - newmann - ads ( kn - ads ) black hole , which is the most general solution of einstein - maxwell field equations in an anti - de sitter background .
the ruppeiner curvature becomes a non - zero function of @xmath1 and @xmath5 and diverges at extremal limits .
this also indicates that classical thermodynamics could not be used for black holes of the size of planck length .
we believe that this is a general property which is related to the nature of fluctuation theory and is applicable for btz black holes that also have a flat ruppeiner curvature in aman s method @xcite .
furthermore , using the quasilocal thermodynamic parameters shows that our results do not contingent upon using of adm parameters .
the structure of the paper is as follows : in section ii , we recall the charged rotating solution of the einstein - maxwell - anti - desitter equations .
we also recall the results of generalized smarr formula for this type of black holes and derive the ruppeiner curvature . in section iii and iv , we discuss the ideas for the rn and kerr families and discuss the phase transition points and their thermodynamic stability . in section
v , we study the ruppeiner geometry of kn and rn black holes by using the quasilocal thermodynamic quantities . throughout this paper
, we use the natural units @xmath6 and also set @xmath7 .
here we consider the usual charged rotating black hole in ads space .
its metric , which is axisymmetric , reads in boyer - lidquist type coordinates @xmath8 ^ 2 + { \rho^2\over\delta_r}\ dr^2+{\rho^2\over\delta_\theta}\ d\theta^2 + { \delta_\theta\sin^2\theta\over\rho^2 } \left[a\ dt-\frac{r^2+a^2}{\xi}\ d\phi\right]^2\ ] ] where @xmath9 @xmath10 here @xmath11 denotes the rotational parameter , @xmath12 is the electric charge , and @xmath13 is defined by @xmath14 , where @xmath15 is the cosmological constant .
the mass @xmath1 , charge @xmath5 and the angular momentum @xmath16 can be defined by means of komar integrals as @xmath17 from these relations , one can obtain a generalized smarr formula for kerr - newmann - ads black holes , which reads @xcite @xmath18 @xmath19 , @xmath16 , @xmath20 and ( [ smarr2 ] ) will be regarded as the complete set of energetic extensive parameters and the black hole thermodynamic fundamental relation , @xmath21 conjugate variables to @xmath22 , @xmath5 and @xmath16 could be obtained , for example , the hawking temperature is defined as : @xmath23 .
extremal black holes have a zero hawking temperature ; so , @xmath24 is used to calculate the extremal surface : @xmath25 .
now , components of the ruppeiner metric could be derived @xmath26 @xmath27 @xmath28 @xmath29 @xmath30 @xmath31 @xmath32 furthermore , by definition of the ricci scalar @xcite , the ruppeiner curvature could be calculated directly by using grtensor ii or grtensor m packages .
the expression is too complicated to present here , so the results are depicted numerically .
the ruppeiner curvature diverges at the extremal limit , @xmath33 , and along the surface @xmath34 where metric changes its sign ( a signature of thermal instability ) . for different combinations of @xmath22 and @xmath5 ,
each @xmath16 function has a double root , so at four physically acceptable values of @xmath16 , the scalar curvature diverges .
the ruppeiner geometry is not flat ( fig .
1 ) , a fact that could also be checked by the cotton - york tensor .
up untill now , we have been considering the rn case which is our main interest , because of the works by aman
_ et al . _
@xcite , indicating that this family of black holes has a trivial thermodynamic geometry ( zero curvature ) and so has a non - interacting underlying statistical system .
this can be an interesting result which may serve as a guide in looking for an appropriate statistical model for black holes in loop quantum gravity or string theory .
on the other hand , it seems that the physical structure of rn black holes could not be reproduced by such a simple underlying system .
furthermore , the absence of the divergent points set it in a different class from other types of black holes .
if rn systems are viewed as the limit of kn - ads black holes by setting @xmath35 and @xmath36 in the ruppeiner curvature calculated for kn - ads case , @xmath37 will takes the form @xmath38 + we now have a new non - zero ruppeiner scalar which diverges at the extremal limits : @xmath39 ( fig .
this observation is in agreement with @xcite , which used the second moments of the fluctuations in the fluxes of energy and angular momentum .
it was shown that phase transitions occur only at the extremal limits , the points at which a black hole changes its nature to a naked singularity , a new phase .
it should be noted that this behavior is related to the nature of the ruppeiner geometry .
if we initially work with spinless black holes @xmath40 we will neglect the fluctuations of parameter @xmath16 in all calculations , which will lead to a wrong flat geometry ( @xmath41 ) .
therefore , the above non - vanishing scalar curvature is a result coming from another dimension specified by @xmath16 which fluctuates even if we set it to zero .
we see that by setting @xmath42 , the metric elements of kn - ads state space ( eqs .
7 - 13 ) reduce to the rn ones ( eq .
15 ) but the curvature does not behave in the same manner .
this behavior originated from the existence of an extra dimension in the parameter space .
when the extremal limit is approached , however , the curvature diverges strongly .
another difference is also observed from reports in certain works on rn black holes phase space : the absence of davies phase transition points. this relates to the difference between ordinary thermodynamics of extensive systems and that of black holes , which constitute non - extensive , non - additive thermodynamic systems because of the well known scaling of the black hole entropy with area rather than with volume .
however , the interpretation of the divergence in specific heats as phase transitions is not settled and has been the subject of much debate @xcite .
ruppeiner formalism relies on the usual thermodynamic properties of extensive systems but it can also be applied to black holes near the divergent points @xcite .
so we can use the ruppeiner method as a probe to find phase transitions .
furthermore , ruppeiner proposed that @xmath43 sets the limiting lower volume in which the classical fluctuation theory provides a good approximation @xcite . for black holes
, it could be obtained from ( 7 ) that even for a non - charged black hole , ruppeiner curvature has a @xmath44 remanent .
replacing @xmath22 by @xmath45 , where @xmath46 is the entropy per volume , and leaving the choice of natural units , one could derive a lower volume , @xmath47 ( in which @xmath48 is the schwarzchild radius ) , above which the classical thermodynamic regime could be used .
this also indicates that if the schwarzchild radius of the black hole reaches the planck length , @xmath49 , the classical thermodynamic description will break down .
still another problem is that of stability . in order to determine the points where a change of thermodynamic stability occurs
, we use poincar turning point method . according to arcioni and lozano - tellechea @xcite , by plotting the conjugate variables ( @xmath50 ) in according to their extensive parameters ( @xmath51 )
one could see the points where black hole changes its stability ( figs .
3a & 3b ) .
there is no turning point and , therefore , no changes of stability are shown along the plot .
the first diagram shows that @xmath2 has a minimum at @xmath52 ( by setting @xmath53 ) .
as described in @xcite , this is not a measure of changing stability , so nothing special happens at this point ( davies point ) .
aman _ et al . _ also calculated the ruppeiner and weinhold s curvatures for rotating black holes against a flat background .
they reported a vanishing weinhold curvature for kerr - type black holes but also noted that the physical meaning of this property was not clear which they referred to as the _ ad hoc _ definition of weinhold geometry @xcite .
using the definition of weinhold s metric @xcite for the complete phase space of parameters ( kn - ads black hole ) , and taking the limits @xmath54 , @xmath35 , a non vanishing weinhold curvature is derived .
the phase transition points could be obtained by calculating the ruppeiner curvature @xmath55 @xmath56 diverges at the extremal limits : @xmath57 ( fig .
further , one could easily check the instabilities by using the poincar diagrams : @xmath58 and @xmath59 .
the plots do not show any turning points ( and thus no instability ) as indicated before in @xcite .
in this section we replace our definition of the internal energy of black holes by quasilocal energy proposed by brown and york @xcite which is derived from the hamiltonian of spatially bounded gravitational systems .
the brown - york derivation of the quasilocal energy , as applied to a four - dimensional ( 4d ) spacetime solution of einstein gravity can be summarized as follows .
the system one considers is a 3d spatial hypersurface @xmath60 bounded by a 2d spatial surface @xmath61 in a spacetime region that can be decomposed as a product of a 3d hypersurface and a real line - interval representing time .
the time - evolution of the boundary @xmath61 is the surface @xmath62 .
one can then obtain a surface stress - tensor on @xmath62 by taking the functional derivative of the action with respect to the 3d metric on @xmath62 .
the energy surface density is the projection of the surface stress - tensor normal to a family of spacelike surfaces such as @xmath61 that foliate @xmath62 .
the integral of the energy surface density over such a boundary @xmath61 is the quasilocal energy associated with a spacelike hypersurface @xmath22 whose orthogonal intersection with @xmath62 is the boundary @xmath61 .
it is assumed that there are no inner boundaries , such that the spatial hypersurfaces @xmath22 are complete . in the case where horizons form , one simply evolves the spacetime inside as well as outside the horizon . under these conditions , the qle ( quasi local energy ) is defined as : @xmath63 where @xmath64 is the determinant of the 2-metric on @xmath61 , @xmath65 is the trace of the extrinsic curvature of @xmath61 , and @xmath66 is a reference term that is used to normalize the energy with respect to a reference spacetime , not necessarily flat . to compute the qle for asymptotically flat solutions , one can choose the reference spacetime to be flat as well . in that case
, @xmath66 is the trace of the extrinsic curvature of a two - dimensional surface embedded in flat spacetime , such that it is isometric to @xmath61 . for spacetimes that are asymptotically flat in spacelike directions , the quasilocal energy and angular momentum defined there agree with the results of arnowitt , deser and misner in the limit that the boundary tends to spatial infinity @xcite .
we have used adm parameters in the above discussions .
now , adm mass ( @xmath1 ) is replaced by quasilocal energy of kn black holes @xmath67+o(\frac{a^2}{r_0})\ ] ] where @xmath68 and @xmath69 is the radius of the bounding surface . here
we only consider the small electrical charge and slow rotating regime @xmath70 so the definition of @xmath5 and @xmath16 do not have to change . inserting @xmath71 in ( 18 ) one could derive a new fundamental equation for quasilocal parameters @xmath72+o(\frac{a^2}{r_0})\ ] ] now the ruppeiner metric components could be calculated @xmath73 for rn black holes ( 2-dimensional state space ) curvature is a function of @xmath69 , such that @xmath74 for kerr - newmann black holes ( full thermodynamic state space ) the calculated curvature is different from @xmath75 for any finite value of @xmath69 ( they are too large to be written here ) and has the following limiting form : @xmath76 as expected this limit equals to ( 14 ) .
the above considerations show that our results do not contingent upon computing all thermodynamic quantities at infinity .
usual statistical mechanics , augmented by renormalization group theory , are used to build up from the microscopic physics and deduce information about the macroscopic world . for the black holes ,
the microscopic description needs a quantum theory of gravity which is still missing .
in contrast , the covariant thermodynamic fluctuation theory builds down from the macroscopic equation of states using the riemannian geometry @xcite . in using this method
one must be careful to consider all possible physical fluctuations because neglecting one parameter may lead to inadequate information about the model .
for the rn case , the ruppeiner curvature could be obtained using the complete set of physical fluctuating parameters .
the resulting ruppeiner curvature is not zero , behaves like other types of black holes , and could be used to set a lower bound , @xmath77 , beyond which the classical thermodynamic description breaks down as expected .
the method proposed here could also be applied to other cases , e.g. , as shown by henneaux and teitelboim @xcite , it is possible to promote the cosmological constant to a thermodynamic state variable which may be the subject of future study . | in some recent studies @xcite , aman _
et al . _ used the ruppeiner scalar as a measure of underlying interactions of reissner - nordstrm black holes , indicating that it is a non - interacting statistical system for which classical thermodynamics could be used at any scale . here , we show that if we use the complete set of thermodynamic variables , a non - flat state space will be produced .
furthermore , the ruppeiner curvature diverges at extremal limits , as it would for other types of black holes . |
in this paper , we consider the following unconstrained integer optimization problem @xmath0 where , @xmath1 , @xmath2 , @xmath3 .
+ generally speaking ,
the above optimization problem is np - hard , which can not be solved in polynomial time ( by enumeration , there exist @xmath4 choices ) .
a direct method is to adopt the so called divide - and - conquer " strategy , which separates the optimization problem into several subproblems and then solve these subproblems step by step .
branch and bound ( b&b ) , branch and cut ( b&c ) , and branch and price ( b&p ) belong to this kind ; however , these methods are essentially in exponential time .
an indirect method is to relax the optimization problem by loosening its integrality constraints to continuity and then solve the continuous relaxation problem or its lagrangian dual problem , including lp - based relaxation , sdp - based relaxation , lagrangian relaxation , etc .
nevertheless , when rounding off the relaxation solution , there may cause some infeasibility or can only get approximate solution , and when using lagrangian dual , there may exist duality gap between the primal and the dual problem @xcite .
+ on the other hand , some stochastic algorithms , such as genetic algorithm ( ga ) @xcite , simulated annealing ( sa ) @xcite , ant colony optimization ( aco ) @xcite , are also widely used for combinatorial optimization problems , which aim to obtain good solutions " in reasonable time . in terms of the concepts of state and state transition , a new heuristic search algorithm called state transition algorithm ( sta ) has been proposed recently , which exhibits fantastic results in continuous function optimization @xcite . in @xcite ,
three intelligent operators ( geometrical operators ) named swap , shift and symmetry have been designed for discrete sta to solve the traveling salesman problem ( tsp ) , and it shows that the discrete sta outperforms its competitors in respect to both time complexity and search ability . to better develop discrete sta for medium - size or large - size discrete optimization problems , in the study , we firstly build the framework of discrete state transition algorithm and propose the five key elements for discrete sta , of which , the representation of a decision variable in discrete sta , the local and global operators and the dynamic adjustment strategy are mainly studied .
four geometrical operators named swap , shift , symmetry and substitute are designed , which are intelligent due to their adaptability in various types of integer programming .
the mixed strategies of greedy criterion " and risk and restore in probability " are proposed , in which , greedy criterion " and restore in probability " are used to guarantee good convergence performance , and
risk a bad solution in probability " aims to escape from local optimality .
some applications ranging from traveling salesman problem to discrete value selection problem are studied .
if a solution to a specific optimization problem is described as a state , then the transformation to update the solution becomes a state transition . without loss of generality ,
the unified form of discrete state transition algorithm can be described as @xmath5 where , @xmath6 stands for a current state , corresponding to a solution of a specific optimization problem ; @xmath7 is a function of @xmath8 and historical states ; @xmath9 , @xmath10 are transformation operators , which are usually state transition matrixes ; @xmath11 is a operation , which is admissible to operate on two states ; @xmath12 is the cost function or evaluation function .
+ the first element we want to mention is the representation of a solution to a discrete optimization problem . in discrete sta , we choose special representations , which can be easily manipulated by the intelligent operators .
why we call the operators intelligent " due to its geometrical property ( swap , shift , symmetry and substitute ) , and a intelligent operator has the same geometrical function for different representations . the big advantage of such representations and operators is that , after each state transformation , the newly created state is always feasible , avoiding the trouble in dealing with constraints .
+ the second important thing in state transition algorithm is that , when a transformation operator is exerted on current state , the next state is not deterministic , that is to say , there are possibly different choices for the next state .
it is not difficult to imagine that all possible choices will constitute a set , or a neighborhood .
then we execute several times of transformation ( search enforcement @xmath13 ) on current state , to sampling in the neighborhood .
sampling is the first key element in state transition algorithm , which can reduce the search space and avoid enumeration .
+ as a intelligent optimization algorithm , the discrete state transition algorithm also comprises the following key elements : + ( 1 ) local exploitation and global exploration . in optimization algorithms
, it is very significant to design good local and global operators .
the local exploitation can guarantee high precision of a solution and convergent performance of a algorithm , and the global exploration can avoid getting trapped into local minima or prevent premature convergence . in discrete optimization , it is extremely difficult to define a good " local optimal solution due to its dependence on a problem s structure , which leads to the same difficulty in the definition of local exploitation and global exploration .
anyway , in the discrete state transition algorithm , we define the slow change to current solution by a transformation as local exploitation , while the big change to current solution by a transformation as global exploration .
+ ( 2 ) self learning and regular communication .
state transition algorithm behaves in two forms , one is individual - based , the other is population - based , which is certainly a extended version .
the individual - based state transition algorithm focuses on self learning , in other words , with emphasis on the operators designing and dynamic adjustment ( details given in the following ) .
undoubtedly , communication among different states is a promising strategy for state transition algorithm , as indicated in @xcite . through communication
, states can share information and cooperate with each other .
however , how to communicate and when to communicate are key issues . in continuous state transition algorithm , intermittent exchange strategy was proposed , which means that states communicate with each other at a certain frequency in a regular way .
+ ( 3 ) dynamic adjustment .
it is a potentially useful strategy for state transition algorithm . in the iteration process of a algorithm ,
the fitness value can decrease sharply in the early stage , but it stagnates in the late stage , due to the static environment . as a result
, some perturbation should be added to activate the environment .
in fact , dynamic adjustment can be understood and implemented in various ways .
for example , the alternative use of different local and global operators is dynamic adjustment to some extent .
then , we can change the search enforcement , vary the cost function , reduce the dimension , etc .
of course , risk a bad solution in probability " is another dynamic adjustment , which is widely used in simulated annealing ( sa ) . in sa ,
the metropolis criterion @xcite is used to accept a bad solution : @xmath14 where , @xmath15 , @xmath16 is the boltzmann probability factor , @xmath17 is the temperature to regulate the process of annealing . in the early stage
, temperature is high , and it has big probability to accept a bad solution , while in the late stage , temperature is low , and it has very small probability to accept a bad solution , which is the key point to guarantee the convergence .
we can see that the metropolis criterion has the ability to escape from local optimality , but on the other hand , it will miss some good solutions " as well .
+ in this study , we focus on the individual - based sta , and the main process of discrete sta is shown in the pseudocode as follows + as for detailed explanations , swap function in above pseudocode is given as follows for example from the pseudocodes , we can find that in the simple sta , only greedy criterion " is adopted to accept a new @xmath18 " , while in dynamic sta , in the whole , greedy criterion " is adopted to keep the incumbent
@xmath18 " , in the partial , a bad solution
@xmath19 " is accepted in each state transformation at a probability @xmath20 , and in the same while , the
@xmath18 " is restored in the process at another probability @xmath21 .
the risk a bad solution in probability " strategy aims to escape from local optimal , while the greedy criterion " and restore the incumbent best solution in probability " are to guarantee a good convergence .
firstly , we define the concept of convergence for discrete sta @xmath22 is a global optimum of a discrete optimization problem , @xmath23 is a small constant , and @xmath24 is a natural number .
if @xmath25 , we can say that the algorithm converges to a @xmath26 ; if @xmath27 , we can say that the algorithm converges to a _
global minimum_. however , if @xmath28 is a local minimum solution , then we can say that the algorithm converges to a @xmath26 and a _ local minimum _ for @xmath25 and @xmath27 , respectively .
+ * theorem 1 * the discrete sta can at least converge to a _
local minimum_. + _ proof_. let suppose the maximum number of iterations ( denoted by _ m _ ) is big enough , considering that the greedy criterion " is used to keep the incumbent best , then there must exist a number @xmath29 , when @xmath30 , no update of better solution will happen .
that is to say , @xmath31 , where @xmath32 is the solution in the _
_ n__th iteration .
denote @xmath32 as the local minimum solution @xmath28 , and then @xmath33 . + * theorem 2 * the discrete sta can converge to a _
global minimum _ in probability . + _ proof_. let suppose @xmath34 is a global minimum solution , and @xmath35 is the _ _ n__th best solution .
if @xmath36 , according to the greedy criterion " , @xmath31 , which means that it converges to @xmath28 . otherwise , there must exist a transformation , either swap , shift , symmetry or substitute , such that @xmath37 , which means that after @xmath38 iterations , @xmath39 will be changed into @xmath40 .
if @xmath41 , the @xmath42 is kept as incumbent best for next iteration , else , @xmath42 is also kept as incumbent best for next iteration in probability . following the similar way , after at most @xmath43 iterations
, @xmath44 will be @xmath45 in probability .
+ from the proof of theorem 2 , we can find that dynamic sta has higher probability than the simple sta to gain the global optimum .
some applications are given to describe the details in the following .
suppose @xmath46 is the set of cities , the traveling salesman problem can be described as : given a set of _ n _ cities and the distance @xmath47 for each pair of cities @xmath48 and @xmath49 , find a roundtrip of minimal total length visiting each city exactly once .
typically , the traveling salesman problem is usually modeled as the following two representations @xcite .
+ ( lp - tsp ) : @xmath50 where , the decision variable @xmath51 is defined by x_ij= 1 , & if city _
i _ is followed by city _
j _ + 0 , & otherwise ( qp - tsp ) : @xmath52 here , the decision variable @xmath51 is defined by x_ij= 1 , & if city _
i _ is in the _ _
j__th position + 0 , & otherwise we can find that the model of linear programming ( lp ) based tsp is different from the model of quadratic programming ( qp ) based tsp in two aspects , one is the definition of the decision variable @xmath51 , the other is that ( lp - tsp ) has one more constraint than ( qp - tsp ) . taking the difficulty of dealing with constraints into consideration , in discrete sta , we use another simple representation which can be easily manipulated by intelligent operators . as for a _
n_-city traveling salesman problem , a permutation of @xmath53 is used to represent a solution to the problem in discrete sta .
based on the representation , three special transformation operators are proposed to illustrate local and global search .
+ ( 1 ) swap transformation @xmath54 where , @xmath55 is called swap transformation matrix , @xmath56 is a constant integer called swap factor to control the maximum number of positions to be exchanged , while the positions are random . if @xmath57 , we call the swap operator local exploitation , and if @xmath58 , the swap operator is regarded as global exploration in this case .
figure [ swaptsp ] gives the function of the swap transformation graphically when @xmath57 .
+ ( 2 ) shift transformation @xmath59 where , @xmath60 is called shift transformation matrix , @xmath61 is a constant integer called shift factor to control the maximum length of consecutive positions to be shifted . by the way , the selected position to be shifted after and positions to be shifted are chosen randomly . similarly , shift transformation is called local exploitation and global exploration when @xmath62 and @xmath63 respectively . to make it more clearly , if @xmath62 , we set position 3 to be shifted after position 5 , as described in figure [ shifttsp ] .
+ ( 3 ) symmetry transformation @xmath64 where , @xmath65 is called symmetry transformation matrix , @xmath66 is a constant integer called symmetry factor to control the maximum length of subsequent positions as center . by the way , the component before the subsequent positions and consecutive positions to be symmetrized are both created randomly . considering that the symmetry transformation can make big change to current solution , it is intrinsically called global exploration . for instance , if @xmath67 , let choose component 3 , then the subsequent position or the center is \{@xmath68 } , the consecutive positions are @xmath69 , and the function of symmetry transformation is given in figure [ symtsp ] . to evaluate the performance of the simple sta and the dynamic sta ( dsta ) ,
4 medium - size tsp instances are used for test .
we set @xmath70 and the maximum number of iterations at 1500 .
programs are run independently for 20 trails for each algorithms in matlab r2010b ( version of 7.11.0.584 ) on intel(r ) core(tm )
i3 - 2310 m cpu @2.10ghz under window 7 environment , and for each run , it costs about 0.5 minutes .
some statistics as well as the error " are computed for comparison , where , error " is defined by @xmath71 where , @xmath72 is the best result achieved by discrete sta , @xmath73 is the incumbent best result in tsplib @xcite .
experimental results for these tsp instances are given in table [ tspinstances ] .
ccccccc _ instance _ & _ optimum _ & _ algorithm _ & _ best _ & _ mean _ & _ s.t .
_ & _ error _
+ gr96.tsp & 512.3094 & sta & 525.0124 & 544.9243 & 13.2005 & 2.48% + & & dsta & 515.5143 & 549.9838 & 17.1709 & 0.63 % + kroa100.tsp & 2.1285e4 & sta & 2.2214e4 & 2.2970e4 & 515.9307 & 4.37% + & & dsta & 2.1910e4 & 2.2740e4 & 700.3377 & 2.94% + kroc100.tsp & 2.0751e4 & sta & 2.1570e4 & 2.2525e4 & 679.3821 & 3.95% + & & dsta & 2.1084e4 & 2.2186e4 & 703.0488 & 1.60% + gr120.tsp & 1.6665e3 & sta & 1.6826e3 & 1.7269e3 & 30.0698 & 0.97% + & & dsta & * 1.6418e3 * & 1.7257e3 & 34.4110 & * -1.48% * + + as can be seen from table [ tspinstances ] , the biggest error is no more than 5% , and the dsta gets much more better results than the sta . especially for the gr120.tsp , as emphasized by the bold font , dsta achieves the better result than the incumbent best result given in the tsplib , with the sequence of ( 82 11 51 23 9 103 119 4 38 7 56 41 87 14 75 44 46 20 50 98 42 17 118 49 13 113 69 65 68 91 58 79 100 33 52 25 108 43 107 18 19 117 31 66 85 22 81 94 86 78 15 59 76 1 16 61 29 120 32 30 92 28 45 74 105 72 40 60 24 111 96 90 54 8 70 116 34 26 71 47 55 6 89 112 110 48 102 101 114 106 104 36 84 35 10 99 62 37 67 83 39 57 73 63 77 95 12 97 88 21 115 2 93 109 64 53 5 27 80 3 ) and length of route at 1.6418e3 , which is illustrated in figure [ gr120 ] .
in boolean integer programming ( bip ) , a solution comprise a series of boolean values ( @xmath74 or @xmath75 ) .
swap , shift and symmetry operators are also proposed for internal transformation ( operators aiming to change the internal components of a sequence ) , and another operator called substitute is designed for external transformation ( operator aiming to bring new components into a sequence ) . it should be noted that @xmath74 is the same to @xmath75 under such circumstances , although there exists a linear transformation relationship between them in other studies . as we said previously , the same intelligent operator has the same geometrical property for different applications .
it is not difficult to imagine that the swap , shift and symmetry operators for boolean integer programming have the same formulation as that of traveling salesman problem .
let @xmath74 , the illustrations of internal transformation are given from figure [ bswap ] to figure [ bsymmetry ] . + now , let introduce the external transformation .
+ ( 4 ) substitute transformation @xmath76 where , @xmath77 is called substitute transformation matrix , @xmath78 is a constant integer called substitute factor to control the maximum number of positions to be substituted . by the way , the positions are randomly created .
if @xmath79 , we call the substitute operator local exploitation , and if @xmath80 , the substitute operator is regarded as global exploration in this case .
figure [ bsubstitute ] gives the function of the substitute transformation vividly when @xmath79 .
let @xmath81 be an undirected graph with edge weight @xmath82 on @xmath83 vertices and @xmath84 edges , for each edge @xmath85 , the maximum cut problem ( max - cut ) is to find a subset @xmath86 of the vertex set @xmath87 such that the total weight of the edges between @xmath86 and its complementary subset @xmath88 is as large as possible .
+ ( lp based max - cut model)@xcite : + considering a variable @xmath89 for each edge @xmath85 , and assuming @xmath89 to be 1 if @xmath90 is in the cut , and 0 otherwise , the max - cut can be modeled as the following linear programming ( lp ) optimization problem : @xmath91 here the incidence vector @xmath92 , where the @xmath93 is the number of edges in the graph .
+ let cut(@xmath94 ) denote the convex hull of the incidence vectors of cuts in @xmath94 .
since maximizing a linear function over a set of points equals to maximizing it over the convex hull of this set of points , we can rewrite ( [ maxcut1 ] ) to the following @xmath95 where @xmath96 .
+ ( sdp based max - cut model)@xcite : + for a bipartition @xmath97 , with @xmath98 if @xmath99 , and @xmath100 otherwise , the max - cut can also be formulated as the following semidefinite programming optimization problem : @xmath101 without loss of generality , if we fix the value of the last variable at 1 , then the problem ( [ maxcut2 ] ) is equivalent to the integer quadratic optimization problem @xmath102 where , @xmath103 is a symmetric matrix with @xmath104 , and @xmath105 .
it is not difficult to find that a optimal solution @xmath106 to problem ( [ maxcut3 ] ) corresponds to a optimal solution @xmath107 of original problem ( [ maxcut2 ] ) .
+ considering that the sdp based max - cut model is easier to manipulate for intelligent operators , it is adopted in discrete sta for simple representation .
the parameters setting is the same to that in tsp instances . under the circumstance
, it takes about 8 seconds for each run . in the same way
, we define the following error " @xmath108 where , @xmath72 is the best result achieved by discrete sta , @xmath73 can be found in @xcite .
experimental results for max - cut instances are given in table [ maxcutinstances ] .
ccccccc _ instance _ & _ optimum _ & _ algorithm _ & _ best _ & _ mean _ & _ s.t .
_ & _ error _
+ gr96 & 105328 & sta & 105328 & 105051 & 567 & 0 + & & dsta & 105328 & 104982 & 614 & 0 + kroa100 & 5897392 & sta & 5897392 & 5864276 & 46302 & 0 + & & dsta & 5897392 & 5887930 & 29122 & 0 + kroc100 & 5890760 & sta & 5890760 & 5860540 & 34277 & 0 + & & dsta & 5890760 & 5870613 & 31573 & 0 + gr120 & 2156667 & sta & 2156667 & 2150378 & 7132 & 0 + & & dsta & 2156667 & 2150378 & 7132 & 0 + + as can be seen from table [ maxcutinstances ] , the discrete sta has the ability to achieve the global minimum for all of the instances , and the difference between simple sta and dsta is small in this case .
typically , the model of discrete value selection ( dvs ) problem is different from the model in ( [ eqn1 ] ) , because the domain is defined as follows @xmath109 by introducing a linear transformation @xmath110 where @xmath111 then the discrete value selection can be rewritten to the equivalent constrained boolean integer programming problem @xcite . + in discrete sta , we only use the index of @xmath112 to represent a solution , for example , a solution @xmath113 is corresponding to @xmath114 , which is easy to be manipulated by the intelligent operators .
the intelligent operators swap , shift , symmetry and substitute for discrete value selection are similar to that in boolean programming problem . as a result , only illustrations of these transformations are given from figure [ dvsswap ] to figure [ dvssubstitute ] .
the continuous rosenbrock function has been widely studied as a benchmark .
considering that the rosenbrock function also has many integer local minima , in this paper , it is the first time to use it as a benchmark for integer programming problem .
the integer rosenbrock function is defined as @xmath115 , x_i \in \{-2,-1,0,1,2\}. \nonumber\end{aligned}\ ] ] it is not difficult to find that @xmath1 for any @xmath116 is a local minimum of the integer rosenbrock function .
+ the parameters setting is the same to that of max - cut instances except the maximum number of iterations , which is specified as 10 , 20 , 100 , 200 , 500 , 2000 for @xmath117 , respectively . under the circumstance , 0.1 , 0.15 , 0.35 , 0.6 , 1.5
, 6.0 seconds are consumed correspondingly .
experimental results for integer rosenbrock function are given in table [ irosenbrock ] .
n _ & _ optimum _ & _ algorithm _ & _ best _ & _ mean _ & _ s.t . _ & _ error _ + 5 & 0 & sta & 0 & 0 & 0 & 0 + & & dsta & 0 & 0 & 0 & 0 + 10 & 0 & sta & 0 & 0 & 0 & 0 + & & dsta & 0 & 0 & 0 & 0 + 20 & 0 & sta & 0 & 0 & 0 & 0 + & & dsta & 0 & 0 & 0 & 0 + 50 & 0 & sta & 0 & 0 & 0 & 0 + & & dsta & 0 & 0 & 0 & 0 + 100 & 0 & sta & 0 & 78.4500 & 46.0669 & 0 + & & dsta & 0 & 0 & 0 & 0 + 200 & 0 & sta & 107 & 155.4500 & 42.1495 & - + & & dsta & 0 & 0 & 0 & 0 + + as can be seen from table [ irosenbrock ] , both the simple sta and the dynamic sta do well in the low dimensional cases . for large scale problem ,
the simple sta begins to have poor performance and even fails to find the global optimum . on the other hand ,
the dynamic sta is more robust and adaptable , which verifies the theorem 2 that dynamic sta has higher probability to gain global optimum than the simple sta .
in this paper , a new intelligent optimization algorithm named discrete state transition algorithm is studied for integer programming problem .
it is the first time to build the framework for discrete state transition algorithm , and five key elements are discussed to better develop the algorithm .
the representation of a feasible solution and the dynamic adjustment strategy are mainly studied .
various applications have shown the adaptability of the designed intelligent operators and experimental results have testified the effectiveness and efficiency of the proposed algorithm and strategies .
ahmed , z.h . : genetic algorithm for the traveling salesman problem using sequential constructive crossoveroperator .
international journal of biometrics & bioinformatics .
3(6 ) , 96105 ( 2010 ) dorigo , m. , gambardella , l.m . :
ant colony system : a cooperative learning approach to the traveling salesman problem .
ieee transaction on evolutionary computation .
1(1 ) , 5366 ( 1997 ) darby dowman , k. , and wilson , jm : developments in linear and integer programming
. journal of the operational research society .
53 , 10651071 ( 2002 ) geoffrion , a.m. , marsten , r.e .
: integer programming algorithms : a framework and state - of - the - art survey .
management science .
18(9 ) , 456491 ( 1972 ) jnger , m. , liebling , th.m . , naddef , d. , nemhauser , g.l . ,
pulleyblank , w.r . ,
reinelt , g. , rinaldi , g. , wolsey , l.a . : 50 years of integer programming 19582008 .
new york : springer ( 2010 ) kirkpatrick , s. , gelatt c.d . and vecchi , m.p .
: optimization by simulated annealing .
220(4590 ) , 671680 ( 1983 ) krishnan , k. , mitchell , j.e . : a semidefinite programming based polyhedral cut and price approach for the maxcut problem .
computational optimization and applications .
35 , 5171 ( 2006 ) li , d. , sun , x.l . : nonlinear integer programming .
new york : springer ( 2006 ) metropolis , n. , rosenbluth , a. , rosenbluth , m. , teller , a. , teller , e. : equation of state calculations by fast computing machines .
journal of chemical physics .
21(6 ) , 10871092 , ( 1953 ) reinelt , g. : tsplib a traveling salesman problem library .
orsa journal on computing .
3(4 ) , 376384 ( 1991 ) schlter , m. egea , j.a . ,
banga , j.r . : extended ant colony optimization for non - convex mixed integer nonlinear programming .
computers & operations research .
36 , 22172229 ( 2009 ) smith , k. : an argument for abandoning the traveling salesman problem as a neural - network benchmark .
ieee transaction on neural network , 7(6 ) , 15421544 ( 1996 ) wang , z.b . ,
fang , s.c . , gao , d.y .
, xing , w.x
. : canonical dual approach to solving the maximum cut problem .
oneline in journal of global optimization .
( 2012 ) yang , c.h . ,
tang , x.l . , zhou , x.j . ,
gui , w.h . : state transition algorithm for traveling salesman problem .
arxiv:1206.0329v1 , to apear in the 31st chinese control conference ( 2012 ) yokota , t. , gen , m. , li , y.x . : genetic algorithm for non - linear mixed integer programming problems and its applications .
computers & industrial engineering .
30(4 ) , 905917 ( 1996 ) yu , c.j . , teo , k.l . ,
bai , y.q . : an exact penalty function method for nonlinear mixed discrete programmming problems .
optimiation letters .
( 2011 ) doi:10.1007/s11590 - 011 - 0391 - 2 zhang , d.f .
, liu y.k . , hallah , r.m . ,
leung , s.c.h .
: a simulated annealing with a new neighborhood structure based algorithm for high school timetabling problems .
european journal of operational research .
203 , 550558 ( 2010 ) zhou , x.j . , yang , c.h . ,
gui , w.h . : initial version of state transition algorithm . international conference on digital manufacturing and automation ( icdma ) .
pp . 644647 ( 2011 )
zhou , x.j . , yang , c.h . ,
gui , w.h . : a new transformation into state transition algorithm for finding the global minimum . international conference on intelligent control and information processing ( icicip ) .
pp . 674678 ( 2011 )
zhou , x.j .
, yang , c.h . , gui , w.h . : state transition algorithm .
arxiv:1205.6548v1 , to appear in journal of industrial and management optimization ( 2012 ) | we focus on a recently new intelligent optimization algorithm called discrete state transition algorithm , for solving integer optimization problems .
firstly , we summarize some key elements for discrete state transition algorithm to guide its well development . several intelligent operators are designed for local exploitation and global exploration .
then , a dynamic adjustment strategy risk and restore in probability " is proposed to gain global convergence with high probability .
finally , numerical experiments are carried out to test the effectiveness of the proposed algorithm , and they show that the similar intelligent operators can be applied to ranging from traveling salesman problem , boolean integer programming , to discrete value selection problem , which indicates the robustness and adaptability of the proposed intelligent elements . |
flux compactifications of string theory ( for reviews see @xcite ) populate large parts of the string landscape , and may describe our universe .
however , the theoretical basis for the construction of these compactifications is still far from rigorous ( see @xcite for criticism ) , and is based on using low - energy supergravity actions in a regime which is different from the flat - space regime where they are usually derived from string theory .
it would be very interesting if a non - perturbative construction of some flux compactifications could be found , providing further support for their consistency , and perhaps leading to new methods for their analysis .
a promising arena for such a construction is in flux compactifications involving four dimensional anti - de sitter ( ads ) space .
such compactifications are dual , by the ads / cft correspondence @xcite , to three dimensional conformal field theories .
thus , understanding the three dimensional conformal field theory dual to some @xmath0 flux compactification would give a non - perturbative definition for that background .
eventually we would like to study the statistics of conformal field theories that are dual to @xmath0 backgrounds , in order to learn about statistics of flux compactifications , and to try to understand how to describe also backgrounds with a positive cosmological constant . for general flux compactifications
, it seems that understanding the dual conformal field theory must be very complicated ( see @xcite for attempts in this direction ) .
this is because the cosmological constant of the resulting background , which is related to the central charge of the dual conformal field theory , depends in a very complicated way on the fluxes , and it seems that extremely complicated dynamics is needed to reproduce this on the field theory side .
the situation seems to be much simpler in the type iia flux compactifications constructed in @xcite ( these solutions were further analyzed from the ten dimensional point of view in @xcite ) .
these backgrounds have a `` large - flux limit '' in which some of the fluxes ( the three four - form fluxes @xmath2 , @xmath3 and @xmath4 ) are taken to be large , such that in that limit the cosmological constant becomes small , the string coupling becomes weak , and the compact space becomes large .
this means that these backgrounds can reliably be studied in the supergravity approximation ( except perhaps near the orientifold where the string coupling may be large ) , and that in the `` large - flux limit '' the properties of the dual conformal field theories depend in a simple way on the fluxes .
one can then hope to reproduce this simple dependence in some field theoretic model .
a first attempt at such an analysis , in a different `` large - flux limit '' which does not lead to a weakly coupled string theory ( not all four - form fluxes are taken to be large ) appeared in @xcite ; we will attempt here to describe the field theories appearing in the generic `` large - flux limit '' , which is described by a weakly coupled string theory .
the naive way to construct a field theory dual for flux backgrounds is to imagine constructing the flux gradually from branes carrying that flux , in a manner similar to that in which the @xmath5 background of string theory is constructed from d3-branes in flat space . in particular , the 4-form fluxes @xmath6 in our background are carried by d4-branes wrapped on 2-cycles in the compact space , and it is natural to imagine building the background from such d4-branes @xcite .
it is certainly possible to go from a background with a large flux ( which is already a weakly coupled weakly curved background ) to a background with an even larger flux by adding such branes , and we will use this in our discussion of the moduli space of the conformal field theory .
however , it is not clear if one can construct the full theory from such branes , since in the limit of a small flux the background becomes not only strongly curved ( this happens also for d3-branes ) but also strongly coupled .
nevertheless , it is still natural to guess that the dual conformal field theory arises from some decoupled low - energy theory living on three sets of d4-branes .
however , we will find that assuming that the degrees of freedom in this theory are weakly coupled open strings ( in adjoint and bi - fundamental representations of the resulting @xmath7 gauge theory ) leads to a contradiction , since the central charge of the dual conformal field theory ( which scales as @xmath8 , as we will compute in section [ sec : dual - cft ] ) grows faster than the number of such degrees of freedom .
thus , the field theory must be more complicated than the naive theory of open strings , perhaps involving a larger gauge group , or @xcite fields in multi - fundamental or other higher representations , or perhaps not coming from any gauge theory at all . in order to find clues about this mysterious field theory we investigate in some detail its moduli space , which can be described using configurations of domain walls in @xmath0 . of course , generic flux backgrounds preserve no supersymmetry so they would not be expected to have a moduli space .
the flux backgrounds of @xcite preserve a four dimensional @xmath9 supersymmetry , so they are dual to three dimensional @xmath9 superconformal field theories .
this amount of supersymmetry is not enough to protect the moduli space from quantum corrections , since generic scalar potentials are consistent with three dimensional @xmath9 supersymmetry .
nevertheless , in our study ( performed in the weak coupling weak curvature limit ) we will find a large moduli space in these backgrounds .
we expect this moduli space to be lifted by quantum corrections ( perhaps non - perturbative ) , but these quantum corrections are small in the `` large flux limit '' , and we expect the existence of a moduli space in this limit to be a useful clue for the construction of the dual field theory .
the moduli space turns out to be very complicated , with many different branches that may be interconnected .
for each such branch we employ some mathematical theorems that count the number of solutions of polynomial equations , in order to compute its dimension .
we will show that for large values of the fluxes , the dimension of the moduli space scales as @xmath10 .
the effective field theory at generic points on the moduli space includes @xmath11 gauge fields , scalars and fermions ; however , in @xmath12 dimensions a @xmath11 gauge field is equivalent to a compact scalar , so the presence of these gauge fields does not necessarily imply that the full theory is related to a @xmath13 gauge theory .
however , there are special submanifolds of the moduli space in which we can see gauge groups corresponding to all subgroups of @xmath13 , suggesting that the conformal field theory may be described as the low - energy limit of some gauge theory which includes this gauge group .
this is further supported by the scaling of the dimension of the moduli space , that is reminiscent of strings in the bi - fundamental representation of each pair of gauge groups ( and such bi - fundamental fields indeed appear on the special submanifolds mentioned above ) .
so far we have not been able to find a simple field theory model that would reproduce all the properties that we find ; in particular it seems hard to explain the large number of degrees of freedom , and the complicated form of the moduli space .
we hope that these properties will provide useful clues for the construction of such a field theory in the future .
we begin in section [ sec : supergravity - model ] with a review of the type iia backgrounds of @xcite that we will be studying and of their supersymmetry equations . in section [ sec : dual - cft ]
we compute various basic properties of the dual field theory , like its central charges and the generic features of its operator spectrum . in section [ sec : domain - walls ] we consider branes spanning domain walls in the @xmath0 space , and find the condition that they preserve supersymmetry .
we then go on in section [ sec : geom - moduli - space ] to study the structure of their moduli space .
we compute the moduli space explicitly for a simple example and find some properties , such as the dimension , for the generic case . in the appendices we include some additional calculations , including an explicit calculation of the supersymmetry in the bulk in appendix [ sec : supersymm - equat - bulk ] . in appendix [
sec : bps - condition ] we show that the domain walls found in section [ sec : domain - walls ] obey the bps condition , and in appendix [ sec : other - domain - walls ] we consider the possibility of additional domain wall brane configurations .
in this section we review the low - energy limit of the background of massive type iia string theory described by an orientifold of type iia string theory on @xmath14 .
this model was studied extensively in @xcite , where it was shown that by turning on generic values for the background fluxes it is possible to stabilize all moduli without the use of non - perturbative effects .
we will start by reviewing the geometrical properties of the compact manifold , and then discuss the possible moduli and the way in which they can be stabilized .
finally we will show that the background satisfies the supersymmetry equations in the bulk .
the compact space is an orbifold of @xmath15 .
we parameterize the torus by the three complex coordinates @xmath16 , with @xmath17 .
we take the complex structure moduli of the tori to be @xmath18 , so that the @xmath19 coordinates are periodic with the identifications @xmath20 at this point in the moduli space of the torus , the @xmath15 has a @xmath21 symmetry , under which the coordinates transform as @xmath22 it is then possible to orbifold by this symmetry .
this gives rise to a singular space , with 27 singular points corresponding to the fixed points of the @xmath21 symmetry @xcite . after this identification
, there is a second @xmath21 symmetry acting freely on the coordinates as @xmath23 this symmetry identifies triplets of fixed points , thus leading , after a second orbifold by the second @xmath21 symmetry , to a singular calabi - yau manifold with only 9 singular points ( that can be locally described as a @xmath24 singularity ) .
the cohomology of this manifold is given by @xmath25 and @xmath26 .
there are therefore no complex structure moduli and 12 khler moduli .
nine of them are associated to blow - up modes of the singular points , while the other three khler moduli describe the volume of the three tori .
these volume moduli @xmath27 appear in the metric as @xmath28 or in the khler form for the manifold as @xmath29 it will be useful to write an explicit basis for the cohomology of the compact space .
there are no one - forms , since the two @xmath21 orbifolds project out all of the one - forms of the torus .
there are three two - forms that form the basis of the untwisted part of @xmath30 .
these are the two - forms that remain invariant under the @xmath31 , and they can be chosen as @xmath32 in an arbitrary normalization ( in which the triple intersection is @xmath33 ) .
their poincar - dual four - forms form the basis for the untwisted part of @xmath34 , @xmath35 where @xmath36 are different elements of the set @xmath37 .
we choose the normalizations such that @xmath38 there are also two - forms and four - forms associated with the blow - up modes of the orbifold fixed points , which we will not write down explicitly . since @xmath25 ,
the only 3-forms in the compact geometry are the holomorphic 3-form @xmath39 and its complex conjugate @xmath40 .
these are normalized such that @xmath41 and can be verified to obey the standard relations @xmath42 as a last step in defining the geometry we quotient by an orientifold action .
we will use the orientifold of @xmath14 presented in @xcite .
the total orientifold action is given by @xmath43 , where @xmath44 is reflection on the worldsheet , @xmath45 is the worldsheet left moving fermion number , and @xmath46 is the spacetime involution @xmath47 under this action there is a 3 dimensional space left fixed , given by @xmath48}=0 $ ] .
thus , the theory contains an @xmath49-plane wrapping this 3-cycle and filling the non compact directions . under the orientifold action the different forms have non trivial transformation properties .
the forms defined above transform as @xmath50 one can write the three - forms in a diagonal basis with respect to the orientifold by decomposing @xmath44 to its real and imaginary parts , @xmath51 .
these transform as @xmath52 in order to stabilize all the moduli we will need to turn on a 10-form ( or 0-form ) rr flux , so that in the low - energy limit we obtain romans massive iia supergravity theory @xcite ( with a mass parameter proportional to the rr 0-form field strength ) , compactified to four dimensions on the @xmath14 orientifold discussed in the previous subsection .
in addition to the background metric and dilaton , the theory includes a ns - ns 2-form @xmath53 ( whose field strength is @xmath54 ) , and a rr 1-form and 3-form , @xmath55 and @xmath56 ( with field strengths @xmath57 and @xmath58 ) . with respect to the standard convention , while working with signs as in @xcite .
so , we use opposite signs for @xmath59 and @xmath60 compared to @xcite . ] before turning on fluxes , the massless spectrum includes the khler parameters from the metric , @xmath27 , and the dilaton @xmath61 .
since @xmath53 is odd under @xmath44 , its zero modes are related to the forms @xmath62 in the @xmath46-odd cohomology @xmath63 , and it can be expanded as @xmath64 the three zero modes @xmath65 combine with @xmath27 to form the bosonic part of a chiral multiplet .
similarly we can expand the rr forms .
since @xmath66 , the one - form has no zero modes .
the three - form , being even under @xmath44 , has one zero mode , related to the unique even three - form , @xmath67 .
thus we have @xmath68 the four dimensional axiodilaton superfield contains the combination of this axion @xmath69 with the dilaton @xmath61 .
all of these moduli can be stabilized by turning on fluxes along the compact directions . in order to preserve poincar invariance ,
the fluxes can be written as @xmath70 where all the indices in @xmath71 and @xmath72 are internal , and they are poincar dual using the 6 dimensional metric , @xmath73 .
the background values for the fluxes can then be written by expanding the fields in the relevant cohomology ( having the correct parity under the orientifold ) : @xmath74 they obey the following integrality condition @xmath75 so that the integer fluxes are related to the ones in ( [ eq:18 ] ) by @xmath76 we will split the field strengths into the background part and an excitations part .
they can then be written as @xmath77 the background values of the fluxes are constrained by the tadpoles of the different fields .
these were analyzed in @xcite , where it was found that there is a unique tadpole for @xmath78 which requires @xmath79 in terms of the integer fluxes ( [ eq:167 ] ) this means @xmath80 , so that there are four different possibilities , @xmath81 .
all other fluxes are not constrained by tadpoles .
the scalar potential was analyzed in detail in @xcite , and it was found that by turning on such fluxes @xmath82 the moduli are stabilized at values given by @xmath83 with @xmath84 ( where @xmath85 ) . from the four dimensional point of view , this solution has a negative cosmological constant @xmath86 and we will consider the maximally symmetric solution of the resulting four dimensional action , which is given by @xmath87 .
there are several things to note here regarding this solution . from supersymmetry
we get ( see the next subsection ) a constraint on the signs of the fluxes @xmath88 which also guarantees that the @xmath27 and @xmath89 are real .
when we take large values for the quantized fluxes , @xmath90 ( without making some of them much larger than the others ) , we get to a regime with large volume and weak coupling where we can trust our computation . throughout this paper
we will work in this regime .
we also note that there is a non - singular solution with @xmath91 , which has no 2-form and 6-form background fluxes .
there are additional moduli localized near the @xmath24 singularities .
one can turn on @xmath92 and @xmath93 fluxes on the corresponding localized cycles , which we denote , respectively , by @xmath94 and @xmath95 ( @xmath96 goes over the different singular points ) .
the blow up khler modes @xmath97 are then stabilized at @xmath98 where we defined @xmath99 , and the integer @xmath100 is the non - trivial triple intersection of the twisted cycles .
the values for @xmath101 and @xmath69 are modified by these additional fluxes ( the dilaton by a small amount when @xmath102 ) : @xmath103}^{1/2 } \cr \xi&=&\frac1p { \left(}e_0+\frac{e_im_i+ \sum_a f_an_a}{m_0}+\frac{6\kappa m_1m_2m_3+{\beta}\sum_an_a^3}{3m_0 ^ 2}{\right)}.\end{aligned}\ ] ]
in this subsection we review how the background described above satisfies the supersymmetry equations .
we will write the background as a warped product of a four - dimensional anti de - sitter space with @xmath14 , with the metric @xmath104 where @xmath105 is the warp factor , @xmath106 is the 4 dimensional ads metric and @xmath107 is the metric on @xmath14 .
we will use the double spinor convention , which in type iia amounts to writing the majorana killing spinor as two majorana weyl spinors with opposite chirality , @xmath108 we can decompose the ten dimensional clifford algebra into the @xmath109 algebras in the following way , @xmath110 where the 4d gamma matrices are real and the 6d are purely imaginary and antisymmetric .
we denote by underlined indices the tangent space flat indices .
the killing spinors also decompose as @xmath111 where @xmath112 is the unique covariantly constant spinor on the calabi - yau , while @xmath113 , @xmath114 ( with @xmath115 ) are the killing spinors on @xmath0 satisfying @xmath116 the complex number @xmath117 is the value of the superpotential , so that the cosmological constant of the @xmath0 space is given by @xmath118 .
the spinor @xmath119 on the calabi - yau gives rise to an @xmath120 structure .
following @xcite we can write the two pure spinors as bispinors of @xmath121 in the following way @xmath122 using the clifford map , there is a one - to - one correspondence between such bispinors and p - forms , given by @xmath123 using this map , the pure spinors can also be represented by the almost complex structure 2-form and the holomorphic 3-form , @xmath124 following these notations , the equations for preserved supersymmetry are given by @xcite @xmath125},\\ e^{-2a+\phi}(d+h\wedge)(e^{2a-\phi}\psi_-)&= & 3i{\,\mathcal{i}\mathrm{m}\!\left[\bar\mu\psi_+\right ] } + da\wedge\bar\psi_- \cr & & + \frac{\sqrt2}{16}e^\phi{\left[}(|a|^2-|b|^2)\hat f + i(|a|^2+|b|^2 ) \tilde f{\right ] } , \cr&&\end{aligned}\ ] ] where @xmath126 are the modified rr fields defined as @xmath127 so that they obey the non - standard bianchi identity @xmath128 .
we solve these equations in appendix [ sec : supersymm - equat - bulk ] , finding that for supersymmetry to be preserved the killing spinors should have @xmath129 , and the moduli should obtain values as in ( [ eq:22 ] ) .
in the previous section we described a solution of supergravity ( and , thus , of string theory ) that includes a four dimensional ads space . according to the ads / cft correspondence @xcite , there is a three dimensional conformal field theory which is the holographic dual of this solution .
many properties of this cft can be calculated in a simple manner from the supergravity solution .
we will discuss these properties in this section , including the central charge , dimensions of operators and the global symmetries of the cft .
we will also discuss d - branes wrapping cycles in the compact space to give particles or strings on @xmath0 . throughout this paper
we will work only in the limit where all 4-form fluxes are large , so that the string coupling is weak and the supergravity approximation is good .
we will begin by finding the central charge of the cft from the curvature of the ads space .
there are various possible definitions of a central charge for three dimensional cfts , including the coefficient of the two - point function of the stress - energy tensor , and the coefficient multiplying the volume times the temperature squared in the entropy of the theory at finite temperature . in the gravity approximation ,
all definitions give answers proportional to @xmath130 , where @xmath131 is the four dimensional newton s constant , since this is the coefficient ( in units of @xmath132 ) of the four dimensional action , so that all correlation functions are proportional to this .
using our formulas from the previous section , we have ( up to constants ) @xmath133 since the 3-form and 0-form fluxes are numbers of order one . in particular , if we take all the fluxes @xmath134 , we find a central charge scaling as @xmath135 ( this was independently noted in @xcite ) . equation ( [ eq:173 ] ) is reminiscent of the formula for the central charge in the case of @xmath136 @xmath137 sym in 2 + 1 dimensions . in that case
the central charge of the theory in the ir ( where it is dual to m theory on @xmath138 ) scales like @xmath139 ( which is not understood in terms of any effective field theory degrees of freedom ) . by analogy
, this suggests that in our case there may be some @xmath140 , namely that if there is a uv description of any sort it should include an order of @xmath141 degrees of freedom .
this is also suggested by the fact that these are the minimal integer powers which are larger than those appearing in the central charge ( [ eq:173 ] ) .
this uv description could be for instance an @xmath142 gauge theory , or an @xmath143 gauge theory with matter in representations whose dimension is of order @xmath144 ( such representations are consistent with asymptotic freedom in @xmath12 dimensions ) .
the analysis in @xcite give some support to this suggestion .
it was argued there that after two t - dualities in the directions of the first 2-torus , and in the limit of @xmath145 , the background should be lifted to m theory , and resembles the near - horizon limit of @xmath2 m2-branes ( at some singularity ) . in this case
we see that the degrees of freedom are renormalized from @xmath146 in the theory on some d2-branes ( at the same singularity ) to @xmath147 in the theory on the m2-branes . below we will use another indicator for the number of branes in the problem which will be the structure ( and in particular the dimensionality ) of different branches of the moduli space .
the moduli space will be made out of holomorphic ( in an appropriate sense ) d4-branes which wrap different 2-cycles of the torus .
our analysis of the moduli space will be performed in the limit where all fluxes are large , but since it preserves some supersymmetry it is natural to expect that the same results for the form and dimension of the moduli space will hold also in other limits ( though we have not verified this directly ) . assuming this , we find ( using our results derived below ) that for the scaling of @xcite the dimension of the largest branch of the moduli space will scale like @xmath2 .
indeed , this branch is described by the motion of d4-branes wrapping the first @xmath148 , which become m2-branes ( or d2-branes ) after 2 t - dualities .
our more general analysis below will show that the dimension of the maximal branch of the moduli space scales like @xmath149 .
the previous case is a special case of this .
note that this might suggest that in a scaling limit in which two of the fluxes ( say , @xmath2 and @xmath3 ) become large while the third remains finite , the theory resembles that of @xmath150 m2-branes . while the dimension of the moduli space and the number of degrees of freedom are consistent with this suggestion , the precise form of the moduli space is very different from what one would obtain from any theory of @xmath151 m2-branes .
as described above , the supergravity solution preserves a four dimensional @xmath9 supersymmetry . by the ads
/ cft correspondence this maps to a three dimensional @xmath9 superconformal symmetry , with two supersymmetry charges and two superconformal charges . in the ads / cft correspondence , the global symmetries of the cft
are related to gauge symmetries of the gravitational theory .
such symmetries arise from reductions of the supergravity fields on the compact space ( or from space - filling d - branes ) .
the simplest gauge fields are related to the ten dimensional metric , and are related to the isometry group of the compactification manifold . in our case
the compact space is a calabi - yau manifold and thus has no isometry group .
so , we do not get any gauge fields from the metric .
in addition to the metric , the rr 1-form and 3-form can also give rise to gauge symmetries .
in our background we have a non - trivial 0-form flux which gives a mass to the 1-form ( it is swallowed by the 2-form @xmath53 which becomes massive ) . thus , there is no gauge symmetry associated with the 1-form . in order to get a 1-form gauge field from the 3-form we need to integrate it over a 2-cycle .
as the compactification manifold contains three such untwisted 2-cycles , we obtain three commuting gauge fields .
however , since the 2-cycles are odd under the orientifolding , these gauge fields are projected out by the orientifold .
the gauge fields arising from the twisted 2-cycles are similarly projected out .
thus , the conformal field theory that we are looking for does not have any global symmetry ( beyond the @xmath9 superconformal algebra , which does not include any continuous r - symmetry group ) .
another basic property of a conformal field theory is the spectrum of operators in the theory .
the simplest operators are related to the supergravity fields , and their dimensions are related to the masses so we can easily find the spectrum .
there are two mass scales for fields in the supergravity .
the first is the mass of the moduli , which can be computed from their potential .
this was written explicitly in @xcite for some of the moduli , and it is easy to see that the others have the same scaling . in units of the four dimensional planck scale @xmath152 the moduli masses
are @xmath153 the other mass scale in supergravity is the mass of the kaluza - klein modes , given by the inverse radii of the compact tori , @xmath154 the dimensions of the corresponding operators are given using the ads / cft correspondence as @xmath155 thus , as in all other conformal field theories dual to theories with a four dimensional supergravity approximation ( implying a separation of scales between the moduli and the kk modes ) , there is a small number of operators with dimensions of order one , and all others have large dimensions .
the order one operators correspond to the eight moduli fields , @xmath156 .
another type of operators in the field theory involves d@xmath157-branes wrapped on @xmath157-cycles in the compact space , giving particles in the @xmath0 .
since our background involves massive type iia string theory , we can not have any d0-branes ( which must have @xmath158 strings ending on them ) or d6-branes ( which must have @xmath158 ns 5-branes ending on them ) ; this is related to the fact that the rr 1-form is swallowed by the ns - ns 2-form . naively we can have wrapped d2-branes or d4-branes on our 2-cycles or 4-cycles , but in fact the orientifold maps these to anti - d - branes , so it is unlikely that any stable configurations of this type would exist .
we can also consider a @xmath157-brane wrapping a @xmath159-cycle , leading to a string in @xmath0 ( mapped to some type of flux tube in the conformal field theory ) .
the only such possible configurations are a d4-brane wrapping a 3-cycle and an ns5-brane wrapped on a 4-cycle .
a d4-brane wrapped around the @xmath67 cycle is mapped to an anti - brane by the orientifold , while a d4-brane wrapping the @xmath160 cycle is not a consistent configuration , since there is @xmath54-flux on that 3-cycle , implying that such d4-branes must have d2-branes ending on them .
the same phenomenon arises for ns5-branes wrapped on the 4-cycles , since these have 4-form flux .
note that the fundamental string is also mapped to a string with opposite orientation by the orientifold .
thus , we do not expect to have any stable extended objects in our theory .
in the next two sections we wish to study the moduli space of the conformal field theory dual to the background described in section [ sec : supergravity - model ] . to describe the moduli space we need to find lorentz - invariant configurations with zero energy which have the same asymptotics as the solution described above , but differ in the interior . usually in the ads
/ cft correspondence such configurations are described by supersymmetric branes sitting at some value of the radial position , giving domain walls in ads along which the flux which the brane is charged under jumps . moving along the moduli space of these configurations
is described in the field theory side as giving non - trivial vacuum expectation values to operators .
such domain walls break half of the supersymmetry in the bulk ; in the conformal field theory they break the superconformal generators and preserve the standard supersymmetry generators .
we will consider here d - brane domain walls , given by d@xmath157-branes wrapping @xmath161-cycles in the compact space , and sitting at fixed radial position in @xmath0 .
for the configuration to be supersymmetric ( which is the same as having zero energy in the field theory ) these must obey some calibration condition @xcite .
we will find the supersymmetric cycles over which d - branes can be wrapped by considering the @xmath33-symmetry equation . in appendix
[ sec : bps - condition ] we will also verify directly that these configurations are bps states by considering the dbi+cs action for the d - branes and checking that there is no force acting on them .
all of these equations are valid in the probe approximation , in which the back - reaction of the d - brane on the background is small
. this approximation will be good in the limit of large four - form fluxes that we are working in .
since in three dimensional @xmath9 theories the moduli space is generally not protected , we expect some potential along the moduli space to be generated by corrections to our leading order approximation ; however , this potential is very small in the limit we are working in , so that there will still be an approximate moduli space in the conformal field theory .
the general supersymmetry condition for a d@xmath157-brane filling time plus @xmath162 dimensions and wrapping a @xmath163-cycle in the compact directions is the @xmath33-symmetry equation @xcite , which in the double spinor notation can be written as in @xcite : @xmath164 where @xmath165 @xmath166+{{\cal f } } ) } } \sum_{2l+s = r } \frac{{\epsilon}^{\alpha_1\ldots\alpha_{2l}\beta_1\ldots\beta_s}}{l!s!2^l } { { \cal f}}_{\alpha_1\alpha_2}\ldots{{\cal f}}_{\alpha_{2l-1}\alpha_{2l}}\hat\gamma_{\beta_1\ldots\beta_s}.\ ] ] here , @xmath167 $ ] indicates the pullback of a bulk field onto the worldvolume of the d - brane , and @xmath168 $ ] where @xmath169 is the field strength of the gauge field on the worldvolume of the d - brane , and we set @xmath170 .
we can split the @xmath33-symmetry equation into an equation in the ads space , @xmath171 for some constant @xmath172 , and an equation in the compact space @xmath173 where @xmath174 is defined to be @xmath175 ( @xmath176 ) for even ( odd ) values of @xmath177 ( and @xmath178 and @xmath179 were defined in ( [ eq:29 ] ) ) . from these
we can see ( using the unitarity of the @xmath180 matrices ) that @xmath172 must be a pure phase , and that the d - brane can be supersymmetric only if @xmath181 , which is indeed the case for our background . for type iia ( even @xmath157 ) the internal equation
can be brought to the form @xmath182 from which one gets , as in @xcite , the following calibration condition on the cycle which the d - brane wraps : @xmath183\wedge e^{{\cal f}}\right\}\right|_{2k } = -a^*\alpha\sqrt{\det(p[g]+{{\cal f}})}\ d\sigma^1\wedge\ldots\wedge d\sigma^{2k}\ ] ] for d - branes wrapping even @xmath184-cycles , and @xmath185\wedge e^{{\cal f}}\right\}\right|_{2k+1 } = a^*\alpha^*\sqrt{\det(p[g]+{{\cal f}})}\ d\sigma^1\wedge\ldots\wedge d\sigma^{2k+1}\ ] ] for d - branes wrapping odd @xmath186-cycles . we denote the @xmath177-form part of an expression by @xmath187 .
we will next use this formalism to describe different configurations of d - branes in this background and study their supersymmetry properties .
we begin by verifying that a d6-brane parallel to the orientifold plane obeys the above equations .
we then continue to study the equation for d4-branes spanning domain walls in space - time . after finding the general supersymmetric solution we will study the special case of linear d - branes . in appendix
[ sec : other - domain - walls ] we show that there are no other types of d - branes that lead to supersymmetric domain walls .
we start by considering a probe d6-brane filling the whole non - compact @xmath0 space - time and wrapping a three - cycle in the compact space .
this is not a domain wall , but we use it to test our equations , since we know that such a configuration carrying the same charges as the o6-plane must be supersymmetric . the @xmath0 part of the @xmath33-symmetry equation ( [ eq:65 ] ) gives @xmath188 so it fixes @xmath189 .
since the orientifold action is @xmath190 the orientifold plane is located on @xmath191 , and we wish to put the d6-branes in the same position , so we can parameterize the three compact coordinates of the d6-brane using the embedding @xmath192 the induced metric on the worldvolume is @xmath193 and the induced 3-form is @xmath194=\sqrt{\gamma_1\gamma_2\gamma_3 } d\sigma_1\wedge d\sigma_2\wedge d\sigma_3.\ ] ] the right hand side of ( [ eq:69 ] ) is @xmath195+{{\cal f}})}\ d\sigma_1\wedge d\sigma_2\wedge d\sigma_3 = a^ * i\sqrt{\gamma_1\gamma_2\gamma_3}\ d\sigma_1\wedge d\sigma_2\wedge d\sigma_3,\ ] ] and the left hand side of the equation is @xmath196\wedge e^{{\cal f}}\right\}\right|_{3 } = -ib \sqrt{\gamma_1\gamma_2\gamma_3}\ d\sigma_1\wedge
d\sigma_2\wedge d\sigma_3,\ ] ] so in order for the configuration to be supersymmetric we must have @xmath129 , precisely as we found from the bulk supersymmetry in section [ sec : supersymmetry ] . next ,
consider a d4-brane extended as a domain wall in the ads space and wrapping a generic untwisted 2-cycle , in the cohomology class of @xmath197 .
on such a domain wall , the fluxes jump by @xmath198 . in order to find the supersymmetric configuration we will solve the @xmath33-symmetry equation , starting as before with the @xmath0 part , @xmath199 we choose the @xmath0 metric @xmath200 where @xmath201 is a flat minkowski metric .
this is just the standard ads metric in the poincar patch , with the redefinition @xmath202 .
the covariant derivative can be written as in @xcite , @xmath203 we are interested in the poincar supercharges , obeying @xmath204 , so using ( [ eq:30 ] ) we get @xmath205 @xmath206 plugging this into ( [ eq:77 ] ) we find @xmath207 . to solve the internal part of the @xmath33-symmetry equation we need to choose how to wrap the d4-brane .
we start with the simplest case where the d4-brane wraps the torus @xmath208 .
we can choose the embedding @xmath209 with the induced metric being @xmath210 and the pullback of @xmath211 given by @xmath212=\gamma_1 d\sigma^1\wedge d\sigma^2.\end{aligned}\ ] ] plugging into the supersymmetry condition ( [ eq:68 ] ) we have on the right - hand side @xmath213+{{\cal f}})}d\sigma^1\wedge d\sigma^2 = i { \rm sign}(p)\frac{a^ * b^*}{b } \gamma_1 d\sigma^1\wedge d\sigma^2 = -{\rm sign}(p)i b^ * \gamma_1 d\sigma^1\wedge d\sigma^2,\ ] ] while on the other side we have @xmath214\wedge e^{{\cal f}}\right\}\right|_{2 } = -ib^*p[j ] = -ib^ * \gamma_1 d\sigma^1\wedge d\sigma^2.\ ] ] we see that when the background value of @xmath157 is positive the configuration is supersymmetric . and @xmath215 are the same ( [ eq:57 ] ) .
] when @xmath157 is negative one can take the same embedding and flip its orientation such that @xmath216 and @xmath217 , to get a supersymmetric configuration .
this is just an anti - d4-brane instead of a d4-brane .
we see that depending on the sign of the background fluxes , the supersymmetric brane is either a d4-brane or an anti - d4-brane .
note that in the above we assumed @xmath218+f=0 $ ] .
as the @xmath33-symmetry equation only depends on @xmath219 , the result wo nt change if we have a non - trivial background @xmath57 ( which generates also a non - trivial background @xmath220 ) as long as we turn on fluxes on the worldvolume @xmath221 $ ] .
if the worldvolume flux is different than this value there will be an additional contribution to both sides of the equation . in the right hand side
@xmath219 appear only inside the square root so it will change only the absolute value while keeping the phase unchanged .
in contrast , the left hand side is proportional to @xmath222 and so will change its phase .
we thus conclude that the configuration is supersymmetric only for @xmath223 .
a different type of cycle the d4-branes can wrap is a twisted cycle at a fixed point .
when we go away from the singular limit by turning on 2-form flux on these cycles , the background fluxes and values of the moduli change , see equation ( [ eq:25 ] ) .
however the @xmath33-symmetry equations are only sensitive to changes in the bulk supercharges , that is to the relation between @xmath178 and @xmath179 which remains unchanged .
thus , by turning on the appropriate worldvolume flux on d4 branes wrapping the twisted cycles such that @xmath223 as before we get additional supersymmetric configurations .
we will not consider these configurations in detail , since their contribution to the dimension of the moduli space is finite in the large flux limit .
since the linear embedding described in the previous subsection can not be realized for generic values of the @xmath224 , we will now analyze the most general supersymmetric case of a d4-brane wrapping a generic ( untwisted ) surface .
we will use the complex coordinates @xmath225 in space - time as in ( [ eq:1 ] ) and define the worldvolume complex coordinate to be @xmath226 with the same complex structure .
the position of the d4-brane can be written as @xmath227 the induced metric is given by @xmath228 so the right - hand side of the @xmath229 equation is proportional to @xmath230 this should be equal to the pullback of the almost complex structure , which gives @xmath231 taking the squares of both sides and equating we get @xmath232 which vanishes if and only if @xmath233 in order to understand the meaning of this result , let s consider @xmath208 and @xmath234 .
we start by defining a new variable @xmath235 .
we have @xmath236 and @xmath237 we now can write @xmath238 which vanishes according to ( [ eq:106 ] ) .
we see that the supersymmetry condition can be understood as the statement that the three coordinates @xmath239 can be written as holomorphic functions of each other .
in other words , supersymmetry is equivalent to the requirement that the worldvolume wraps a cycle that can be written as the zero locus of two holomorphic functions of the coordinates
. we will study now a simple class of configurations , in which the embedding of the d - brane can be chosen to be a linear map .
we can write the embedding as @xmath240 two of the six parameters @xmath241 can be absorbed into a shift in @xmath242 , while the others parameterize the moduli of the position of the d4-brane .
we also need to check that this embedding keeps the periodicity of the tori .
the identifications on @xmath243 are @xmath244 and similarly for the @xmath245 pairs . under the first transformation
, we get @xmath246 for these two points to be identified we must have @xmath247 and @xmath248 for some integers @xmath249 .
the second transformation acts as @xmath250 which gives us the restrictions @xmath251 and @xmath252 ( with integers @xmath253 ) .
we are now able to express @xmath254 in terms of four integers @xmath255 .
the wrapping numbers @xmath224 are given by @xmath256}{\right)}\frac{\int_{\sigma^1,\sigma^2}d\sigma^1d\sigma^2 } { \int_{x^i , y^i}dx^idy^i } = a^id^i - b^ic^i = n^i\tilde m^i - \tilde n^i m^i.\ ] ] plugging the embedding into the supersymmetry equations ( [ eq:106 ] ) we get @xmath257 we can solve the first two equations for @xmath258 , and plugging into the third we get @xmath259 it turns out that not all charges @xmath224 may be realized by a single linear d4-brane of the type described above , since there is not always an integer solution to ( [ eq:98 ] ) . to see this , we will now prove some things about this ratio .
first , note that @xmath260 so we can write the equation as @xmath261 with integer @xmath262 . next , we will show that a number that can be written as @xmath263 has an even power for the factor of @xmath264 in its prime decomposition .
then , the ratio of two such numbers must also have an even power for the @xmath264 in its prime decomposition ( if the ratio is a fraction its prime decomposition is the one coming from the prime decompositions of the numerator and denominator ) .
we will prove this by induction , showing that if @xmath265 is divisible by @xmath266 then it is also divisible by @xmath267 .
for @xmath268 , if both @xmath269 are even , @xmath265 is obviously divisible by @xmath270 .
else for @xmath265 to be even both @xmath269 have to be odd , i.e. of the form @xmath271 .
we then get @xmath272 which is divisible by 4 .
we next consider general @xmath177 . again , since @xmath265 is even , @xmath269 are both even or both odd .
in the first case we can divide the entire equation by 4 and reduce it to the case with @xmath273 . for the latter case , we can write again @xmath274 , which after division by 4 is an odd number , specifically it is not a multiple of @xmath266 , so this case can not arise .
we have seen that the background of section [ sec : supergravity - model ] allows for supersymmetric domain walls , described by d4-branes wrapped on 2-cycles . over each domain wall
the 4-form fluxes jump according to the number of times the domain wall is wrapped over each cycle .
when we go far away from the domain walls , we arrive at a background with specific values for the 4 form fluxes .
however there are many different configurations of domain walls which result in the same background in the interior of ads space ( beyond all the domain walls ) .
for example , we can take one d - brane wrapped @xmath224 times over the @xmath275th cycle , or several branes whose total wrapping number is @xmath224 . from this
we see that the moduli space may be composed of many different branches .
the parameterization of each branch includes the radial position of the domain walls , so each branch is a cone , and all the branches are connected at the origin ( when all the domain walls go to the horizon of ads space ) .
naively , the full moduli space is made out of all configurations of d4-branes carrying total wrapping numbers equal to the total fluxes @xmath6 ( some of the d4-branes can of course sit at the origin ) .
however , it is not completely clear that this is true , since our approximations break down when the 4-form fluxes @xmath6 become small ( and it is certainly not clear if there is an @xmath0 solution when one of the fluxes vanishes ) . nevertheless , we expect that this naive approach will be a useful tool for counting the dimension of the moduli space at large @xmath6 .
note that often configurations made out of different sets of d4-branes ( with the same total wrapping number ) can be connected without the need to send some of the branes to the origin of the moduli space .
when two d4-branes intersect , new light degrees of freedom arise at their intersection point which may deform the configuration and smooth it into a configuration of a single d4-brane with the same total flux .
we will begin by considering a simple branch of the moduli space where there is only one d4-brane wrapping a simple cycle .
we will study it in detail and describe its global structure .
we will then go on to describe some properties of the general moduli space . specifically , we will parameterize the different branches , and estimate the dimension of a generic branch .
we start with the simplest branch of the moduli space , which includes branes with wrapping numbers @xmath276 . for these values we can have only one possible configuration of domain walls , which consists of a single d4-brane wrapping the first @xmath148 inside the compact space .
the geometry of its moduli space can be simply read from its low - energy effective action .
we consider the d4-brane to be located at specific values of @xmath277 and embedded as @xmath278 we begin by assuming that the d4-brane is away from all fixed points of the orbifold and orientifold .
the dbi action is given ( up to quadratic order in the fields ) by @xmath279}\end{aligned}\ ] ] where @xmath280 is the induced metric on the d - brane , and @xmath281 is the ten dimensional metric which we now write in the form ( with @xmath282 ) @xmath283 we use @xmath284 to denote the worldvolume indices , @xmath285 are transverse coordinates and @xmath286 denote full ten dimensional indices . reducing the action on the torus
we take the fields @xmath277 to depend only on @xmath287 .
the 5 dimensional gauge field @xmath288 can be expanded as we need to take the gauge field to have non vanishing background flux so that @xmath223 .
this does nt change the rest of the analysis . ]
@xmath289 giving rise to two wilson lines @xmath290 , and a three dimensional gauge field , @xmath291 , which can be dualized to another scalar @xmath292 . using @xmath293
we get the three dimensional action @xmath294,\end{aligned}\ ] ] with indices raised and lowered using the flat metric .
the chern - simons term is @xmath295 where @xmath296 , with @xmath297 .
this can be written as an integral of a 6-form , @xmath298 , over the volume bounded by the d4-brane . in our background only @xmath60 contributes , and using the calculation in appendix [ sec : bps - condition ] , we have @xmath299},\ ] ] which becomes @xmath300}\ ] ] when we use our specific embedding of the d4-brane .
when we compactify , we assume that no fields depend on the compact coordinates , so the only term that contributes to the low - energy effective action is the constant , which is canceled with the constant term in the dbi part
. we can also redefine the radial coordinate to be @xmath301 so that the action is given by @xmath302.\end{aligned}\ ] ] this describes an 8 dimensional moduli space which is a cone ( with radial coordinate @xmath303 ) over a 7 dimensional space parameterized by @xmath304 . to study the global structure of the moduli space we will consider now each of the scalar fields . starting with the dual scalar , @xmath61 , one can see that it is actually periodic . in a 3d ym theory ,
whose action is given by @xmath305 the electric charge inside an @xmath306 is given by @xmath307 since the field values @xmath308 and @xmath309 are the same , and @xmath310 are integers we have @xmath311 in our case @xmath312 .
the wilson lines are also periodic fields . performing a gauge transformation @xmath313 with @xmath314 , on a torus of complex structure @xmath315 ,
shifts the wilson lines by @xmath316 . since @xmath317 must be periodic under the identifications of the coordinates given by @xmath318},v^1+{\,\mathcal{i}\mathrm{m}\!\left[\tau\right]})$
] , we need @xmath319}+c_2{\,\mathcal{i}\mathrm{m}\!\left[\tau\right]}=2\pi n_2,\ ] ] for integers @xmath320 and @xmath321 .
these are solved for integral linear combinations of @xmath322}}{{\,\mathcal{i}\mathrm{m}\!\left[\tau\right]}}\ } \cr & \{c_1=0,\qquad\ \
c_2=2\pi\frac{1}{{\,\mathcal{i}\mathrm{m}\!\left[\tau\right]}}\}.\end{aligned}\ ] ] under the corresponding gauge transformation the fields do not change so we must identify these points on the moduli space of the wilson lines . therefore we get ( using @xmath323 ) @xmath324 this is a torus with complex structure @xmath323 in the coordinate @xmath325 .
thus , we see that the moduli space has the structure of a cone with a seven dimensional base @xmath326 ( before imposing the orbifold and orientifold identifications ) , where the circumference of the @xmath306 is @xmath327 and the complex structure of the three tori are all @xmath323 , while their volumes are @xmath328 the above analysis was for a d4-brane located at a generic point , where it is separated from its images . however
, we can consider also a d4-brane located at the fixed points of the @xmath14 .
we start with the non fractional brane ( and obtain the fractional ones from it via higgsing ) .
the d - brane wraps @xmath329 , and it can sit at a fixed point on the other two tori .
there are three such points , distinct after all identifications . in the covering space of the orbifold action , @xmath15
, the d4-brane has nine copies , which are divided into three separate groups of three coincident branes . to study the moduli space we need to consider the transformation of the chan - paton indices . under the first @xmath21
the fields transform as @xmath330 where @xmath331 .
the invariant fields are then @xmath332 the moduli space is determined by considering commuting matrices , since the scalar potential contains terms with commutators .
we then find two branches . on one branch
the fields @xmath333 vanish , and @xmath334 can have any value , giving rise to 3 scalars each .
this describes the d4-brane and its images at the fixed point as fractional branes with the corresponding gauge group of @xmath335 , each at a different radial position .
the second is when @xmath333 are generic and @xmath61 and @xmath336 are proportional to the identity matrix , in which case the d - brane is away from the fixed points and has some non - trivial wilson line . here
the gauge group is broken back to a single @xmath11 .
the position and wilson lines are given by @xmath337{z_{01}^iz_{12}^iz_{20}^i}$ ] and @xmath337{a_{01}a_{12}a_{20}}$ ] , respectively .
this just spans locally a @xmath21 singularity .
the global identifications are just as in the case away from the fixed points .
we also need to consider the effect of the orientifold action , @xmath338 , where @xmath339 is the spacetime involution @xmath340 , @xmath45 is the worldsheet left moving fermion number and @xmath44 is the worldsheet parity reversal .
the action on our fields is @xmath341 we then get the following degrees of freedom : @xmath342 the d - brane position is now @xmath343{z^i_{01}z^i_{12}z^i_{20}}|$ ] so it can move only along the o - plane . to move out of this plane
the d - brane must meet its image and so we need a pair of such d - branes .
similarly , the wilson line is @xmath344{a_{01}a_{12}a_{20}}|$ ] . in the previous section we found that supersymmetric domain wall configurations are described by a holomorphic curve . here
we will provide a more detailed description of a generic branch of this type , and explain how to count its dimension ( in the limit of large charges ) .
the main tools that we will use are the bezout and bernstein theorems , which we will review , which will be used to calculate the wrapping numbers of a generic branch
. we will be interested primarily in the branch of largest dimension , and examine how this maximal dimension scales with the wrapping numbers .
we will now introduce some mathematical theorems that will help us count the number of solutions for a system of generic polynomial equations .
more details can be found in @xcite . the basic theorem that answers this question is bezout s theorem : [ cols= " < " , ] for the two cases in @xmath345 described in figure [ fig : polytopes ] it is simple to calculate the mixed volume .
@xmath346 with all terms up to order @xmath347 have triangular newton polytopes , as in figure [ fig : polytopes](a ) , and their mixed volume is given by @xmath348 while polynomials with terms up to order @xmath349 in @xmath350 and @xmath351 in @xmath352 have square newton polytopes as in figure [ fig : polytopes](b ) , for which @xmath353
next we will use bernstein s theorem to calculate the properties of the moduli space for a generic d4-brane ( or several d4-branes ) wrapping a 2-cycle on the compact space .
we have seen that the supersymmetry condition requires the embedding of the d4-brane to be holomorphic , so the 2-cycle is given by a set of two holomorphic equations in the @xmath19 .
since the @xmath19 are doubly periodic the holomorphic equations should be periodic as well .
the most general elliptic function over a torus with complex structure @xmath315 can be written in terms of the periodic weierstrass functions @xmath354 and their derivatives @xmath355 .
for the purposes of berenstein s theorem we will treat these variables as independent and add to the set of polynomials @xmath356 the relations @xmath357 a general supersymmetric d - brane will thus be located at the zeros of ( [ eq:138 ] ) and of two holomorphic polynomials of the form @xmath358 we can restrict the polynomials to have terms only up to first order in @xmath359 , since higher powers can be removed using the relations ( [ eq:138 ] ) .
we will take the highest degree of the variable @xmath360 in @xmath361 and @xmath362 to be @xmath363 , respectively . on the torus side , in the computations below
, we can always avoid such points . ] .
given a set of such polynomials , which describe a d4-brane , we will use bernstein s theorem , applied to subsets of this set , to count the wrapping number of the d - brane on the different cycles .
consider for example @xmath364 the number of times the brane wraps the @xmath208 cycle .
we will evaluate @xmath364 by fixing a value of @xmath208 and then counting how many solutions there are to the equations for this @xmath208 ( for a generic @xmath208 ) .
fixing @xmath208 means that we fix both @xmath365 and @xmath366 which satisfy the constraint ( [ eq:138 ] ) for @xmath367 .
this leaves us with 4 polynomials in the variables @xmath368 , on which we apply bernstein s theorem .
the number of solutions to these equations is @xmath369 and similarly for permutations of @xmath37 .
recall that we fix @xmath370 and count the dimensionality of the moduli space for this set of @xmath265 s .
we are interested in finding the values of @xmath371 and @xmath372 for which we obtain the largest dimensionality .
the dimension of the moduli space for a given set of @xmath371 and @xmath372 can be estimated by the number of different monomials in the two polynomials , which is @xmath373 . however , the actual dimension of the moduli space is smaller , since different pairs of polynomials might have the same zero locus .
if we assume @xmath374 , then any multiple of @xmath362 with degree smaller than @xmath371 can be removed from @xmath361 .
we are thus left with a moduli space of dimension @xmath375 to summarise , we have found that the moduli space describing a domain wall across which the flux jump by @xmath376 units of flux , consists of different branches each parametrized by a set of 2 polynomials with degrees satisfying ( [ eq:140 ] ) .
the dimension of such a branch is given by ( [ eq:141 ] ) .
it is interesting from the point of view of the dual field theory to understand how the dimension of the moduli space scales as we take large wrapping numbers , @xmath377 ( which should still be much smaller than the fluxes since we are using the probe approximation ) . for this we will find the dimension of the maximal branch with given wrapping numbers
. we can use ( [ eq:140 ] ) to solve for @xmath372 in terms of the @xmath371 for given values of the fluxes , @xmath378 and similarly for @xmath379 .
since the @xmath372 s are positive , this give some non - trivial condition on the @xmath371 s . the requirement
@xmath374 then leads to the inequalities @xmath380 which can be brought to the form @xmath381 and its permutations . in the same way we can get @xmath382 and its permutations .
we can now use our results in the equation ( [ eq:141 ] ) for @xmath383 .
the term with three @xmath371 s cancels . for the terms of the form
@xmath384 we can use ( [ eq:160 ] ) to see that they scale like @xmath385 , since @xmath371 can be just as large as @xmath386 ( @xmath387 ) .
next , we have @xmath388 so these terms are also smaller than @xmath389 . finally , the term with three @xmath372 s is @xmath390 and scales as the other terms .
terms with less than three @xmath157 s or @xmath162 s are smaller for the same reasons .
we thus conclude that for large fluxes the dimension of moduli space behaves as @xmath391 we can actually find a configuration which saturates this bound on the dimensionality of moduli space .
for instance , if all @xmath224 are of the same order , then by choosing all @xmath392 we get @xmath393 , in which case we get @xmath394 .
the previous analysis was done under the assumption that each @xmath372 is smaller than @xmath371 so that we can eliminate terms in the polynomial @xmath361 using @xmath362 thus reducing the dimension of moduli space .
however it is possible that this is not the case .
if one of the @xmath372 s is larger we need to take all monomials , and the dimension of the moduli space is @xmath395 .
we will assume that @xmath396 is the large @xmath162 so that @xmath397 .
we find @xmath398 from the first inequality we get that @xmath399 and from the second one we get that also @xmath400 so that we arrive again to the same conclusion ( [ eq:163 ] ) as before .
in addition to the directions in the moduli space that change the two polynomials and control the embedding of the d - brane , there are additional dimensions of the moduli space related to wilson lines .
the number of wilson lines is related to the 1 dimensional homology group of the riemann surface the d - brane is wrapping .
we can try to estimate this number as follows .
we can think of the polynomials @xmath401 as polynomials in a projective space by adding a new variable @xmath402 and making all terms have the same weight of @xmath403 and @xmath404 .
the relations ( [ eq:138 ] ) are then of weight @xmath405 .
it is then possible using algebraic geometry methods to calculate the euler characteristic of the complete intersection of these 5 polynomials to be @xmath406 as before , we have @xmath407 and @xmath408 so that we have @xmath409 .
since the number of wilson lines is just the genus , it scales as the euler characteristic , and we get that @xmath410 as before . we thus conclude that the total dimension of the moduli space with given wrapping numbers scales in the same fashion , @xmath411 we note that this behavior may point us towards an @xmath143 gauge theory , as the dimension of the moduli space can than be viewed as coming from the degrees of freedom of strings sitting in the bifundamental representation of any two @xmath137 factors .
we would like to thank k. blum , g. engelhard , y. hochberg , s. kachru , a. lawrence , j. louis , r. minasian , a. sever , e. silverstein , j. simon , a. tomasiello and t. volanski for useful discussions .
this work was supported in part by the israel - u.s .
binational science foundation , by a center of excellence supported by the israel science foundation ( grant number 1468/06 ) , by a grant ( dip h52 ) of the german israel project cooperation , by minerva , by the european network mrtn - ct-2004 - 512194 , and by a grant from g.i.f . , the german - israeli foundation for scientific research and development .
in this appendix we solve the equations for supersymmetry in the bulk for the background discussed in section [ sec : supergravity - model ] .
we find the unbroken spinors and the values for the stabilized moduli .
the equations for preserved supersymmetry are given by @xcite @xmath125},\\ e^{-2a+\phi}(d+h\wedge)(e^{2a-\phi}\psi_-)&= & 3i{\,\mathcal{i}\mathrm{m}\!\left[\bar\mu\psi_+\right ] } + da\wedge\bar\psi_- \cr & & + \frac{\sqrt2}{16}e^\phi{\left[}(|a|^2-|b|^2)\hat f + i(|a|^2+|b|^2 ) \tilde f{\right ] } , \cr&&\end{aligned}\ ] ] where @xmath126 are the modified rr fields defined as @xmath127 so that they obey the non - standard bianchi identity @xmath128 . plugging our background into the first equation we get @xmath412 } \cr & & \rightarrow -p\beta_0\wedge \frac{a\bar b}{8 } e^{-ij } = 2\mu { \,\mathcal{r}\mathrm{e}\!\left[-\frac{iab}{8}\omega\right ] } \cr & & \rightarrow -pa\bar b \beta_0 = 2\mu { \,\mathcal{i}\mathrm{m}\!\left[ab\omega\right ] } = \sqrt2\mu\frac{\sqrt{\gamma_1\gamma_2\gamma_3}}{3^{1/4 } } ( { \,\mathcal{r}\mathrm{e}\!\left[ab\right ] } \beta_0+{\,\mathcal{i}\mathrm{m}\!\left[ab\right ] } \alpha_0 ) , \cr&&\end{aligned}\ ] ] where @xmath413}$ ] and @xmath414}$ ] .
the solution is @xmath415}=0 $ ] and @xmath416 in the second equation we use for the left hand side @xmath417 this equation can be split according to the rank of the forms that appear in it .
the zero - form part of the equation is @xmath418}+\frac{\sqrt2}{16}e^\phi { \left[}(|a|^2-|b|^2)\hat f_0-i(|a|^2+|b|^2)*_6\hat f_6{\right]}.\ ] ] the first term is proportional to @xmath415}$ ] so it vanishes .
the real part implies @xmath419 assuming a non - vanishing value for @xmath420 , while the imaginary part of this equation requires @xmath421 which gives , using ( [ eq:35 ] ) , @xmath422 with @xmath36 being summed over all permutations of @xmath37 . the two - form part is @xmath423 } + \frac{\sqrt2}{16}e^\phi{\left[}(|a|^2-|b|^2)\hat f_2+i(|a|^2+|b|^2)\tilde f_2{\right]},\ ] ] with @xmath424 where we used @xmath425 again , ( [ eq:43 ] ) splits into real and imaginary parts .
the real part vanishes since @xmath181 , while the imaginary part reduces to @xmath426 which gives @xmath427 ( with no summation over @xmath275 ) .
this can be solved to give @xmath428 where we used @xmath415}=0 $ ] and @xmath181 to get @xmath429 or @xmath430 .
we will later see that we must take the minus sign for the background to be supersymmetric .
the 4-form part of the equation is @xmath431 } + \frac{\sqrt2}{16}e^\phi{\left[}(|a|^2-|b|^2)\hat f_4+i(|a|^2+|b|^2)\tilde f_4{\right]}.\ ] ] just as before , the first term vanishes , and since we have @xmath432 we get @xmath433 } f_2=-m_i+m_0 b_i \
\rightarrow\ b_i=\frac{m_i}{m_0}.\ ] ] plugging this back into ( [ eq:45 ] ) we get @xmath434 . finally , the 6-form part is @xmath435 } + \frac{\sqrt2}{16}e^\phi{\left[}(|a|^2-|b|^2)\hat f_6+i(|a|^2+|b|^2)\tilde f_6{\right]}.\ ] ] we use ( [ eq:39 ] ) and @xmath436 to get @xmath437 } + \frac{\sqrt2}{16}e^\phi{\left[}(|a|^2-|b|^2)\hat f_6+i(|a|^2+|b|^2)\tilde f_6{\right]}\cr&= & \frac{ipab}{8\sqrt2 } \frac{\sqrt{\gamma_1\gamma_2\gamma_3}}{3^{1/4 } } \alpha_0\wedge\beta_0 + 3i\frac{\bar\mu a\bar b}{8}\frac18\frac{\gamma_1\gamma_2\gamma_3}{\sqrt3 } \alpha_0\wedge\beta_0 -\frac1{16}e^\phi i2|a|^2\sqrt2m_0\frac18\frac{\gamma_1\gamma_2\gamma_3}{\sqrt3 } \alpha_0\wedge\beta_0 \cr&= & \frac{ipab}{8\sqrt2 } \frac{\sqrt{\gamma_1\gamma_2\gamma_3}}{3^{1/4 } } \alpha_0\wedge\beta_0 -3i \frac{p}{\sqrt2}\frac{3^{1/4}}{\sqrt{\gamma_1\gamma_2\gamma_3 } } \frac{ab}{8}\frac18\frac{\gamma_1\gamma_2\gamma_3}{\sqrt3 } \alpha_0\wedge\beta_0 \cr & & -\frac1{16}e^\phi i2|a|^2\sqrt2m_0\frac18\frac{\gamma_1\gamma_2\gamma_3}{\sqrt3 } \alpha_0\wedge\beta_0 \cr&= & \frac58\frac{ipab}{8\sqrt2 } \frac{\sqrt{\gamma_1\gamma_2\gamma_3}}{3^{1/4 } } \alpha_0\wedge\beta_0 -\frac1{16}e^\phi i2|a|^2\sqrt2m_0\frac18\frac{\gamma_1\gamma_2\gamma_3}{\sqrt3 } \alpha_0\wedge\beta_0 , \ ] ] which gives us @xmath438 using ( [ eq:50 ] ) we can solve for @xmath101 and @xmath439 .
we know that the o6-plane generates a tadpole that is canceled by the fluxes @xmath440 and @xmath157 according to ( [ eq:21 ] ) , so that we find @xmath441 .
we thus must have , for supersymmetry to hold , @xmath129 as stated in the main text .
the results we got for the moduli agree with ( [ eq:22 ] ) .
we note that equations ( [ eq:49 ] ) , ( [ eq:56 ] ) give us the following conditions on the signs of the background fluxes , @xmath442
in this appendix we will consider the supersymmetric domain wall solutions found in section [ sec : domain - walls ] .
such a supersymmetric configuration should be a bps state and therefore feel no radial force .
we will verify this fact directly by considering the d - brane effective action .
( this was independently verified in @xcite . ) in the supersymmetric configuration the gravitational force coming from the dbi term will be canceled against the rr force coming from the wz term , related to the charge of the d - brane .
the d - branes extend along a @xmath12 dimensional surface in @xmath0 parallel to the boundary at constant @xmath336 , and wrap a @xmath161-cycle in the compact space .
their world - sheet action in the string frame is given by @xmath443 where @xmath444 is a dirac - born - infeld type action in the string frame , @xmath445 is the d - brane tension and @xmath446~.\ ] ] @xmath447 is the following wess - zumino ( wz ) type action @xmath448 where @xmath449 and @xmath440 is the massive type iia mass parameter .
the @xmath450 is due to the different normalization of the rr fields we use ( following @xcite ) . as in @xcite ,
the brane action has two contributions which depend on the radial location of the brane in @xmath0 .
one contribution is proportional to the brane area @xmath288 and comes from the dbi action , the other is proportional to the volume enclosed by the brane @xmath451 and comes from the wz action .
next , we are going to evaluate the different terms for wrapped d4-branes .
we will assume @xmath223 .
the @xmath452-brane domain walls wrap a two - cycle in the compact space . in our background
there is a 3-dimensional basis for the untwisted 2-cycles given by @xmath453,\ i=1,2,3.$ ] for simplicity we consider wrapping the two - cycle @xmath454 $ ] in @xmath455 .
* dbi * [ sec : dbi ] in this case we have @xmath456 so we have @xmath457 where we define @xmath458}du^1 dv^1~.\ ] ] * wz * [ sec : wz ] the only non - zero contribution to the wz term is given by @xmath459 where @xmath460 is the d4-brane worldvolume wrapping a 2-cycle in the compact space and spanning a surface of constant @xmath336 in the ads space .
@xmath461 is the two cycle times the volume in ads bounded by the surface of constant @xmath336 .
the other boundary of the volume , at @xmath462 , gives a contribution @xmath463 so it does not contribute .
the supergravity fields obey @xmath464 integrating over @xmath461 we get that the last two terms vanish , since there are no such background fields with indices in the non - compact space .
we can then write @xmath465 the right - hand side is what we want to calculate , while the left - hand side is proportional to the integration of @xmath466 over the dual cycle which we can calculate . in massive type iia supergravity
we have @xmath467 since the @xmath53 and @xmath55 are only the fluctuations and vanish in the background , we can replace @xmath466 in the integral by @xmath58 .
in addition to a boundary term we are left with @xmath468 and the wz term can now be written as @xmath469 now @xmath470 where @xmath471 is the volume form in @xmath472 .
we find @xmath473 thus , the gravitational force due to the dbi term is canceled exactly by the force from the wz term , as must be the case for a bps configuration .
the analysis is very similar for a more general cycle , and we will not write it down explicitly here .
here we consider d2 and d6-branes in domain wall configurations and study their supersymmetry equations .
we show that a d2-brane can never be supersymmetric .
a d6-brane can classically be supersymmetric , however due to flux quantization there are generically no such solutions with integer values for the flux .
the @xmath33-symmetry equation ( [ eq:62 ] ) takes a simple form when we consider d2-branes , since they are not extended along any compact dimension . in order for these domain walls to be supersymmetric
we need @xmath474 this can be brought to the form @xmath475 the @xmath0 part of the equation is the same as for d4-branes , which results in the equation @xmath476 which contradicts our supersymmetric condition in the bulk , @xmath129 .
therefore we conclude that d2-branes can not be supersymmetric domain walls in this background .
consider now a d6-brane which extends as a domain wall in the ads and wraps ( for instance ) the 4-torus spanned by @xmath477 .
its embedding may be chosen as @xmath478 with the induced metric being @xmath479 and the pullback of @xmath211 given by @xmath480=\gamma_1d\sigma^1\wedge d\sigma^2 + \gamma_2d\sigma^3\wedge d\sigma^4.\end{aligned}\ ] ] plugging into the supersymmetry condition ( [ eq:68 ] ) and taking @xmath223 as for the d4-branes , we have on the right - hand side @xmath481+{{\cal f } } ) } d\sigma^1\wedge d\sigma^2\wedge d\sigma^3\wedge d\sigma^4 = -{{\rm sign}}(p)ib^ * \gamma_1\gamma_2 d\sigma^1\wedge d\sigma^2\wedge d\sigma^3\wedge d\sigma^4\ ] ] while on the left - hand side we have @xmath482)\wedge e^{{\cal f}}\}_{4 } = b^ * \frac12(-ip[j])\wedge(-ip[j ] ) = b^ * \frac12(p[j])\wedge(p[j ] ) ) \cr & = b^ * \gamma_1\gamma_2 d\sigma^1\wedge d\sigma^2\wedge d\sigma^3\wedge d\sigma^4.\end{aligned}\ ] ] since the first is purely imaginary and the second is real they can not be equal , and the domain walls are not supersymmetric .
a more generic embedding will not be able to compensate for the factor of @xmath275 , and so even the general case is not supersymmetric .
however as we have seen for the d4-branes , adding non trivial @xmath219 can add a relative phase between the two sides . with @xmath483 the @xmath33-symmetry equation becomes @xmath484 writing @xmath485 we get that for positive @xmath157 @xmath486 which has a solution for @xmath487 .
for negative @xmath157 the right hand side should be @xmath488 so there are solutions for @xmath489 .
classically such supersymmetric configurations exist , however generically there are no such configurations consistent with flux quantization .
a. r. frey , `` warped strings : self - dual flux and contemporary compactifications , '' arxiv : hep - th/0308156 .
e. silverstein , `` tasi / pitp / iss lectures on moduli and microphysics , '' arxiv : hep - th/0405068 .
m. grana , `` flux compactifications in string theory : a comprehensive review , '' phys .
rept . * 423 * , 91 ( 2006 ) [ arxiv : hep - th/0509003 ] .
j. polchinski , `` the cosmological constant and the string landscape , '' arxiv : hep - th/0603249
. m. r. douglas and s. kachru , `` flux compactification , '' rev .
phys . * 79 * , 733 ( 2007 ) [ arxiv : hep - th/0610102 ] .
f. denef , m. r. douglas and s. kachru , `` physics of string flux compactifications , '' ann .
nucl . part .
sci . * 57 * , 119 ( 2007 ) [ arxiv : hep - th/0701050 ] .
t. banks , `` landskepticism or why effective potentials do nt count string models , '' arxiv : hep - th/0412129 . j. m. maldacena , `` the large n limit of superconformal field theories and supergravity , '' adv .
* 2 * , 231 ( 1998 ) [ int .
j. theor .
phys . * 38 * , 1113 ( 1999 ) ] [ arxiv : hep - th/9711200 ] .
e. witten , `` anti - de sitter space and holography , '' adv .
theor . math .
* 2 * , 253 ( 1998 ) [ arxiv : hep - th/9802150 ] .
s. s. gubser , i. r. klebanov and a. m. polyakov , `` gauge theory correlators from non - critical string theory , '' phys .
b * 428 * , 105 ( 1998 ) [ arxiv : hep - th/9802109 ] .
m. fabinger and e. silverstein , `` d - sitter space : causal structure , thermodynamics , and entropy , '' jhep * 0412 * ( 2004 ) 061 [ arxiv : hep - th/0304220 ] .
e. silverstein , `` ads and ds entropy from string junctions or the function of junction conjunctions , '' arxiv : hep - th/0308175 .
o. dewolfe , a. giryavets , s. kachru and w. taylor , `` type iia moduli stabilization , '' jhep * 0507 * , 066 ( 2005 ) [ arxiv : hep - th/0505160 ] .
d. lust and d. tsimpis , `` supersymmetric ads(4 ) compactifications of iia supergravity , '' jhep * 0502 * , 027 ( 2005 ) [ arxiv : hep - th/0412250 ]
. b. s. acharya , f. benini and r. valandro , `` fixing moduli in exact type iia flux vacua , '' jhep * 0702 * ( 2007 ) 018 [ arxiv : hep - th/0607223 ] .
t. banks and k. van den broek , `` massive iia flux compactifications and u - dualities , '' jhep * 0703 * ( 2007 ) 068 [ arxiv : hep - th/0611185 ] . c. kounnas , d. lust , p. m. petropoulos and d. tsimpis , `` ads4 flux vacua in type ii superstrings and their domain - wall solutions , '' jhep * 0709 * , 051 ( 2007 ) [ arxiv:0707.4270 [ hep - th ] ] .
l. j. dixon , j. a. harvey , c. vafa and e. witten , `` strings on orbifolds , '' nucl .
b * 261 * , 678 ( 1985 ) .
a. strominger , `` topology of superstring compactification , '' nsf - itp-85 - 109 _ presented at santa barbara workshop on unified string theories , santa barbara , ca , jul 29 - aug 16 , 1985 _
r. blumenhagen , l. gorlich and b. kors , `` supersymmetric 4d orientifolds of type iia with d6-branes at angles , '' jhep * 0001 * , 040 ( 2000 ) [ arxiv : hep - th/9912204 ] . l. j. romans , `` massive n=2a supergravity in ten - dimensions , '' phys .
b * 169 * , 374 ( 1986 ) .
t. w. grimm and j. louis , `` the effective action of type iia calabi - yau orientifolds , '' nucl .
b * 718 * , 153 ( 2005 ) [ arxiv : hep - th/0412277 ] .
m. grana , r. minasian , m. petrini and a. tomasiello , `` supersymmetric backgrounds from generalized calabi - yau manifolds , '' jhep * 0408 * , 046 ( 2004 ) [ arxiv : hep - th/0406137 ] . m. grana , r. minasian , m. petrini and a. tomasiello , `` generalized structures of n=1 vacua , '' jhep * 0511 * , 020 ( 2005 ) [ arxiv : hep - th/0505212 ] . l. martucci and p. smyth , `` supersymmetric d - branes and calibrations on general n = 1 backgrounds , '' jhep * 0511 * , 048 ( 2005 ) [ arxiv : hep - th/0507099 ] .
p. koerber and l. martucci , `` d - branes on ads flux compactifications , '' jhep * 0801 * , 047 ( 2008 ) [ arxiv:0710.5530 [ hep - th ] ] .
p. koerber , `` stable d - branes , calibrations and generalized calabi - yau geometry , '' jhep * 0508 * , 099 ( 2005 ) [ arxiv : hep - th/0506154 ] .
e. bergshoeff and p. k. townsend , `` super d - branes , '' nucl .
b * 490 * , 145 ( 1997 ) [ arxiv : hep - th/9611173 ] . h. lu , c. n. pope and p. k. townsend , `` domain walls from anti - de sitter spacetime , '' phys .
b * 391 * , 39 ( 1997 ) [ arxiv : hep - th/9607164 ] .
h. lu , c. n. pope and j. rahmfeld , `` a construction of killing spinors on s**n , '' j. math .
* 40 * , 4518 ( 1999 ) [ arxiv : hep - th/9805151 ] .
n. seiberg and e. witten , `` the d1/d5 system and singular cft , '' jhep * 9904 * , 017 ( 1999 ) [ arxiv : hep - th/9903224 ] .
david cox , john little , donal oshea , using algebraic geometry ( graduate texts in mathematics ) . | we study the conformal field theory dual of the type iia flux compactification model of dewolfe , giryavets , kachru and taylor , with all moduli stabilized .
we find its central charge and properties of its operator spectrum .
we concentrate on the moduli space of the conformal field theory , which we investigate through domain walls in the type iia string theory .
the moduli space turns out to consist of many different branches .
we use bezout s theorem and bernstein s theorem to enumerate the different branches of the moduli space and estimate their dimension . to on the conformal field theory duals of type iia + @xmath0 flux compactifications + ofer aharony@xmath1 , yaron e. antebi@xmath1 , micha berkooz@xmath1 + _ department of particle physics , + weizmann institute of science , rehovot 76100 , israel _
e - mails : [email protected] , [email protected] , [email protected] |
the vela pulsar ( b0833 - 45 ) , with a period of 89 millisecond and a characteristic age of 11,000 years , it is the prototype of the young `` vela - like '' pulsars .
the excellent angular resolution of _ chandra _ allows us to separate the pulsar from its synchrotron nebula and study its x - ray properties .
we extracted a total of 16000 counts in the range 0.252.0 kev from two hrc - s / letg observations of 25 ks each .
no deviations of counts in the individual bins from the mean value of neighboring 1020 bins were found with statistical significance higher than 2.7@xmath7 .
this may indicate that there are no elements heavier than h present on the ns surface .
the spectral fits with thermal models show excess counts at the high energy end . to get a handle on the harder component
, we used the publicly available acis / hetg data obtained in the continuous clocking ( cc ) mode .
we used only the zero order image , which is equivalent to about 3 ks observation without the grating .
we applied manual corrections for the dither and the sim motion for both timing and the event position in sky plane ( see zavlin et al .
2000 for details ) .
the background was estimated by interpolating between the neighboring regions in the 1d image of the pulsar and its nebula .
since the acis response is poorly known at e @xmath8 0.5 kev , and the background dominates above 8.0 kev in the cc mode , we use the events in the energy range 0.58.0 kev .
lcccccc thermal & @xmath9 & t@xmath10 & r@xmath10 & l@xmath11 & @xmath12 & l@xmath13 + model & ( @xmath14 @xmath15 ) & ( mk ) & ( km ) & ( erg s@xmath16 ) & & ( erg s@xmath16 ) + blackbody & 2 & 1.5 & 2 & 1.5@xmath17 & 2.7 & @xmath18 + h atmosphere & 3 & 0.7 & 12 & 3.8@xmath17 & 1.5 & 1.5@xmath19 + the combined spectral fit to the hrc - s / letg and acis / hetg - cc data clearly shows two components a soft thermal component , which fits equally well with a blackbody or a magnetic neutron star atmosphere model , and a harder component , which we interpret as a power - law ( pl ) .
the parameters of the fit depend on the model chosen for the soft thermal component .
table 1 shows the parameters of fits to the combined data ( see pavlov et al .
2001 for details ) .
figure 1 shows the broadband spectrum of the vela pulsar .
the soft x - ray spectral parameters are for the atmosphere model .
the extrapolation of the pl component matches both the optical and the hard x - ray/@xmath12-ray flux .
this is the first conclusive observation of both the thermal and the non - thermal radiation from the vela pulsar .
the vela pulsar was observed twice with the hrc - i and twice with the hrc - s / letg .
both of the hrc - i observations and the first letg observation suffered from the hrc timing problem .
we corrected the times by shifting to the time of the next event in the level 1 event list taking only the events with time correction less than 4 ms .
the second letg observation was in the timing mode , so we could recover all the event times .
[ fig2 ] we used the radio ephemeris of the vela pulsar based on the observations at the hobart and parkes radio observatories .
the phase of each event was determined and folded to get the pulse profiles .
the left panel of figure 2 shows the observed pulse profiles of the hrc observations .
the pulse profiles show three peaks separated by about 1/3 of phase , with a total pulsed fraction of @xmath20 ( see also helfand et al . 2001 ) .
the pulsed fluxes in the three peaks are shown in figure 1 .
for the acis / hetg - cc observation , the recorded times are the event readout times .
we corrected the times for the dither and sim motion , and the delay from the event to its readout .
the correction for absolute time depends on the actual position of the source on the ccd , which is poorly known at present .
therefore , we are unable to get the absolute times for the cc observation . the relative times , on the other hand , are accurate .
the right panel of figure 2 shows the acis pulse profiles in two energy bands ( @xmath21 kev and @xmath6 kev ) .
the difference in the profile shapes and the pulsed fraction is striking .
we estimate the intrinsic pulsed fraction of the pulsar to be @xmath22 and @xmath23 in the @xmath21 kev and @xmath6 kev bands , respectively .
* the hrc - s / letg spectrum shows no conclusive evidence for lines in the 0.252.0 kev range .
the spectrum shows two components ( thermal + non - thermal ) , similar to middle - aged pulsars .
the thermal component is consistent with the emission from a magnetic hydrogen atmosphere .
extrapolation of the pl component matches the optical and the hard x - ray points . *
the thermal luminosity of the vela pulsar is lower than that predicted by the `` standard '' neutron star cooling models ( tsuruta 1998 ) .
* hrc data show three peaks with total pulsed fraction of about @xmath20 .
energy - resolved acis pulse profiles show energy - dependent shape and pulsed fraction .
the pulse profile at @xmath21 kev is similar to that observed with hrc . at @xmath6 kev , the pulse profile shows only two peaks with estimated intrinsic pulsed fraction of @xmath23 . | we report the results of the spectral and timing analysis of observations of the vela pulsar with the _ chandra _ x - ray observatory .
the spectrum shows no statistically significant spectral lines in the observed 0.258.0 kev band .
it consists of two distinct continuum components .
the softer component can be modeled as either a magnetic hydrogen atmosphere spectrum with @xmath0 ev , @xmath1 km , or a standard blackbody with @xmath2 ev , @xmath3 km ( the radii are for a distance of 250 pc ) .
the harder component , modeled as a power - law spectrum , gives photon indices depending on the model adopted for the soft component : @xmath4 for the magnetic atmosphere soft component , and @xmath5 for the blackbody soft component .
timing analysis shows three peaks in the pulse profile , separated by about 0.3 in phase .
energy - resolved timing provides evidence for pulse profile variation with energy . the higher energy ( @xmath6 kev ) profile shows significantly higher pulsed fraction .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in |
the n - body problem is a classical and important problem in celestial mechanics and mathematics .
it consists of determining the orbits of n bodies interacting in accordance with the gravitational law of newton .
that is two bodies attracting each other , the force of attraction being directed along the line joining them , proportional to the product of the masses , and inversely proportional to the square of the distance between them . for given @xmath0 bodies ,
the preceding n - body problem can be described by the following nonlinear system of second order differential equations : @xmath1 where @xmath2 is the mass of @xmath3th body , @xmath4 is the position of @xmath3th body in @xmath5 , @xmath0 is the number of bodies , and @xmath6 is the universal gravitational constant , which is always taken to be 1 for convenience in mathematics .
hence we take @xmath7 in this paper .
poincar showed that it is impossible for @xmath8 to find an explicit expression for the general solution .
so even for the three - body problem , it is impossible to describe all the solutions .
till now , for planar three - body problem , there are only three orbits are found clearly for three equal masses , including the existence and the shape of the orbits .
they are euler s solution ( see fig.[euler ] ) , lagrange s solution ( see fig.[lagrange ] ) and figure - eight solution ( see fig.[eight ] ) .
the euler s solution was discovered by euler in 1765 .
it described a collinear configuration , in which one body locates on the center of masses of the three bodies , and the other two rotate around the center of the masses on the two endpoints of the diameter of the circle orbit .
the lagrange solution was discovered by lagrange in 1772 .
the three bodies form an equilateral triangle on a circle orbit and the center of masses is the origin all the time .
the figure - eight solution is a new , surprisingly and remarkable solution after the two solutions , which was found numerically by moore first in 1993 ( see @xcite ) . in 2000 , chenciner and montgomery rediscovered it and proved its existence at the same time .
it describes three equal masses chasing one by one on an eight - shaped curve in the plane .
we refer the reader to the following articles and references therein for more detailed description : chenciner et al @xcite , montgomery @xcite and sim etc .
there is no other specific result for @xmath8 .
hence the numerical solution is more and more important in finding and describing new solutions vividly . in recent years
, some new periodic solutions of the n - body problem have been found by minimizing the action of cycles with some special symmetric or choreographic constraint ( e.g. @xcite,@xcite etc ) , in which fourier approach is used by most authors for the numerical solutions of n - body problem . in this paper
, we propose the n - body type difference equations , and from which we can discover some new periodic orbits by minimizing the corresponding action functional .
the n - body type problems have been studied by bahri , rabinowitz @xcite , zelati @xcite among others . in @xcite , zelati proved the existence of periodic solutions under the radial symmetry assumption . in this paper
we prove the existence of periodic solutions for the difference equations with multi - radial symmetric constraint , which includes the radial symmetry . in the section on numerical examples ,
we propose a new variational difference method for the periodic solutions numerically .
the periodic solutions for n - body type problems ( including n - body problems ) can be described by @xmath9 @xmath10 where @xmath0 , @xmath2 and @xmath4 are defined as above , @xmath11 in @xmath12 , @xmath13 is the period , and generally the singular potential function @xmath14 is @xmath15 with assumptions : @xmath16 and @xmath17 , we have @xmath18 @xmath19 where @xmath20 is usually described by @xmath21
in order to formulate the difference equations for n - body type problems ( [ n])-([p ] ) , we discretize time within one period @xmath22 $ ] with @xmath23 points @xmath24 and denote the set of these discretized points as @xmath25 , i.e. @xmath26 , and introduce a difference scheme by @xmath27 where @xmath28 is an abbreviation of @xmath29 , @xmath30 is the @xmath31th discretized point of @xmath22 $ ] , @xmath32 , and @xmath33 is the discretized step .
moreover , the traditional forth difference scheme is used to approximate the derivative of @xmath34 , which is used in @xcite .
then the first order difference of @xmath34 is @xmath35 and the second order difference of @xmath34 is @xmath36 hence the derivatives of @xmath37 can be approximated by @xmath38 by the difference approach above , ( [ n ] ) can be transformed into @xmath39 where @xmath40 and @xmath41 for convenience , we set the following notations : @xmath42 @xmath43 and @xmath44 where @xmath45 @xmath46 it is easy to verify that @xmath47 is a positive definitive matrix . then the difference equation of ( [ n ] ) becomes @xmath48 and it is obviously that the difference boundary conditions can be described by @xmath49 where @xmath50 is the set of positive integer .
hence the lagrangian functional of ( [ n1])-([p1 ] ) is the following @xmath51 [ p0 ] there exists a small number @xmath52 for all @xmath53 , @xmath54 , @xmath55 , such that @xmath56 .
then the solution @xmath57 of ( [ n1])-([p1 ] ) approximates the solution @xmath34 of ( [ n])-([p ] ) when @xmath58 . since @xmath59
, we can deduce the following equation from ( [ uil1 ] ) and ( [ u ] ) , @xmath60 where @xmath61 is a constant . similar to ( [ ds ] ) , we introduce the difference scheme to @xmath62 , i.e. we have @xmath63 where @xmath64 $ ] and @xmath65 . notice that @xmath66 then when we let @xmath67 the system ( [ n])-([p ] ) can be rewritten as follows : @xmath68 where @xmath69 satisfies ( [ ds ] ) and @xmath70 , that is @xmath71 from ( [ ud ] ) , we have @xmath72 for some constant @xmath73 , and for @xmath74 , there exist @xmath75 , such that @xmath76 when @xmath77 . from ( [ ds ] ) , for @xmath78 $ ] , @xmath79 , clearly we have @xmath80 therefore , there exists a @xmath81 such that @xmath82 the error within one period can be estimated by @xmath83 hence @xmath84 from ( [ eq0 ] ) , noting that the discretized step is @xmath85 , we have @xmath86 this means @xmath87 , @xmath88 , i.e. @xmath89 when @xmath58 . the theorem is proved .
let @xmath90 then the _ multi - radial symmetry _ can be described as follows : @xmath91 where @xmath92 , @xmath93 , @xmath94 is a common multiple of @xmath95 , and @xmath96 means @xmath97,\quad \tau\in[t_{s-1},t_{s}].\ ] ]
let @xmath98 is a hilbert space , @xmath99,(r^{d})^{n})$ ] , i.e. @xmath100,r^{d})$ ] , @xmath40 , and we define the norm of @xmath57 in @xmath98 by @xmath101 where @xmath102 the following lemma describes the relationship between the critical points of @xmath103 and the periodic solutions of ( [ n1])-([p1 ] ) .
[ l1 ] @xmath104 is a critical point of @xmath103 if and only if @xmath105 is a solution of ( [ n1])-([p1 ] ) .
since the difference equations transform the infinite dimensional dynamics into finite dimensional dynamics , it is reasonable to view @xmath103 as a continuously differentiable functional on @xmath98 .
let @xmath106 is the critical point of @xmath103 if and only if the first order derivative of @xmath103 on @xmath57 is equal to zero . that is @xmath107 notice that @xmath108 this satisfies ( [ n1])-([p1 ] ) obviously .
thus the proof is completed .
actually , @xmath109 is a @xmath110-polygon in @xmath5 as a simple case with the difference scheme ( [ ds ] ) .
hence we can call @xmath111 a ` cycle ' in @xmath12 , which means obviously that @xmath112 , @xmath40 , are ` cycles ' in @xmath5 .
we denote @xmath113 \ | \ \hat{q}_{i}(t)=\hat{q}_{l}(t),\ \mbox{for some}\
i \neq l \}.\ ] ] we call @xmath57 is a _ non - collision solution _ of ( [ n1 ] ) and ( [ p1 ] ) , if @xmath114 .
we call @xmath57 is a
_ generalized solution _ of ( [ n1 ] ) and ( [ p1 ] ) , if \a ) @xmath115 ; \b ) @xmath57 satisfy ( [ n1 ] ) ; \c ) @xmath116 where @xmath117\backslash \delta$ ] and @xmath61 is a constant .
it is easy to verify that the integral of ( [ def2 ] ) with respect to @xmath118 approximates @xmath119 since @xmath120 for convenience , we suppose @xmath121 in ( [ n1 ] ) has only one singular point , and without loss of generality , let it be @xmath122 .
hence we have let some @xmath123 , @xmath124 ( @xmath125 ) and @xmath126 where @xmath127 is small .
then we say @xmath20 satisfies _ strong force ( sf ) condition_. the condition ( [ sf ] ) is very efficient to overcome the obstacle coming from singular potential . in @xcite , ( [ sf ] )
is used to bound @xmath34 away from @xmath25 , where @xmath25 is a closed nonempty set , and @xmath128 as @xmath129 . here , we introduce lemma 2 based on the discussions above for n - body type difference equations .
let @xmath130 be the family of homotopic ` cycles ' in @xmath131 , ( r^{d}\backslash\{0\})^{n})$ ] , then we have [ l2 ] assume condition ( [ sf ] ) is satisfied , and there exists a constants @xmath132 such that @xmath133 for all @xmath134 , then @xmath57 is bounded away from @xmath135 .
that is , there exists @xmath52 such that no cycles in @xmath130 intersect the @xmath136-neighborhood of @xmath135 . from ( [ sf ] ) ,
it is easy to verify that @xmath20 are all negative , then from ( [ u ] ) we have the following estimate for @xmath121 : @xmath137 let @xmath138 @xmath139 and let @xmath140 be the graph of @xmath141 in ( [ sf ] ) .
suppose the binary relation is @xmath142 whose arc length is @xmath143 .
then we have @xmath144 now let @xmath145 be a sequence in @xmath130 .
if @xmath146 are not bounded away from @xmath135 , for @xmath74 , the @xmath147-neighborhood of @xmath135 in @xmath5 , @xmath148 , must intersect with infinite many @xmath146 . by passing to a subsequence
, we can suppose that @xmath149 this implies that the corresponding @xmath150 ascend infinitely far up to the skylight of @xmath6 at @xmath135 as @xmath151 .
suppose there exists @xmath52 , @xmath152 is @xmath136-neighborhood of @xmath135 , such that portions of @xmath146 fall outside the ball @xmath152 and the others ( infinite many ) @xmath146 in @xmath152 .
then the variation of @xmath153 would become infinite as @xmath151 , but @xmath154 is greater than the variation of @xmath153 , a contradiction to the fact @xmath155 . on the other hand ,
if there is no such @xmath136 , @xmath146 will collapse to the point @xmath135 , which implies that @xmath156 this contradicts to the hypotheses at the beginning .
hence @xmath121 is bounded away from @xmath135 . now , suppose @xmath146 converge uniformly to @xmath157 in the norm in @xmath98 . by the above argument , if @xmath146 are not bounded away from @xmath135 , @xmath157 will be attached to @xmath135 .
the proof is complete .
( a ) for the case that the potential @xmath14 has more than one singular points has the same result in lemma 2 .
\(b ) lemma 2 means that there is no @xmath134 intersects the @xmath136-neighborhood of @xmath135 under ( [ sf ] ) condition .
\(c ) the converse - negative proposition of lemma 2 shows that if @xmath158 , @xmath159 weakly in @xmath98 and strongly in @xmath160 , then @xmath161 . with respect to the periodic solutions of ( [ n1])-([p1 ] ) , we have the following theorem : [ main ] let @xmath20 and @xmath121 in ( [ u ] ) satisfy ( [ a1])-([a3 ] ) and @xmath57 satisfy ( [ ms ] )
. then ( [ n1])-([p1 ] ) has infinitely many generalized solutions ; furthermore , if ( [ sf ] ) holds , then ( [ n1])-([p1 ] ) has infinitely many non - collision solutions .
it is obvious that @xmath162 is a subset of @xmath163 , and @xmath164 is an even function on @xmath162 for @xmath165 , so we can easily check that the critical points of ( [ jd ] ) on @xmath162 are also the critical points on @xmath163 .
since @xmath20 does not satisfy ( [ sf ] ) , we can modify it as @xmath166 where @xmath136 is small and @xmath167 then we get a new lagrangian functional with condition ( [ sf ] ) from ( [ j ] ) , @xmath168 where @xmath169 let @xmath170 have infimum @xmath171 consider minimizing sequence @xmath172 such that @xmath173 when @xmath174 .
then , when @xmath175 is large enough , we have @xmath176 since @xmath177 , we have that @xmath178 @xmath179 where @xmath180 .
this illuminates that there exists a @xmath181 and a subsequence @xmath182 such that @xmath183 @xmath184 then @xmath185 is a critical point of ( [ jd ] ) .
now , from the remarks of lemma [ l2 ] and lemma [ l1 ] , there exist a subsequence @xmath186 and @xmath187 in @xmath188 , such that @xmath187 is a non - collision periodic solution of n - body problem ( [ n1])-([p1 ] ) . along the line of theorem1.1 in @xcite , we can find a generalized solution @xmath157 of ( [ n1])-([p1 ] ) by constructing compact nested interval cover of @xmath22\backslash \delta$ ] to show @xmath189 and @xmath190 when @xmath191 .
we can also prove @xmath192 is a @xmath13-periodic solution of ( [ n1])-([p1 ] ) if @xmath193 is , where @xmath61 is a positive integer , then we can find infinitely many @xmath13-periodic solutions of ( [ n1])-([p1 ] ) .
this completes the proof .
in this section , we present some numerical solutions of n - body problem ( [ nb])-([p ] ) as an example of n - body type . in this case , @xmath194 in ( [ uil1 ] ) , which is usually called newtonian potential .
the method of steepest descent is adopted in finding the minimum of the functional @xmath103 since this method is simple , easy to apply , and each iteration is fast .
if the minimum points exist , the method guarantees that we can locate them after at least an infinite number of iterations .
we implement it in matlab7.0 for getting the following orbits . and from the process in deducing @xmath103 , we named the method _ variational difference method_. the following examples show that it is at least very valid for our problem . when we set the multi - radial symmetry with @xmath195 which is a simple radial symmetric constraint in the plane .
we get the following two figures with different initial values for three - body problems with equal masses . in the following tables , @xmath109 in the columns is the initial value of @xmath3th body , where @xmath196 . and the @xmath197th raw is corresponding to the initial value of the @xmath197th difference point , where @xmath198 .
we set the values at random by matlab . [ cols="^,^,^",options="header " , ] the following orbits are also found for the planar n - body problems , where @xmath199 shows the number and the masses of the bodies , @xmath103 is the corresponding value of lagrangian functional action .
[ p4_e_2_050_0127 ] [ p5_e_1_040_0127 ] [ p4_n_1_040_0209 ] [ p5_n_1_040_0209 ] [ p4_n_2_040_0330 ] [ p5_n_3_040_0508 ] for spacial n - body problems , @xmath200 i.e. @xmath201 in ( [ ms ] ) , it is also a simple radial symmetric constraint but in space .
we find the orbits in the following figures . in each figure , the picture on the top left corner is the projection of the orbits on @xmath69-@xmath202 plane , the picture on the top right corner is the projection of the orbits on @xmath202-@xmath203 plane , the picture on the bottom left corner is the projection of the orbits on @xmath203-@xmath69 plane , and the picture on the bottom right corner is the orbits in space .
[ s3_e_1_050_0126 ] [ s3_n_1_040_0209 ] [ s4_e_1_040_0202 ] [ s4_n_1_030_0208 ] [ s5_n_2_040_0330 ] [ s4_n_2_030_0330 ] [ s5_e_1_040_0209 ] [ s5_n_1_040_0209 ] when we set @xmath204 in ( [ ms ] ) , @xmath205 , we have the following multi - radial symmetric constraint @xmath206 @xmath207 in this case , the following orbits are found .
[ s3_e_1_sym4t_30pt_0330 ] [ s3_n_1_sym4t_30pt_0330 ] [ multi_n_3_192_060730 ] [ multi_n_3_240_060730 ]
the n - body type difference equation is first introduced in this paper with a new difference scheme . to prove the existence of periodic solutions for n - body type problem , the variational approach and the method of minimizing the lagrangian functional action , both of which are typical in studying n - body dynamical systems ,
are employed in this paper . in finding the generalized solutions and non - collision solutions numerically ,
we use the method of minimizing the functional action of n - body type difference equation .
the presented numerical solutions of n - body problem illuminate that the variational difference method , which is combined with the traditional steepest method , is rather effective in finding the periodic solutions for n - body type problem numerically . with these numerical examples
, we can conclude that there exist some interesting orbits under multi - radial symmetric constraint ; and surprisingly , even under radial symmetric constraint we have obtained some interesting orbits .
furthermore , difference equations transform infinity dimension dynamical systems into finite dimension dynamical systems , and it can be used in the studies for other solutions under corresponding symmetric or choreographic constraints . | this paper introduces a new difference scheme to the difference equations for n - body type problems . to find the non - collision periodic solutions and generalized periodic solutions in multi - radial symmetric constraint for the n - body type difference equations , the variational approach and the method of minimizing the lagrangian action
are adopted and the strong force condition is considered correspondingly , which is an efficient method in studying those with singular potentials . and
the difference equation can also be taken into consideration of other periodic solutions with symmetric or choreographic constraint in further studies . |
quantum technology relies on the utilization of resources , like quantum coherence and entanglement , which allow quantum information and computation processing @xmath1
. however , the quantum entanglement is so fragile and undergoes either an asymptotic decay or a sudden death @xmath2 .
this is due to decoherence , whereby the unavoidable interaction between any real quantum system with its surrounding environment alters the quantum system and consequently disentanglement occurs .
therefore , it is very important to investigate the physical systems and physical effects that may lead to effective long - time entanglement preservation .
although there are so - called decoherence - free states whose initial entanglement is invariant over time , however , only a certain kind of entangled state can be decoupled from the influence of the environment in this way @xmath3 .
so far , a lot of researches have been devoted to entanglement manipulation and protection .
for instance , the quantum zeno effect ( qze ) is a promising way to avoid the decaying behaviour of the entanglement in dissipative systems .
this effect which refers to the inhibition of the temporal evolution of a quantum system by repeated projective measurements during a defined period of time , has been discussed in @xmath4 .
but , this scheme is relatively difficult since one has to perform a series of measurements to the system during the course of the evolution .
the other scheme for protecting the entanglement is detuning modulation that discussed in @xmath5 , which in this approach the quantum entanglement is not preserved in the long - time limit .
another proposed scheme , without the disadvantages of the aforementioned schemes , focuses on the protection of the long time limit of quantum entanglement through the addition of qubits .
the protection of entanglement between two qubits via the additional qubits scheme , was first observed in @xmath6 , where the bipartite entanglement is sustained by the addition of a third qubit . improving the preservation of entanglement by much more additional qubits
was performed in @xmath7 , in such way that all of the entangled qubits and the additional ones were contained in a common environment . distantly noninteracting
two - level quantum systems each of them interacts with their independent environments are the preferable elements of a quantum hardware in order to accomplish the individual control required for quantum information processing @xmath8 .
therefore , finding strategies to protect quantum resources , such as coherence and entanglement , is an essential task in these configurations @xmath9 . in this paper
, we consider a system consisting of @xmath10 non - interacting qubits in a common zero - temperature thermal lorentzian reservoir and show how quantum coherence of a given single - qubit in this system can be preserved by increasing the number of additional @xmath11 qubits . at long time limit ( @xmath0 ) , we show that , apart from markovian and non - markovian dynamics , the coherence measure for the single - qubit approaches asymptotically to a non - zero steady value depending on the number of additional qubits and the initial value of qubit coherence .
also , we see that , as @xmath12 , the coherence measure reaches to its initial value . in the next step , we extend this scheme for protecting bipartite and tripartite entanglements among two and three distant qubits each of them
is contained in independent reservoirs .
it is observed that adding another qubits to each of the reservoirs leads to protecting entanglement from sudden death . at the limit @xmath0
, we obtain simple analytical expressions for each of the bipartite concurrence @xmath13 and tripartite lower bound of concurrence ( lbc ) @xmath14 in terms of number of additional qubits contained in each of the reservoirs and the respective initial state entanglement .
moreover , increasing the number of additional qubits in each of the reservoirs preserves completely the initial entanglement from the dissipative processes in the reservoirs .
as illustrated in the text , this mechanism of entanglement preservation works even for different markovian and non - markovian dynamics occurred individually in each of the related subsystems . after this introduction , in sec .
ii we formulate our scheme for preserving the quantum coherence of a single - qubit . sec .
iii is devoted for extending the scheme proposed in sec ii for protection of bipartite and tripartite entanglements . and
finally , the paper is ended with conclusion in sec .
in this section , we consider an system consisting of @xmath10 independent qubits ( two - level atoms ) immersed in a common zero - temperature thermal reservoir , as depicted in fig .
the hamiltonian @xmath15 of the system contains two parts ( @xmath16 ) @xmath17 with @xmath18 the free hamiltonian and @xmath19 describing the interaction hamiltonian , @xmath20 @xmath21 where , @xmath22 is the raising ( lowering ) operator of the @xmath23 qubit with transition frequency @xmath24 and @xmath25 ( @xmath26 ) is the annihilation ( creation ) operator of the @xmath27 field mode with frequency @xmath28 .
the strength of coupling between the @xmath23 qubit and the @xmath27 field mode is described by @xmath29 and the dimensionless real constants @xmath30 are introduced to individualize the qubits . it is convenient to work in the interaction picture where the state @xmath31 of the system obeys the schrodinger equation as @xmath32 and the hamiltonian in this picture is given by @xmath33 since the total hamiltonian commutes with the number of excitations ( i. e. @xmath34=0 $ ] ) , therefore , any initial state of the form @xmath35 evolves after time @xmath36 into the state @xmath37 where @xmath38 , which means all of the qubits are in the ground state @xmath39 except the @xmath23 qubit that is in the excited state @xmath40 and we have @xmath41 ( that is , tensor products of @xmath39 performed @xmath10 times ) .
also , we denote @xmath42 being the vacuum state of the reservoir and @xmath43 the state of it with only one excitation in the @xmath44th field mode . substituting eqs .
( 5 ) and ( 7 ) to eqs . ( 4 ) ,
gives the probability amplitudes that is governed by a system of @xmath45 differential equations @xmath46 where @xmath47 .
it is clear that @xmath48 , then we have @xmath49 . integrating eqs .
( 9 ) , and substituting it into the eqs .
( 8) , gives the following set of @xmath10 closed integro - differential equation @xmath50 and the correlation function @xmath51 is related to the spectral density @xmath52 of the reservoir by @xmath53 the exact form of @xmath54 thus depends on the particular choice for the spectral density of the reservoir . we take a lorentzian spectral density for the reservoir as @xmath55 where @xmath56 is the detuning of @xmath57 and the central frequency of the reservoir that we have assumed to be @xmath24 . the parameter @xmath58 defines the spectral width of the coupling and the parameter @xmath59 is the coupling constant . by using the spectral density @xmath52 given by eqs .
( 12 ) , we can then obtain the exact solutions by laplace transform @xmath60 where @xmath61 and @xmath62 . here , there are two regimes for the system environment coupling ; weak coupling regime ( @xmath63 ) and strong coupling regime ( @xmath64 ) . in the weak coupling regime
the behaviour of the system is markovian and irreversible decay occurs .
however , in the strong coupling regime non - markovian dynamics occurs accompanied by an oscillatory reversible decay . after tracing out the zero - temperature thermal reservoir and
the qubits except @xmath23 one , the reduced density matrix of the @xmath23 qubit in the basis @xmath65 , at time @xmath36 , becomes @xmath66 furthermore , if we let @xmath67 for @xmath68 and @xmath69 with @xmath70 , eqs .
( 13 ) reduces to @xmath71 where @xmath72 and @xmath73 . the reduced density matrix of the @xmath23 qubit in eqs . ( 14 ) , can be rewritten as @xmath74 the @xmath23 qubit dynamics thus depends only on the function @xmath75 ( @xmath76 ) , which depends on the spectral density and the number of qubits .
also , as discussed in @xmath77 , by differentiating of eqs .
( 17 ) with respect to time and comparing it with an exact master equation , decay rate of the @xmath23 qubit can be obtained as @xmath78 baumgratz @xmath79 introduced an intuitive measure of quantum coherence based on the off - diagonal elements of density matrix of the desired quantum state @xmath80 , as @xmath81 where @xmath82 are the off - diagonal elements of the system density matrix .
indeed , it has been recently shown that @xmath83 satisfies the physical requirements which make it a proper coherence measure .
suppose the @xmath23 qubit is initially prepared in the state @xmath84 ( @xmath85 ) , the other qubits are prepared in the state @xmath39 and the reservoir is in its respective vacuum state , so in this way , @xmath86 .
therefore , at time @xmath87 , the coherence of the @xmath23 qubit becomes @xmath88 .
the dynamics of coherence as a function of the dimensionless time @xmath89 is shown in fig .
2 , where we have assumed that @xmath90 . in fig .
2 ( a , b ) , the coherence measure of the qubit exhibits a markovian dynamics with an asymptotical decay in the absence of additional qubits ( @xmath91 ) .
however , it is readily observed that it can be greatly inhibited by increasing the number of additional qubits ( @xmath92 ) .
moreover , in the non - markovian regime and in the absence of additional qubits ( @xmath91 ) , the measure of coherence oscillatory undergoes to sudden death and in the presence of additional qubits ( @xmath92 ) , undergoes to a non - zero steady value , as depicted in fig .
2(c , d ) .
the oscillations of coherence in the non - markovian regime constitute a sufficient condition to signify the presence of memory effects in the system dynamics , being due to information backflow from the environment to the system .
we point out that , in contrast to the markovian regime , the existence of detuning ( @xmath56 ) tend to slow down the decay process of the coherence in the non - markovian regime .
it should be noted that , in the long time limit , the measure of coherence reduces to the following expression @xmath93 indeed , the qubit coherency can be determined only by knowing the number of additional @xmath11 qubits in the reservoir and the initial state of the qubit .
also , as @xmath94 , @xmath95 reaches to @xmath96 , i.e. to the initial coherency of the @xmath23 qubit . on the other hand ,
these observations are confirmed by the behavior of the related decay rate ( 18 ) in terms of the scaled parameter @xmath89 , as shown in fig .
to achieve our scheme of entanglement preservation , we consider a composite system consisting of two subsystems @xmath97 and @xmath98 contained in two independent lorenzian reservoirs .
each of the subsystems has @xmath99 and @xmath100 qubits respectively .
let s consider the @xmath23 qubit of subsystem @xmath97 and the @xmath101 qubit of subsystem @xmath98 prepared initially in an epr - type entangled state as follows @xmath102 as depicted in fig . 1 ( b ) . according to the ref .
@xmath103 , the complete dynamics of the above two - qubit system can be obtained easily ( see appendix ) . therefore , in the standard computational basis set such as @xmath104 , the explicit forms of the matrix elements of above two - qubit density operator at time @xmath36 becomes @xmath105 where @xmath106 are similar to the eqs .
the parameter @xmath107 defines the spectral width of the coupling of the qubits of subsystem @xmath97 ( subsystem @xmath98 ) to its respective reservoir .
we assume that the coupling constants of the qubits to their respective reservoirs are equal , i.e. @xmath108 . to quantify amount of entanglement of the state ( 22 ) , we use concurrence as a measure of two - qubit entanglement .
the following analytical closed form for the concurrence is obtained @xmath109 the time dependency of the concurrence has been shown in fig . 4 in the markovian and non - markovian regimes .
notice that in the case of @xmath110 ( in the absence of additional qubits in the subsystems a , b ) , the concurrence ultimately decays into zero but there is a steady non - zero value for it by increasing @xmath99 and @xmath100 in both the markovian and non - markovian regimes .
an interesting result for the concurrence occurs at the asymptotical limit @xmath111 , as follows @xmath112 indeed , in the long time limit , the concurrence can be determined only by knowing the number of qubits in the reservoirs and the initial conditions of the two - qubit system .
it is clear that when @xmath113 and @xmath114 or @xmath115 and @xmath116 , or each of them equal to one , then the concurrence decays into zero in the long time limit . on the other hand , as @xmath117
, the concurrence approaches to its initial value ( 2@xmath118 ) .
also , if we let , for instance , @xmath119 and rename @xmath120 then the eqs . ( 24 ) becomes @xmath121 the concurrence evolution in ( 23 ) can be evaluated under different non - markovian behaviors of the related subsystems as shown in fig . 5 with @xmath122 .
at the long time limit , it is interesting to note that , apart from different non - markovian natures of the subsystems , the asymptotic concurrences are completely coincided to each others . in this section ,
we extend the results of the previous section to three subsystems @xmath97 , @xmath98 and @xmath123 each of them contained in three independent lorenzian reservoirs .
each of subsystems has @xmath99 , @xmath100 and @xmath124 qubits respectively
. a w - type entangled state of @xmath23 qubit of subsystem @xmath97 , @xmath101 qubit of subsystem @xmath98 and @xmath125 qubit of subsystem @xmath123 , at time @xmath126 , can be written as follows @xmath127 as depicted in fig . 1 ( c ) . as discussed in appendix , in the standard computational basis set such as @xmath128 ,
the matrix elements of the three - qubit density operator at time @xmath36 are @xmath129 and zero for the remained ones and @xmath106 and @xmath130 are similar to the eqs .
the parameter @xmath131 ( @xmath132 and @xmath133 ) is the spectral width of the coupling of @xmath99 ( @xmath100 and @xmath124 ) qubits to its respective reservior .
also , we assume that the coupling constants of the subsystems to their respective reservoirs are equal @xmath134 . to assess entanglement analytically the so - called lower bound of concurrence ( lbc ) can be used .
any separable states have a vanishing lbc but its inverse is not true .
however , a positive lbc detects entanglement with certainty yet .
thus , use of lbc for the tendency of entanglement dynamics is acceptable . according to ref .
@xmath135 , the lbc of a three qubit mixed state @xmath136 is defined by @xmath137 ^ 2+[c_{r}^{(23|1)}(\rho)]^2+[c_{r}^{(31|2)}(\rho)]^2\}},\end{aligned}\ ] ] where @xmath138 where @xmath139 are the eigenvalues , in decreasing order , of the matrix @xmath140 , @xmath141 the six generators of group @xmath142 acting on qubits @xmath143 and @xmath144 the y - component pauli matrix acting on qubit @xmath145 . for the density matrix of our concern , its lbc can be obtained as @xmath146 fig .
6 , shows the lbc in markovian and non - markovian regimes .
notice that similar to the previous section , in the case of @xmath147 ( without the presence of additional qubits in the subsystems a , b and c ) , the lbc eventually decays to zero but there is a steady value for lbc being retained by increasing @xmath99 , @xmath100 and @xmath124 in both markovian and non - markovian regimes . in the limit @xmath111 , the lbc reduces to @xmath148 consider , for example , @xmath149 then the lbc in ( 31 ) becomes @xmath150 and in the same way , if @xmath115 and @xmath149 , then lbc is equal to zero
. ultimately , in the limits @xmath151 , the lbc reaches @xmath152 and also for the lbc in ( 32 ) , as @xmath153 , we have @xmath154 the lbc evolution in ( 30 ) can be evaluated under different non - markovian behaviors of the related subsystems as shown in fig . 7 with @xmath155 .
obviously , at long time limit , it is interesting to note that , apart from different non - markovian behaviors for the subsystems , the asymptotic lbcs are completely coincided to each others .
we investigated preservation of quantum coherence of a single - qubit interacting with a zero - temperature thermal reservoir throughout the addition of non - interacting qubits in the reservoir . also , we discussed extension of this scheme for entanglement protection of two and three distant non - interacting qubits each of them individually has been contained in an independent reservoir . at limit @xmath0 ,
explicit dependence of the coherence measure , bipartite and tripartite concurrences on the number of additional qubits have been derived .
it was pointed out that , as the number of additional qubits in each reservoir grows up , the initial coherence and the respective entanglements are completely protected in both markovian and non - markovian regimes .
interestingly it was shown that , for preserving initial entanglement , the dynamics of each subsystem is not necessarily similar to the dynamics of other ( others ) subsystem ( subsystems ) ( from markovian and non - markovian point of views ) .
on the other hand , from the experimental point of view , the proposed scheme can be realized using lossy ( i.e. , non - perfect ) cavities at zero temperature with the lorentzian spectral density as our reservoirs . as illustrated , for preserving the entanglement
, it is not important for the cavities to have the same spectral density .
finally , it should be noted that the scheme can be extended for protecting higher order multipartite entanglement of qubits distantly contained in the lorentzian reservoirs such as cavities . + * dynamics of two and three independent qubits * : as discussed in ref .
@xmath103 , a single - qubit dynamics has the form @xmath156 where @xmath157 with @xmath158 .
the matrix whose entries are the values @xmath159 is said to form a matrix representation of @xmath160 . by using eqs .
( 17 ) , the matrix representation of @xmath160 in the standard computational basis @xmath161 can be obtained as @xmath162 where @xmath163 .
we now consider a system consisting of two independent qubits , each locally interacting with its own reservoir .
the complete dynamics of the two - qubit system can be obtained by knowing the single - qubit dynamics which has been obtained in eqs .
( 17 ) . under these conditions , given the dynamics of each qubit as @xmath164 and @xmath165 , the dynamics of the two - qubit system is expressed by @xmath166 where @xmath167 and @xmath168 with @xmath169 . since the qubits are in general in different environments so that their evolution is characterized by the different functions @xmath170 and @xmath171 as @xmath172 where @xmath173 and @xmath174 can be considered as @xmath175 and @xmath176 with @xmath177 .
in the following , we extend this procedure to explore the dynamics of three independent qubits , each locally interacting with its own reservoir . by considering the dynamics of the third qubit as @xmath178 , the dynamics of the three - qubit system
is simply given by @xmath179 where @xmath180 and also @xmath181 with @xmath182 . by knowing the matrix representations of @xmath170 and @xmath171 in eqs .
( 4 ) , the function @xmath183 can be obtained as @xmath184 where @xmath173 , @xmath174 and @xmath130 can be considered as @xmath175 and @xmath176 with @xmath185 .
fig . 1 . ( a ) system under consideration consisting of @xmath10 noninteracting qubits contained in a common zero - temperature thermal reservoir .
( b ) composite system consisting of two subsystems @xmath97 and @xmath98 each of them contained in two independent reservoirs with @xmath99 and @xmath100 qubits respectively .
( c ) composite system consisting of three subsystems @xmath97 , @xmath98 and @xmath123 each of them contained in three independent reservoirs with @xmath99 , @xmath100 and @xmath124 qubits respectively .
coherence @xmath95 as a function of the scaled time @xmath89 with @xmath91 ( solid line ) , @xmath186 ( dashed line ) , @xmath187 ( dotted line ) and @xmath188 ( dotted dashed line ) , ( a , b ) markovian regime with @xmath189 and ( c , d ) non - markovian regime with @xmath190 .
the @xmath23 qubit has been initially prepared in the state @xmath191 .
panels ( a , c ) display the coherence without detuning ( @xmath192 ) and ( b , d ) display with detuning ( @xmath193 ) . fig .
decay rate @xmath194 as a function of the scaled time @xmath89 with @xmath91 ( solid line ) , @xmath186 ( dashed line ) , @xmath187 ( dotted line ) and @xmath188 ( dotted dashed line ) , ( a , b ) markovian regime with @xmath189 and ( c , d ) non - markovian regime with @xmath190 .
the @xmath23 qubit has been initially prepared in the state @xmath191 .
panels ( a , c ) display the coherence without detuning ( @xmath192 ) and ( b , d ) display with detuning ( @xmath193 ) . fig .
concurrence as a function of the scaled time @xmath89 with @xmath195 , ( a ) markovian regime with @xmath196 and ( b ) non - markovian regime with @xmath197 .
the initial entanglement is determined by @xmath198 with @xmath199 for @xmath200 and @xmath201 for @xmath202 .
lbc as a function of the scaled time @xmath89 with @xmath205 , ( a ) markovian regime with @xmath206 and ( b ) non - markovian regime with @xmath207 .
initial entanglement is determined by @xmath208 with @xmath199 for @xmath200 , @xmath201 for @xmath202 , @xmath209 for @xmath210 . | in this paper , we investigate preservation of quantum coherence of a single - qubit interacting with a zero - temperature thermal reservoir throughout the addition of non - interacting qubits in the reservoir .
moreover , we extend this scheme to preserve quantum entanglement between two and three distant qubits which each of them interacting independently with separated reservoirs . at the limit @xmath0 , we obtain analytical expressions for the coherence measure and the concurrence of two and three qubits in terms of the number of additional qubits .
it is observed that , as the number of additional qubits in each reservoir becomes large , the initial coherence and the respective entanglements are completely protected in both markovian and non - markovian regimes .
interestingly , the protection of entanglements occurs even under the individually different behaviors of the reservoirs .
+ + * pacs nos : * + * keywords : * quantum coherence preservation , entanglement preservation , measure of coherence , concurrence , lower bound of concurrence , thermal reservoir , additional qubits , markovian , non - markovian . |
it has become a clich to say that coherent states abound in quantum physics @xcite .
moreover , it turns out that they can also be applied in the theory of quantum deformations @xcite and even in the theory of classical dynamical systems @xcite . in spite of the fact that the problem of the quantization of a particle motion on a sphere is at least seventy years old
, there still remains an open question concerning the coherent states for a particle on a sphere .
indeed , the celebrated spin coherent states introduced by radcliffe @xcite and perelomov @xcite are labelled by points of a sphere , i.e. , the elements of the configuration space .
on the other hand , it seems that as with the standard coherent states , the coherent states for a particle on a sphere should be marked with points of the phase space rather than the configuration space .
the aim of this work is to introduce the coherent states for a quantum particle on the sphere @xmath0 , labelled by points of the phase space , that is the cotangent bundle @xmath1 .
the construction follows the general scheme introduced in @xcite for the case of the motion in a circle , based on the polar decomposition of the operator defining via the eigenvalue equation the coherent states . from the technical point of view
our treatment utilizes both the barut - girardello @xcite and perelomov approach @xcite .
namely , as with the barut - girardello approach the coherent states are defined as the eigenvectors of some non - hermitian operators . on the other hand , in analogy to the perelomov formalism
those states are generated from some `` vacuum vector '' , nevertheless in opposition to the perelomov group - theoretic construction , the coherent states are obtained by means of the non - unitary action . in section 2
we recall the construction of the coherent states for a particle on a circle .
sections 36 are devoted to the definition of the coherent states for a particle on a sphere and discussion of their most important properties . for an easy illustration of the introduced approach we study in section 7 the case with the free motion on a sphere .
in this section we recall the basic properties of the coherent states for a particle on a circle introduced in @xcite .
consider the case of the free motion in a circle .
for the sake of simplicity we assume that the particle has unit mass and it moves in a unit circle .
the classical lagrangian is @xmath2 so the angular momentum canonically conjugate to the angle @xmath3 is given by @xmath4 and the hamiltonian can be written as @xmath5 evidently , we have the poissson bracket of the form @xmath6 implying accordingly to the rules of the canonical quantization the commutator @xmath7 = i,\ ] ] where we set @xmath8 .
the operator @xmath9 does not take into account the topology of the circle and ( 2.5 ) needs very subtle analysis . the better candidate to represent
the position of the quantum particle on the unit circle is the unitary operator @xmath10 @xmath11 indeed , the substitution @xmath12 does not change @xmath10 , i.e. @xmath10 preserves the topology of the circle .
the operator @xmath10 leads to the algebra @xmath13 = u,\ ] ] where @xmath10 is unitary . consider the eigenvalue equation @xmath14 using ( 2.7 ) and ( 2.8 ) we find that the operators @xmath10 and @xmath15 are the ladder operators , namely @xmath16 demanding the time - reversal invariance of representations of the algebra ( 2.7 ) we conclude @xcite that the eigenvalues @xmath17 of the operator @xmath18 can be only integer ( boson case ) or half - integer ( fermion case ) .
we define the coherent states @xmath19 for a particle on a circle by means of the eigenvalue equation @xmath20 where @xmath21 is complex . in analogy to the eigenvalue equation satisfied by the standard coherent states
@xmath22 @xcite with complex @xmath23 , of the form @xmath24 where @xmath25 is the standard bose annihilation operator and @xmath26 and @xmath27 are the position and momentum operators , respectively , we set @xmath28 hence , making use of the baker - hausdorff formula we get @xmath29 we remark that the complex number @xmath21 should parametrize the cylinder which is the classical phase space for the particle moving in a circle .
the convenient parametrization of @xmath21 consistent with the form of the operator @xmath30 such that @xmath31 arises from the deformation of the circular cylinder by means of the transformation @xmath32 the coherent states @xmath19 can be represented as @xmath33 the coherent states satisfy @xmath34 where the maximal error arising in the case @xmath35 is of order @xmath36 per cent and we have the exact equality in the case with @xmath37 integer or half - integer .
therefore , @xmath37 can be identified with the classical angular momentum .
furthermore , we have @xmath38 it thus appears that the average value of @xmath10 in the normalized coherent state does not belong to the unit circle . on introducing the relative average of @xmath10 of the form @xmath39 where @xmath19 and @xmath40 are the normalized coherent states , we find @xmath41 from ( 2.21 ) it follows that that the relative expectation value @xmath42 is the most natural candidate to describe the average position of a particle on a circle and @xmath3 can be regarded as the classical angle .
we remark that the coherent states on the circle have been recently discussed by gonzles _
et al _ @xcite .
in spite of the fact that they formally generalize the coherent states described above , the ambiguity of the definition of those states manifesting in their dependence on some extra parameter , can be avoided only by demanding the time - reversal invariance mentioned earlier , which leads precisely to the coherent states introduced in @xcite . since the time - reversal symmetry seems to be fundamental one for the motion of the classical particle in a circle and makes the quantization unique , therefore the generalization of the coherent states discussed in @xcite which does not preserve that symmetry is of interest rather from the mathematical point of view .
having in mind the properties of the standard coherent states one may ask about the minimalization of the heisenberg uncertainty relations by the introduced coherent states for a particle on a circle . in our opinion , in the case with the compact manifolds the minimalization of the heisenberg uncertainty relations is not an adequate tool for the definition of the coherent states .
a counterexample can be easily deduced from ( 2.7 ) , ( 2.8 ) and ( 2.9 ) . indeed , taking into account ( 2.8 ) and ( 2.9 )
we find that for the eigenvectors @xmath43 s of the angular momentum @xmath18 the equality sign is attended in the heisenberg uncertainty relations implied by ( 2.7 ) such that @xmath44 more precisely , for these states ( 2.22 ) takes the form @xmath45 . on the other hand ,
the vectors @xmath43 s are clearly rather poor candidate for the coherent states . in our opinion
the fact that the coherent states are `` the most classical '' ones is better described by the following easily proven formulae : @xmath46 where the approximations are very good ones .
in fact , these relations mean that the quantum variables @xmath18 and @xmath10 are at practically constant `` distance '' from their classical counterparts @xmath47 and @xmath48 , respectively , and therefore the quantum observables and the corresponding expectation values connected to the classical phase space are mutually related .
we point out that in the case with the standard coherent states for a particle on a real line we have the exact formulae @xmath49 it seems to us that the approximative nature of the relations ( 2.23 ) and ( 2.24 ) is related to the compactness of the circle .
our experience with the case of the circle discussed in the previous section indicates that in order to introduce the coherent states we should first identify the algebra adequate for the study of the motion on a sphere . the fact that the algebra ( 2.7 ) referring to the case with the circle @xmath51 is equivalent to the @xmath52 algebra , where @xmath53 is the group of the plane consisting of translations and rotations , @xmath54={\rm i}\varepsilon_{\alpha\beta}x_\beta , \qquad [ x_\alpha , x_\beta]=0,\qquad \alpha,\,\beta=1,\,2,\ ] ] realized in a unitary irreducible representation by hermitian operators @xmath55 where the casimir is @xmath56 and @xmath57 is the anti - symmetric tensor , indicates that the most natural algebra for the case with the sphere @xmath0 is the @xmath50 algebra such that @xmath58={\rm i}\varepsilon_{ijk}j_k,\qquad [ j_i , x_j]={\rm i } \varepsilon_{ijk}x_k,\qquad [ x_i , x_j]=0,\qquad i,\,j,\,k=1,\,2,\,3.\ ] ] indeed , the algebra ( 3.4 ) has two casimir operators given in a unitary irreducible representation by @xmath59 where dot designates the scalar product . therefore , as with the generators @xmath60 , @xmath61 , describing the position of a particle on the circle , the generators @xmath62 , @xmath63 , can be regarded as quantum counterparts of the cartesian coordinates of the points of the sphere @xmath0 with radius @xmath64 .
we point out that unitary irreducible representations of ( 3.4 ) can be labelled by @xmath64 and the new scale invariant parameter @xmath65 .
it is clear that @xmath66 is simply the projection of the angular momentum @xmath67 on the direction of the radius vector of a particle .
since we did not find any denomination for such an entity in the literature , therefore we have decided to call @xmath66 the _ twist _ of a particle .
let us now recall the basic properties of the unitary representations of the @xmath50 algebra .
the @xmath50 algebra expressed with the help of operators @xmath68 , @xmath69 , @xmath70 and @xmath71 , takes the form @xmath72 & = & 2j_3,\qquad [ j_3,j_\pm]=\pm j_\pm,\\ { } [ j_\pm , x_\mp ] & = & \pm 2x_3,\qquad [ j_\pm , x_\pm]=0,\qquad [ j_\pm , x_3]=\mp x_\pm,\\ { } [ j_3,x_\pm ] & = & \pm x_\pm,\qquad [ j_3,x_3]=0,\\ { } [ x_+,x_- ] & = & [ x_\pm , x_3]=0.\end{aligned}\ ] ] consider the irreducible representation of the above algebra in the angular momentum basis spanned by the common eigenvectors @xmath73 of the operators @xmath74 , @xmath68 , @xmath75 and @xmath76 @xmath77 where @xmath78 . recall that @xmath79 the operators @xmath80 and @xmath70 act on the vectors @xmath73 in the following way : @xmath81 an immediate consequence of ( 3.9 ) is the existence of the minimal @xmath82 satisfying @xmath83 thus , it turns out that in the representation defined by ( 3.9 ) the twist @xmath66 can be only integer or half integer .
we finally write down the orthogonality and completeness conditions satisfied by the vectors @xmath73 such that @xmath84 is the identity operator .
now , an experience with the circle indicates that one should identify by means of the @xmath50 algebra an analogue of the unitary operator @xmath10 ( 2.6 ) , representing the position of a particle on a sphere . to do this ,
let us recall that a counterpart of the `` position '' @xmath85 on the circle @xmath51 is a unit length imaginary quaternion which can be represented with the help of the pauli matrices @xmath86 , @xmath63 , as @xmath87 where @xmath88 .
notice that @xmath89 is simply an element of the @xmath90 group and it is related to the @xmath91 quotient space .
therefore the most natural choice for the `` position operator '' of a particle on a sphere is to set @xmath92 where @xmath62 , @xmath63 obey ( 3.4 ) and ( 3.9 ) and we have omitted for convenience the imaginary factor i. furthermore , let us introduce a version of the dirac matrix operator @xcite @xmath93 observe that @xmath94 making use of the operators @xmath95 and @xmath96 we can write the relations defining the @xmath50 algebra in the space of the unitary irreducible representation introduced above as @xmath97_+ & = & { \rm tr}kv,\\ v^2&=&i,\end{aligned}\ ] ] where @xmath98 , and the subscript `` + '' designates the anti - commutator . in particular , @xmath99 it should also be noted that in view of ( 4.4 ) and ( 4.5_c _ ) @xmath95 satisfies the unitarity condition @xmath100 .
we now introduce the vector operator @xmath101 generating , via the eigenvalue equation analogous to ( 2.10 ) , the coherent states for a particle on a sphere @xmath0 .
the experience with the circle ( see eq . ( 2.13 ) ) suggests the following form of the `` polar decomposition '' for the matrix operator counterpart @xmath30 of the operator @xmath101 : @xmath102 indeed , it is easy to see that in the case of the circular motion in the equator defined semiclassically by @xmath103 and @xmath104 , @xmath30 reduces to the diagonal matrix operator with @xmath30 given by ( 2.13 ) and its hermitian conjugate on the diagonal .
furthermore , using ( 4.5_b _ ) we find @xmath105 motivated by the complexity of the problem we now restrict to the simplest case of the twist @xmath106 when ( 4.8 ) takes the form @xmath107 in the following we confine ourselves to the case @xmath106 .
the general case with arbitrary @xmath108 will be discussed in a separate work . besides ( 4.9 ) we have also remarkably simple relation ( 4.5_b _ ) referring to @xmath106 such that @xmath109_+ = 0.\ ] ] notice that the case @xmath106 is the `` most classical '' one .
indeed , the projection of the angular momentum onto the direction of the radius vector should vanish for the classical particle on a sphere .
it should also be noted that in view of ( 3.10 ) @xmath17 s and @xmath110 s labelling the basis vectors @xmath73 are integer in the case of the twist @xmath111 .
we finally point out that the condition @xmath106 ensures the invariance of the irreducible representation of the @xmath50 algebra under time inversions and parity transformations which change the sign of the product @xmath112 . clearly demanding the time - reversal or the parity invariance when @xmath108 one should work with representations involving both @xmath113 and @xmath114 .
we now return to ( 4.7 ) . making use of ( 4.10 ) and the fact that the matrix operator @xmath95 in view of ( 4.2 ) is traceless one we obtain for @xmath106 @xmath115 hence , @xmath116 taking into account ( 4.9 ) we get from ( 4.12 ) @xmath117 and @xmath118=0,\qquad i , j=1,\,2,\,3.\ ] ] as with ( 4.2 ) describing in the matrix language the position of a quantum particle on a sphere
, the matrix operator ( 4.12 ) can be only interpreted as a convenient arrangement of the operators @xmath119 generating the coherent states , simplifying the algebraic analysis of the problem .
accordingly , we define the coherent states for a quantum mechanics on a sphere in terms of operators @xmath119 , as the solutions of the eigenvalue equation such that @xmath120 where in view of ( 4.13 ) @xmath121
. what is @xmath101 ? using ( 4.7 ) , ( 4.2 ) , ( 4.3 ) and setting @xmath106 , we find after some calculation @xmath122 we remark that @xmath119 have the structure resembling the standard annihilation operators .
in fact , one can easily check that it can be written as a combination @xmath123 of the `` position operator '' @xmath124 and the `` momentum '' @xmath125 , where the coefficients @xmath126 and @xmath127 are functions of @xmath128 .
we finally point out that derivation of the operator @xmath101 ( 4.16 ) without the knowledge of the matrix operator @xmath30 seems to be very difficult task .
in this section we construct the coherent states specified by the eigenvalue equation ( 4.15 ) . on projecting ( 4.15 ) on the basis vectors @xmath129 and using ( 3.7_a _ ) , ( 3.8 ) and ( 3.9 ) with @xmath106 we arrive at the system of linear difference equations satisfied by the fourier coefficients of the expansion of the coherent state @xmath130 in the basis @xmath131 .
the direct solution of such system in the general case seems to be difficult task .
therefore , we adopt the following technique .
we first solve the eigenvalue equation for @xmath132 , and then generate the coherent states from the vector @xmath133 using the fact ( see ( 4.16 ) ) that @xmath101 is a vector operator . as demonstrated in the next section the case with @xmath134 refers to @xmath135 and @xmath136 , where @xmath137 is the position and @xmath138 the angular momentum , respectively , i.e. , the particle resting on the `` north pole '' of the sphere .
let us write down the eigenvalue equation ( 4.15 ) for @xmath134 @xmath139 using the following relations which can be easily derived with the help of ( 4.16 ) , ( 3.7_a _ ) , ( 3.8 ) and ( 3.9 ) with @xmath111 : @xmath140 it can be easily checked that the solution to ( 5.1 ) is given by @xmath141 now , using the commutator @xmath142=-{\rm i}{\bi w}\times{\bi z},\ ] ] where @xmath143 , we generate the complex rotation of @xmath101 @xmath144 taking into account ( 5.5 ) and ( 4.15 ) we find that the coherent states can be expressed by @xmath145 is given by @xmath146 it thus appears that the coherent states can be written as @xmath147 generated analogously as in the case of the circle described by the equation ( 2.16 ) .
the formula ( 5.8 ) can be furthermore written in the form @xmath148 where @xmath149 finally , eqs .
( 5.9 ) , ( 5.3 ) , ( 3.7_a _ ) and ( 3.8 ) taken together yield the following formula on the coherent states : @xmath150 where @xmath151 and @xmath152 are expressed by ( 5.10 ) and @xmath153 . taking into account the identities @xmath154 @xmath155 where @xmath156 is the hypergeometric function , @xmath157 are the gegenbauer polynomials and
@xmath158 is the gamma function , we obtain @xmath159 where @xmath160 is the sign of @xmath110 .
let us recall in the context of the relations ( 5.14 ) that the polynomial dependence of the projection of coherent states onto the discrete basis vectors , on the complex numbers parametrizing those states is one of their most characteristic properties .
clearly , the polynomials ( 5.14 ) should span via the `` resolution of the identity operator '' the fock - bargmann representation .
we recall that existence of such representation is one of the most important properties of coherent states .
the problem of finding the fock - bargmann representation in the discussed case of the coherent states for a particle on a sphere is technically complicated and it will be discussed in a separate work . finally , notice that the coherent states @xmath130 are evidently stable under rotations .
we now show that the introduced coherent states for a quantum particle on a sphere are labelled by points of the classical phase space , that is @xmath1 . referring back to eq .
( 4.16 ) and taking into account the fact that the classical limit corresponds to large @xmath17 s , we arrive at the following parametrization of @xmath161 by points of the phase space : @xmath162 where the vectors @xmath163 , fulfil @xmath164 and @xmath165 , i.e. , we assume that @xmath138 is the classical angular momentum and @xmath137 is the radius vector of a particle on a sphere . in accordance with the formulae ( 4.15 ) and ( 4.13 ) the vector @xmath161 satisfies @xmath121 .
thus , the vector @xmath161 is really parametrized by the points @xmath166 of the classical phase space @xmath1 .
consider now the expectation value of the angular momentum operator @xmath67 in a coherent state .
the explicit formulae which can be derived with the help of ( 3.7_a _ ) , ( 3.8 ) , ( 3.12 ) and ( 5.14 ) are too complicated to reproduce them herein . from computer simulations it follows that @xmath167 nevertheless , in opposition to the case with the circular motion , the approximate relation ( 6.2 ) does not hold for practically arbitrary small @xmath168 .
namely , we have found that whenever @xmath169 , then ( 6.2 ) is not valid .
note that returning to dimension entities in the formulae like ( 3.6 ) we measure @xmath168 in the units of @xmath170 , so in the physical units we deal rather with @xmath171 .
for @xmath172 the relative error @xmath173 , @xmath63 , is small .
more precisely , if @xmath174 , then @xmath1751 per cent .
in other words , in the case of the motion on a sphere , the quantum fluctuations are not negligible for @xmath1761 @xmath170 and the description based on the concept of the classical phase space is not adequate one .
however , it must be borne in mind that the condition @xmath177 10 @xmath170 , when ( 6.2 ) holds is not the same as the classical limit @xmath178 .
we only point out that @xmath179 .
it thus appears that the parameter @xmath138 in ( 6.2 ) can be identified with the classical angular momentum divided by @xmath170 .
we now study the role of the parameter @xmath137 in ( 6.1 ) . as with the momentum operator
@xmath67 the explicit relations obtained by means of ( 3.9 ) with @xmath106 , ( 3.12 ) and ( 5.14 ) are too complicated to write them down herein .
the computer simulations indicate that @xmath180 it seems that the formal resemblance of the formula ( 6.3 ) and ( 2.19 ) referring to the case with the circular motion is not accidental one .
the range of application of ( 6.3 ) is the same as for ( 6.2 ) , i.e. , @xmath181 .
because of the term @xmath182 , it appears that the average value of @xmath124 does not belong to the sphere with radius @xmath64 .
proceeding analogously as in the case of the circle we introduce the relative average value of @xmath124 of the form @xmath183 where @xmath184 is a coherent state with @xmath185 where @xmath186 is the unit vector along the @xmath187 coordinate axis and @xmath138 is the same as in ( 6.1 ) . in view of ( 6.3 ) and ( 6.4 ) we have @xmath188 therefore , the relative expectation value @xmath189 seems to be the most natural one to describe the average position of a particle on a sphere . we have thus shown that the parameter @xmath137 can be immediately related to the classical radius vector of a particle on a sphere .
as with the case of the circular motion ( see formulae ( 2.18 ) and ( 2.21 ) ) , we interpret the relations ( 6.2 ) and ( 6.6 ) as the best possible approximation of the classical phase space . in this sense the coherent states labelled by points of such deformed phase space are closest to the classical ones .
the quantum fluctuations which are the reason of the approximate nature of ( 6.2 ) and ( 6.6 ) are in our opinion a characteristic feature of quantum mechanics on a sphere .
we finally remark that the discussion of the heisenberg uncertainty relations analogous to that referring to the circle ( see section 2 ) can be performed also in the case with the coherent states for a particle on a sphere .
for example a counterpart of the formula ( 2.22 ) is @xmath190 where according to eq .
( 4.2 ) we have @xmath191 .
such discussion as well as the detailed analysis of the heisenberg uncertainty relations for the quantum mechanics on a compact manifold will be the subject of a separate paper which is in preparation .
we now illustrate the actual treatment by the example of a free twist 0 particle on a sphere , i.e. the rotator . the corresponding hamiltonian is given by @xmath192 by ( 3.7_a _ ) the normalized solution of the schrdinger equation @xmath193
can be expressed by @xmath194 determined by ( 5.14 ) and ( 6.1 ) , which gives the probability of finding the system in the state @xmath131 , when the system is in the normalized coherent state @xmath195 , has the following properties . for fixed integer
@xmath196 the function @xmath197 has a maximum at @xmath198 coinciding with the integer nearest to the positive root of the equation @xmath199 ( see fig . 1 ) .
thus , it turns out that the parameter @xmath200 can be regarded as the energy of the particle .
further , for fixed integer @xmath17 in @xmath201 ( see fig . 2 ) , such that ( 7.5 ) holds , the function @xmath197 has a maximum at @xmath202 coinciding with the integer nearest to @xmath203 .
it thus appears that the parameter @xmath203 can be identified with the projection of the momentum on the @xmath204 axis .
in this work we have introduced the coherent states for a quantum particle on a sphere .
an advantage of the formalism used is that the coherent states are labelled by points of the classical phase space .
the authors have not found alternative constructions of coherent states for a quantum mechanics on a sphere preserving this fundamental property of coherent states . as pointed out in sec .
6 , the quantum fluctuations arising in the case of the motion on a sphere are bigger than those taking place for the circular motion .
this observation is consistent with the appearance of the additional degree of freedom for the motion on a sphere .
we remark that as with the particle on a circle , we deal within the actual treatment with the deformation of the classical phase space expressed by the approximate relations ( 6.2 ) and ( 6.6 ) .
we also point out that besides ( 6.2 ) and ( 6.6 ) the quasi - classical character of the introduced coherent states is confirmed by the behaviour of the distribution of the energies investigated in section 7 .
it seems that the approach introduced in this paper is not restricted to the study of the quasi - classical aspects of the quantum motion on a sphere .
for example , the results of this work would be of importance in the theory of quantum chaos .
in fact , in this theory the kicked rotator is one of the most popular model systems . because of the well known difficulties in the analysis of the heisenberg uncertainty relations occuring in the case with observables having compact spectrum like the position operator @xmath124 satisfying the @xmath50 algebra ( 3.4 ) we have not studied them herein .
the analysis of the heisenberg uncertainty relations as well as the discussion of the case with a nonvanishing twist will be performed in future work .
vv klauder j r and skagerstam b s 1985 _ coherent states applications in physics and mathematical physics _ ( world scientific : singapore ) kowalski k and rembieliski j 1993 _ j. math .
phys . _ * 34 * 2153 kowalski k 1994 _ methods of hilbert spaces in the theory of nonlinear dynamical systems _
( world scientific : singapore ) radcliffe j m 1971 _ j. phys . a _ * 4 * 313 perelomov a m 1972 _ commun .
* 26 * 222 ; perelomov a m 1986 _ generalized coherent states and their applications _ ( springer : berlin ) kowalski k , rembieliski j and papaloucas l c 1996 _ j. phys . a _ * 29 * 4149 barut a o and girardello l 1971 _ commun .
* 21 * 41 glauber r j 1963 _ phys . rev _ * 130 * 2529 ; * 131 * 2766 klauder
j r 1963 _ j. math .
phys . _ * 4 * 1055 gonzles j a and del olmo m a 1998 _ j. phys . a _
* 31 * 8841 biedenharn l c and louck j d 1981 _ angular momentum in quantum physics . theory and application . _
( addison - wesley : massachusetts ) | the coherent states for a particle on a sphere are introduced .
these states are labelled by points of the classical phase space , that is the position on the sphere and the angular momentum of a particle . as with the coherent states for a particle on a circle discussed in kowalski k _ et al _ 1996 _ j. phys .
a _ * 29 * 4149 , we deal with a deformation of the classical phase space related with quantum fluctuations .
the expectation values of the position and the angular momentum in the coherent states are regarded as the best possible approximation of the classical phase space .
the correctness of the introduced coherent states is illustrated by an example of the rotator . 8.8 in = .5 cm = .5 cm |
as well - known , nonlinear born - infeld electrodynamics has become , over the past years , a focus of intense activity after its connection with brane physics has been elucidated .
specifically , the low - energy dynamics of d - branes can be described by a nonlinear born - infeld - type action @xcite .
we further recall that born and infeld @xcite suggested to modify maxwell electrodynamics to get rid of infinities in the theory .
also , it is important to mention that , in addition to the string interest , the born - infeld theory has also attracted attention from different viewpoints .
for example , in connection with noncommutative field theories @xcite , also in magnetic monopoles studies @xcite , and possible experimental determination of the parameter that measures the nonlinearity of the theory @xcite .
more recently , it has been shown that born - infeld electrodynamics , due to its non - linearity , predicts the existence of real and well - behaved magnetostatic field associated with a electric charge at rest @xcite .
it is important to emphasize that this result is a non - linear effect restricted to four space - time dimensions only , and it naturally raises the question about its generalization to lower dimensions .
indeed , it is not quite evident that the same phenomenon will take place in three space - time dimensions .
the present work specifically deals with this problem , where we examine the effect of the space - time dimensionality on magnetostatic fields associated with electric charges at rest . on the other hand , it is also important to recall here that systems in three - dimensional theories have been extensively discussed over the last few years @xcite
. this is primarily due to the possibility of realizing fractional statistics , where the physical excitations obeying it are called anyons , which continuously interpolate between bosons and fermions . in this connection ,
three - dimensional chern - simons gauge theory sets out as a standard description , so that wilczek s charge - flux composite model of anyon can be implemented @xcite .
also , three - dimensional theories are interesting due to its connection with the high - temperature limit of four - dimensional theories @xcite , as well as for their applications to condensed matter physics @xcite . in this perspective
, it is of interest to improve our understanding of the physical consequences of born - infeld electrodynamics .
as already mentioned , of special interest will be to further explore the impact of the space - time dimensionality on magnetostatic fields associated with electric charges at rest .
as we shall see , in the case of three - dimensional born - infeld electrodynamics , unexpected features are found .
in fact , a magnetostatic field associated with a electric charge at rest not appears in this case .
interestingly , the addition of a topological term ( chern - simons ) to born - infeld electrodynamics results in the occurrence of novel non - trivial solutions for the magnetic field , which clearly shows the key role played by the chern - simons term .
it is worth mentioning that these solutions agree with those of ref .
furthermore , our analysis goes on to explore what is the three - dimensional analogue of a magnetic monopole .
in other words , what kind of electric charge - like is responsible for electrostatic fields with non - vanishing rotational .
the general outline of our paper is now described .
this work is a sequel to ref .
@xcite where we provide a analysis of three - dimensional born - infeld electrodynamics by studying the respective equations of motion . in section
ii , we are most concerned with properties of electrostatic and magnetostatic fields .
we now examine the equations of motion for born - infeld electrodynamics in three dimensions , that is , @xmath0 @xmath1 where the corresponding constitutive relations are given by @xmath2 @xmath3 it is worthwhile mentioning at this point that unlike to our previous analysis @xcite , equation ( [ 5 ] ) requires that the magnetic field be always constant , leading to drastic consequences for the magnetic sector , as it can be seen below .
accordingly , the magnetic induction , @xmath4 , present in the constitutive relation ( [ 7 ] ) can be written as @xmath5 we admit radial and/or angular dependence in the region of space where the nonlinearity dominates .
however , further analysis shows that , far enough from the electric charge , so that the linear regime can be established , @xmath6 and @xmath7 .
since @xmath4 must tend to zero when @xmath8 , this implies that @xmath9 must always be zero , because it is constant .
we remark that the new feature of the present model is that the magnetic field generated from an standstill electric charge is ruled out in three dimensions .
this situation is exactly the opposite to that described for born - infeld electrodynamics in four dimensions .
the result above may suffer a drastic change whenever we allow that a topological term is added to born - infeld electrodynamics .
thus , for a standstill point - like electric charge of intensity @xmath10 , at @xmath11 , the equations of motion are given by @xmath12 and @xmath13 these equations , together with the constitutive relations ( [ 6 ] ) and ( [ 7 ] ) , describe the static and planar born - infeld electrodynamics coupled with a term chern - simons .
here , @xmath14 is the coefficient of the chern - simons theory .
we also recall that it is an integer multiple of some basic unit in the quantum hall effect ( qhe ) , while in the fractional quantum hall effect ( qhe ) it is a rational multiple . considering the purely radial dependence and only a single radial component for the fields @xmath15 and @xmath16 ,
these two equations can be written as @xmath17 @xmath18 moreover , the constitutive relations ( [ 6 ] ) and ( [ 7 ] ) , for the @xmath19 and @xmath4 fields read @xmath20 @xmath21 with the help of ( [ 9.1 ] ) and ( [ 9.2 ] ) , equations ( [ 8 ] ) and ( [ 9 ] ) can be rewritten as @xmath22 @xmath23 we observe that this system has no analytical solution all over the space . in the region
far from the electric charge , the fields are weak and the solution is already known @xmath24 @xmath25 close to the electric charge , the fields are strong and the nonlinearity is present . the fields @xmath19 and @xmath26 , for example , are no longer proportional .
it is important to realize that equation ( [ 11 ] ) states that the derivative of the scalar magnetic field , @xmath9 , is always negative , if @xmath26 and @xmath14 are positive .
therefore is a decreasing function . at this point
, we make the assumption that the field @xmath26 is singular at @xmath11 .
thus , the term inside the root ( @xmath27 ) will be dominated by @xmath28 and near the origin the derivative of the field @xmath9 will reduces to @xmath29 by assuming that @xmath30 ( value to be determined ) , the field @xmath31 when @xmath32 . putting this result into equation ( [ 10 ] ) ( near the origin ) , we arrive at the following differential equation for the field @xmath26 @xmath33 this allows us to find the following solution for @xmath32 @xmath34 in other terms , this confirms the @xmath35-singularity for the @xmath26-field .
the constant @xmath36 is related to the electric charge if one takes the @xmath37 . in order to estimate the behavior of the @xmath4-field in this region
, we examine equation ( [ 9.1 ] ) , that is , @xmath38 here , we see the strong nonlinear effect of the fields near the electric charge . while @xmath9 is finite and nonzero , the field @xmath4 vanishes due to the singular behavior of the field @xmath26 .
equation ( [ 9.2 ] ) we can evaluate the electric field @xmath19 near the origin .
the field @xmath19 goes to @xmath39 , as the square root of the denominator is dominated by the field @xmath26 . as a result , the electric field is not only finite but it also assumes the maximum born - infeld field intensity , @xmath39 , totally different from the @xmath26 singular field behavior .
@xmath40 finally , we can determine the value of the magnetic field @xmath9 at the origin using the constitutive relation ( [ 7 ] ) : @xmath41 a summary of the behavior of fields will provide an overview of their spatial distribution and guide us to the correct boundary values on numerical solution of the problem .
thus , we have the following picture : for @xmath42 , the behavior is linear and well established @xmath43 @xmath44 for @xmath45 , the fields radically differ , showing the effects of the non - linearity @xmath46 @xmath47 @xmath48 @xmath49 we have the following challenge : the fields @xmath19 and @xmath9 at @xmath11 is unknown because we ignore the maximum field strength @xmath39 .
we must use a different strategy . far away from the electric charge , where the linear regime dominates ,
the fields are well described by analytical functions ( [ 11aa ] ) and ( [ 11ab ] ) .
the bessel functions describe the shape of the fields .
the constant @xmath50 describes the field amplitude and it can be estimated with good accuracy . the intention is to use that field as a boundary condition . for this purpose , the phenomenon of superconductivity can be very useful .
it has a quasi - planar structure and can be described by a three - dimensional model .
the analogy is present also in its linear equation for the magnetic field .
many physical properties of high - temperature superconductors are two - dimensional phenomena derived from their square planar crystal structure .
interest in planar field theories is mainly motivated by the possibility of understand the critical phenomena and extend them to four - dimensions . how we know , the hallmark of superconductivity is the expulsion of the @xmath4 field from the inner region of the superconductor , called meissner effect . the london penetration depth , @xmath51 , is well experimentally measured for several super - conducting materials .
its value ranges from @xmath52 to @xmath53 nm . therefore , these experimental data are of great importance for an indirect estimate of limits or ranges for the bi maximum field strength @xmath39 , if it exist . our equations , in the linear regime , are exactly the same if the cs - coefficient be associated with this penetration length , @xmath54 .
our task is to solve numerically the system of equations ( [ 8 ] ) and ( [ 9 ] ) , look for fields spatial distribution until nonlinear region , satisfying the previously established limits at the @xmath11 and finally , take the @xmath39 directly from the curve . to calculate the amplitude , @xmath50 , of the electric field
, we take a typical value involving the charge and the characteristic length as @xmath55 , where @xmath56 is the permittivity of free space and is numerically equal to @xmath57 .
the charge @xmath10 has @xmath58 units and will be set as @xmath59 , where @xmath60 is the elementary charge . the maximum experimental characteristic length ,
@xmath51 , which is around @xmath61 , will set up the lower limit for maximum field @xmath39 of born - infeld theory . with these data set the estimated value for @xmath50 is equal to @xmath62 .
the result can be seen in the figure @xmath63 .
[ h ] _ the fundamental fields near the electric charge._,scaledwidth=80.0% ] these curves confirm the prior analysis predictions for the fields far away and near of the electric charge .
an alternative setting was also performed in @xcite .
the possibility of bianchi identity violation is considered .
this section explores that consequences in the context of born - infeld with topological chern - simons term added . in the static and planar version , it acquires the following form : @xmath64 this equation requires that the scalar potential @xmath65 be non regular , in analogy to the four dimensional case where the potential vector , @xmath66 , exhibits a string of singularities , allowing the introduction of magnetic charges . writing equation ( [ 20 ] ) in cartesian coordinates , @xmath67 so that the solution is well - known and is written as : @xmath68 changing to polar coordinates the solution acquires an immediate interpretation : @xmath69 as a consequence , we have a tangential electric field that arises from an scalar current : @xmath70 taking the bics equation for @xmath9 and @xmath15 : @xmath71 @xmath72 the solution is immediate : @xmath73 inverting the constitutive relation of bi we have : @xmath74 assuming @xmath75 , and substituting ( [ 1 ] ) and ( [ 2 ] ) in ( [ 2.1 ] ) , we arrive at : @xmath76 this equation for the field @xmath4 is dependent on @xmath77 and @xmath78 . far away the electric charge , the field @xmath4 behaves like : @xmath79 assuming @xmath80
, we restore the limit @xmath81 .
however , the argument of the square root @xmath82 , when close to the electric charge , can become negative leaving the field @xmath4 complex .
in addition , it exhibits the same discontinuity of the field @xmath9 when the angle @xmath78 is close to @xmath83 .
our main effort in the present paper has been to understand how non - linear effects of planar electrodynamics in born - infeld formulation may induce magnetic effects out of simple electrostatic systems .
lower dimensionality actually reveals special properties if compared with the same problem in its @xmath84-d formulation @xcite .
however , we would like to formulate as much as possible how to understand the results of this work in connection with the dimensional reduction of @xmath84-d electrodynamics to @xmath85-d .
both planar situations studied here , @xmath86 and @xmath87 , can be mapped into the ( spatial ) @xmath88-dimensional problem of calculating the magnetostatic field on an infinite wire placed along the @xmath89-axis with a steady current density , @xmath90 . if @xmath91 and @xmath92 stand for the electric and magnetic fields in @xmath88 spatial dimensions ( where @xmath93 and @xmath94 denote vectors in @xmath95 space dimensions @xmath96 ) , planar maxwell electrodynamics rules out @xmath19 and @xmath97 and we have maxwell equations for @xmath93 and @xmath4 in @xmath85-d : @xmath98 @xmath99 @xmath100 where the @xmath101-operator , @xmath102 and @xmath103 are planar vectors and @xmath4 behaves as a pseudo - scalar in @xmath104 .
planar maxwell electrodynamics singles out the fields above and discard the following equations for @xmath19 and @xmath94 : @xmath105 and @xmath106 where @xmath107 , @xmath108 , @xmath109 and @xmath19 and @xmath110 are scalars in @xmath104 .
then , a point - like charge at the origin in the planar case is not the dimensional reduction of the corresponding situation in @xmath111 .
it rather corresponds to the case of an infinite wire with a steady current in @xmath111 : @xmath112 . thus , the planar equations @xmath113 and @xmath87 can be both re - expressed in terms of the equation @xmath114 , ( @xmath115 ) which can be written in @xmath95 spatial dimensions as @xmath116 , whose solution ( in polar coordinates ) is @xmath117 , so that @xmath118 .
we stress that this @xmath94-field is ruled out by planar maxwell electrodynamics .
it is the @xmath95-dimensional sector of the @xmath88d magnetic field .
so , in the case @xmath119 , we see that @xmath93 replaces @xmath120 in @xmath116 , whereas in the case @xmath121 , it is @xmath122 that plays the rle of @xmath123 .
then , we may say that the electrostatic fields we have in planar electrodynamics correspond to the @xmath94-field ruled out by the low dimensionality .
we can now confirm why planar electrostatics can not support a @xmath4-field even in the non - linear case , as we have worked out in section ii - a .
if the electric charge at rest in @xmath95d corresponds to the infinite wire with current in @xmath88d , the born - infeld equation for h reads @xmath124 , @xmath125 . but @xmath126 , since only @xmath127 is non - trivial .
also , @xmath128 , due to the translational symmetry along the @xmath89-axis .
then , @xmath129 and so @xmath130 , since it must be constant and there are no sources at infinity .
since in @xmath88d @xmath131 , and planar electrodynamics rules out @xmath94 and @xmath19 , the vanishing of @xmath9 is equivalent to the vanishing of @xmath4 .
this is how we justify the absence of a @xmath4-field in non - linear electrodynamics for a charge at rest , contrary to what happens in @xmath88d .
we shall exploit further this interplay between @xmath85-d and @xmath84-d born - infeld electrodynamics focusing our attention on the problem of magnetic monopoles and vortices .
one of us ( pg ) wants to thank the field theory group of the cbpf for hospitality and pci / mct for support .
this work was supported in part by fondecyt ( chile ) grant 1080260 ( pg ) .
wish to thanks to cbpf and ime for academic support and ctex by the financial help .
l.p.g.a is grateful to faperj - rio de janeiro for his post - doctoral fellowship .
n . expresses his gratitude to cnpq for financial help . | in the framework of three - dimensional born - infeld electrodynamics , we pursue an investigation of the consequences of the space - time dimensionality on the existence of magnetostatic fields generated by electric charges at rest in an inertial frame , which are present in its four - dimensional version .
our analysis reveals interesting features of the model .
in fact , a magnetostatic field associated with an electric charge at rest does not appear in this case .
interestingly , the addition of the topological term ( chern - simons ) to born - infeld electrodynamics yields the appearance of the magnetostatic field .
we also contemplate the fields associated to the would - be - magnetic monopole in three dimensions . |
finite many - body systems ( e.g. , nuclei , small metallic grains , metallic clusters ) robustly maintain similar regularities , despite their different binding interactions .
for example , they all present the odd - even staggering on their binding energies , which are , however , attributed to various mechanisms @xcite . particularly in nuclear systems , the nucleon - nucleon interactions numerically exhibit a random " pattern with no trace of symmetry groups , whereas nuclear spectra follow some robust dynamical features : the nuclear spectral fluctuation is universally observed @xcite ; low - lying spectra of even - even nuclei are orderly and systematically characterized by seniority , vibrational and rotational structures @xcite , beyond @xmath5 ground states without exception . to demonstrate the insensitivity of these robust regularities to the interaction details , and to reveal its underlying origin , random interactions are employed to simulate ( or even introduce ) the variety and chaos into a finite many - body system .
thus , the predominant behaviors in a random - interaction ensemble correspond to dynamical features in a realistic system .
many efforts have been devoted along this direction @xcite .
for instance , similarly to realistic even - even nuclei , the predominance of the @xmath6 ground states @xcite and collective band structures @xcite have been observed in random - interaction ensembles . however , there are only few attempts to study the robustness of nuclear quadrupole collectivity against the random interaction .
this is partly because a random - interaction ensemble potentially gives weaker e2 transitions than a shell - model calculation with
realistic " interactions @xcite .
even so , some robust correlations about the e2 collectivity can be expected .
for example , the alaga ratio between the quadrupole moment ( @xmath7 ) of the @xmath8 state and b(e2 , @xmath9 ) highlights both near - spherical shape and well deformed rotor in random - interaction ensembles @xcite ; ratios of e2 transition rates between yrast @xmath10 , @xmath8 and @xmath11 states are also correlated to the ratio of @xmath8 and @xmath11 excitation energies @xcite . and @xmath12 matrix elements from table i of ref .
@xcite ( i.e. , [ @xmath13 , @xmath14 plots scaled by @xmath15 ) . the @xmath16 parameterization defined in eq .
( [ theta ] ) is illustrated in the red sector .
the @xmath17 correlation is obvious along the @xmath18 direction ( the black dash diagonal line).,scaledwidth=45.0% ] this work further studies the robust correlation between @xmath7 values of the first two @xmath8 states , inspired by a recent experimental survey @xcite . as shown in fig .
[ q_exp ] , this survey demonstrated a global @xmath17 correlation across a wide range of masses , deformations , and @xmath8 energies .
we will make use of random - interaction ensembles to provide an interacting - particle vision to this correlation , and search for other underlying @xmath7 correlations . the statistic analysis based on the elliott su(3 ) model @xcite and the mean - field hartree - bose theory @xcite
is applied .
in our random - interaction calculations , the single - particle - energy degree of freedom is switched off to avoid the interference from the shell - structure detail .
the two - body interaction matrix element , on the other hand , is denoted by @xmath19 as usual , where @xmath20 , @xmath21 , @xmath22 and @xmath23 represent the angular momenta of single - particle orbits ( half integer for fermions and integer for bosons ) , and the superscript @xmath24 labels the total angular momentum of the two - body configurations involved the interaction element . in our calculations , @xmath19 is randomized independently and gaussianly with @xmath25 , which insures the invariance of our random two - body interactions under arbitrary orthogonal transformations @xcite .
all the possibilities of random interactions and their outputs via microscope - calculations construct the two - body random ensemble ( tbre ) @xcite .
obviously , in the tbre , diagonal interaction elements potentially have larger magnitudes . for the shell - model tbre in this work , four model spaces with
either four or six valence protons in either @xmath3 or @xmath26 shell are considered , correspondingly to four nuclei : @xmath27si , @xmath28s , @xmath29cr and @xmath30fe . for the ibm1 tbre
, @xmath3-boson spaces are constructed for nuclei with valence boson numbers @xmath3112 , 13 , 14 and 15 , where @xmath32 and @xmath33 represents @xmath34 and @xmath35 bosons , respectively .
it is noteworthy that a single calculation with random interactions does not match , and does not intend to match , to a realistic nucleus .
it only presents a pseudo nucleus in the computational laboratory .
thus , in this article , model spaces described above are named as corresponding pseudo nuclei for convenience .
for example , the model space with four protons in the @xmath3 shell corresponds to pseudo @xmath27si .
statistic properties of many random - interaction calculations for pseudo nuclei can be related to the robustness of dynamic features in realistic nuclear systems . to insure the statistic validity of our conclusions , 1 000 000 sets of random interactions
are generated for each pseudo nucleus , and inputted into the shell - model or ibm1 calculations .
if one calculation produces a @xmath6 ground state , @xmath7 matrix elements of @xmath8 and @xmath36 states will be calculated and recorded for the following statistic analysis .
in the shell model , the @xmath7 matrix element of one @xmath37 state , @xmath38 , is defined conventionally as @xmath39 where @xmath40 and @xmath41 are single - particle creation and time - reversal operators at orbits @xmath42 and @xmath43 , respectively . a proportional @xmath7 correlation between the first two @xmath8 states
is normally characterized by the ratio of @xmath44 .
geometrically , such correlation also corresponds to a straight line with the polar angle , @xmath45 across the origin in the [ @xmath13 , @xmath14 plane .
for example , the experimental @xmath46 correlation suggested by ref .
@xcite can be illustrated by a diagonal @xmath47 line as expected in fig .
[ q_exp ] .
we also visualize the polar - angle scheme of the [ @xmath13 , @xmath14 plane in fig .
[ q_exp ] . in this work ,
we prefer the statistic analysis based on the polar angle @xmath16 over the @xmath44 ratio because of two reasons .
firstly , the distribution of the @xmath44 ratio spreads widely , so that the statistic detail about @xmath48 correlation may be concealed . in particular , there robustly exists 8% probability of @xmath49 due to the predominance of weak quadruple collectivity , i.e. small @xmath50 , in the shell - model tbre @xcite .
however , we intend as comprehensively as possible to present the statistic detail about the experimental @xmath51 correlation .
the wide statistic range of the @xmath44 ratio may fails this intention . by converting the @xmath44 ratio to the @xmath16 value
, the statistic range is limited between @xmath52 , and a clearer vision around the @xmath48 correlation can be obtained around @xmath53 .
secondly , the @xmath16 parameterization intuitively provides a reasonable geometric standard of symmetric sampling . taking the @xmath48 correlation for example
, there is actually no ( pseudo ) nucleus following exact @xmath48 relation in experiments or our tbre , and yet we can take @xmath54 as the sampling range to represent this correlation .
one sees this sampling range indeed covers a symmetric area related to the @xmath48 correlation in the [ @xmath13 , @xmath14 plane . with the @xmath44 statistic , the determination of symmetric sampling range for one specifically @xmath7 correlation can be controversial or simply another representation of the @xmath16 parameterization .
therefore , all the statistic , analyses , and discussions in this work are based on the @xmath16 value
. distributions from the experimental survey ( exp ) @xcite and the shell - model tbre .
@xmath55 peaks are highlighted correspondingly to @xmath56 correlations , respectively .
error bars correspond to statistic error.,scaledwidth=45.0% ] in fig .
[ q_sm ] , we present @xmath16 distributions of four pseudo nuclei in the shell - model tbre compared with the experimental distribution from ref . @xcite
. the experimental @xmath46 correlation is represented by the main peak around @xmath53 , which is also reproduced by the tbre .
furthermore , several weak peaks around @xmath57 are also observed in both experimental data and random - interaction systems , corresponding to the @xmath58 correlation . as proposed by ref .
@xcite , nuclear rotor models can give the @xmath53 correlation , even although such correlation experimentally occurs in both rotational or non - rotational nuclei .
therefore , we will further examine whether @xmath59 correlations is the symbol of the underlying rotational collectivity in tbre .
firstly , we verify whether @xmath60 correlations accompany rotational spectra in the tbre .
secondly , we search statistic signature of the random - interaction elements that provides the @xmath60 correlations , and trace such signature back to the microscopical hamiltonian of nuclear rotor model , namely the elliott su(3 ) hamiltonian . distributions around @xmath60 correlations ( red circles and blue triangles , respectively ) compared with those in the whole shell - model tbre ( black squares ) .
error bars correspond to statistic error.,scaledwidth=45.0% ] following previous random - interaction studies @xcite , potential rotational spectra with @xmath60 correlations can be characterized by the energy ratio @xmath61 , where @xmath62 and @xmath63 correspond to the excitation energy of yrast @xmath37 and @xmath64 states , respectively .
thus , we plot the @xmath65 distributions with @xmath66 and @xmath67 , respectively , in fig .
[ r_sm ] , and compare them with that in the whole tbre . except for @xmath28s , @xmath68 distributions in both @xmath66 and @xmath67 regions are identical to those in the whole tbre within statistic error . for @xmath28s ,
the @xmath65 distribution has an observable enhancement at @xmath69 with @xmath67 .
namely , the @xmath70 correlation seems to partially originate from the seniority - like level scheme in @xmath28s space .
this observation explains why the @xmath70 peak for @xmath28s is stronger shown in fig .
[ q_sm ] , given the dominance of pairing - like behaviors in the tbre @xcite . nevertheless , there is no special favor on rotational spectra from @xmath60 correlations in the shell - model tbre , consistently with the survey on the realistic nuclear system @xcite .
.@xmath71 values around @xmath72 correlations and @xmath73 elements [ su(3 ) " column ] in the @xmath3 shell .
the index " column presents the integer , @xmath74 , to identify two - body interaction elements .
all the data is organized in an increasing order of the index column . [ cols="^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^ " , ] in table [ peak - fit ]
, we list the best - fit peak positions and intensities for the three sharp peaks in fig .
[ theta_ibm ] .
the @xmath75 and @xmath76 peaks are very close to @xmath77 , i.e.
@xmath56 " correlations , as labeled in fig .
[ theta_ibm ] and table [ peak - fit ] .
the @xmath78 peak is located around @xmath79 , and thus gives @xmath80 , the typical ibm1 @xmath7 ratio at the u(5 ) limit regardless of the boson number .
therefore , we believe the @xmath78 peak may correspond to the vibrational u(5 ) collectivity , and denote it as
u(5 ) " in following analysis . to identify or confirm the collective patterns corresponding to the three sharp peaks in fig .
[ theta_ibm ] , we firstly investigate their @xmath65 distribution , i.e. the predominance of low - lying collective excitations , similarly to our @xmath65 analysis for the shell - model tbre with fig .
[ r_sm ] ; secondly , we adopt the @xmath3-boson mean - field theory to observe dominant nuclear sharps of these peaks .
distributions around three peaks in fig .
[ theta_ibm].,scaledwidth=48.0% ] for the analysis of @xmath65 distributions , we firstly collect all the random interactions , which produce ( @xmath81 , @xmath82 ) points within @xmath83 from peaks in fig . [ theta_ibm ] .
secondly , all @xmath65 values from these interactions are calculated .
thirdly , @xmath65 distributions of these peaks are calculated and presented in fig .
[ r_ibm ] , respectively .
@xmath56 peaks always have large probabilities at rotational limit @xmath84 , which agrees with the rotor - model description .
on the other hand , @xmath65 distributions of u(5 ) peaks are dominated by @xmath85 , corresponding to a typical u(5 ) vibrational spectrum , which supports our u(5 ) assignment for this peak .
our analysis with the @xmath3-boson mean - field theory starts with the @xmath3-boson coherent state for the ground band as @xmath86 value , is determined by minimizing the hamiltonian expectation value of this coherent state as @xmath87 where @xmath88 reaches the minimum of this equation ; @xmath89 , @xmath90 , @xmath91 and @xmath92 are linear combination of @xmath3-boson two - body interaction matrix elements as formulated in ref .
we calculate @xmath93 values for all the interactions with spin-0@xmath94 ground states in the tbre , and perform frequency counting for calculated @xmath93 values .
thus , the ensemble - normalized @xmath95 distribution for the @xmath96th peaks is given by @xmath97 where @xmath98 is the counting number with @xmath99 and @xmath100 , and @xmath101 is that with @xmath99 in the whole ibm1 tbre . around three peaks in fig .
[ theta_ibm ] , as defined in eq .
( [ palpha]).,scaledwidth=48.0% ] fig . [ alpha ]
presents calculated @xmath102s .
the u(5 ) peak only has a significant probability at @xmath103 , corresponding to the @xmath32-boson condensation .
thus , @xmath37 states for the u(5 ) peak are constructed by replacing @xmath32 bosons with @xmath33 bosons in the @xmath32-boson condensation , which agrees with the quadrupole vibration described by the u(5 ) limit .
this further confirms our u(5 ) characterization of this peak . on the other hand , the @xmath56 peaks both have large probabilities for @xmath104 , corresponding to the axially symmetric rotor at the su(3 ) limit .
considering that the @xmath56 peaks also favor su(3 ) rotational spectra with @xmath105 in fig .
[ r_ibm ] , we conclude that both @xmath56 correlations in ibm1 are strongly related to the su(3 ) limit .
conversely , we also derive @xmath56 correlations from the su(3 ) limit of the ibm1 . at the su(3 )
limit , the @xmath8 state is from the ground band with ( @xmath106 , @xmath107 ) and @xmath108 ; yet , the @xmath36 state belongs to the ( @xmath109 , @xmath110 ) representation , which generates @xmath111 and @xmath112 bands with @xmath108 and 2 , respectively @xcite .
thus , the @xmath36 is from either @xmath111 or @xmath112 band , which leads to two phase - different @xmath7 correlations @xmath113 for @xmath114 , @xmath115 and @xmath116 are achieved , corresponding to @xmath58 and @xmath17 correlations , respectively . in the shell - model tbre
, the @xmath117 correlation has a larger probability than @xmath58 one ( see fig .
[ q_sm ] ) . yet , in the ibm1 tbre , these two correlations have roughly equal peak intensities , i.e. probabilities , as shown in fig .
[ theta_ibm ] and table [ peak - fit ] .
this is a major difference between behaviors of the @xmath7 correlations in shell - model and ibm1 tbres .
this difference can be explained according to the @xmath118 assignment of the su(3 ) scheme . in the shell model ,
i.e. the elliott model , the @xmath117 correlation normally emerges with two @xmath37 states from a single @xmath118 representation , which empirically provides a larger probability .
however , in the @xmath3-boson space , both @xmath56 correlations require @xmath8 and @xmath36 states from two different @xmath118 representations , and thus have similar probabilities .
as shown in table [ peak - fit ] , @xmath119 and @xmath120 of @xmath56 correlations are systematically smaller than @xmath121 .
this observation can be explained with eq .
( [ q_nb ] ) . for large but finite @xmath122
, the magnitude of @xmath123 is always smaller that that of @xmath13 , which drives the @xmath119 and @xmath120 value smaller than @xmath121 .
therefore , we attribute the systematical derivation of @xmath124 peak positions from the exact su(3 ) prediction to the finite - boson - number effect , as proposed by ref .
@xcite with consistent-@xmath7 calculations .
to summarize , we observe three proportional correlations between @xmath7 values of the first two @xmath0 states in the tbre .
@xmath1 correlations robustly and universally exists in both shell - model and @xmath3 spaces , consistently with experiments . in the ibm1 tbre , the @xmath2 correlation is also reported . by using the elliot model and the @xmath3-boson mean - field theory
, we can microscopically assign @xmath1 correlations to the rotational su(3 ) symmetry , and the @xmath2 correlation to the quadrupole vibrational u(5 ) limit .
phenomenologically , the anharmonic vibration may also provide the @xmath4 correlation , although its spectral behavior is not observed in the shell - model tbre .
in particular , the invariance of under orthogonal transformation intrinsically provides the shell - model tbre more opportunity to restore part of su(3 ) properties , i.e. @xmath1 correlations , even though these @xmath7 correlations are insensitive to the su(3 ) rotational spectrum as expected based on the experimental survey @xcite . on the other hand
, ibm1 @xmath1 correlations always favor low - lying rotational spectra , which indicates that the ibm is more strongly governed by the dynamic symmetry .
the su(3 ) group reduction rule also qualitatively explains why the shell model more obviously favors the @xmath4 correlation compared with the ibm1 .
low - lying @xmath7 correlations represent intrinsic nuclear collectivity , and thus are more sensitive to the wave - function detail than the spectrum .
therefore , the nuclear quadrupole collectivity may maintain in a far more deep level than the common realization based on the orderly spectral pattern .
the discussion with prof .
y. m. zhao and prof .
n. yoshida is greatly appreciated .
we also thank dr .
z. y. xu for his careful proof reading .
this work was supported by the national natural science foundation of china under grant no . | in the random - interaction ensembles , three proportional correlations between quadrupole moments of the first two @xmath0 states robustly emerge , including @xmath1 correlations consistently with realistic nuclear survey , and the @xmath2 correlation , which is only observed in the @xmath3-boson space .
these correlations can be microscopically characterized by the rotational su(3 ) symmetry and quadrupole vibrational u(5 ) limit , respectively , according to the elliott model and the @xmath3-boson mean - field theory .
the anharmonic vibration may be another phenomenological interpretation for the @xmath4 correlation , whose spectral evidence , however , is insufficient . utf8 |
very recently , babar collaboration reported the measurement of the baryonic @xmath6 decay @xmath7 and observed a new structure in the @xmath8 invariant mass spectrum at @xmath9 @xcite . for simplicity , one could name the new structure as @xmath0 .
soon after the experimental observation , he
_ et al . _ have suggested that @xmath0 could be a @xmath1 molecular state from an effective lagrangian calculation @xcite . theoretically , the molecular concept is well and truly not a new topic but with a history .
it was put forward nearly 40 years ago in ref .
@xcite and was predicted that molecular states have a rich spectroscopy in ref .
the possible deuteron - like two - meson bound states were studied in ref . @xcite . in recent years , some of x " , y " , and
z " new hadrons are ranked as possible molecular candidates . such as , @xmath10 could be a @xmath11 molecular state @xcite ; @xmath12 is interpreted as a @xmath13 state @xcite ; @xmath14 is proposed to be a @xmath15 @xcite or an @xmath16 state @xcite ; @xmath17 is deciphered as a @xmath18 state @xcite ; @xmath19 and @xmath20 could be @xmath21 and @xmath22 states , respectively @xcite . especially , there already have a lot of works discussing baryon resonances with meson - baryon molecular structures , e.g. @xcite . if molecular states can be completely confirmed by experiment , qcd will be further testified and then one will understand the qcd low - energy behaviors more deeply .
therefore , it is interesting to study whether the newly observed @xmath0 state could be a @xmath1 molecular state . in the real world , quarks are confined inside hadrons and the strong interaction dynamics of hadronic systems is governed by nonperturbative qcd effect completely .
many questions concerning dynamics of the quarks and gluons at large distances remain unanswered or understood only at a qualitative level .
it is quite difficult to extract hadronic information quantitatively from the basic theory of qcd .
the qcd sum rule method @xcite is a nonperturbative formulation firmly based on the first principle of qcd , which has been successfully applied to conventional hadronic systems , i.e. mesons or baryons ( for reviews see @xcite and references therein ) . for multiquark states ,
there have appeared fruitful results from qcd sum rules these years ( for a review on multiquark qcd sum rules one can see @xcite and references therein ) . in particular for hadrons containing five quarks , some authors began to study light pentaquark states in refs .
the application of qcd sum rules to heavy pentaquark states was performed in ref .
@xcite for the first time . in this work ,
we devote to investigating that whether the newly observed resonance @xmath0 could be a @xmath1 molecular state ( @xmath23 has a quark content @xmath24 ) in the framework of qcd sum rules . as a byproduct
, the mass for its bottom counterpart @xmath4 is also predicted on the assumption that it could exist ( @xmath25 has a quark content @xmath26 and one should note that it has not been observed ) . in theory , one could expect that the @xmath25 meson should be assigned the same spin parity @xmath27 as @xmath23 .
the rest of the paper is organized as three sections .
we discuss qcd sum rules for molecular states in sec .
[ sec2 ] utilizing similar techniques as our previous works @xcite . the numerical analysis and discussions
are presented in sec .
[ sec3 ] , and masses of @xmath1 and @xmath4 molecular states are extracted out . the sec .
[ sec4 ] includes a brief summary and outlook .
the qcd sum rules for molecular states are constructed from the two - point correlation function @xmath28|0\rangle.\end{aligned}\ ] ] in full theory , the interpolating current for @xmath29 or @xmath25 meson can be found in ref .
@xcite , and the one for nucleon @xmath30 has been listed in ref .
. one can construct the @xmath1 or @xmath4 molecular state current from meson - baryon type of fields @xmath31 where @xmath32 is heavy quark @xmath33 or @xmath34 , and @xmath35 , @xmath36 , @xmath37 , as well as @xmath38 denote light quarks .
the index @xmath39 means matrix transposition , @xmath40 is the charge conjugation matrix , with @xmath41 , @xmath34 , @xmath33 and @xmath42 are color indices .
one should note that meson - baryon molecules in the real world are long objects in which the meson and the baryon are far away from each other .
the currents in this work and in most of the qcd sum rule works are local and the five field operators here act at the same space - time point .
it is a limitation inherent in the qcd sum rule disposal of the hadrons since the bound states are not point particles in a rigorous manner .
lorentz covariance implies that the two - point correlation function in eq .
( [ correlator ] ) has the form @xmath43 according to the philosophy of qcd sum rules , the correlator is evaluated in two ways .
phenomenologically , the correlator can be expressed as a dispersion integral over a physical spectral function @xmath44 where @xmath45 is the mass of the hadronic resonance , and @xmath46 gives the coupling of the current to the hadron @xmath47 . in the ope side , short - distance effects
are taken care of by wilson coefficients , while long - distance confinement effects are included as power corrections and parameterized in terms of vacuum expectation values of local operators , the so - called condensates .
one can write the correlation function in the ope side in terms of a dispersion relation @xmath48 where the spectral density is given by the imaginary part of the correlation function @xmath49 technically , one works at leading order in @xmath50 and considers condensates up to dimension @xmath2 . to keep the heavy - quark mass finite
, one can use the momentum - space expression for the heavy - quark propagator @xcite @xmath51\nonumber\\ & & { } -\frac{i}{4}g^{2}t^{a}t^{b}g^{a}_{\alpha\beta}(0)g^{b}_{\mu\nu}(0)\frac{\rlap / p+m_{q}}{(p^{2}-m_{q}^{2})^{5 } } [ \gamma^{\alpha}(\rlap / p+m_{q})\gamma^{\beta}(\rlap / p+m_{q})\gamma^{\mu}(\rlap / p+m_{q})\gamma^{\nu}\\ & & { } + \gamma^{\alpha}(\rlap / p+m_{q})\gamma^{\mu}(\rlap / p+m_{q})\gamma^{\beta}(\rlap / p+m_{q})\gamma^{\nu}+ \gamma^{\alpha}(\rlap / p+m_{q})\gamma^{\mu}(\rlap / p+m_{q})\gamma^{\nu}(\rlap / p+m_{q})\gamma^{\beta}](\rlap / p+m_{q})\nonumber\\ & & { } + \frac{i}{48}g^{3}f^{abc}g^{a}_{\gamma\delta}g^{b}_{\delta\varepsilon}g^{c}_{\varepsilon\gamma}\frac{1}{(p^{2}-m_{q}^{2})^{6}}(\rlap / p+m_{q } ) [ \rlap / p(p^{2}-3m_{q}^{2})+2m_{q}(2p^{2}-m_{q}^{2})](\rlap / p+m_{q}).\nonumber\end{aligned}\ ] ] the light - quark part of the correlation function can be calculated in the coordinate space , with the light - quark propagator @xmath52 which is then fourier - transformed to the momentum space in @xmath53 dimension .
the resulting light - quark part is combined with the heavy - quark part before it is dimensionally regularized at @xmath54 . equating the two sides for @xmath55 and assuming quark - hadron duality yield the sum rules , from which masses of hadrons can be determined . after making a borel transform and transferring the continuum contribution to the ope side ,
the sum rules can be written as @xmath56 @xmath57 where @xmath58 indicates the borel parameter . to eliminate the hadron coupling constant @xmath59 and extract the resonance mass @xmath45
, one can take the derivative of eq .
( [ sumrule1 ] ) with respect to @xmath60 , divide the result by itself and deal with eq .
( [ sumrule2 ] ) in the same way to get @xmath61 @xmath62 where @xmath63 as a matter of fact , many terms of @xmath64 are approximate to zero because they are proportional to light quarks masses in the calculations .
thereby , we merely present the spectral densities resulted from @xmath65 here . concretely
, they can be written as @xmath66,\nonumber\\ \rho_{2}^{\langle\bar{q}q\rangle^{3}}(s)&=&\frac{m_{q}\langle\bar{q}q\rangle^{3}}{3\cdot2^{4}\pi^{2}}\int_{\lambda}^{1}d\alpha(1-\alpha),\nonumber\\ \rho_{2}^{\langle\bar{q}q\rangle\langle g\bar{q}\sigma\cdot g q\rangle}(s)&=&-\frac{\langle\bar{q}q\rangle\langle g\bar{q}\sigma\cdot g q\rangle}{2^{8}\pi^{4}}s\int_{\lambda}^{1}d\alpha(1-\alpha)^{2},\nonumber\\ \rho_{2}^{\langle g\bar{q}\sigma\cdot g q\rangle\langle g\bar{q}\sigma\cdot g q\rangle}(s)&=&\frac{\langle g\bar{q}\sigma\cdot
g q\rangle^{2}}{2^{10}\pi^{4}}\int_{\lambda}^{1}d\alpha(1-\alpha),\nonumber\\ \rho_{2}^{\langle\bar{q}q\rangle\langle g^{2}g^{2}\rangle}(s)&=&\frac{m_{q}\langle\bar{q}q\rangle\langle g^{2}g^{2}\rangle}{3^{2}\cdot2^{13}\pi^{6}}\int_{\lambda}^{1}d\alpha\frac{(1-\alpha)^{2}}{\alpha^{3}}[6\alpha^{2}(\alpha s - m_{q}^{2})+(1-\alpha)^{2}(3\alpha s-4m_{q}^{2})],\nonumber\\ \rho_{2}^{\langle\bar{q}q\rangle\langle g^{3}g^{3}\rangle}(s)&=&-\frac{m_{q}\langle\bar{q}q\rangle\langle g^{3}g^{3}\rangle}{3\cdot2^{14}\pi^{6}}\int_{\lambda}^{1}d\alpha\frac{(1-\alpha)^{4}}{\alpha^{3}},\nonumber\\ \rho_{2}^{\langle g^{2}g^{2}\rangle\langle g\bar{q}\sigma\cdot g q\rangle}(s)&=&-\frac{m_{q}\langle g^{2}g^{2}\rangle\langle g\bar{q}\sigma\cdot g q\rangle}{3\cdot2^{13}\pi^{6}}\int_{\lambda}^{1}d\alpha\frac{(1-\alpha)^{3}}{\alpha^{2}},\nonumber\end{aligned}\ ] ] and @xmath67e^{-m_{q}^{2}/(\alpha m^{2})}\nonumber\\ & & + \frac{m_{q}^{3}\langle g^{2}g^{2}\rangle\langle g\bar{q}\sigma\cdot g q\rangle}{3^{2}\cdot2^{13}\pi^{6}}\int_{0}^{1}d\alpha\frac{(1-\alpha)^{3}}{\alpha^{3}}e^{-m_{q}^{2}/(\alpha m^{2})}\nonumber\\ & & + \frac{m_{q}\langle g\bar{q}\sigma\cdot g q\rangle\langle g^{3}g^{3}\rangle}{3^{2}\cdot2^{14}\pi^{6}}\int_{0}^{1}d\alpha\frac{(1-\alpha)^{3}}{\alpha^{2}}\bigg(\frac{3}{\alpha}-\frac{m_{q}^{2}}{\alpha^{2}m^{2}}\bigg)e^{-m_{q}^{2}/(\alpha m^{2})}\end{aligned}\ ] ] for @xmath1 or @xmath4 state . the lower limit of integration is given by @xmath68 .
in this section , the sum rule ( [ sum rule q ] ) is numerically analyzed .
the input values are taken as @xmath69 , @xmath70 , @xmath71 , @xmath72 , @xmath73 , @xmath74 , and @xmath75 @xcite . in order to ensure the quality of qcd sum rule analysis
, it is known that one can analyze the ope convergence and the pole contribution dominance to determine the conventional borel window for @xmath58 in the standard qcd sum rule approach : on the one hand , the lower constraint for @xmath76 is obtained by considering that the perturbative contribution should be larger than each condensate contribution to have a good convergence in the ope side ; on the other hand , the upper bound for @xmath76 is obtained by the consideration that the pole contribution should be larger than the continuum state contributions . meanwhile ,
the threshold @xmath77 is not arbitrary but characterizes the beginning of continuum states .
therefore , one naturally expects to find conventional borel windows for studied states to make qcd sum rules work commendably .
however , things go contrary to one s wishes in some cases and it may be difficult to find a conventional work window rigidly satisfying both of two rules , which has been discussed in some works ( e.g. refs .
@xcite ) . referring to the present work , there also arises some similar problem .
concretely , some condensates are very large and play an important role in the ope side , which makes the standard ope convergence ( i.e. the perturbative at least larger than each condensate contribution ) happen only at very large values of @xmath58 .
the consequence is that it is difficult to find a conventional borel window where both the ope converges well ( the perturbative at least larger than each condensate contribution ) and the pole dominates over the continuum . to obtain some useful hadronic information from qcd sum rules
, one could try releasing the rigid convergence criterion of the perturbative contribution larger than each condensate contribution in some case .
the comparison between pole and continuum contributions from sum rule ( [ sumrule2 ] ) for @xmath1 state for @xmath78 is shown in the left panel of fig . 1 , and its ope convergence by comparing the perturbative with other condensate contributions is shown in the right panel .
not too bad for the present plight , there are four main condensates ( i.e. @xmath79 , @xmath80 , @xmath81 , and @xmath82 ) and they could cancel out each other to some extent since they have different signs . besides , most of other condensates calculated are very small and almost negligible .
thus , one could try releasing the rigid ope convergence criterion ( i.e. the perturbative larger than each condensate contribution ) and restrict the ratio of the perturbative to the total ope contribution " ( the sum of the perturbative and other condensates calculated ) at least larger than one half , for example @xmath83 or more . in other words , here we consider the perturbative dominating over the sum of condensates instead of the perturbative larger than each condensate .
furthermore , it is also very important that we have examined that condensates higher than dimension @xmath2 are quite small and the ratio of the perturbative to the total ope contribution " does not change much even adding them ( in the total ope contribution ) , which means that condensates higher than dimension @xmath2 could not radically influence the character of ope convergence here .
all the above factors bring that the ratio of the perturbative to the total ope contribution " can be bigger than @xmath83 at relatively low values of @xmath76 in this work . by way of parenthesis
, one could also visually see that there exist very stable plateaus from the borel curves for the @xmath1 state shown in fig .
= 5.8truecm=5.8truecm = 5.8truecm = 5.8truecm all in all , to test the ope convergence , we have considered the ratio of the perturbative to the total ope contribution " instead of the ratio of the perturbative to each condensate contribution .
this treatment is not freewheeling but has some definite constraints ( i.e. there are merely few important condensates and they could cancel out each other to some extent ; other condensates are almost negligible ) . in this sense
, one could expect that the ope convergence is still under control .
we must truthfully admit that it is not a so good ope convergence as the conventional case , but then one could find a comparatively reasonable work window and extract the hadronic information from qcd sum rules reliably .
thus , we choose some transition range @xmath84 as a compromise borel window and take the continuum thresholds as @xmath85 , and arrive at @xmath86 for @xmath1 state .
considering the uncertainty rooting in the variation of quark masses and condensates , we gain @xmath87 ( the first error reflects the uncertainty due to variation of @xmath77 and @xmath76 , and the second error resulted from the variation of qcd parameters ) or @xmath3 for @xmath1 state . on account of the difficulty encountered in finding a conventional borel window
, one may suppose the nonexistence of @xmath88 molecule itself .
as one possibility , the assumption of its nonexistence indeed should be drawn attention . however , in the present work , we are inclined to make a premise that the @xmath88 molecular state could exist and then study whether it could act as one potential explanation of @xmath0 in view of two mian points : i ) the possibility for the existence of @xmath1 molecule and the molecular interpretation of @xmath0 are not entirely fabricated without any grounds . by an effective lagrangian calculation @xcite ,
_ found that @xmath29 and nucleon can form a loosely bound state with the small binding energy , and @xmath0 can be well explained as the @xmath1 molecular hadron , which is supported by both the analysis of the mass spectrum and the study of its dominant decay channel .
moreover , the observed @xmath89 can also be reasonably described .
ii ) we believe the present result from qcd sum rules could provide another support to the @xmath1 explanation to @xmath0 .
certainly , we must confess to a weakness that it is difficult to find the conventional borel window in the present case . just as we have stated above
, one could try releasing the rigid ope convergence criterion and eventually find the ope convergence is still under control in the present case .
although it is not a so good ope convergence as the conventional case , one could find a comparatively reasonable work window and safely extract the hadronic information from qcd sum rules
. there comes forth the same problem for @xmath4 as the above case for @xmath1 , and we treat it similarly .
the mass of @xmath4 state as a function of @xmath76 from sum rule ( [ sum rule q ] ) is shown in fig .
3 . graphically , one can see there have very stable plateaus for borel curves .
we choose a compromise borel window @xmath90 and take @xmath91 for @xmath4 state . in the work windows ,
we obtain @xmath92 for @xmath4 state .
varying input values of quark masses and condensates , we attain @xmath93 ( the first error reflects the uncertainty due to variation of @xmath77 and @xmath76 , and the second error resulted from the variation of qcd parameters ) or @xmath5 for @xmath4 state .
assuming the newly observed structure @xmath0 by babar collaboration as a @xmath1 molecular state , we calculate its mass value in the framework of qcd sum rules .
technically , contributions of operators up to dimension @xmath2 are included in the ope .
we find that it is difficult to find the conventional ope convergence in this work . via trying releasing the rigid ope convergence criterion
, one could find that the ope convergence is still under control in the present work and the final numerical result for @xmath1 state is @xmath3 , which coincides with the experimental value @xmath9 . in view of
that the conventional ope convergence is not obtained here , thus only weak conclusions can be drawn regarding the explanation of @xmath0 in terms of a @xmath1 molecular state .
meanwhile , one should note that the @xmath88 molecular state is just one possible theoretical interpretation of @xmath0 and there may have some other different explanations for its configuration .
one could expect that contributions from both future experimental observations and theoretical analysis will further reveal the nature structure of @xmath0 .
additionally , we have also studied the bottom counterpart @xmath4 state and predicted its mass to be @xmath5 . by analogy with @xmath1 state , this bottom counterpart state
could be searched in the @xmath94 invariant mass spectrum in future experiments .
k. p. khemchandani , a. martnez torres , and e. oset , eur .
j. a * 37 * , 233 ( 2008 ) ; a. martnez torres , k. p. khemchandani , and e. oset , phys .
c * 77 * , 042203(r ) ( 2008 ) ; d. gamermann , c. garca - recio , j. nieves , and l. l. salcedo , phys .
d * 84 * , 056017 ( 2011 ) ; o. romanets , l. tolos , c. garca - recio , j. nieves , l. l. salcedo , and r. g. e. timmermans , phys . rev .
d * 85 * , 114032 ( 2012 ) .
m. a. shifman , a. i. vainshtein , and v. i. zakharov , nucl .
b147 * , 385 ( 1979 ) ; * b147 * , 448 ( 1979 ) ; v. a. novikov , m. a. shifman , a. i. vainshtein , and v. i. zakharov , fortschr . phys .
* 32 * , 585 ( 1984 ) .
colangelo and a. khodjamirian , in at the frontier of particle physics : handbook of qcd , edited by m. shifman , boris ioffe festschrift vol .
3 ( world scientific , singapore , 2001 ) , pp . 1495 - 1576 ; a. khodjamirian , continuous advances in qcd 2002/arkadyfest .
s. l. zhu , phys .
lett * 91 * , 232002 ( 2003 ) ; r. d. matheus , f. s. navarra , m. nielsen , r. rodrigues da silva , and s. h. lee , phys .
b * 578 * , 323 ( 2004 ) ; j. sugiyama , t. doi , and m. oka , phys .
b * 581 * , 167 ( 2004 ) ; m. eidemuller , phys . lett .
b * 597 * , 314 ( 2004 ) ; z. g. wang , w. m. yang , and s. l. wan , phys .
d * 72 * , 034012 ( 2005 ) .
j. r. zhang and m. q. huang , phys .
d * 77 * , 094002 ( 2008 ) ; phys .
d * 78 * , 094007 ( 2008 ) ; phys .
b * 674 * , 28 ( 2009 ) ; phys .
d * 80 * , 056004 ( 2009 ) ; j. phys .
phys . * 37 * , 025005 ( 2010 ) ; j. r. zhang , m. zhong , and m. q. huang , phys .
b * 704 * , 312 ( 2011 ) . | we present a qcd sum rule analysis for the newly observed resonance @xmath0 by assuming it as a @xmath1 molecular state .
technically , contributions of operators up to dimension @xmath2 are included in the operator product expansion ( ope ) .
we find that it is difficult to find the conventional ope convergence in this work . by trying releasing the rigid ope convergence criterion
, one could find that the ope convergence is still under control in the present work and the numerical result for @xmath1 state is @xmath3 , which is in agreement with the experimental data of @xmath0 . in view of
that the conventional ope convergence is not obtained here , thus only weak conclusions can be drawn regarding the explanation of @xmath0 in terms of a @xmath1 molecular state . as a byproduct , the mass for the bottom counterpart @xmath4 state is predicted to be @xmath5 . |
at certain stages of stellar evolution , some stars show peculiarity of spectral lines in their spectra that argues in favour of an enhanced or depleted abundance of several chemical species in their stellar atmospheres with respect to the solar abundance . abnormally strong lines of silicon in @xmath1 cvn were first reported by antonia maury @xcite , while the strong lines of ionised silicon and ionised strontium were found by annie cannon @xcite in some bright southern stars in the process of determining their spectral class .
the first systematic study and classification of stars of spectral classes b - f with abnormally strong lines of various chemical species was performed by morgan @xcite . to explain the mystery of the observed abnormally strong lines , shapley @xcite suggested the possibility of abundance abnormalities in the atmospheres of these stars .
several decades later , this idea was confirmed by the results of a differential coarse analysis using the curves of growth @xcite .
it should be noted , that the observed peculiar abundance of chemical species is only related to the stellar atmosphere and does not reflect the chemical composition of the entire star .
we turn our attention to two types of stars with abundance anomalies : blue horizontal - branch ( bhb ) stars that burn helium in their core and hydrogen in a shell @xcite , and main sequence stars that burn hydrogen in their core .
bhb stars are found mostly in globular clusters .
comprehensive surveys show that the hot bhb stars have abundance anomalies as compared to the cool bhb stars in the same cluster @xcite .
hot bhb stars show enhanced abundances of iron , magnesium , titanium and phosphor , and depleted helium @xcite .
surveys of rotational velocities of bhb stars indicate that stars with @xmath2 11500 k possess modest rotation with v@xmath3 10 km s@xmath4 , while the cooler stars rotate more rapidly on average @xcite .
slow rotation is an indicator of a hydrodynamically stable atmosphere and the observed abundance peculiarities as well as the brighter @xmath5-magnitudes ( as compared to the theoretical prediction for @xmath5-magnitudes based on stellar evolution models ) of hot bhb stars @xcite can be explained in the frame of atomic diffusion mechanism @xcite at play in their atmospheres .
competition between the gravitational and radiative forces in a hydrodynamically stable atmosphere can cause accumulation or depletion of chemical elements at certain optical depths and lead to a vertical stratification of element abundances .
hui - bon - hoa et al .
@xcite have constructed stellar atmosphere models of hot bhb stars with vertical stratification of elements , that successfully reproduced some of the anomalies mentioned above .
analysing the available high resolution spectra of some hot bhb stars , khalack et al .
@xcite have detected vertical stratification of the abundance of several chemical elements , including iron .
the stellar atmosphere models of hot bhb stars calculated self - consistently with abundance stratifications predict that the iron abundance stratification decreases as a function of @xmath6 and becomes negligible for bhb stars hotter than @xmath7 14000 k @xcite .
this theoretical result is consistent with the detected iron abundance slopes found by khalack et al .
@xcite from spectral line analysis of hot bhb stars with different effective temperatures .
a significant portion of upper main sequence stars show abundance peculiarities of various chemical species and are commonly named chemically peculiar ( cp ) stars @xcite .
george preston @xcite introduced this collective term and divided the cp stars in four groups : cp1 metallic - line stars ( type am - fm ) , cp2 magnetic peculiar b and a - type stars ( type bp - ap ) , cp3 mercury - manganese stars ( type hg - mn ) and cp4 helium weak stars . following a large number of extensive studies of cp stars , new types of abundance peculiarities
have been discovered and an extension of preston s classification was proposed @xcite . in this new scheme ,
the cp4 group includes only magnetic he - weak stars ( sr - ti - si branch ) , while the non - magnetic he - weak stars ( p - ga branch ) are placed in the group cp5 .
the he - rich magnetic stars ( usually having spectral class b2 ) are placed in the group cp6 .
the cp7 group is reserved for the non - magnetic he - rich stars . in this classification
the odd numbers correspond to non - magnetic stars , while groups with the even numbers contain stars that possess a detectable magnetic field .
more details about the history of discovery and study of cp stars can be found in @xcite .
an extensive list of known and suspected ap , hgmn and am stars was recently published by renson & manfroid @xcite , while a list of known magnetic cp stars has been compiled by bychkov et al .
some upper main sequence cp stars show variability of absorption line profiles in their spectra with the period of stellar rotation due to horizontal inhomogeneous distributions of elements abundance in their stellar atmosphere @xcite .
one can see the periodic ( days , months , years ) variation of spectral line profiles with the axial rotation of a cp star due to the variation of contribution from different parts of the stellar atmosphere to these lines .
ap stars also show signatures of a strong magnetic field and its structure correlates with the patches of overabundances ( or underabundances ) of certain elements @xcite .
observationally , it was shown that the properties of the line profile variability due to abundance patches as well as the magnetic field structure of ap stars do not change during several decades @xcite and , therefore , it is usually assumed that stellar atmospheres of ap stars are hydrodynamically stable . in a hydrodynamically stable atmosphere
the presence of a magnetic field can intensify accumulation or depletion of chemical elements at certain optical depths @xcite . from the analysis of line profiles of ionised iron and ionised chromium in the spectra of @xmath8 crb , ryabchikova et al . @xcite
have shown that iron and chromium abundances increase towards the deeper atmospheric layers .
this can be explained in terms of the mechanism of atomic diffusion @xcite .
it appears that the light and the iron - peak elements are concentrated in the lower atmospheric layers of some ap stars ( hd 133792 and hd 204411 ) @xcite .
meanwhile the rare - earth elements ( for example , pr and nd ) are usually pushed into the upper atmosphere of ap stars @xcite .
detection of vertical stratification of element abundances in stellar atmospheres of stars is an indicator of the effectiveness of the atomic diffusion mechanism that may be responsible for the observed abundance peculiarities .
the vertical stratification of element abundances in a hydrodynamically stable atmosphere can be estimated through the analysis of multiple line profiles that belong to the same ion of the studied element @xcite using the zeeman2 code @xcite , for instance .
this approach has been successfully employed to study the vertical abundance stratification in the atmospheres of several bhb @xcite and hgmn stars @xcite . to search for the signatures and to study the abundance stratification of chemical species with optical depth in the atmospheres of cp stars , we have initiated project veselka , which means rainbow in ukrainian and stands for vertical stratification of element abundances . slowly rotating ( v@xmath9 < 40 km / s )
cp stars of the upper main sequence were selected for our study using the catalogue of ap , hgmn and am stars of @xcite .
this limitation for the rotational velocity is imposed with the aim to increase the probability for a star to have a hydrodynamically stable atmosphere , where the atomic diffusion mechanism can produce vertical stratification of element abundances .
a low value of v@xmath9 also leads to narrow and mostly unblended line profiles in the observed spectra , which is beneficial for our abundance analysis .
.cp stars observed with espadons in the frame of project veselka . [ cols= " < , > , > , < " , ] + @xmath10known spectroscopic binary stars [ tab1 ]
the selected slowly rotating cp stars and two normal main sequence stars ( used as reference stars ) have been observed during 2013 - 14 with espadons ( echelle spectropolarimetric device for observations of stars ) at the cfht ( canada - france - hawaii telescope ) employing the deep - depletion e2v device olapa .
the instrument performances as well as the optical characteristics of the spectropolarimetre are described in @xcite .
espadons acquires high resolution ( r= 65000 ) stokes iv spectra throughout the spectral range from 3700 to 10500 in a single exposure @xcite . in order to be able to find convincing signatures of vertical stratification of element abundances from the spectral analysis
, we require spectra with high signal - to - noise ratio ( close to one thousand per bin in the spectral region around 5150 ) . table [ tab1 ] presents a list of the observed slowly rotating chemically peculiar stars and some normal upper main sequence stars .
the first , second and third columns provide respectively the name of a star , its apparent visual magnitude and v@xmath9 value , while the fourth column contains information about its spectral type .
the spectral classes of stars selected for the project veselka are given in table [ tab1 ] taking into account the data from @xcite , simbad astronomical database and the estimates of effective temperature and surface gravity obtained by khalack & leblanc @xcite .
some of the observed program stars are spectroscopic binaries .
our sample also includes several hgmn stars and magnetic bp - ap stars . for each star ,
we have obtained at least two spectra to verify if it is a binary star , or if the observed line profiles are variable with the phase of stellar rotation due to horizontal stratification of element abundances .
the obtained spectra have been reduced using the dedicated software package libre - esprit @xcite which yields both the stokes i spectrum and the stokes v circular polarisation spectrum .
the stokes v spectra are required to estimate and study the stellar magnetic field and its potential contribution to the atomic diffusion mechanism @xcite .
the level of continuum ( free from absorbtion and emission lines ) in each echelle order of the stokes i spectra was approximated with a polynomial function and its coefficients were derived . using this polynomial function ,
the stokes iv spectra were normalized resulting in a continuum with variations no larger than two to three percent . to estimate the effective temperature and gravity of the observed stars ( see next section )
, we have used their non - normalized spectra @xcite .
the abundance analysis of normalized polarimetric spectra has been carried out employing a modified version @xcite of the zeeman2 radiative transfer code @xcite , which requires atomic data and a synthetic stellar atmosphere model ( local temperature , pressure , electronic density , etc . relative to atmospheric optical depth ) characterized by the effective temperature , surface gravity and metallicity to simulate the synthetic line profiles .
the atomic data provided by the vald-2 @xcite and nist @xcite databases have been used for our simulations . to determine the parameters of stellar atmospheres
, one can fit the observed balmer line profiles with synthetic spectra from a grid of stellar atmosphere models with different values of @xmath6 , @xmath11 and metallicity .
the stellar atmosphere models used here were calculated with the phoenix atmospheric code @xcite . with the aim to estimate the effective temperature , surface gravity and metallicity of stars observed in the frame of the project veselka ,
we have generated a new library of high resolution ( 0.05 in the visible range from 3700 to 7700 ) synthetic spectra @xcite using version fifteen of the phoenix code @xcite .
a grid of stellar atmosphere models and corresponding fluxes has been calculated for 5000 k @xmath12 9000 k with a step of 250 k , for 9000 k @xmath13 15000 k with a step of 500 k and for 3.0 @xmath14 4.5 with a step of 0.5 .
the grids of models have been produced for the solar metallicity @xcite as well as for the metallicities [ m / h]= -1.0 , -0.5 , + 0.5 , + 1.0 , + 1.5 assuming a nil microturbulent velocity .
nine balmer line profiles observed in the non - normalized spectra of several cp stars selected for project veselka were fitted using the code fitsb2 @xcite to determine the fundamental parameters of their stellar atmosphere .
an example of the best fit of balmer lines in hd 110380 obtained for the solar metallicity is presented in fig .
this is a f2 m star ( see table .
[ tab1 ] ) with @xmath6 = 6980 k , @xmath11 = 4.19 @xcite , for which our best fit approximates quite well also the caii 3934 and 3968 line profiles in the left side of the balmer @xmath15 line and in its left wing respectively ( see fig . [ fig1 ] ) .
we have found that our grids provide almost the same values of fundamental parameters of stellar atmospheres for the studied cp stars as do the atlas9 grids calculated by castelli & kurucz @xcite .
however , while studying the sensitivity of the determined values of @xmath6 and @xmath11 to the set of balmer lines used , we have found that the use of the atlas9 grids may produce some ambiguity in the determination of fundamental stellar parameters if the effective temperature is close to 10000 k depending on which set of balmer lines is used .
meanwhile , the phoenix-15 grids are not sensitive to the choice of balmer lines in the range of effective temperatures from 9700 k to 12000 k @xcite .
from the analysis of observed line profiles while using the modified zeeman2 code @xcite , we can determine the radial velocity of the studied star , the projection of its rotational velocity to the line of sight v@xmath9 and the abundance of the chemical element responsible for this absorption line .
fig [ fig2 ] shows an example of the feii 4635 line profile observed in hd 95608 that is well fitted by the theoretical profile @xcite .
preliminary analysis has shown that iron lines are not variable in the hd 95608 spectra obtained in a time span of few weeks . therefore , in the case of hd 95608 , we can use the line profile composed from the data of several spectral observations for our analysis ( see fig [ fig2 ] ) . for the abundance analysis
, we usually select unblended ( not contaminated by a contribution from other chemical species ) absorbtion line profiles that belong to a particular ion . in our study , we assume that the core of the line profile is formed mainly at line optical depth @xmath16=1 , which corresponds to a particular layer of the stellar atmosphere.therefore , from the simulation of each line profile that belongs to a particular ion , we can obtain its abundance at a particular layer of the stellar atmosphere ( that corresponds to a particular continuum optical depth @xmath17 ) .
taking into account that the analysed lines usually have different lower energetic levels and oscillator strengths , their cores are generally formed at different optical depths @xmath17 . in this way , we can study the vertical distribution of an element s abundance from the analysis of a large number ( at least ten or more ) line profiles that belong to one or two ions of this element that are detected in the analyzed spectrum @xcite .
a statistically significant vertical stratification of the element s abundance is considered when the abundance change with optical depth exceeds 0.5 dex .
more details on the fitting procedure are given in @xcite .
this method was successfully used by khalack et al .
@xcite to study the vertical abundance stratification of chemical species in bhb stars and by thiam et al .
@xcite in hgmn stars .
using our grids of stellar atmosphere models and synthetic spectra calculated with the phoenix-15 code @xcite we have estimated the values of effective temperature , surface gravity and metallicity for sixteen stars observed in the frame of veselka project @xcite .
our results on @xmath6 and @xmath11 obtained for twelve of these stars are consistent with the previously published data . meanwhile , for the four other stars ( hd 23878 , hd 83373 , hd 95608 and hd 164584 ) we have for the first time reported the estimates of their effective temperature , gravity and metallicity in @xcite .
estimation of average abundances and detection of vertical abundance stratification of some chemical species were carried out for several cp stars selected for project veselka @xcite .
clear evidences of vertical stratification of iron and chromium abundances were found in the stellar atmospheres of hd 95608 and hd 116235 ( see fig .
[ fig3 ] for an example ) . for the analysed chemical species ,
no evidence of vertical stratification was found in the atmospheres of hd 71030 and hd 186568 @xcite .
the radial and rotational velocities determined for the four stars under consideration are consistent with the values found in previous studies . according to leblanc et al .
@xcite , hd 71030 does not show large abundance anomalies and it might be a normal main sequence star .
two normal b - type stars hd 186568 and hd 209459 @xcite have been selected for project veselka as the reference stars ( see table [ tab1 ] ) .
we plan to use them to test the applied method for the detection of vertical abundance stratification and to verify our estimates of average abundance of chemical species well represented by a large number of spectral lines in the analysed spectra .
as was mentioned above , we have not found signatures of vertical abundance stratification from the analysis of hd 186568 spectra .
the solar abundances of titanium and iron were also obtained for hd 209459 , which is known to have no abundance peculiarities @xcite .
the average abundances of oxygen , silicon , iron and chromium were obtained for different rotational phases of hd 22920 that shows clear signatures of the presence of a moderate magnetic field .
all the analyzed elements show variability of their line profiles with the rotational phase .
this argues in favour of a non - uniform horizontal distribution of their abundances . among the studied elements ,
the siii line profiles show the strongest variability with rotational phase @xcite . to study the vertical abundance stratification of chemical species ,
the observed line profiles were fitted with the synthetic ones , simulated assuming a homogeneous horizontal distribution of element abundances and no magnetic field .
this simplification is not physically correct , but for a given rotational phase it provides an estimate for the average abundances of chemical species .
meanwhile , the contribution of the magnetic field is accounted for by the slightly larger value of v@xmath9 obtained as the final result of fitting . among the chemical species represented by a number of unblended spectral lines ,
only silicon and chromium appear to show strong signatures of vertical stratification of their abundance in the stellar atmosphere of hd 22920 @xcite .
[ fig4 ] shows that chromium has a tendency to increase its abundance towards the deeper atmospheric layers in hd 22920 .
similar behaviour with atmospheric depth was also found for the silicon abundance @xcite .
accumulation or depletion of chemical elements at certain optical depths brought about by atomic diffusion @xcite can modify the structure of stellar atmospheres of cp stars @xcite and lead to the observed horizontal @xcite and vertical @xcite stratification of element abundances . for the bp - ap stars
, it is believed that the substantial magnetic field suppresses convection and contribute significantly to the mechanism of atomic diffusion @xcite . in this way , the magnetic field can stabilise the structure of abundance peculiarities in atmospheres of magnetic cp stars , so that they remain stable over more than fifty years of spectral observations @xcite . therefore , for a comprehensive study of the structure of stellar atmospheres in cp stars , it is important to detect and observationally estimate the intensity of vertical abundance stratification of different chemical species @xcite .
such detections are the main goal of project veselka .
these results may lead to improved calculations of the self - consistent stellar atmosphere models with vertical stratification of elements using the phoenix code @xcite , or other atmospheric models such as those of stift & alecian @xcite . for several program cp stars ,
we have estimated their effective temperature , surface gravity and metallicity through the fitting of nine balmer line profiles with synthetic spectra calculated for a grid of stellar atmosphere models . for this aim ,
a new library of grids of stellar atmosphere models and corresponding fluxes has been generated and tested @xcite .
the obtained stellar atmosphere parameters have been used to calculate homogeneous stellar atmosphere models , that were employed to perform an abundance analysis of hd 22920 , hd 23878 , hd 95608 , hd 116235 and hd 186568 @xcite with the help of the modified zeeman2 code @xcite . among the studied cp stars , signatures of vertical stratification of silicon and
chromium abundances were found in hd 22920 @xcite , and of iron and chromium abundances in hd 95608 and hd 116235 @xcite . in order to improve the precision of measurement of the abundance stratification with atmospheric depths for different chemical elements we need to analyse more unblended spectral line profiles that belong to each element .
therefore , we plan to expand our study towards the near infrared spectral region , where absorbtion lines of many metals can be found for the cool upper main sequence stars ( spectral classes a0-f5 ) . for this aim
, we look forward to analyse the spectra from the near ir spectropolarimetre / velocimetre spirou @xcite proposed to be installed in 2017 at the cfht .
the canada foundation for innovation ( cfi ) has approved the production of a copy of spirou that is planned to be installed at a telescope with high aperture located in the southern hemisphere .
the spectropolarimetre spirou will allow the acquisition of high resolution ( r>70000 ) stokes ivqu spectra throughout the spectral range from 0.98 to 2.5 @xmath18 m ( yjhk bands ) in a single exposure @xcite .
such high spectral resolution in the near ir will provide the measurement of radial velocities with a precision higher than one meter per second @xcite .
taking into account the aforementioned characteristics of spirou , we consider it to be very useful to search for signatures of vertical abundance stratification in stars .
the high resolution near ir spectra of cp stars , even with the relatively low signal - to - noise ratio , in combination with their optical spectra could significantly improve the precision of the statistical evaluation of vertical stratification of element abundances .
the abundance analysis of other cp stars selected for project veselka is underway .
the use of available high - resolution and high signal - to - noise spectra of cp stars obtained by other astronomers will be analysed to accumulate an extensive database for vertical stratification of element abundances in these stars .
the database will be used to search for a potential dependence of the vertical abundance stratification of different chemical species relative to the effective temperature similar to the one that we have found for bhb stars @xcite .
we are also going to study the spectra of some magnetic bp - ap stars with the aim to detect vertical stratification of element abundances and to search for possible correlation of the abundance peculiarities with the magnetic field structure .
special attention will be paid to the cp stars in binary systems , because due to the tidal interaction its members usually have a slow axial rotation @xcite , which can lead to a hydrodynamically stable atmosphere .
[ fig5 ] shows an example of two overlapping spectra corrected for radial velocity of the main component in the spectroscopic binary hd 159082 .
one can see there , that the fei and feii line profiles are relatively stable and can be used for the analysis of vertical abundance stratification .
meanwhile , the cri , moi , nii and hei line profiles change significantly during a period of time shorter than one day .
the spectra of the binary stars found in table [ tab1 ] have been incorporated in the database of the binamics project @xcite that provides high quality spectra of magnetic cp stars in binary systems .
we are thankful to the facult des tudes suprieures et de la recherche de luniversit de moncton and nserc for research grants .
the grids of stellar atmosphere models and corresponding fluxes have been generated using the supercomputer _ briarree _ of luniversit de montral , that operates under the guidance of calcul qubec and calcul canada .
the use of this supercomputer is funded by the canadian foundation for innovation ( cfi ) , nanoqubec , rmga and research fund of qubec - nature and technology ( frqnt ) .
this paper has been typeset from a tex / latex file prepared by the authors .
abt h.a . , levato h. , grosso m. 2002 , apj , 573 , 359 alecian g. , & stift m.j .
2010 , a&a , 516 , 53 alecian e. , neiner c. , wade g.a .
2015 , in _new windows on massive stars : asteroseismology , interferometry and spectropolarimetry_ , proc . of iaus 307 ,
g. meynet , c. georgy , j.h .
groh & ph .
stee , 330 barrick g.a .
, vermeulen t. , baratchart s. , et al .
2012 , in _software and cyberinfrastructure for astronomy ii._ , proc . of spie , 8451 , 15 , [ article i d .
84513j ] behr b.b .
2003a , apjs , 149 , 67 behr b.b .
2003b , apjs , 149 , 101 burbidge g.r . & burbidge e.m .
1955 , astrophys .
, 1 , 431 bychkov v.d .
, bychkova l.v . , & madej j. 2003 , a&a 407 , 631 cannon a.j .
& pickering e.c .
1901 , annals of the astronomical observatory of harvard college , 28 , 129 castelli f. , & kurucz r.l . , 2003 , in _modelling of stellar atmospheres_ , proc .
of iaus 210 , eds . :
n. piskunov , w.w .
weiss , & d.f .
gray , a20 delfosse x. , donati j .- f .
, kouach d. et al .
2013 , in proc . of the annual meeting of the french soc . of astronomy and astrophysics , sf2a-2013 ,
l. cambresy , f. martins , e. nuss , & a. palacios , 497 donati j .- f . , semel m. , carter b.d . , et al . , 1997 , mnras , 291 , 658 donati j .- f . ,
catala c. , wade g.a . , et al . , 1999 , a&ass , 134 , 149 donati j .- f . , catala c. , landstreet j.d . , petit p. , & espadons team 2006 , in _solar polarization 4_ , asp conf .
series , 358 , eds .
: r. casini , & b.w .
lites , 362 dworetsky m.m . , &
budaj j. , 2000 , mnras , 318 , 1264 glaspey j.w .
, michaud g. , moffat a.f.j . , & demers s. 1989 , apj , 339 , 926 grevesse n. , asplund m. , suaval a.j .
, & scott p. 2010 ,
ap&ss , 328 , 179 grundahl f. , vandenberg d.a , andersen m.i .
1998 , apj , 500 , l179 grundahl f. , catelan m. , landsman w.b .
, stetson p.b .
, & andersen m. 1999 , apj , 524 , 242 hauschildt p.h . , shore s.n . ,
schwarz g.j . , et al . , 1997 , apj , 490 , 803 hearnshaw j.b .
2014 , _the analysis of starlight : two centuries of astronomical spectroscopy_ , cambrige university press , new york hubrig s. , & castelli f. , 2001 , a&a , 375 , 963 hui - bon - hoa a. , leblanc f. , hauschildt p.h .
2000 , apj , 535 , l43 khalack v. , & wade g. 2006 , a&a , 450 , 1157 khalack v. , leblanc f. , bohlender d. , wade g.a . , behr b.b .
2007 , a&a , 466 , 667 khalack v.r . , leblanc f. , behr b.b . ,
wade g.a . , bohlender d. 2008 , a&a , 477 , 641 khalack v.r . , leblanc f. , behr b.b .
2010 , mnras , 407 , 1767 khalack v. , yameogo b. , thibeault c. , leblanc f. 2014 , in _magnetic fields throughout stellar evolution_ , proc . of iaus 302 ,
eds . : p. petit , m. jardine , & h.c .
spruit , 272 khalack v. , & poitras p. 2015
, in _new windows on massive stars : asteroseismology , interferometry and spectropolarimetry_ , proc . of iaus 307 , eds .
: g. meynet , c. georgy , j.h .
groh & ph .
stee , 383 khalack v.r .
, leblanc f. 2015 , aj , accepted , [ arxiv:1505.08158 ] khochukhov o. , piskunov n. , ilyin i. , ilyina s. , tuominen i. 2002 , a&a , 389 , 420 khokhlova v.l .
1975 , astron .
, 52 , 950 kramida a. , ralchenko yu . , reader j. , & nist asd team 2013 , nist atomic spectra database ( ver .
available : http://physics.nist.gov/asd .
national institute of standards and technology , gaithersburg , md kupka f. , piskunov n. , ryabchikova t.a .
, stempels h.c . , weiss w.w . , 1999 , a&as , 138 , 119 landstreet j.d .
1988 , apj , 326 , 967 leblanc f. , monin d. , hui - bon - hoa a. , hauschildt p.h . , 2009 , a&a , 495 , 937 leblanc f. , hui - bon - hoa a. , khalack v. 2010 , mnras , 409 , 1606 leblanc f. , khalack v. , yameogo b. , thibeault c. , gallant i. 2015 , mnras , submitted maitzen h.m .
1984 , a&a , 138 , 493 mashonkina l. , ryabchikova t. , ryabtsev a. 2005 , a&a , 441 , 309 mathys g. , hubrig s. 1997 , a&as , 124 , 475 maury ac . & pickering e.c .
1897 , annals of the astronomical observatory of harvard college , 28 , 1 michaud g. 1970 , apj , 160 , 641 moehler s. , 2004 , in _the a - star puzzle_ , proc . of iaus 224 ,
: zverko j. , weiss w.w . , iovsk j. , & adelman s.j . , 119 morgan w.w .
1931 , astrophys .
j. , 73 , 104 morgan w.w .
1933 , astrophys .
j. , 77 , 77 napiwotzki r. , yungelson l. , nelemans g. , et al .
2004 , in _spectroscopically and spatially resolving the components of the close binary stars_ , asp conf .
series , 318 , eds .
hilditch , h. hensberge & k. pavlovski , 402 peterson r.c .
, rood r.t . , &
crocker d.a .
1995 , apj , 453 , 214 preston g.w . 1974 , astron .
12 , 257 romanyuk i.i .
, semenko e.a .
, kudryavtsev d.o .
2014 , astrophysical bulletin , 69 , 4 , 427 ryabchikova t. , wade g.a . ; leblanc f. 2003 , in _modelling of stellar atmospheres_ , proc . of iaus 210 ,
n. piskunov , w.w .
weiss , & d.f .
gray , 301 ryabchikova t. , leone f. , kochukhov o. , bagnulo s. 2004 , in _the a - star puzzle_ , proc . of iaus 224 ,
: zverko j. , weiss w.w . ,
iovsk j. , & adelman s.j .
, 580 ryabchikova t. , kochukhov o. , bagnulo s. 2008 , a&a , 480 , 811 recio - blanco a. , piotto g. , aparicio a. , & renzini a. 2004 , a&a , 417 , 597 renson p. , manfroid j. 2009 , a&a 498 , 961 shapley h. 1924 , harvard coll .
observ . bull .
798 , 2 shavrina a.v .
, glagolevskij yu.v .
, silvester j. , et al .
2010 , mnras , 401 , 1882 smith k.c .
1996 , asptrophys . & space sci .
, 237 , 77 song h.f . , maeder a. , meynet g. , et al .
2013 , a&a , 556 , 100 stift m.j . , alecian g. 2012 , mnras , 425 ,
2715 thiam m. , leblanc f. , khalack v. , wade g.a .
2010 , mnras , 405 , 1384 | a new research project on spectral analysis that aims to characterize the vertical stratification of element abundances in stellar atmospheres of chemically peculiar ( cp ) stars is discussed in detail .
some results on detection of vertical abundance stratification in several slowly rotating main sequence cp stars are presented and considered as an indicator of the effectiveness of the atomic diffusion mechanism responsible for the observed peculiarities of chemical abundances .
this study is carried out in the frame of project veselka ( vertical stratification of elements abundance ) for which 34 slowly rotating cp stars have been observed with the espadons spectropolarimeter at cfht . + * key words : * diffusion , line : profiles , stars : abundances , atmospheres , chemically peculiar , magnetic field , rotation @xmath0dpartement de physique et dastronomie , universit de moncton , moncton , n .- b . , canada e1a 3e9 + [email protected] + [email protected] |
density functional theory ( dft ) has dominated atomic - scale materials modeling for the past three decades and will continue to be extremely important . however , there are plenty of problems where the accuracy of dft falls short of requirements .
there are clear reasons to expect dft to struggle in strongly - correlated solids , but these are not the only systems for which dft is insufficient .
take , for example , the problem of distinguishing between molecular crystal phases and competing low - energy polymorphs .
even in relatively simple molecular solids such as crystalline benzene and its polymorphs under pressure , the energy differences of interest are less than a few kj / mol .
the most successful calculations based on dft are only reliable to @xmath010 kj / mol,@xcite and it has recently been shown that the use of _ ab initio _ many - electron wavefunction methods , such as quantum monte carlo ( qmc ) , is essential to tackle this problem successfully.@xcite unlike dft , many - electron wavefunction methods can , in principle , be improved systematically until the required convergence is obtained .
another important example is the adsorption of molecules on surfaces , where dft is sometimes unable to give predictions of useful accuracy .
dft values of surface formation energies of simple paradigmatic materials such as silicon and magnesium oxide depend strongly on the assumed exchange - correlation functional , and there is usually no way of knowing in advance which functional to trust .
the well known problem of calculating electronic band gaps could also be mentioned .
the urgent practical need to go beyond dft in these and other areas is driving current efforts to develop more accurate methods .
there is abundant evidence that there are large classes of problems for which qmc techniques , in particular diffusion monte carlo ( dmc ) , are considerably more accurate that dft.@xcite just recently , using qmc methods , chemically accurate ionization potentials have been obtained for the first - row transition - metal atoms with a mean absolute error of only 0.126 kcal / mol.@xcite by combining high accuracy with good efficiency and scalability , qmc methods promise to bring high accuracy to computational materials science as a matter of routine.@xcite qmc simulations of extended systems are carried out using finite simulation cells subject to periodic boundary conditions .
practical and computational constraints restrict the size of the simulation cell and so introduce finite - size ( fs ) errors , which can be large and are one of the main problems holding back the application of accurate qmc techniques to solids.@xcite quantifying and correcting these errors is an essential part of all qmc simulations of extended systems , particularly when high accuracy is required .
fs errors affect independent - particle approaches such as dft as well as many - body approaches such as qmc . for calculations of perfect crystals , dft fs errors
can be reduced simply by improving the accuracy of the brillouin zone integration , although the errors arise when periodic supercells are used to model aperiodic systems are less easily dealt with .
independent - particle fs errors also affect many - body calculations of perfect crystals@xcite and can be reduced using twist averaging , which is the many - electron equivalent of brillouin zone integration . even after twist averaging
, however , `` many - body '' fs errors with no independent - particle analogues remain .
another important contribution to the fs error in many - body simulations of extended systems arises from the treatment of the potential energy .
the @xmath1 coulomb interaction is inconsistent with the periodicity of the simulation cell and has to be replaced by the ewald interaction , which is the green s function of poisson s equation subject to periodic boundary conditions . unlike the coulomb interaction , the ewald interaction depends on the size and shape of the simulation cell , leading to additional finite - size errors.@xcite one approach to circumventing
this problem is to use a different periodic function , the `` model periodic coulomb '' ( mpc ) interaction , in place of the ewald interaction.@xcite variational quantum monte carlo ( vmc ) simulations using the mpc interaction suffer from smaller fs effects than simulations using the standard ewald interaction .
it has also been shown that using the mpc interaction reduces the fs errors in dmc calculations of ground and excited states.@xcite a drawback of the mpc approach is that it only reduces fs errors arising from the use of the ewald interaction .
the charge density and exchange - correlation hole in a finite simulation cell are often very similar to those of an infinite solid , so the errors in the coulomb energy are indeed primarily due to the errors in the interaction ; but the imposition of periodic boundary conditions also affects the many - electron wavefunction and thus the electronic kinetic energy .
the non - interacting part of the fs error in the kinetic energy may be eliminated by twist averaging , but the many - body contributions are not negligible . under the assumption that the low-@xmath2 behaviour of the structure factor is independent of the choice of simulation cell , chiesa _ _
et al.__@xcite proposed a method to estimate the many - body contributions to the fs errors in both the potential and kinetic energies without abandoning the ewald interaction . employing this correction , which is based on the random - phase approximation at long wavelength
, one can calculate fs corrections within a single simulation .
another approach to the treatment of many - body finite - size errors is provided by the kwee - zhang - krakauer ( kzk ) functional,@xcite which adds a correction computed from the difference between the dft energy evaluated using the local density approximation ( lda ) for an infinite system and the dft energy evaluated using a modified lda specifically designed to reproduce the total energy of the finite simulation cell , including fs errors .
both the standard lda and the kzk lda are parameterized on the basis of dmc simulations of cells of uniform electron gas subject to periodic boundary conditions , but the standard lda uses dmc energies that have been extrapolated to infinite cell size while the kzk lda does not . in this paper
, we systematically study the problem of eliminating fs errors from qmc calculations of real metallic systems , taking lithium and aluminum as examples .
we analyze twist - averaged dmc energies obtained using the ewald interaction and the mpc interaction,@xcite and finite - size corrections based on the chiesa formalism@xcite and the kzk functional.@xcite we also investigate dft - based corrections designed to improve imperfectly twist - averaged results and consider how best to combine the use of twist - averaged boundary conditions@xcite with the kzk functional.@xcite
the diffusion quantum monte carlo method is a stochastic technique for obtaining the ground - state energy of a many - electron system .
dmc has been described in many previous papers@xcite and will not be discussed in detail here , but since this work is focused on technical aspects of dmc simulations we start with a brief explanation .
the dmc algorithm solves the imaginary - time schrdinger equation , @xmath3 where @xmath4 is a @xmath5-dimensional vector defining the positions of all @xmath6 electrons in the simulation cell , @xmath7 is the imaginary time ( a real variable despite its name ) , @xmath8 is the potential energy including electron - electron interactions , and @xmath9 is a constant energy offset .
( we work in hartree atomic units , where the numerical values of @xmath10 , @xmath11 , @xmath12 , and @xmath13 are all equal to 1 . )
the imaginary - time schrdinger equation resembles a @xmath5-dimensional diffusion equation with diffusion constant @xmath14 .
the potential energy term causes the diffusers to `` branch '' ( multiply or die out ) at a position - dependent rate .
the wavefunction @xmath15 is the number density of diffusers , which are normally known as walkers or configurations and are points in the @xmath5-dimensional configuration space , not individual electrons .
the dmc algorithm uses this simple physical interpretation to simulate the imaginary - time evolution of the wavefunction using a finite population of diffusing and branching walkers .
solving the imaginary - time schrdinger equation is useful because it projects out the ground state as @xmath16 .
if the initial wavefunction is expanded as a linear combination of energy eigenfunctions , @xmath17 , the solution of the imaginary - time schrdinger equation @xmath18 is @xmath19 thus , as long as @xmath20 , the wavefunction @xmath21 becomes proportional to @xmath22 as @xmath16 . by gradually adjusting @xmath9 to maintain the normalization of the solution in the large @xmath7 limit , we can find the ground - state energy @xmath23 .
an obvious difficulty with this approach is that the wavefunction @xmath15 , which is not necessarily positive , is interpreted as a walker density , which must be positive .
in fact , a naive application of the dmc algorithm to a many - electron system yields a totally symmetric many - boson ground state of no physical interest .
the fixed - node approximation introduces a trial many - electron ground - state wavefunction , @xmath24 , and forbids walker moves that cause @xmath25 to change sign .
as long as @xmath25 is properly antisymmetric , this is sufficient to ensure that a fermionic solution is obtained .
it may be shown @xcite that energies calculated within the fixed - node approximation are variational : the result is greater than or equal to the many - fermion ground - state energy and tends to the exact energy as the @xmath26-dimensional nodal surface on which @xmath27 tends to the ground - state nodal surface .
the fixed - node approximation is required for dmc simulations of large systems but is the only fundamental limitation of the method .
other approximations , such as the use of a finite time - step or the representation of ions by pseudopotentials , can be made negligible or avoided given sufficient computer time .
fixed - node dmc energies are in most cases comparable in accuracy to energies calculated using the ccsd(t ) method,@xcite which is often known as the `` gold standard '' of quantum chemistry .
the diffusion / branching simulation described above is unstable in practice because the potential energy @xmath8 diverges whenever electrons approach nuclei or each other , leading to uncontrollable branching .
this problem can be overcome using an importance - sampling technique .
the imaginary - time schrdinger equation is re - expressed in terms of the quantity @xmath28 to obtain @xmath29 - [ e_l(\mathbf{r } ) - e_t]f(\mathbf{r},\tau),\ ] ] where @xmath30 is the @xmath5-dimensional gradient operator , @xmath31 is the corresponding laplacian , @xmath32 is the @xmath5-dimensional drift velocity vector , and @xmath33 is the local energy .
the importance - sampled imaginary - time schrdinger equation describes a diffusion process similar to that discussed above , except that the walkers now drift with velocity @xmath34 as well as diffusing and branching .
the branching rate is determined by the shifted local energy @xmath35 instead of the shifted potential energy @xmath36 .
if the trial function is a good approximation to the ground state , the local energy is a smooth function of @xmath37 and the numerical difficulties caused by divergences in @xmath8 are avoided .
the fixed - node approximation is imposed by rejecting walker moves that change the sign of @xmath24 .
our dmc simulations used the casino qmc code @xcite and a trial function of slater - jastrow ( sj ) form , @xmath38 \det[\psi_{n}(\mathbf{r}_i^{\uparrow } ) ] \det[\psi_{n}(\mathbf{r}_j^{\downarrow } ) ] , \label{eq6}\ ] ] where @xmath39 is the position of the ith spin - up electron , @xmath40 is the position of the jth spin - down electron , @xmath41 $ ] is the jastrow factor , and @xmath42 $ ] and @xmath43 $ ] are slater determinants of spin - up and spin - down one - electron orbitals .
these orbitals were obtained from dft calculations using the plane - wave - based quantum espresso code.@xcite a norm - conserving pseudopotential constructed within dft using the perdew - zunger parameterization of the local density approximation@xcite was employed .
we chose a very large basis - set cut - off of 300 ry to guarantee converge to the complete basis - set limit.@xcite the one - electron orbitals , originally expressed as linear combinations of plane waves , were transformed into a blip polynomial basis for efficiency.@xcite the jastrow function @xmath44 consisted of polynomial one - body electron - nucleus ( en ) and two - body electron - electron ( ee ) terms , the parameters of which were optimized by variance minimization at the variational monte carlo ( vmc ) level.@xcite qmc fs errors are conventionally separated into one - body and many - body contributions .
one - body ( independent - particle ) errors arise from the non - interacting kinetic , potential and hartree energies and include shell - filling effects .
these errors can be removed by twist averaging .
@xcite many - body errors arise from the effects of exchange and correlation on the coulomb and kinetic energies and are not removed by twist averaging .
as explained in the introduction , various techniques may be used to reduce or cancel these errors , but none is entirely successful and care is required .
the oldest approach is extrapolation , which remains useful .
the use of the modified periodic coulomb interaction @xcite reduces the coulomb errors but not the kinetic energy errors and must therefore be combined with other techniques . the lda - based kwee - zhang - krakauer ( kzk ) approach @xcite applies corrections obtained from dft calculations carried out using a modified exchange - correlation functional explicitly designed to mimic the dmc many - body errors . to remove single - particle errors and eliminate shell effects in the kinetic energies of metallic systems , we use twist - averaged boundary conditions.@xcite a twist @xmath45 is imposed by insisting that the many - electron wavefunction @xmath46 obeys the bloch boundary condition @xmath47 for all electrons @xmath48 , where @xmath49 is any simulation - cell lattice vector .
expectation values of observables are obtained by averaging over twist vectors @xmath45 uniformly distributed over the simulation - cell brillouin zone : @xmath50 the twists can be chosen from a uniform monkhorst - pack grid,@xcite preferably offset from @xmath51 , or can be chosen randomly , as in this work . the number of twists should be as large as computational resources allow .
as the twist @xmath45 varies , the energies of some of the one - electron states appearing in the slater determinants may cross the fermi level . in the canonical approach
to twist averaging , the electron number is kept constant and the fermi level is allowed to vary with twist .
this makes the twist - averaged total energy slightly too large,@xcite but the bias reduces as the simulation - cell increases in size and is normally negligible . in the grand canonical approach
to twist averaging , the fermi energy is fixed and the electron number is allowed to vary with @xmath45 . as was demonstrated in ref . , energies obtained using grand - canonical twist averaging exhibit much larger fluctuations than energies obtained using canonical twist averaging with the same number of twists .
furthermore , these fluctuations die away very slowly as the number of twists is increased . to save effort
, applications of qmc to real systems normally use the smallest number of twists possible , so canonical twist averaging is preferable despite the fact that it suffers from small systematic errors .
this work only considers twist averaging within the canonical ensemble . in metallic systems , even when substantial computational resources are expended , the fs errors due to incomplete twist averaging are substantial .
we therefore define an incomplete - twist - averaging correction as follows : @xmath52 where @xmath53 is the dft energy computed using a fully converged @xmath2-point mesh and @xmath54 is the twist - averaged dft energy obtained using the same simulation cell and set of twists as the dmc simulation .
the incomplete - twist - averaging correction tends to zero as the dmc twist averaging is improved and works well if the independent - particle finite - size errors are well approximated by their dft equivalents . in practice , this approach allows accurate results to be obtained with surprisingly small sets of dmc twists , even in metals .
we analyze three different methods for correcting the many - body fs errors in dmc results .
two of these use the structure - factor - based corrections proposed by chiesa _
et al . _
@xcite the first employs the standard ewald form of the periodic coulomb interaction and chiesa corrections for both the kinetic and potential energies ; the second uses the mpc@xcite to deal with the coulomb errors and a chiesa correction for the kinetic energy only .
results obtained with both of these methods are expected to be similar in quality.@xcite the third fs - correction method considered here is the kzk approach,@xcite which uses a system - size - dependent local density approximation fitted to the results of dmc simulations of _ finite _ cubic simulation cells of uniform electron gas .
dft energies calculated using the kzk functional incorporate dmc fs errors within an approximation analogous to the lda . to estimate the fs error in the dmc total energy of a given simulation cell , the dft total energy of exactly the same simulation cell is calculated using the kzk functional .
the difference between this value and the dft energy of an infinite simulation cell calculated using the standard lda provides an estimate of the dmc fs error .
the kzk functional was not originally combined with twist averaging , but the combination is easy to implement ( see below ) and very successful . in the following
we explain how to combine twist averaging and kzk corrections . in general ,
any fs correction can be written as @xmath55 where @xmath56 includes contributions from the hartree energy , the electron - nuclear coulomb interaction energy , and the one - body component of the kinetic energy , while @xmath57 is a many - body term that includes contributions from the exchange - correlation energy and the many - body part of the kinetic energy .
more precisely , @xmath56 may be defined as that part of the total finite - size error that is also present in a dft calculation for the same simulation cell and can be corrected using dft results . in their original paper,@xcite kwee
, zhang and krakauer considered the finite - size errors affecting a qmc simulation carried out in a supercell of @xmath58 primitive unit cells with twist @xmath59 .
the corresponding one - particle finite - size error is @xmath60 where @xmath61 is the dft energy obtained using a fully converged @xmath2-point mesh and @xmath62 is the @xmath51-point dft energy of the supercell .
this , of course , can be calculated using an @xmath63 monkhorst - pack grid of @xmath2 points in the primitive brillouin zone .
since @xmath59 , the monkhorst - pack grid includes the origin .
the kzk approximation to the many - body finite - size error is @xmath64 where @xmath65 is the @xmath51-point dft energy of the supercell computed using the kzk functional instead of the standard lda .
the kzk approximation to the total finite - size error is @xmath66 if twist averaging is used , the dmc energy becomes a function of the twist @xmath45 , which lies in the small brillouin zone corresponding to the simulation supercell .
the slater determinants appearing in the twisted trial wavefunction are built using orbitals from a monkhorst - pack grid of @xmath63 @xmath2 points within the larger primitive brillouin zone , offset by @xmath45 from the origin . in a dft context , carrying out a supercell calculation at a non - zero twist @xmath45 is equivalent to approximating the integration over the primitive brillouin zone by a quadrature over this offset grid of @xmath2 points . to help analyze the fs errors , we write the dmc ground - state energy of the infinite simulation cell as @xmath67 where @xmath68 is the twist - averaged dmc energy of the @xmath63 simulation cell and @xmath69 is the required fs correction . in the spirit of kzk
, this is approximated using the formula @xmath70 where @xmath61 is the dft energy computed within the lda using a fully converged @xmath2-point mesh ( which in this work means @xmath71 ) and @xmath72 is the twist - averaged kzk energy for the supercell , computed using the same set of @xmath73 twists employed in the dmc simulations .
the fs correction , @xmath69 , accounts both for the many - body fs errors and for any one - body fs errors not removed by the limited twist averaging employed in the dmc simulations .
the use of twisted boundary conditions requires the use of complex trial wavefunctions and increases the computational cost a little because complex arithmetic is slower than real arithmetic . in the vmc and wavefunction optimization algorithms , since the expectation values of hermitian operators must be real , only the real parts of the local - energy components need to be calculated and collected .
the run - time and programming - time costs of twist averaging are therefore small .
the use of a complex trial function in dmc requires the replacement of the fixed - node approximation , in which the dmc wavefunction is constrained to have the same sign as the trial wavefunction , by the fixed - phase approximation,@xcite in which the dmc and trial wavefunctions are constrained to have the same phase . in practice , however , the fixed - node and fixed - phase algorithms are very similar and little extra coding is required : the real part of the drift vector is used when proposing trial electron moves ; it is neither necessary nor possible to reject node - crossing electron moves ; and , as in vmc , only the real parts of the local energies are gathered .
another important issue in twist averaging is the jastrow factor .
this work uses the same optimized jastrow for each twist vector , as we found that re - optimizing the jastrow factor at every twist provides negligible improvements in the final results.@xcite we note , finally , that the vmc or dmc runs at each twist can be relatively short and need not be fully converged .
the idea is that we collect enough data to achieve an acceptable error bar when the data are averaged over all twist vectors .
if a normal run without twist averaging takes @xmath74 moves to arrive at an acceptable error bar , each twist angle need only be run for around @xmath75 moves .
this section presents dmc results obtained using the three different fs - correction methods explained in the sec .
[ subsec : fse ] . as simple example metals , we have studied lithium ( li ) and aluminum ( al ) , with one and three valence electrons , respectively .
the frozen ionic cores are represented by non - local norm - conserving lda pseudopotentials .
the kzk functional is essentially an lda , so the use of lda pseudopotentials allows us to obtain a consistent comparison of all three finite - size - correction approaches considered .
there is evidence that hartree - fock pseudopotentials may produce more accurate results than lda pseudopotentials when used in dmc simulations , but since our aim is to investigate fs errors , and since these are almost independent of pseudopotential , no advantage would be gained by using another pseudopotential type . to check the accuracy and convergence of the dmc energy as a function of the number of atoms @xmath74 in the simulation cell
, we have performed calculations for a range of different values of @xmath74 ( and thus also different values of @xmath76 ) .
we first investigate the effects of applying chiesa and mpc corrections to the results of @xmath51-point dmc simulations with no twist averaging .
table [ gamli ] shows the @xmath51-point dmc results for li obtained using various different fs correction methods . as expected,@xcite results ( denoted ewald + @xmath77ke + @xmath77pe ) obtained by adding chiesa kinetic and potential energy corrections to the dmc energy calculated using the ewald interaction are in good agreement with results ( denoted mpc + @xmath77ke ) obtained using the modified periodic coulomb interaction with a chiesa correction for the kinetic energy only .
note that the higher - order kinetic energy corrections defined according to equation ( 55 ) in ref .
, are included in @xmath77ke . because of the lack of twist averaging , the single - body fs errors are large and the calculated ground - state energies depend strongly on the size of the simulation cell .
the choice of the @xmath51-point , @xmath78 , maintains the symmetry of the system but usually increases shell - filling effects in metallic systems , making the independent - particle fs errors even worse .
it is clear , however , that no calculation carried out at a single twist vector will yield satisfactorily accurate results .
the use of twist averaging is essential in metals . .
dmc energies of metallic lithium for different numbers @xmath74 of atoms in the simulation cell .
every energy appearing in the table is the outcome of a single @xmath51-point dmc simulation for the appropriate simulation cell .
results obtained by applying chiesa kinetic ( @xmath77ke ) and potential ( @xmath77pe ) finite - size corrections to dmc energies calculated using the ewald interaction agree well with results obtained by applying only the chiesa kinetic correction to dmc energies calculated using the modified periodic coulomb interaction . despite the application of many - body finite - size corrections , the calculated energy depends strongly on the size of the simulation cell .
this indicates that the single - particle finite - size errors are large .
energies are in ev per atom . [ cols="^,^,^,^,^,^,^",options="header " , ] figure [ likzk ] shows how the fs - corrected dmc energies of metallic li depend on system size , allowing an easy comparison of the three different many - body fs - correction methods considered in this work .
all dmc energies are averaged over 24 randomly chosen twists and include @xmath77bz corrections .
red squares indicate dmc results calculated using the ewald interaction with chiesa corrections for the kinetic and potential energies ( ewald + @xmath77ke + @xmath77pe + @xmath77bz ) ; green circles indicate dmc results obtained using the modified periodic coulomb interaction with chiesa corrections for the kinetic energy only ( mpc + @xmath77ke + @xmath77bz ) ; and blue circles indicate kzk - corrected dmc results , again incorporating @xmath77bz corrections . even for the smallest simulation cell considered , with just 32 atoms , the errors in the ewald + @xmath77ke + @xmath77pe + @xmath77bz and mpc + @xmath77ke
+ @xmath77bz total energies are only 15.7 and 12 mev / atom , respectively . the errors in ewald dmc energies corrected using the kzk scheme are even smaller , at approximately 3 mev / atom .
we emphasize the importance of the success of the twist - averaged kzk method from a practical point of view .
it is known , for example , that the cheap and widely used dft approach often fails to provide accurate enough results@xcite to understand the behavior of materials at high pressure .
therefore , to study the very interesting phase diagram of li,@xcite it will be necessary to perform full many - body computations , most likely using dmc .
the drawback is that dmc calculations are typically at least a thousand times more expensive than dft calculations .
the twist - averaged kzk approach allows one to investigate a large number of possible crystal structures and construct the li phase diagram whilst keeping the cost of the dmc simulations within reasonable bounds .
total dmc energies of li as function of the number of particles in the simulation cell .
all energies are averaged over 24 randomly chosen twists and include @xmath77bz corrections .
red squares indicate dmc results calculated using the ewald interaction with chiesa corrections for the kinetic and potential energies ( ewald + @xmath77ke + @xmath77pe + @xmath77bz ) ; green circles indicate dmc results obtained using the modified periodic coulomb interaction with chiesa corrections for the kinetic energy only ( mpc + @xmath77ke + @xmath77bz ) ; and blue circles indicate kzk - corrected dmc results , again incorporating @xmath77bz corrections . ]
figure [ alkzk ] shows how the fs - corrected dmc energy of metallic al depends on the number of atoms in the simulation cell .
all dmc energies are averaged over @xmath79 randomly chosen twists and include @xmath77bz corrections .
red squares indicate fs - corrected results obtained using the ewald interaction with chiesa corrections for the kinetic and potential energies ( ewald + @xmath77ke + @xmath77pe + @xmath77bz ) ; green circles indicate dmc results obtained using the modified periodic coulomb interaction ( mpc ) with chiesa corrections for the kinetic energy only ( mpc + @xmath77ke + @xmath77bz ) ; and blue circles indicate kzk - corrected dmc results , again incorporating @xmath77bz corrections . for all simulation - cell sizes ,
the ewald + @xmath77ke + @xmath77pe + @xmath77bz and kzk results are in almost perfect agreement .
the difference between the mpc + @xmath77ke + @xmath77bz energies and those obtained using the other two fs - correction methods is about 7 mev / atom for a cell containing just 48 atoms and decreases rapidly with increasing simulation - cell size .
the results of this section have shown that the addition of an incomplete twist - averaging correction , @xmath77bz , allows accurate results to be obtained with a remarkably small number of twists , even for metals . in most of the cases studied ( except for the 96-atom li simulation cell ) ,
the @xmath77bz correction lowers the total dmc energy .
the values of @xmath80 for li simulation cells containing 32 , 48 , 72 , 96 , and 144 atoms are -0.0171 , -0.0091 , -0.0006 , + 0.0013 , and -0.0069 ev , respectively . in the case of al simulation cells containing 24 , 32 , 48 , and 72 atoms , the values of @xmath77bz are -0.092 , -0.066 , -0.047 , and -0.050 ev , respectively . a recent paper by shulenberger and mattsson @xcite provided accurate benchmark dmc results for a wide range of different bulk materials .
they required 216 and 64 twists to obtain converged results for li and al supercells containing 28 and 108 atoms , respectively . because we use incomplete twist - averaging corrections , @xmath77bz , we are able to obtain similarly accurate results with considerably fewer twists .
total dmc energies of al as function of the number of particles in simulation cell .
all energies are averaged over 24 randomly chosen twists and include @xmath77bz corrections .
red squares indicate dmc results calculated using the ewald interaction with chiesa corrections for the kinetic and potential energies ( ewald + @xmath77ke + @xmath77pe + @xmath77bz ) ; green circles indicate dmc results obtained using the modified periodic coulomb interaction with chiesa corrections for the kinetic energy only ( mpc + @xmath77ke + @xmath77bz ) ; and blue circles indicate kzk - corrected dmc results , again incorporating @xmath77bz corrections . ]
we have systematically analyzed and compared the various schemes that have been proposed for correcting fs errors in qmc simulations of real metallic systems . we have explained how to combine the use of twist - averaged boundary conditions with the kzk functional and shown the value of incomplete - twist - averaging corrections based on dft .
the reassuring news is that all of the commonly used approaches work well .
we believe that the use of dft - based incomplete - twist - averaging corrections will have an important role to play in dmc simulations of real metallic systems .
the reliance on dft could be considered a drawback , but it is important to bear in mind that _ any _ valid fs - correction method must yield the same total energy in the limit of a large enough simulation cell .
the important question is not whether the unattainable limiting value is correct but how rapidly it is approached .
the use of incomplete - twist - averaging corrections significantly improves this convergence .
furthermore , energies calculated using the twist - averaged kzk scheme ( which implicitly incorporates a @xmath77bz incomplete - twist - averaging correction ) often settle down to a system - size - independent constant more quickly than energies calculated using other methods incorporating @xmath77bz corrections .
we believe that this paper will provide a useful guide and benchmark for researchers using qmc and other many - body electronic structure methods such as ccsd(t ) and the @xmath81 approximation to study metallic systems .
this work was supported by the uk engineering and physical sciences research council under grant ep / k038141/1 , and made use of computing facilities provided by archer , the uk national supercomputing service , and by the high performance computing service of imperial college london .
we acknowledge support from the thomas young centre under grant tyc-101 . | we present a systematic and comprehensive study of finite - size effects in diffusion quantum monte carlo calculations of metals .
several previously introduced schemes for correcting finite - size errors are compared for accuracy and efficiency and practical improvements are introduced . in particular , we test a simple but efficient method of finite - size correction based on an accurate combination of twist averaging and density functional theory .
our diffusion quantum monte carlo results for lithium and aluminum , as examples of metallic systems , demonstrate excellent agreement between all of the approaches considered . |
in table [ tab : optical ] , we list all pulsars in binaries outside of globular clusters .
one sees that their properties vary widely , but one can identify different types on the basis of the spin and orbital properties . for instance , systems separate in clusters by inferred companion mass and orbital period , as can be seen in fig .
[ fig : porbvsmass ] .
below , we briefly describe the different groups and their evolutionary histories ( for reviews , see phinney & kulkarni 1994 ; stairs 2004 ) .
pulsars with massive stellar companions , which formed in binaries in which one star went supernova .
the pulsars are like young , isolated pulsars . probably , there are many more psr+ob(e ) systems in which the pulsar is hidden by the companion s stellar wind .
there should also be pulsars with lower - mass companions , but as yet no secure identifications have been made ( a candidate is psr b1820@xmath011 ; phinney & verbunt 1991 ) .
pulsars formed second in massive binaries , with the first - formed neutron star as a companion . for the one system
known , the first - born neutron star is a pulsar as well , but recycled ( see below ) . in binaries with two stars just below the critical mass required to form a neutron star , the originally more massive star will evolve first and leave a white dwarf . in the process , it may transfer sufficient amounts of matter to make the originally lighter star massive enough to explode , leaving a newly formed pulsar in an eccentric orbit around a massive white dwarf . in binaries with a massive star and a neutron - star companion , the star will eventually evolve . unless the orbit is very wide , it will overflow its roche lobe , and unstable mass transfer to the neutron star will ensue , leading to a spiral in .
the accretion of matter on the neutron star spins it up and , by mechanisms unknown , reduces the magnetic field .
the resulting ` recycled ' pulsar is left in a binary with another neutron star when the remaining helium core is heavy enough to go supernova .
the orbit will be eccentric .
these systems are also called ` high - mass binary pulsars ' ( hmbp ) . for one system ,
the companion neutron star is a pulsar as well .
if , in the above scenario , the helium core is not massive enough to explode , it will form a white dwarf with a co or onemg core and a helium envelope .
a massive white dwarf can also be left if the companion of the pulsar evolved up to the agb before overflowing its roche lobe .
the latter scenario may lead to more accretion , and thus a neutron star spun up to faster periods and with a more strongly reduced magnetic field .
the white dwarf might still have a hydrogen envelope .
in either case , the orbit will be circular .
these systems are also referred to as ` intermediate mass binary pulsars ' ( imbp ) . if the companion of a neutron star is a low - mass star , the mass transfer that ensues when it evolves and overfills its roche lobe is stable .
a lot of mass is transferred , leading to fast spin periods and low magnetic fields , as well as , presumably , greatly increased mass .
if mass transfer started before the helium flash , a helium - core white dwarf will be left , with a hydrogen envelope .
if it started afterwards , a white dwarf with a co core will be formed , and the atmosphere may be either hydrogen or helium . in either case ,
one expects a circular orbit .
these systems are also called ` low - mass binary pulsars ' ( lmbp ) .
recycled pulsars are also found with companions with masses more similar to brown dwarfs .
these companions likely originally had higher mass , but lost most of it in a long x - ray binary phase .
the systems are also called ` very low - mass binary pulsars ' ( vlmbp ) or , because of the strong irradiation and evaporation observed , ` black widow pulsars . '
the various evolutionary scenarios lead to predictions for the properties of the binary pulsars we observe . in particular for the recycled pulsars with low - mass white - dwarf companions , which evolved via stable mass transfer , the predictions seem secure : there should be a relation between companion mass and orbital period , one between eccentricity and orbital period , and the neutron - star masses should have increased ( for a review , phinney & kulkarni 1994 ) . of these predictions ,
the second has been verified ( phinney & kulkarni 1994 ) , but the lack of accurate masses prevented stringent tests of the other two . with continuing high - precision timing , however , the situation has changed .
for instance , nice , splaver , & stairs ( 2004 , these proceedings ) have uncovered clear evidence that the neutron stars are more massive than the typical 1.35@xmath1 inferred from double neutron - star binaries ( thorsett & chakrabarty 1999 ; these should have accreted little ) .
timing measurements , as well as optical studies ( see below ) , have also yielded a number of accurate companion masses , which allow a much improved test of the relation between companion mass and orbital period . in fig .
[ fig : porbmwd ] , we show these masses , with model predictions of tauris & savonije ( 1999 ) overdrawn .
the agreement is impressive , especially when one considers that the models were produced before the accurate masses became available .
in the table with all binary pulsars ( table [ tab : optical ] ) , we also summarise what is known optically .
one sees that for the pulsars with massive stellar companions , which should be bright , two out of three are identified ; the third ( psr j1740@xmath03052 ) is highly obscured . also for the two ` back widow ' pulsars , the two strongly irradiated , bloated , brown dwarf mass companions have been identified .
@lr@ r@r@ rl@ + name & & & & & optical information + & & & & + + + j0045@xmath07319 & 926.3 & 105 & 6.52 & 51.17 & b1v , @xmath2 + b1259@xmath063 & 47.8 & 147 & 5.52&1236.72 & b2e , @xmath3 + j1740@xmath03052 & 570.3 & 741 & 5.55 & 231.03 & + + j0737@xmath03039&2773.5&49&7.70&0.10 + + j1141@xmath06545 & 393.9 & 116 & 6.16 & 0.20 & @xmath4 + b2303@xmath546 & 1066.4 & 62 & 7.47 & 12.34 & @xmath6 , @xmath7 + + j0737@xmath03039&22.7&49&8.32&0.10 + j1518@xmath54904 & 40.9 & 12 & 10.37 & 8.63 & @xmath8 , @xmath9 + b1534@xmath512 & 3.8 & 12 & 7.39 & 0.42 + j1811@xmath01736 & 104.2 & 477 & 8.96 & 18.78 + b1820@xmath011@xmath10 & 279.8 & 429 & 6.51 & 357.76 + b1913@xmath516 & 59.0 & 169 & 8.03 & 0.32 + + j0621@xmath51002 & 28.9 & 37 & 9.98 & 8.32 & @xmath11 + b0655@xmath564 & 195.7 & 9 & 9.66 & 1.03 & dq7 , @xmath12 + j1022@xmath51001 & 16.5 & 10 & 9.78 & 7.81 & @xmath13 , @xmath14 + j1157@xmath05112 & 43.6 & 40 & 9.68 & 3.51 & @xmath15 : + j1435@xmath06100 & 9.3 & 114 & 9.78 & 1.35 & @xmath16 + j1454@xmath05846 & 45.2 & 116 & 8.94 & 12.42 & @xmath17 + j1528@xmath03146 & 60.8 & 19&@xmath18 & 3.18 & @xmath19 : + j1757@xmath05322 & 8.9 & 31 & 9.73 & 0.45 & @xmath20 : + j2145@xmath00750 & 16.1 & 9 & 9.93 & 6.84 & @xmath21 , @xmath22 + + note : optical information without reference refers to unpublished results of ourselves .
colons indicate insecure photometry . for an overview of white - dwarf spectral types ,
see wesemael et al .
briefly , ` d ' is for degenerate dwarf , ` q ' indicates the presence of carbon features in the spectrum , ` a ' the presence of hydrogen , ` c ' the absence of any spectral features .
the subtype @xmath23 is a measure of temperature , @xmath24 .
+ @xmath10 b1820@xmath011 may have a low - mass star as companion rather than a neutron star ( phinney & verbunt 1991 ) .
to @lr@ r@ r@ rl@ + name & & & & & optical information [ references ] + & & & & + + + j0034@xmath00534 & 1.9 & 14 & 9.78 & 1.59 & @xmath25 , @xmath26 + j0218@xmath54232 & 2.3 & 61 & 8.68 & 2.03 & da6 , @xmath27 + j0437@xmath04715 & 5.8 & 3 & 9.20 & 5.74 & dc12 , @xmath28 + j0613@xmath00200 & 3.1 & 39 & 9.71 & 1.20 & brightish star nearby@xmath29 + j0751@xmath51807 & 3.5 & 30 & 9.85 & 0.26 & @xmath30 , @xmath31 + b0820@xmath502@xmath10 & 864.9 & 24 & 8.12&1232.40 & da3 , @xmath32 + j1012@xmath55307 & 5.3 & 9 & 9.69 & 0.60 & da6 , @xmath33 + j1045@xmath04509 & 7.5 & 58 & 9.83 & 4.08 & @xmath34 : + j1232@xmath06501 & 88.3 & 239 & 9.24 & 1.86 & @xmath11 + j1420@xmath05625 & 34.1 & 65 & 9.90 & 40.29 & crowded field + j1455@xmath03330 & 8.0 & 14 & 9.72 & 76.17 & @xmath11 + j1600@xmath03053 & 3.6 & 52 & 9.78 & 14.35 & @xmath11 + j1603@xmath07202 & 14.8 & 38 & 10.18 & 6.31 & @xmath11 ( v.faint ctpt ? ) + j1618@xmath03921 & 12.0 & 118 & 9.55 & 22.80 & @xmath11 + j1640@xmath52224 & 3.2 & 18 & 10.25 & 175.46 & @xmath35 , @xmath36 + j1643@xmath01224 & 4.6 & 62 & 9.60 & 147.02 & @xmath37 : : + j1709@xmath52313 & 4.6 & 25 & 10.31 & 22.71 & @xmath11 + j1711@xmath04322 & 102.6 & 192 & 6.89 & 920.2 & pos .
too unc .
+ j1713@xmath50747 & 4.6 & 16 & 9.93 & 67.83 & @xmath35 , @xmath38 + j1732@xmath05049 & 5.3 & 57 & 9.79 & 5.26 & @xmath11 + j1738@xmath50333 & 5.8 & 34 & 9.61 & 0.35 & da6 , @xmath39 + j1745@xmath00952 & 19.4 & 64 & 9.51 & 4.94 & @xmath11 + b1800@xmath027 & 334.4 & 166 & 8.49 & 406.78 & crowded field + j1804@xmath02717 & 9.3 & 25 & 9.56 & 11.13 & @xmath11 + j1810@xmath02005 & 32.8 & 240 & 9.54 & 15.01 & @xmath11 + b1831@xmath000 & 521.0 & 89 & 8.89 & 1.81 & @xmath40 , @xmath41 + b1855@xmath509 & 5.4 & 13 & 9.68 & 12.33 & @xmath42 , @xmath43 + j1904@xmath50412 & 71.1 & 186 & 10.01 & 14.93 & @xmath11 + j1909@xmath03744 & 2.9 & 10 & 9.52 & 1.53 & da6 , @xmath39 + j1911@xmath01114 & 3.6 & 31 & 9.61 & 2.72 & @xmath11 ( v. faint ctpt ? ) + j1918@xmath00642 & 7.6 & 27 & 9.70 & 10.91 & @xmath11 + b1953@xmath529 & 6.1 & 105 & 9.51 & 117.35 & brightish star nearby@xmath29 + j2016@xmath51948 & 64.9 & 34&@xmath18&635.04 + j2019@xmath52425 & 3.9 & 17 & 9.95 & 76.51 & @xmath44 , @xmath45 + j2033@xmath517 & 5.9 & 25 & 9.93 & 56.31 & pos .
too unc .
+ j2129@xmath05721 & 3.7 & 32 & 9.45 & 6.63 & @xmath11 + j2229@xmath52643 & 3.0 & 23 & 10.51 & 93.02 & @xmath46 : + j2317@xmath51439 & 3.4 & 22 & 10.35 & 2.46 & @xmath11 + + b1957@xmath520 & 1.6 & 29 & 9.18 & 0.38 & @xmath47 + j2051@xmath00827 & 4.5 & 21 & 9.75 & 0.10 & @xmath48 + + @xmath10 b0820 + 02 has a co - wd companion .
j1603@xmath07202 may be an imbp , given its high mass function .
b1831@xmath000 is hardly recycled ; it may have formed differently .
+ @xmath29 these stars are likely unassociated , but prevent deep searches . for the pulsars likely to have white - dwarf companions ,
22 out of 49 have been identified optically ( see also fig . [
fig : porbvsmass ] ) . of these , 7 have spectral types and a further 9 have some colour information . as expected from evolutionary considerations , the spectral type of the one more massive companion indicates a helium atmosphere , while a hydrogen atmosphere is found for companions inferred to be low - mass white dwarfs ( e.g. , van kerkwijk & kulkarni 1995 ) . for most of the 27 unidentified pulsar ,
white dwarf binaries , there are upper limits of roughly 24th magnitude .
these are predominantly the result of a systematic campaign to find all objects bright enough to do spectroscopy , which , as will become clear below , allows one to obtain the most interesting results .
for the counterparts bright enough for spectroscopy , one can model the spectrum , and determine a precise temperature and surface gravity for the white dwarf .
combined with white - dwarf mass - radius relations from cooling models , these yield the white - dwarf mass , which can be used to verify the predictions from binary evolution .
if the orbit is short enough , one can also measure radial velocities and determine the radial - velocity amplitude .
combined with the precise radial - velocity amplitude of the pulsar ( derived from timing ) , this yields the mass ratio and thus , with the mass of the white dwarf , the neutron - star mass .
for the first relatively bright counterpart discovered , that of psr j0437@xmath04715 , the result was disappointing : the spectrum was featureless ( danziger et al .
the temperature of @xmath49k is too low for any features to appear . for the even brighter counterpart of psr j1012
+ 5307 , however , strong hydrogen lines were present in the spectrum , and a model - atmosphere and radial - velocity analysis were done by two groups ( van kerkwijk , bergeron , & kulkarni 1996 ; callanan , garnavich , & koester 1998 ) .
unfortunately , the results were less constraining than hoped .
first , the radial - velocity amplitudes found by the two teams were different .
it turned out this was related to a reduction error by van kerkwijk et al .
( 1996 ) ; a re - reduction of the original data , complemented with more recent results , yields a radial - velocity amplitude of @xmath50 , consistent with the @xmath51 found by callanan et al .
this result should still improve , once small remaining systematic effects have been taken into account .
from the two estimates , the current best estimate of the mass ratio is @xmath52 .
a second discrepancy between the two studies was the value of the surface gravity inferred : van kerkwijk et al .
( 1996 ) found @xmath53 , while callanan et al .
( 1998 ) inferred @xmath54 .
for this difference , the underlying cause turned out to be two different sets of model atmospheres used ( by p. bergeron and d. koester , respectively ) . when the spectra of van kerkwijk et al
. were fitted using the koester models , a surface gravity consistent with that of callanan et al .
is found ( d. koester , pers.comm . ) .
finally , it was found that for the very low masses involved , the mass - radius relation was less securely known than expected .
coincidentally , the slightly different approaches taken by the two teams compensated the differences in surface gravity , leading to the same final white dwarf mass , @xmath55 .
the above mass and mass ratio correspond to a neutron star mass of @xmath56 , where the uncertainty is dominated by the uncertainty in the white - dwarf mass .
if instead we use the white - dwarf mass expected from the models of tauris & savonije ( 1998 ) , @xmath57 ( see fig . [
fig : porbmwd ] ) , we infer a neutron star mass of @xmath58 .
this is a large mass , but similar to is found for psr j0751 + 1807 by nice et al.(2004 , these proceedings ) . clearly , it will be worthwile to try to determine the white - dwarf mass more accurately . the above procedure
was also tried on another short - period pulsar binary , psr j0218 + 4232 .
this system is substantially fainter ( @xmath59 ; table [ tab : optical ] ) , and , unfortunately , it turned out to be beyond the capabilities of the then - available instrumentation : while an accurate temperature could be determined , no useful constraints could be derived on the surface gravity and radial - velocity amplitude ( bassa , van kerkwijk , & kulkarni 2003a ) .
obviously , the current large uncertainties are somewhat discouraging .
fortunately , on all fronts improvements are being made .
first , two more bright counterparts to recycled pulsars have been discovered , to psr j1909@xmath03744 and j1738 + 0333 ( see table [ tab : optical ] ) .
both have spectra similar to j1012 + 5307 , and thus similar temperatures and surface gravities . of the two , psr j1909@xmath03744 is particularly interesting , as the orbit is sufficiently edge - on to allow for an accurate determination of the white - dwarf mass from shapiro delay ( jacoby et al .
combined with much improved white - dwarf mass - radius relations ( from detailed cooling models ; see below ) , this implies we will know the surface gravity of the white dwarf , which we can use to calibrate the model - atmosphere analysis .
the second improvement is that more blue - sensitive and more stable spectrographs are now available on large telescopes , and hence more precise radial - velocity curves can be measured .
it should be possible to obtain mass ratios to better than @xmath60% accuracy .
observing campaigns of both new pulsar binaries are underway .
we conclude by mentioning one last system , psr b0820 + 02 , for which it has been possible to derive an accurate companion mass , @xmath61 ( koester & reimers 2000 ) .
this mass is similar to the masses found for isolated white dwarfs , for which the white - dwarf model atmospheres and mass - radius relations are well understood .
hence , it should be reliable .
unfortunately , the system is unsuitable for determining the neutron star mass , as the orbit is too wide to determine an accurate radial - velocity curve .
however , as we will see below , the accurate white - dwarf mass and temperature allow for interesting constraints on the cooling age .
after the last bit of mass is transferred from the progenitor of the white dwarf to the neutron star , two independent clocks start running .
the first clock is the millisecond pulsar , which will turn on and start slowing down . assuming a spin - down torque @xmath62 , the pulsar age is given by @xmath63 , \label{eq : tpsr}\ ] ] where @xmath64 is the current spin period
, @xmath65 is its rate of change , @xmath66 is the period when the pulsar began spinning down following cessation of mass transfer , and @xmath67 is the `` braking index , '' equal to 3 under the assumption of magnetic dipole radiation . for @xmath68 and @xmath69 , @xmath70 , where @xmath71 is the pulsar `` characteristic age . ''
the second clock is the white dwarf .
after the remaining envelope has been burned off , the white dwarf can only radiate its internal heat , making it cool down as time goes by . in principle , it is fairly easy to estimate the cooling , as the thermal structure of the white dwarf is simple .
the heat is stored in the non - degenerate ions in the white dwarf interior , which is kept nearly isothermal due to the efficient heat conduction by degenerate electrons .
the cooling rate is determined by the much less efficient radiative heat transport near the white dwarf s atmosphere . in practice , there are complications . apart from difficulties in modelling the radiative opacities and dealing with convection
, there are two additional physical processes that play a role .
first , at low temperatures , the ion gas in the core starts to crystallise .
the latent heat released temporarily keeps the white dwarf warmer , but once gone , allows for much more rapid cooling .
this effect is particularly important for more massive , carbon - oxygen white dwarfs .
second , for white dwarfs with relatively thick residual hydrogen envelopes , the pressure at the bottom can be sufficiently high for significant pycno - nuclear fusion , keeping the white dwarf warm longer .
we will return to this below .
the possible use of comparing pulsar and white - dwarf ages was realized immediately upon the first detection of optical emission from white - dwarf counterparts ( kulkarni 1986 ) .
for the white dwarf accompanying psr b0655 + 64 , a relatively low temperature , @xmath72k , was inferred , which implied an age of @xmath73gyr .
this meant that pulsar magnetic fields could not decay completely on this timescale , as had been common wisdom at the time .
the first systematic comparison between spin - down and cooling ages was done by hansen & phinney ( 1998a , b ) . since they had to rely on estimated masses ,
their results were uncertain , but two clear discrepancies stood out : for psr j1012 + 5307 , the inferred cooling age was far shorter than the characteristic age , while for psr b0820 + 02 , the cooling age was significantly longer . we discuss both discrepancies in turn . already in their discovery paper of psr j1012 + 5307 s optical counterpart , lorimer et al .
( 1995 ) noted that the white dwarf was much hotter than expected given the pulsar s characteristic age .
they suggested that the problem might lie in the pulsar age : if the initial period to which the pulsar was spun up was similar to the current one ( i.e. , @xmath74 instead of @xmath75 in eq . 1 )
, then the pulsar age could be equal to a short cooling age .
alberts et al . ( 1996 ) , however , suggested the white - dwarf age had been underestimated : given the low mass , the hydrogen layer on the white dwarf could be quite thick , and residual hydrogen burning could keep the white dwarf hot . indeed , this effect had already been found by webbink ( 1975 ) , in a general study of the evolution of helium white dwarfs in close binaries .
a flurry of modelling followed , confirming the likely presence of a thick hydrogen layer ( driebe et al .
1998 , 1999 ) , and adding complications , such as the duration of the semi - detached phase in which the companion is becoming a white dwarf ( sarna , antipova , & muslimov 1998 ) . at first
, the new models seemed to suggest that most low - mass white - dwarf companions should be fairly bright , but this was disproven by the discovery of a very faint counterpart to psr b1855 + 09 ( van kerkwijk et al .
this led to more careful considerations of the effects of shell flashes during the formation ( e.g. , schnberner , driebe , & blcker 2000 ) , and to the effects of element diffusion on these flashes ( sarna , ergma , & gerkevit - antipova 2000 ; althaus , serenelli , & benvenuto 2001a , b ) .
the current consensus appears to be that below a certain critical mass , the hydrogen layer should be thick and white dwarfs should be relatively hot , @xmath76k even when several gyr old . above the critical mass ,
shell flashes occur , and the hydrogen layer will be too thin to sustain significant residual fusion ; after several gyr , those more massive white dwarfs will have cooled to below 5000k .
the precise value of the critical mass is not known , but , since the shell flashes result from cno burning , almost certainly depends on metallicity ( sarna et al .
2000 ; serenelli et al .
2002 ) .
@lrrrrclr@ + name & & & & & & & h layer + & & & & & & + + j0751@xmath51807 & 2.0&1.1 & 9.85 & 0.26 & 0.160.21 & 3 & thin + j1738@xmath50333 & 1.9&1.4 & 9.61 & 0.35 & & 9 & thick + j1012@xmath55307 & 0.5&0.4 & 9.69 & 0.60&0.120.20 & 8.5 & thick + j0613@xmath00200 & 2.2&1.7 & 9.71 & 1.20 + j1909@xmath03744 & & 9.52 & 1.53&0.190.22 & 8 & thick + j0034@xmath00534 & 1.0&0.5 & 9.78 & 1.59 & & 4 & thin + j1232@xmath06501 & 10&6.2 & 9.24 & 1.86 + j0218@xmath54232 & 5.9&2.6 & 8.68 & 2.03 & & 8 & thin + j2317@xmath51439 & 1.9&0.8 & 10.35 & 2.46 & & 4 : & thin + j1911@xmath01114 & 1.6&1.2 & 9.61 & 2.72 & & 5 : & thin + j0437@xmath04715 & & 9.20 & 5.74&0.200.27 & 4 & thin + b1855@xmath509 & & 9.68 & 12.33&0.240.40 & 5 & thin + j1713@xmath50747 & & 9.93 & 67.83 & & 3.5 & thin + b0820@xmath502@xmath77 & & 8.12&1232.40&0.440.76 & 5 & thin + + @xmath10 distances are inferred from the dispersion measure using the taylor & cordes ( 1993 ) and the cordes & lazio ( 2002 ) models of the galactic electron distribution
. distances with uncertainties ( 1@xmath78 ) are from parallax measurements ( j1909 : jacobi et al . , in prep .
; j0437 : [ vsbb+01 ] ; b1855 : [ ktr94 ] ; j1713 : [ cfw94 ] ) and from modelling the white - dwarf spectrum ( b0820 : [ kr00 ] )
. + @xmath29 masses are from shapiro delay ( j0751 , b1855 , j1713 : [ nss04 ] ; j1909 : jacobi et al . , in prep . ; j0437 : [ vsbb+01 ] ) , and from modelling the white - dwarf spectrum ( j1012 : [ vkbk96,cgk98 ] ; b0820 : [ kr00 ] ) .
+ @xmath79 temperatures are from colours or spectra , except for j2317 + 1439 and j1911@xmath01114 , for which the limits were derived from the magnitude limit , combined with the cordes & lazio distance .
these are marked with a colon to indicate they are uncertain .
+ @xmath77 b0820 + 02 has a co wd companion , but formed like a lmbp . until recently , there was not much supporting observational evidence
indeed , except for psr j1012 + 5307 , all identified white dwarf counterparts were rather cool , implying thin hydrogen envelopes .
could it be that thick envelopes did not occur at all , and that psr j1012 + 5307 was re - born with @xmath74 after all ( bassa et al.2003a ) ?
additional evidence for the presence of thick hydrogen envelopes was uncovered with the identification of rather bright and hot white - dwarf counterparts to two new pulsars with large characteristic ages ( psr j1909@xmath03744 and psr j1738 + 0333 ; see table [ tab : optical ] ) . with those , it has become possible to try to determine empirically the critical mass below which hydrogen layers will be thick
unfortunately , we do not have accurate masses for most white - dwarf companions .
however , we can use the orbital period as a proxy , as this should correlate fairly tightly with the companion mass ( see fig .
[ fig : porbmwd ] and section 1 ) . in table
[ tab : hlayer ] , we list the properties for those low - mass binary pulsars that either have periods less than 3 days or have measured companion masses .
the final column lists whether a thick or a thin hydrogen envelope is required to understand the observed temperature ( assuming the white - dwarf cooling age is similar to the characteristic age ) .
one sees that for all systems with orbital periods in excess of 1.55d , thin hydrogen envelopes are inferred , while below that period they are likely thick ( with one exception ; see below ) . for the system just below the critical period , psr j1909@xmath03744
, we have a mass estimate from shapiro delay . the preliminary results from jacoby et al .
( 2004 , in prep . ) , is @xmath80 .
this is slightly more massive than found from the models of , e.g. , althaus et al .
( 2001b ) , which suggest that @xmath81 white dwarfs should still have thin envelopes . however , given the uncertainties ( e.g. , in metallicity ; serenelli et al .
2002 ) , the agreement seems satisfactory .
while the above paints a consistent picture , there is one exception : for psr j0751@xmath01807 , a low temperature is measured ( bassa , van kerkwijk , & kulkarni 2004 ) .
a low temperature had already been indicated by earlier limits of lundgren et al .
( 1996a ) , and ergma , sarna , & gerkevit - antipova ( 2001 ) suggested the hydrogen envelope might have been reduced shortly after the mass transfer ceased , due to irradiation by the pulsar ( which would be particularly effective in such a close binary ) .
intriguingly , however , our photometry seems to indicate that the white dwarf has no hydrogen whatsoever , but instead a helium atmosphere .
the second discrepancy identified by hansen & phinney ( 1998b ) was that the cooling age for psr b0820 + 02 was significantly longer than the characteristic age .
this was confirmed by koester & reimers ( 2000 ) , who analysed spectra of the optical companion and inferred @xmath82k and @xmath83 .
combined with cooling models , this yields a mass of @xmath84 .
this mass implies a co core , and is typical for an isolated white dwarf , not unexpected given the very wide orbit , and the fact that the pulsar appears to have been recycled only very mildly ( @xmath85s ; @xmath86 g ) .
for a mass of @xmath87 , the implied cooling age is @xmath88myr , which is significantly above the pulsar s characteristic age , @xmath89myr .
koester & reimers ( 2000 ) argue that the white - dwarf cooling models for these masses are secure , and hence that the pulsar age must be larger than the characteristic age . from eq . 1
, one sees that the only way to do that is to decrease the breaking index @xmath23 ; consistency with the cooling age would then require @xmath90 .
if true , the above implies that pulsar spin - down is different from magnetic dipole braking not just for young pulsars ( e.g. , lyne 1996 ) , but also for older ones . as yet , however , the conclusion is less firm than it might appear : the uncertainty in the age listed by koester & reimers ( 2000 ) does not seem to include the uncertainty in the mass of the object . for a less massive white dwarf , the cooling age would be reduced ( e.g. , 150myr for a @xmath91 white dwarf ; see also schnberner et al .
this can be verified with better spectra . with the cooling models in quantitative agreement with the observations , it has become possible to use them .
one particularly interesting case is psr j1911@xmath05958a in the globular cluster ngc 6752 .
this binary is puzzling as it is very far , @xmath92 half - mass radii outside the core ( damico et al .
2002 ; also possenti et al . 2004 , these proceedings ) , and it is unclear how it could have gotten there .
colpi et al .
( 2002 ) investigated different possibilities , and found it was difficult to produce the system either from a primordial binary or by a scattering or exchange event .
instead , they suggested that the binary may have been scattered by a binary composed of two fairly massive black holes . in this scenario ,
the binary would already have formed before the scattering event .
an age estimate could be used to distinguish between the various possibilities . for a primordial origin or an older binary being scattered by a binary black hole
, the white dwarf would most likely be rather old .
if the white dwarf was formed during or shortly after an exchange , however , it could not be much older than the @xmath93gyr the binary can be expected to stay in the outskirts if it is currently on a highly eccentric orbit in the cluster ( colpi et al .
2002 ) .
it turned out that archival eso wide field imager and _ hst _
wfpc2 images were available of the field , and those allowed us to identify the white - dwarf companion of psr j1911@xmath05958a ( bassa et al .
2003b ; the source was also identified in new vlt observations by ferraro et al .
it is relatively bright and hot , with a best - fit temperature of @xmath94k .
as the distance is known , we have a measure of the radius ; the implied mass is low , about @xmath95 , as expected from the pulsar mass function .
the high temperature implies that the white dwarf is rather young , @xmath96gyr , even if it has a thick hydrogen layer ( at the low cluster metallicity , the critical mass is about @xmath97 ; serenelli et al .
this suggests that the system was formed in a simple exchange interaction with another star or binary in the core .
independent of its formation , the system is interesting simply because it is sufficiently bright , @xmath98 , to allow spectroscopy , and measure the white dwarf and neutron star masses using the procedure outlined in section 3 .
an advantage over systems in the field is that the distance is known , which means one can measure the white - dwarf radius using the flux and temperature , and thus have a second constraint on the white - dwarf mass .
alberts , f. , savonije , g. j. , van den heuvel , e. p. j. , & pols , o. r. 1996 , nature , 380 , 676 althaus , l. g. , serenelli , a. m. , & benvenuto , o. g. 2001a , , 323 , 471 althaus , l. g. , serenelli , a. m. , & benvenuto , o. g. 2001b , , 324 , 617 bassa , c. g. , van kerkwijk , m. h. , & kulkarni , s. r. 2003a , , 403 , 1067 [ bvkk03 ] bassa , c. g. , verbunt , f. , van kerkwijk , m. h. , & homer , l. 2003b , , 409 , l31 bassa , c. g. , van kerkwijk , m. h. , & kulkarni , s. r. 2004 , in preparation [ bvkk04 ] bell , j. f. , bessell , m. s. , stappers , b. w. , bailes , m. , & kaspi , v. m. 1995 , , 447 , l117 [ bbs+95 ] callanan , p. j. , van paradijs , j. , & rengelink , r. 1995 , , 439 , 928 [ cvpr95 ] callanan , p. j. , garnavich , p. m. , & koester , d. 1998 , , 298 , 207 [ cgk98 ] camilo , f. , foster , r. s. , & wolszczan , a. 1994 , , 437 , l39 [ cfw94 ] colpi , m. , possenti , a. , & gualandris , a. 2002 , , 570 , l85 cordes , j. m. , & lazio , t. j. w. 2002 , astro - ph/0207156 damico , n. , possenti , a. , fici , l. , manchester , r. n. , lyne , a. g. , camilo , f. , & sarkissian , j. 2002 , , 570 , l89 danziger , i. j. , baade , d. , & della valle , m. 1993 , , 276 , 382 [ dbdv93 ] driebe , t. , schnberner , d. , blcker , t. , & herwig , f. 1998 , , 339 , 123 driebe , t. , blcker , t. , schnberner , d. , & herwig , f. 1999 , , 350 , 89 ergma , e. , sarna , m. j. , & gerkevit - antipova , j. 2001 , , 321 , 71 ferraro , f. r. , possenti , a. , sabbi , e. , & damico , n. 2003 , , 596 , l211 fruchter , a. s. , bookbinder , j. , & bailyn , c. d. 1995 , , 443 , l21 [ fbb95 ] hansen , b. m. s. & phinney , e. s. 1998a , , 294 , 557 hansen , b. m. s. & phinney , e. s. 1998b , , 294 , 569 jacoby , b. a. , bailes , m. , van kerkwijk , m. h. , ord , s. , hotan , a. , kulkarni , s. r. , & anderson , s. b. 2003 , , 599 , l99 [ jbvk+03 ] johnston , s. , manchester , r. n. , lyne , a. g. , bailes , m. , kaspi , v. m. , qiao , g. , & damico , n. 1992 , , 387 , l37 [ jml+92 ] kaspi , v. m. , johnston , s. , bell , j. f. , manchester , r. n. , bailes , m. , bessell , m. , lyne , a. g. , & damico , n. 1994 , , 423 , l43 [ kjb+94 ] kaspi , v. m. , taylor , j. h. , & ryba , m. f. 1994 , , 428 , 713 [ ktr94 ] koester , d. & reimers , d. 2000 , , 364 , l66 [ kr00 ] kulkarni , s. r. 1986 , , 306 , l85 [ kul86 ] lorimer , d. r. , lyne , a. g. , festin , l. , & nicastro , l. 1995 , nature , 376 , 393 [ llfn95 ] lundgren , s. c. , cordes , j. m. , foster , r. s. , wolszczan , a. , & camilo , f. 1996a , , 458 , l33 [ lcf+96 ] lundgren , s. c. , foster , r. s. , & camilo , f. 1996b , asp conf .
ser . 105 : iau colloq . 160 : pulsars : problems and progress , 497 [ lfc96 ] lyne , a. g. 1996 , asp conf .
ser . 105 : iau colloq . 160 : pulsars : problems and progress , 73 nice , d. j. , sayer , r. w. , & taylor , j. h. 1996 , , 466 , l87 [ nst96 ] nice , d. j. , splaver , e. m. , & stairs , i. h. 2004 , these proceedings [ nss04 ] phinney , e. s. & kulkarni , s. r. 1994 , , 32 , 591 phinney , e. s. & verbunt , f. 1991 , , 248 , 21p possenti , a. , damico , n. , corongiu , a. , manchester , r. n. , sarkissian , j. , camilo , f. , & lyne , a. g. 2004 , these proceedings sarna , m. j. , antipova , j. , & muslimov , a. 1998 , , 499 , 407 sarna , m. j. , ergma , e. , & gerkevit - antipova , j. 2000 , , 316 , 84 sch " onberner , d. , driebe , t. , & bl " ocker , t. 2000 , , 356 , 929 serenelli , a. m. , althaus , l. g. , rohrmann , r. d. , & benvenuto , o. g. 2002 , , 337 , 1091 stairs , i. h. 2004 , science , 304 , 547 stairs , i. h. , et al . 2001 , , 325 , 979 [ st++01 ] stappers , b. w. , van kerkwijk , m. h. , bell , j. f. , & kulkarni , s. r. 2001 , , 548 , l183 [ svkbk01 ] tauris , t. m. & savonije , g. j. 1999 , , 350 , 928 taylor , j. h. & cordes , j. m. 1993 , , 411 , 674 thorsett , s. e. & chakrabarty , d. 1999 , , 512 , 288 van kerkwijk , m. h. & kulkarni , s. r. 1995 , , 454 , l141 [ vkk95 ] van kerkwijk , m. h. & kulkarni , s. r. 1999 , , 516 , l25 [ vkk99 ] van kerkwijk , m. h. , bergeron , p. , & kulkarni , s. r. 1996 , , 467 , l89 [ vkbk96 ] van kerkwijk , m. h. , bell , j. f. , kaspi , v. m. , & kulkarni , s. r. 2000 , , 530 , l37 [ vkbkk00 ] van straten , w. , bailes , m. , britton , m. , kulkarni , s. r. , anderson , s. b. , manchester , r. n. , & sarkissian , j. 2001 , nature , 412 , 158 [ vsbb+01 ] webbink , r. f. 1975 , , 171 , 555 wesemael , f. , greenstein , j. l. , liebert , j. , lamontagne , r. , fontaine , g. , bergeron , p. , & glaspey , j. w. 1993 , , 105 , 761 | optical observations of the companions of pulsars can help determine the properties of the binaries , as well as those of their components , and give clues to the preceding evolution . in this review ,
we first describe the different classes of binary pulsars , and present a table with a summary what is known about their optical counterparts .
next , we focus on the class of pulsars that have low - mass , helium - core white dwarf companions .
we discuss attempts to determine the masses of both components using optical spectroscopy , and compare the pulsar spin - down ages with cooling ages of the white dwarfs .
we confirm that for a given age , the lowest - mass white dwarfs are much hotter than the more massive ones , consistent with recent evolutionary models , although with one glaring exception .
we discuss the case of psr b0820 + 02 , where the cooling age indicates a braking index less than 3 , and conclude by describing how cooling ages can be used to test formation scenarios for psr j1911@xmath05958a , a pulsar binary in the outskirts of ngc 6752 . |
with the rapid development of advanced data acquisition technology , processing and analyzing data sampled on 3d shapes or even higher dimensional geometric objects becomes ubiquitous tasks such as those used in a 3d camera , medical imaging , protein structuring , social network analysis etc .
partial differential equation ( pde ) and variational pde based methods have made great success to handle problems in signal and image processing which can be viewed as data on euclidean domains .
it is natural to consider pde based methods to analyze and process signals on a general manifold and to understand geometric structures hidden in the data .
besides the classical implicit method @xcite , finite difference methods @xcite and finite element methods @xcite for solving differential equations on surfaces in @xmath0 , there has been increasing interests of solving pdes on general @xmath1-dimensional manifold in @xmath2 and their applications to data analysis .
for instance , a diffusion geometry framework is developed to investigate the geometric structure of data based on solving laplace - beltrami ( lb ) eigenproblem using integral kernel methods @xcite .
more recently , a moving least square method and a local mesh method are considered to intrinsically solve different types of pdes on manifolds , and then applications to the geometric understanding of ground manifolds have been considered @xcite .
all the above methods of solving pdes on a general manifold @xmath3 are commonly considered @xmath4 is sampled as a set of points @xmath5 , referred as a _ point cloud _ , and discretization of differential operators or approximation of integral equations are relied on available coordinates information of @xmath6 . however , there are many applications whose input information has no point coordinates but only represents an incomplete inter - point distance @xmath7 .
examples include incomplete distance information from sensor network localization @xcite , protein structuring from nmr spectroscopy @xcite and global position from local distance configuration of cities @xcite .
a well - known distance geometry problem @xcite is to find the global position of data based on the incomplete distance information .
the objective in this paper is different from the canonical distance geometry problem .
we would like to develop numerical methods for solving pdes on manifolds represented as incomplete inter - point distance .
one naive way to approach this problem is first to apply a global reconstruction algorithm to obtain a point cloud representation of the input manifold , then established numerical methods for pdes on point clouds can be directly applied .
however , the global coordinate reconstruction might be very time consuming as it involves with semi - positive definite programming whose problem size depends on the number of points and could be very large in practice @xcite . in this paper , we propose a different strategy to solve pdes on manifolds represented as incomplete inter - point distance without conducting global coordinates reconstruction .
our idea is based on two intrinsic properties of differential operators on manifolds .
namely , the definition of a differential operator is only point - wisely depending on local information of the manifold and is invariant under different choices of local coordinates .
this motivates us to only consider to conduct point - wisely local coordinate reconstruction for the associated local neighborhood , then we can point - wisely approximate differential operators based on the reconstructed local coordinates . after that , numerical solver for differential equations can be constructed .
inspired by the classical multi - dimensional scaling @xcite , the full distance matrix is one - to - one corresponding to the gram matrix which can be further used to determine coordinates by its eigen - decomposition .
more importantly , the gram matrix can be essentially viewed as an inner product matrix after certain coordinates shift .
as long as the matrix size , namely the size of a given local neighborhood , is larger than the embedding dimension of the local neighborhood , then the corresponding gram matrix has to be a low - rank matrix .
this suggests us to use low rank as a prior knowledge to reconstruct the gram matrix based on constraints of available distance information .
adapted from the recent advances of the low - rank matrix completion theory @xcite , we consider a nuclear norm regularized convex optimization problem to reconstruct local coordinates based on available distance information .
once local coordinates can be obtained , we apply the intrinsic methods of approximating differential operators on point clouds developed in @xcite to discretize the desired differential equation .
these intrinsic methods can be used discretize different types of pdes including parabolic , elliptic and hyperbolic pdes .
our method can be viewed as natural extensions of these two methods to a new data structure where no coordinate information but only partial inter - point distance information of point clouds is provided .
the outline of our strategy can be as summarized as follows : step 0 : : for the @xmath8-th point , chose its k - nearest neighborhood ( knn ) @xmath9 based on the given incomplete distance or from the prior information if its available .
step 1 : : applying the matrix completion method discussed in section [ sec : coordinaterecon ] to construct local coordinates for knn of the @xmath8-th point .
step 2 : : applying the mls method or the local mesh method in ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) to approximate the desired differential operators at the @xmath8-th point .
this provides the @xmath8-th row of the discretized matrix representation of the desired differential equation .
an immediate advantage of this approach is to save computation time by avoiding the global coordinate reconstruction , which can reduce the complexity quadratically to linearly scaling with the total number of points .
more details about this method will be discussed in section [ sec : solvepdes ] and will be verified in our numerical experiments .
furthermore , this approach enables us to conduct geometric understanding of data without global coordinates reconstruction .
examples include global pattern extraction , comparison and classification as many existing methods conducted using results of differential equations . as a byproduct of global information from pdes
, we also propose a new method for reconstructing manifolds by stitching its local patches .
this new method is much more efficient than the way of direct global reconstruction using matrix completion method .
moreover , it can also overcome possible reconstruction failure using global matrix completion due to the coherent missing information .
the rest of this paper is organized as follows . in section
[ sec : coordinaterecon ] , we propose a low - rank matrix completion model to reconstruct local coordinates and design an algorithm to solve the proposed convex optimization problem based on operator splitting and the augmented lagrangian method .
after that , section [ sec : solvepdes ] is devoted to discuss point - wisely approximate differential operators based on the intrinsic methods proposed in @xcite .
we also discuss our two model pdes , the laplace - beltrami ( lb ) eigenvalue problem and the eiknoal equation for solve geodesic distance on manifolds . as an application of global information using solutions of lb eigenvalue problem
, we propose a new method for reconstructing manifolds by stitching local patches in the lb frequency space in section [ sec : mfdstitching ] .
all numerical experiments about the effectiveness of our local coordinates reconstruction model , the accuracy and robustness of the pdes solvers , and demonstration of manifold reconstruction are discussed in section [ sec : experiments ] .
finally , we conclude our work in section [ sec : conclusion ] .
in this section , we first review the concept of the classical multidimensional scaling ( mds ) . inspired by the classical mds
, we propose a method of reconstructing local coordinates from the given incomplete inter - point distance matrix using the matrix completion theory @xcite . a numerical algorithm based on the augmented lagrangian
is also designed to solve the proposed convex optimization problem .
as we described before , our strategy is to conduct local coordinates reconstruction for the knn @xmath9 of the @xmath8-th point . for convenience ,
we write @xmath10 as @xmath11 points for @xmath9 as the following model does not rely on a typical choice of @xmath9 .
the classical mds is to find a configuration of a set of points @xmath10 in @xmath2 from the given distance matrix @xmath12 is specified by the square of distance . ]
such that the euclidean distance matrix @xmath13 among these points is as close as possible to the given distance matrix @xmath14 @xcite .
[ def : eucli ] a distance matrix @xmath15 is called euclidean if there exist some points @xmath16 such that @xmath17 .
let s write @xmath18 as an @xmath19 matrix and denote the centering matrix by @xmath20 .
a gram matrix @xmath21 is defined as @xmath22 , equivalently , @xmath23 .
note that the euclidean distance matrix is translation invariant , thus , the centering matrix @xmath24 essentially translates the point set centering at the origin .
it is well - known that @xmath14 is euclidean if and only if @xmath25 is positive semidefinite .
furthermore , @xmath26 .
it is clear that @xmath27 . in our case
, we consider the number of points is much larger the dimension of points .
this suggests the low - rank property of @xmath28 as its prior knowledge .
if @xmath14 is euclidean , in other words , @xmath28 is positive semidefinite , then the eigen - decomposition of @xmath29 provides the coordinate reconstruction @xmath30 , where @xmath31 is a diagonal matrix formed by the largest @xmath32 eigenvalues of @xmath33 and @xmath34 are the corresponding eigenfunctions .
different from the mds described as the above , one of the major challenges in our assumption is that only a proportion of the distance matrix @xmath14 is available .
generally speaking , it is impossible to recover the whole matrix only from its small portion of information .
we here assume the information @xmath14 is randomly missing .
different models of semi - definite programming have been considered @xcite in this senario . here , we consider a different model inspired by the low - rank prior information of the gram matrix and the recent advances from matrix completion theory @xcite .
in other words , we are seeking for a symmetric positive - semidefinite low - rank matrix @xmath28 satisfying @xmath35 for @xmath36 in the available portion .
mathematically , let s write @xmath37 as a random portion of the index set and write @xmath38 .
we further denote the available part of @xmath14 as a vector @xmath39 and define a restriction operator @xmath40 by @xmath41 with @xmath42 .
our problem becomes to find a positive semidefinite matrix @xmath43 satisfying @xmath44 .
it is clear that this problem is underdetermined and there are infinite many solutions , most of which are far from the ground truth of @xmath28 and thus not applicable .
regularization method is most common to handle this type of situation .
based on the prior knowledge of @xmath28 is a low rank matrix , we here consider the following problem to reconstruct @xmath28 .
@xmath45 inspired by the matrix completion theory @xcite , we relax the above np - hard problem by replacing @xmath46 as the nuclear norm @xmath47 .
in addition , the input distance matrix is invariant under translation of the reconstructed coordinates @xmath48 .
this translation ambiguity can be fixed by requiring @xmath49 .
this condition is equivalent to row sum of @xmath28 is zero , which can also be obtained from the definition of the gram matrix @xmath28 .
therefore , we propose the following matrix completion model : @xmath50 where @xmath51 denotes as the @xmath52-dimensional column vector with all elements constant @xmath53 . once the above gram matrix is obtained , coordinates can be reconstructed by the eigen - decomposition of @xmath28 as the method used in the classical mds described in section [ subsec : mds ] .
we would like to remark that the constraints and the restriction operator are different from the setting in @xcite , therefore , theoretical analysis conducted in @xcite can not be directly applied in our problem . by considering a dual basis method , we can also theoretically show the exact completion under certain coherence condition .
more details about the theoretical analysis of the model will appear in @xcite .
additionally , in a related work @xcite , the authors propose a semidefinite embedding ( sde ) model as a kernel learning method for nonlinear dimension reduction , where no random missing but only knn information of @xmath14 is considered , and the rank minimization is not considered but maximization of @xmath54 is proposed for maximizing the variance in feature space . for convenience of developing a numerical algorithm of model , we first introduce some notations .
note that the essential freedom of the linear space @xmath55 is @xmath56 , which can be viewed as @xmath57 due to a natural isomorphism @xmath58 .
we denote @xmath59 and further merge the two linear constraints together by defining the following linear operator : @xmath60 and extend the vector @xmath61 as @xmath62 .
note that @xmath47 is the same as @xmath63 as @xmath28 is symmetric and positive semi - definite . in summary ,
the proposed model [ eqn : reconmodel_b2 ] can be written as @xmath64 the model is a semi - definite programming ( sdp ) problem .
it is well - known that the sdp problem can be solved by interior - points method @xcite as a canonical choice .
however , when the number of constraint is with the order @xmath65 as in , the computational time can be with the complexity @xmath66 ( @xcite ) .
thus , we use the augmented lagrangian method @xcite to save both the time and memory consumption .
we first introduce an auxiliary variable @xmath67 to split the model to a multi - variable optimization .
then , the above problem is equivalent to the following problem : @xmath68 the splitting of variable enables us to alternatingly optimize the @xmath69 and @xmath70 .
note that the model is convex , and therefore convergence can also be guaranteed @xcite .
then we introduce an augmented lagrangian of model as follows : @xmath71 where @xmath72 and @xmath73 are two dual variables .
as the desired problem is convex , the saddle point of @xmath74 will be the solution which can be obtained by the following iterative procedure : @xmath75 the objective function in the first step is convex and differentiable . from its first order optimality condition , @xmath76 solves the following linear system : @xmath77 where @xmath78 is the adjoint operator of @xmath79 defined as follows .
it is clear that for any given @xmath80 with @xmath81 , the conjugate of @xmath79 is given by @xmath82 .
here we define symmetric matrices @xmath83 and @xmath84 by:@xmath85 and @xmath86 for any @xmath87 . as @xmath88 is a symmetric and positive definite matrix ,
there are numerous numerical solvers to approach the solution of the above problem . here , we typically chose the conjugate gradient method to solve . the second step of is to find the nearest point of @xmath89 in the positive semi - definite cone .
its closed - form solution can be simply realized by truncating all the information in the negative eigenvalue part .
more precisely , for a symmetric matrix @xmath33 with eigen - decomposition @xmath90 , such eigenvalue hard thresholding ( evht ) operator can be defined as follows : @xmath91 with this definition , the second step of can be calculated as follows : @xmath92 we summarize the above iterative method in algorithm [ alg : reconmodel_b ] . after finding an inner - product matrix by algorithm [ alg : reconmodel_b ] , eigen - decomposition of @xmath93
can ultimately reconstruct the coordinates of points @xmath10 from the incomplete distance @xmath61 .
set @xmath94 .
set @xmath95 .
solve @xmath96 . @xmath97 .
@xmath98 .
@xmath99 .
let @xmath100 . in algorithm [ alg :
reconmodel_b ] , step 2 is usually time consuming because the full eigen - decomposition has time complexity @xmath101 . because ultimately the solution of @xmath28 and @xmath102 should have low rank , it is safe to only carry out the first @xmath103 eigenfunctions by setting @xmath103 is relatively big enough comparing with the rank of @xmath28 but also small in terms of the matrix size .
for instance , in our later numerical experiments for 3d point clouds , we use the matlab function `` eigs '' to find the largest @xmath104 eigenvalues and truncate all the other parts if @xmath105 . in this way
, the time consumption of step 2 can be apparently decreased , and numerical results suggest no difference between this simplification and calculating full eigen - decomposition . [
rk : lowrankd ] in addition , we would like to remark that the distance matrix is also a low rank matrix provided by @xmath106 due to the relationship between @xmath28 and @xmath14 satisfying @xmath107 , where @xmath108 is the linear operator to project diagonal component of @xmath28 to a column vector .
therefore , we can also directly reconstruct @xmath14 based on its available information .
this leads to a low rank minimization model as follows : @xmath109 the above model is exactly the same as the general matrix completion theory discussed in @xcite , which also indicates that the minimum required amount of @xmath14 for successful reconstruction increases is proportion to the matrix rank . not that the rank of @xmath14 is higher than the rank of matrix @xmath28 . plus , the less flexibility of @xmath28 due to the positive semi - definite condition can also help to reduce the requirement of valid distance information for successful reconstruction .
therefore , we would expect that model requires a higher rate of valid distance than the proposed model .
in fact , our numerical experiments in section [ sec : experiments ] also verify this observation .
in this section , after a brief review of the moving least square ( mls ) method @xcite and the local mesh ( lm ) method @xcite for solving pdes on manifolds represented as point clouds , we demonstrate our proposed methods of solving an elliptic eigenvalue problem and a nonlinear hyperbolic equation on manifold represented as incomplete distance information .
different from traditional implicit or triangle mesh representation of surfaces in @xmath110 , which the implicit methods @xcite , finite difference methods and finite element methods @xcite have been proposed to solve pdes on surfaces . however , it is not straightforward to use these methods in our case as the incomplete distance data structure has no global mesh structure and the ground manifold could be with dimension higher than 2 in a high co - dimensional ambient space .
as the definitions of differential operators on manifolds are coordinate invariant , we can approximate differential operators point - wisely based on its local coordinate reconstruction obtained from the matrix completion method proposed in section [ sec : coordinaterecon ] .
namely , each given point , its reconstructed local neighborhood can be viewed as a point cloud , therefore , we apply moving least square ( msl ) method @xcite or local mesh ( lm ) method @xcite to approximate the differential operators . these methods can achieve high order accuracy and enjoy more flexibility .
they can be applied to manifolds with arbitrary dimensions and codimensions . to make the paper self - sufficient , we briefly discuss these two methods in this section .
we denote the index set of the knn of the @xmath8-th point as @xmath9 and write @xmath111 as the reconstructed euclidean coordinates of the neighborhood @xmath9 . based on @xmath112 ,
its tangent space and normal space can be determined by standard principle component analysis ( pca ) @xcite , which is provided by eigensystem of the covariance matrix @xmath113 of @xmath9 defined by : @xmath114 where @xmath115 is the centroid of @xmath112 .
if the intrinsic dimension of the manifold is @xmath1 , then the jump of @xmath113 s eigenvalues guides the splitting @xmath116 . here
@xmath117 represents the tangent space spanned by @xmath118 corresponding to the @xmath1 largest eigenvectors of @xmath113 , and @xmath119 represents the normal space spanned by the rest of the eigenvectors of @xmath113 . as @xmath120 form orthogonal axes near the point @xmath121 , a new _ virtual coordinates _ of @xmath122
can be obtained by @xmath123 .
therefore , the manifold structure near @xmath121 can be approximated by a degree 2 polynomial map @xmath124 whose coefficients are given by the following moving least square problem : @xmath125 where the standard multi - index notation is used for @xmath126 and @xmath127 is a weight function whose typical choice can be @xmath128 and @xmath129 .
similarly , any function @xmath130 or vector filed @xmath131 defined @xmath112 can also be interpolated as a degree 2 polynomial @xmath132 or @xmath133 .
overall , we obtain a local polynomial interpolation of @xmath112 as @xmath134 and a local polynomial interpolation of function defined on @xmath112 as @xmath132 .
therefore , the metric tensor @xmath135 near @xmath121 can be otained by @xmath136 .
this enables us to approximate the following three intrinsic differential operators at @xmath121 : @xmath137 where @xmath138 , @xmath139 and the einstein summation is used in the above formulas .
other differential operators can also be approximated using similar method .
once these differential operators can be approximated , then a finite difference type of method can be applied to solve differential equations on the given data .
we refer @xcite for more details about this approach of approximating differential operators on manifold represented as point clouds . the local mesh method proposed in @xcite
is another method to approximate the above differential operators from the virtual local coordinates .
namely , at @xmath8-th point , a local connectivity can be constructed for the projection image of @xmath112 on @xmath117 through the standard delaunay triangulation .
thus , this connectivity can be directly inherited on @xmath112 .
if we denote the simplex of the first ring of @xmath121 by @xmath140 , the differential operators at each point can be approximated by weighted average as follows : @xmath141 moreover , a finite element type of method can be applied to estimate the mass matrix and stiffness matrix .
therefore , differential equations like laplace - beltrami eigenvalues problems can be solved .
we refer @xcite for more detailed discussion about construction of mass matrix and stiffness matrix .
we remark that the above constructions are conducted point - wisely . as long as a differential operator is well - defined on a manifold ,
namely , its definition does not depend on the choice of local coordinate , then the above procedure can consistently approximate the desired differential operators . a different strategy can also be considered is to reconstruct coordinates first using the proposed gram matrix completion algorithm , then approximate differential operators based on the global reconstructed coordinates .
we would like to point out that the proposed methods of approximating differential operators based on local reconstruction enjoys advantages of computation efficiency and memory consumption .
in fact , the most time - consuming part of the coordinate reconstruction algorithm [ alg : reconmodel_b ] is the eigenvalue hard thresholding .
consider a incomplete distance data with @xmath52 nodes , the complexity of each eigenvalue thresholding step is @xmath142 for global reconstruction if only the largest @xmath103 eigenvalues are computed . in our local reconstruction strategy ,
we only reconstruct coordinates of @xmath143 nearest points near @xmath121 with complexity @xmath144 . when @xmath52 is very large ( such as @xmath145 used in our experiments ) and @xmath143 is small ( @xmath146 30 in our experiments ) , the local reconstruction strategy is apparently much more efficient than global reconstruction strategy .
our numerical experiments discussed in section [ sec : experiments ] also support this point ( see table [ tab : timelocalvsglobal ] ) .
moreover , the local reconstruction method supports parallel computation which can further reduce time consumption to @xmath101 . additionally , the memory consumption of global reconstruction is @xmath147 and the matrix is not sparse , which often crashes the program due to insufficient memory for large size data .
however , our local reconstruction strategy only requires the memory @xmath65 which avoids the problem of memory limit exceeding .
the first pde we consider is an elliptic eigenvalue problem of the laplace - beltrami ( lb ) operator @xmath148 .
the lb operator is self - adjoint and elliptic , so its spectrum is discrete .
we denote the eigenvalues of @xmath149 as @xmath150 and the corresponding eigenfunctions as @xmath151 satisfying the following equations @xcite : @xmath152 the set of lb eigenfunctions @xmath153 forms an orthonormal basis of the space of @xmath154 functions on @xmath4 .
the set @xmath155 is called lb eigensystem of @xmath4 . due to the intrinsic definition of the lb operator @xmath156 , the induced lb
eigensystem @xmath157 is also completely intrinsic to the manifold geometry and provides an intrinsic and systematic characterization of the underlying geometry @xcite .
recently , there have been increasing interests in using the laplace - beltrami ( lb ) eigensystem for 3d shape analysis as well as point clouds analysis @xcite .
therefore , it would be also important to consider solve the lb eigensystem for manifolds represented as incomplete inter - point distance , then all existing methods of conducting data analysis using the lb eigensystem can also be adapted to the incomplete distance data structure .
our numerical solvers are based on the numerical solvers for point clouds .
first , we apply the matrix completion model for point - wise local coordinate reconstruction .
either mls or lm can be then applied to obtain the discretized matrix form of the equation , where the only step we need to conduct is to use the reconstructed local coordinates for the lb operator discretization as we discussed in the section [ subsec : reviewmlslm ] . as
the lb operator is invariant under different choices of the local coordinates , the rigid motion ambiguity from the local coordinate reconstruction will not introduce inconsistency across different coordinate patches , which guarantee our method can provide satisfactory numerical results as we illustrate in the numerical section .
we would like to remark that it is relatively straightforward to have the local consistency of the differential operator as long as the available distance information satisfying certain incoherence condition as we discussed in our theoretical validation work @xcite .
however , the stability of the discretization is hard to obtain , which will be considered in our future work .
the second equation we consider is the eikonal equation , a special type of nonlinear hyperbolic pde , on manifold which is used to characterize the geodesic distance , an intrinsic measurement between two points on manifold .
the eikonal equation for the distance map @xmath1 to a given set @xmath158 on a manifold @xmath4 can be stated as follows : @xmath159 when @xmath158 only includes a single point @xmath32 , the distance map can be denoted as @xmath160 . in practice , provided points coordinates and mesh structure , model can be solved by fast marching and fast sweeping methods @xcite . for manifolds represented as point clouds ,
a local mesh method based on fast marching is proposed in @xcite .
as long as the local coordinate reconstruction is obtained using the proposed matrix completion model , then a local delaunay mesh structure can be constructed .
after that , we repeat the local mesh method to conduct the fast marching as this approach only depends on the distance information of the first ring structure of the local mesh , which is again rigid motion invariant and will not be affected due to the rigid motion ambiguity from the local coordinate reconstruction .
as we mentioned before , solutions of differential equations on manifolds can provide global information for understanding data structure .
thus , pdes can be viewed as bridges " linking between local information and global information .
this shares the same sprit as think globally , fit locally " discussed in @xcite although no pdes are involved . therefore , without global reconstruction , some global analysis of point clouds such as pattern recognition , comparisons and classification can be further considered for data represented as incomplete distance . in this section
, we illustrate results on applications of manifold reconstruction and dimension reduction based on solutions of the laplace - beltrami eigenvalue problem on incomplete distance data .
the patch stitching problem is to reconstruct the coordinate of a point cloud @xmath161 from coordinates of @xmath162 subsets ( referred as patches ) @xmath163 , where @xmath164 denotes the index set for patch @xmath165 . in practice , for each patch @xmath164 , the local coordinates @xmath166 , which are provided as input information or are reconstructed from distance information , have the same pair - wise distance and geometry as the restricted global coordinate @xmath167 with possible rigid motion difference . in other words , @xmath168 , where @xmath169 is an orthogonal matrix for rotation and reflection
, @xmath170 is a vector with all element equal to 1 , and @xmath171 is the translation vector .
a straightforward approach to solve @xmath172 and @xmath173 is to minimize the quadratic loss : @xmath174 such model is a non - convex problem since @xmath172 are required to optimize over the non - convex domain of orthogonal transforms . in @xcite , the authors proposed a three - stage patch stitching method from the local coordinate by synchronizing the reflection , rotation and translation successively .
however , the quality of synchronization is strongly relies on the proportion of the overlapping index .
more recently , a spectral relaxation method is proposed to solve the above problem in @xcite .
this method essentially relax the set of orthogonality constraints and perform well if neighborhood patches has reasonable enough overlapping points . here
, we propose an alternative method to reconstruct the global coordinate from local coordinate , or merely incomplete distance information of each patch .
based on solution of lb eigen - problem incomplete distance information discussed in section [ subsec : solvepde ] , we consider the lb eigen - system as a `` bridge '' to connect the information of different patches .
this idea of this new method is essentially align each local patch in the spectral domain instead of the original euclidean coordinate domain . as the global information has been captured from the lb eigen - system ,
the overlapping of different patches is not directly required here .
the idea of this approach is inspired by approximating coordinates using lb eigenfunctions .
for a given point cloud @xmath161 , it can be approximated by the linear combination of the lb eigenfunctions as @xmath175 .
here , @xmath176 represents the first @xmath177 lb eigenfunctions @xmath178 and @xmath179 represents coefficients of coordinates as column vectors . since the mass matrix @xmath180 is generally unknown without the prior knowledge of the structure information , the coefficients matrix @xmath181 can be obtained in the following least square sense : @xmath182 consider the relationship between restricted global coordinate @xmath183 and the input local coordinate @xmath184 , we already have that for any @xmath185 , @xmath186
. therefore the norm @xmath187 in local patch @xmath188 can be equivalently written as @xmath189 .
since @xmath190 is still a translation vector in @xmath191 , the norm can be simplified as @xmath192 without loss of generality . as a result , with the given @xmath193 and all the local coordinate @xmath184
, we propose to the following model for stitching point clouds @xmath194 : @xmath195 once @xmath172 and @xmath173 are obtained , we can find all the adjusted coordinates @xmath168 , which leads to the global coordinate @xmath196 ultimately .
different from the synchronization method and the spectral relaxation method considered in @xcite , we design the following method to solve the nonconvex problem by updating @xmath181 , @xmath172 and @xmath173 iteratively : @xmath197 where we update @xmath181 and @xmath173 using newton s method and it is straightforward to check @xmath198 we next describe a method of updating the orthogonality constrained variable @xmath172 .
the operator @xmath199 proposed in @xcite is designed for searching the gradient descent direction on the stiefel manifold , the geometric description of the set of orthogonal matrices . to realize the operator @xmath200
, we first define the skew - symmetric operator as @xmath201 then the new trial point @xmath202 which satisfies @xmath203 can be generated by @xmath204 where the step size @xmath205 can be obtained by setting fixed value or line search methods .
more detailed discussion about the orthogonality preserving property and convergence of this approach can be found in @xcite . with the above explanation of , we summarize an algorithm of solving as algorithm [ algorithm stitching ] .
set the initial values such that @xmath206 . by roughly estimating the orthogonal transform matrix @xmath207 between neighbourhood patches @xmath208 and @xmath164 with overlapping points
, one can set the initial guess of @xmath209 . using the broad first search ( bfs ) scheme ,
the rotation matrix of all connected patches can be estimated . if a patch @xmath208 is isolated to any visited patches in the bfs algorithm , set @xmath210 .
@xmath211 .
@xmath212 where @xmath213 , @xmath214 and @xmath205 is the step size which can be solved by line search method .
@xmath215 , calculate @xmath216 for the stopping criteria .
apply @xmath217 and combine all the @xmath167 to find the @xmath196 . based on the above method ,
we therefore propose a global coordinate reconstruction algorithm from merely random missing distance in each local patches .
given the local patches @xmath218 with patch size not larger than @xmath143 , we assume the local distance information include @xmath219 of the local distance of each patches . the full reconstrution procedure can be summarized as follows : step 1 : : using the method discussed in section [ subsec : solvepde ] with local mesh reconstruction method to compute the first @xmath177 laplace - beltrami eigenfunctions @xmath193 .
step 2 : : using algorithm [ alg : reconmodel_b ] to generate the virtual local coordinate @xmath220 for each patch .
step 3 : : using algorithm [ algorithm stitching ] to find the global coordinate @xmath196 from @xmath193 and @xmath184 .
in this section , numerical tests are presented to illustrate the proposed methods for solving pdes on manifold represented as incomplete inter - point distance information and the application to manifold stitching .
first , we demonstrate the effectiveness of the proposed matrix completion method for coordinate reconstruction .
we also illustrate the successful reconstruction curve .
second , we test our method of computing the lb eigen - problem and illustrate that the proposed method of solving pdes through local coordinate reconstruction is much more efficient than solve pdes from global coordinate reconstruction . in addition
, we also test the problem of solving eiknoal equations and illustrate the accuracy improvement of this approach to the canonical dijkstra method . as potential applications
, we also demonstrate preliminary results on point clouds reconstruction using global information from lb eigenfunctions and its potential application to dimension reduction problem .
all numerical experiments are implemented by matlab in a pc with a 32 g ram and 32 cpus of basic frequency 2.7 ghz . [ cols="^,^ " , ]
in this paper , we proposed a framework for discretizing pdes on manifolds represented as incomplete inter - point distance information .
our approach is to conduct pde discretization point - wisely from a local coordinate reconstruction .
this strategy successfully avoids a time - consuming global coordinate reconstruction and leads to a method with complexity linearly scaling to the number of sample size .
our local reconstruction model is inspired from the recent advances of low - rank matrix completion theory , which guarantee accuracy local coordinate reconstruction by only requiring a very small portion of distance information .
our method can be viewed as natural extensions of the moving least square method and the local method method for solving pdes on point clouds to data structure of manifolds represented as incomplete inter - point distance information . as an application of using solutions of pdes , we propose a new manifold reconstruction model by stitching local patches on the spectrum domain .
intense numerical experiments indicate effectiveness and efficiency of our methods for solving lb eigenvalue problem , eikonal equations and manifold reconstruction using stitching methods based on global information from the lb eigensystem .
we would like to thank prof .
xiangxiaong zhang s suggestions about optimizing distance matrix directly as we discussed in remark [ rk : lowrankd ] . , _ sensor positioning in wireless ad - hoc sensor networks using multidimensional scaling _ , in infocom 2004 .
twenty - third annualjoint conference of the ieee computer and communications societies , vol . 4 , ieee , 2004 , pp . | solutions of partial differential equations ( pdes ) on manifolds have provided important applications in different fields in science and engineering .
existing methods are majorly based on discretization of manifolds as implicit functions , triangle meshes , or point clouds , where the manifold structure is approximated by either zero level set of an implicit function or a set of points . in many applications ,
manifolds might be only provided as an inter - point distance matrix with possible missing values .
this paper discusses a framework to discretize pdes on manifolds represented as incomplete distance information . without conducting a time - consuming global coordinates reconstruction
, we propose a more efficient strategy by discretizing differential operators only based on point - wisely local reconstruction .
our local reconstruction model is based on the recent advances of low - rank matrix completion theory , where only a very small random portion of distance information is required .
this method enables us to conduct analyses of incomplete distance data using solutions of special designed pdes such as the laplace - beltrami ( lb ) eigen - system . as an application , we demonstrate a new way of manifold reconstruction from an incomplete distance by stitching patches using the spectrum of the lb operator .
intensive numerical experiments demonstrate the effectiveness of the proposed methods . |
the existence of inactive supermassive black holes in the nuclei of massive nearby galaxies ( magorrian et al .
1998 ) is a sure sign that most galaxies underwent substantial active phases .
furthermore , the correlation between the amount of material accreted ( black hole mass ) and the galaxy mass indicates important links between galaxy formation and agn activity .
one implication is that the most massive black holes live in the most massive galaxies which reside in the most massive dark matter halos .
the @xmath0 quasars being discovered in the sloan digital sky survey ( see fan , this volume ) have exceptionally high luminosities ( @xmath1 ) .
the available evidence suggests that these luminosities are not amplified by gravitational lensing ( fan et al .
2003 ) or beaming ( pentericci et al . 2002
; willott et al .
2003 ) . assuming that these quasars are radiating at the eddington limit
gives a lower limit on their black hole masses of several billion solar masses .
these are comparable to the black holes inside the largest dominant cluster ellipticals at the present time . applying local calibrations from this black hole mass to galaxy and
halo masses suggests that these quasars reside in halos with mass @xmath2 ( e.g. fan et al .
lcdm simulations show that structure grows ` bottom - up ' with the largest halos typically collapsing at the latest epochs .
halos with mass @xmath3 at @xmath4 are therefore extremely rare ( the fraction of mass in such halos is of order @xmath5 ; sheth & tormen 1999 ) there is evidence that at least some of these quasars are located in massive galaxies .
sub - millimetre photometry of several of them show very high far - infrared luminosities , implying extreme starbursts of @xmath6 ( bertoldi et al .
2003 ; priddey et al .
the most distant quasar , sdssj1148 + 5251 at @xmath7 , contains @xmath8 of molecular gas , again indicative of a massive , primeval galaxy ( walter et al .
these distant quasars , which are thought to be residing in massive galaxies , are therefore ideal places to search for the first clusters of galaxies .
the most straightforward method for identifying galaxies at such high redshifts is the lyman - break technique a large discontinuity in the spectral slope due to absorption by neutral hydrogen ( e.g. steidel et al . 1996 ) .
the increase in the optical depth of hydrogen absorption at such high redshifts ( e.g. songaila & cowie 2002 ) makes the lyman - break technique even more effective at @xmath9 than at lower redshifts . to search for galaxies in these putative mass overdensities
, we carried out a search for lyman - break galaxies in fields of three @xmath10 sdss quasars .
we have imaged three quasar fields with the gmos - north imaging spectrograph on the gemini - north telescope : sdssj1030 + 0524 at @xmath11 , sdssj1048 + 4637 at @xmath12 and sdssj1148 + 5251 at @xmath7 .
the imaging field - of - view is 5.5 arcmin on a side , equivalent to a co - moving size of 13 mpc at @xmath13 .
observations were carried out in queue mode during november and december 2003 .
the typical seeing fwhm of the observations is in the range 0.5 to @xmath14 .
typical exposure times are @xmath15 hours in the @xmath16-band and @xmath17 hours in the @xmath18-band . the relative exposure times were designed to give similar sensitivity in the two bands for very red objects with colours of @xmath19 .
full details of these observations and their reduction will be published elsewhere .
-band magnitude derived from recovery of simulated galaxies .
the curves are quite similar for the three different quasar fields .
dotted lines indicate the location of a completeness ratio of 0.8 and the adopted complete magnitude limit of @xmath20 . ]
detection of objects in the reduced images was performed using the sextractor software ( bertin & arnouts 1996 ) .
the @xmath16-band was selected as the primary detection waveband since @xmath0 galaxies are expected to have @xmath21 and may therefore be undetected at @xmath18 .
sextractor was run in `` double - image '' mode to determine @xmath18-band measurements for objects detected in @xmath16 .
magnitudes on the ab system were measured in circular apertures of diameter @xmath22 .
aperture corrections were applied statistically to the @xmath16-band magnitudes by fitting a linear function to the difference between the total magnitude and aperture magnitude as a function of aperture magnitude .
typical magnitude limits ( @xmath23 limits in @xmath24 apertures ) are @xmath25 and @xmath26 . to assess the completeness of the @xmath16-band catalogues we consider both the observed number counts and the recovery of simulated objects .
the number counts in the three fields do not differ significantly from each other .
they agree well with the @xmath16-band counts determined from a much larger area survey ( 0.2 square degrees or thirty times the gmos - north field - of - view ) by capak et al .
the number counts in the quasar fields begin to change slope at @xmath27 and turn over at @xmath28 , indicating this is where the sample becomes incomplete .
the source recovery as a function of magnitude was determined by populating the images with artificial galaxies and then using sextractor to attempt to detect these objects .
about 10000 artificial galaxies with magnitudes in the range @xmath29 were placed into copies of the @xmath16 images of each quasar .
regions of the images already occupied by objects were masked out of the process to eliminate incompleteness due to blending .
the resulting completeness ratio is plotted as a function of magnitude for the three quasar fields in fig.[fig : comp ] .
the completeness in all the fields is fairly flat at close to 1 up to @xmath30 and then begins to decline .
the rapid decline occurs at @xmath27 and the completeness drops to 0.5 by @xmath31 .
all the fields have completeness @xmath32 at @xmath20 and we adopt this as the magnitude at which completeness begins to become an issue .
this analysis with simulated objects agrees well with the results for the number counts discussed previously .
the @xmath0 sdss quasars have @xmath33 and colours of @xmath34 respectively for j1030 , j1048 and j1148 .
quasars and star - forming galaxies have broadly similar spectra over the rest - frame wavelength range @xmath35 nm probed by the @xmath18 and @xmath16 filters .
their spectra are dominated by a large break due to absorption by neutral hydrogen .
therefore one would expect companion galaxies to have comparable colours to the quasars . a colour - magnitude diagram for objects in the field of sdssj1030 + 0524
is shown in fig.[fig : colmags ] .
most objects have colours in the range @xmath36 as is well known from previous surveys ( dickinson et al .
2004 ; capak et al .
2004 ) . also plotted are curves showing the colour - magnitude relation for two different types of galaxy .
model galaxy spectra were generated from the bruzual & charlot ( 2003 ) spectral synthesis code with lyman forest absorption evolution matching the observations of songaila & cowie ( 2002 ) .
the upper curve is a present - day l@xmath37 elliptical which formed all of its stars in a starburst starting at @xmath38 with a characteristic timescale of 1gyr and evolved since without merging .
note that there is no dust extinction assumed for this model , but in reality the dust extinction would increase with redshift ( due to evolution and @xmath39-correction ) making the galaxy fainter than plotted at higher redshifts .
the lower curve is a l@xmath37 lyman - break galaxy model where the galaxy is observed 0.5gyr into a constant star formation rate starburst .
a search was made for objects which could plausibly be high - redshift galaxies .
the @xmath18-band dropout selection criteria adopted were snr in the @xmath16-band @xmath40 and a colour of @xmath41 .
possible candidates were inspected and magnitudes checked to ensure their unusual colours are not spurious .
a total of four objects satisfying these criteria were found ; two in the field of sdssj1030 + 0524 , shown in fig.[fig : colmags ] with filled symbols , and one in each of the other two fields . only one of these four galaxies has a magnitude brighter than the completeness limit of @xmath20 .
this object has @xmath42 which is quite close to the colour selection value and the size of the uncertainty means it is quite plausible that photometric errors have scattered the colour into the dropout range .
the measured colour is about @xmath43 away from the colours of the sdss quasars , which suggests that if it is a high - redshift galaxy , then it is most likely foreground to the quasars with a redshift in the range @xmath44 .
the other three objects have @xmath45 and only lower limits on their @xmath46 colours which lie in the range @xmath47 .
-band in the field of the @xmath11 quasar sdssj1030 + 0524 .
the completeness limit of @xmath20 is shown with a dot - dashed line .
the high - redshift galaxy selection criterion of @xmath48 is marked with the dashed line .
the dotted line represents the colour of an object which is just detected at the @xmath23 level in the @xmath18-band as a function of the @xmath16-band magnitude .
most objects with measured zero or negative flux in the @xmath18-band are plotted at @xmath49 .
the exceptions to this are objects which pass the @xmath18-band dropout selection criteria discussed in the text .
these are shown with star symbols and lower limits on the @xmath18-band @xmath23 line .
the dots at @xmath48 indicate sources which are detected at less than @xmath50 at @xmath16-band and hence have very large uncertainties on their magnitudes and colours and in some cases may be spurious .
the labelled curves show the colour and magnitude as a function of redshift for an evolving l@xmath37 elliptical ( upper curve ) and a non - evolving l@xmath37 lyman - break galaxy ( lower curve ) see text for more details . ]
our @xmath16-band images are complete at the greater than 80% level to @xmath20 . at @xmath51 , every single @xmath16-band object detected on our images
has a counterpart at the @xmath52 level in the @xmath18-band image . at fainter magnitudes ,
this is no longer true and our constraints on the @xmath46 colours of objects at these magnitudes becomes very weak due to uncertainty in both the @xmath16 and @xmath18 magnitudes .
we now consider the number of @xmath18-band dropouts we could have expected to find under the assumption that the quasar fields are `` random '' and show no enhancement due to the existence of the quasars .
the best comparison dataset which goes deep enough over a wide area is the _ hubble space telescope _
acs imaging of the goods regions ( giavalisco et al .
these observations give a surface density of objects with @xmath53 and @xmath48 of 0.02 arcmin@xmath54 ( dickinson et al 2004 ; bouwens et al .
the total sky area we have surveyed with gmos - north is 82 arcmin@xmath55 . therefore on the basis of the goods observations we would expect 1.6 @xmath18-band dropouts in our total area .
our finding of one dropout is entirely consistent with the expectations for blank fields .
our observations show that these quasar fields do not exhibit an excess of luminous companion galaxies .
the magnitude limit of @xmath20 corresponds to a uv luminosity of @xmath56 at a redshift of @xmath13 .
this is equivalent to an unobscured star formation rate of @xmath57 , assuming the conversion given in madau , pozzetti & dickinson ( 1998 ) .
the few known galaxies at redshifts @xmath58 discovered in narrow - band surveys have star formation rates derived from their uv luminosities comparable to this limit ( hu et al .
2002 ; kodaira et al .
for comparison , the millimeter detections of dust in j1048 and j1148 imply total star formation rates @xmath59 in the host galaxies of the quasars ( bertoldi et al .
2003 ) .
how do we explain the lack of companion galaxies with high star formation rates ?
one possibility is that galaxies undergoing intense star formation do exist but they are heavily extinguished by dust .
such galaxies may be detectable via their far - infrared emission and we are conducting a program of sub - mm imaging around these quasars to identify such galaxies .
deep observations with the irac camera on the _ spitzer space telescope _ would also be able to reveal these obscured galaxies .
another possibility is that massive companion galaxies exist but they are not observed during a period of intense star formation .
this possibility seems unlikely since the youth of the universe at this redshift means galaxies would likely be gas - rich and forming stars , especially if located in a dense environment which is not yet virialized .
finally , we raise a question mark over the notion that these rare massive black holes actually reside in the rarest , densest peaks in the dark matter distribution .
perhaps the correlation between black hole mass and dark matter halo mass ( or halo circular velocity see wyithe & loeb 2004 ) does not apply to these quasars and they are actually located inside less massive dark matter halos .
bertin , e. , arnouts , s. 1996 , a&as , 117 , 393 bertoldi , f. , et al .
2003 , a&a , 406 , l55 bouwens , r. j. , et al .
2004 , apj , 606l , 25 bruzual , g. , charlot , s. 2003 , mnras , 344 , 1000 capak , p. , et al .
2004 , aj , 127 , 180 dickinson , m. , et al .
2004 , apj , 600 , l99 fan , x. , et al .
2001 , aj , 122 , 2833 fan , x. , et al .
2003 , aj , 125 , 1649 giavalisco , m. , et al .
2004 , apj , 600 , l93 hu , e. m. , et al .
2002 , apj , 568 , l75 kodaira , k. , et al .
2003 , pasj , 55l , 17 madau , p. , pozzetti , l. , dickinson , m. 1998 , apj , 498 , 106 magorrian j. , et al .
1998 , aj , 115 , 2285 pentericci , l. , et al .
2002 , aj , 123 , 2151 priddey , r. s. , et al .
2003 , mnras , 344l , 74 sheth , r. k. , tormen , g. 1999 , mnras , 308 , 119 songaila , a. , cowie , l. l. , 2002 , aj , 123 , 2183 steidel , c. c. , et al . 1996 , apj , 462l , 17 walter , f. , et al .
2003 , nature , 424 , 406 willott , c. j. , mclure , r. j. , jarvis , m. j. 2003 , apj , 587 , l15 wyithe , j. s. b. , loeb , a. 2004 , apj , submitted , astro - ph/0403714 | we have obtained deep , multi - band imaging observations around three of the most distant known quasars at redshifts @xmath0 .
standard accretion theory predicts that the supermassive black holes present in these quasars were formed at a very early epoch .
if a correlation between black hole mass and dark matter halo mass is present at these early times , then these rare supermassive black holes will be located inside the most massive dark matter halos .
these are therefore ideal locations to search for the first clusters of galaxies .
we use the lyman - break technique to identify star - forming galaxies at high redshifts .
our observations show no overdensity of star - forming galaxies in the fields of these quasars .
the lack of ( dust - free ) luminous starburst companions indicates that the quasars may be the only massive galaxies in their vicinity undergoing a period of intense activity . |
ferromagnetic nanowires possess interesting properties that might be exploited in spintronic devices and more specifically in non - volatile memory ( mram ) and magnetic logic devices @xcite .
the mermin - wagner @xcite theorem forbids ( heisenberg - type ) magnetism in systems of dimension ( @xmath1 ) with short - range interactions .
hence , ferromagnetic nanowires being a quasi - one dimensional system displaying magnetic properties represent an interesting system from the fundamental point of view .
in addition to their interest in fundamental magnetism , they have many applications in microwave devices such as circulators , superconducting single - photon ghz detectors and counters @xcite , mass information storage ( perpendicular recording ) as read - write recording heads , magneto - electronics ( wire bending angle dependent gmr ) as well as in quantum computing and telecommunication .
they are simpler than nanotubes since their physical properties do not depend on chirality and they can be grown with a variety of methods : molecular beam epitaxy , electrochemical methods ( template synthesis , anodic alumina filters ) , chemical solution techniques ( self - assembly , sol - gel , emulsions ... ) etc ... ordered nanowire arrays may be of paramount importance in areas such as extremely high - density information recording of patterned media such as the quantum magnetic disk @xcite and in novel high - frequency communication or signal - processing devices based on the exploitation of spin - waves ( magnonic devices ) @xcite to transfer and process information of spin - currents with no disspative joule effect . in this work ,
fmr measurements are performed to detect the uniform resonance mode and extract the effective anisotropy field @xmath2 versus angle from arrays of ni nanowires with diameters ranging from 15 nm to 100 nm using conventional angular - dependent ferromagnetic resonance fmr in the x - band ( 9.4 ghz ) as a function of temperature ranging from liquid helium ( 4.2 k ) to room temperature .
we find that magneto - elastic effects play an important role in those systems that might be exploited in novel storage concepts .
this work is organised as follows : in section 2 , fmr measurements are presented and later analysed in section 3 .
we conclude the work in section 4 and appendix a details the fmr angular fitting procedure while appendix b covers dipolar effects .
individual ni wires inside the array are aligned parallel to each other within a deviation of a few degrees .
they have a length of 6 @xmath3 m and are characterized by a cylindrical shape with a typical variation in diameter of less than 5% and with a low - surface roughness . ,
applied field @xmath4 and corresponding angles @xmath5 they make with the nanowire axis that can be considered as an ellipsoid - shaped single domain with characteristic lengths @xmath6 and @xmath7 .
when the aspect ratio @xmath8 is large enough the ellipsoid becomes an infinite cylinder.,width=240 ] previous experiments on magnetization reversal mechanisms in ni arrays using alternating gradient - force magnetometry and superconducting quantum interference device ( squid ) magnetometry reveal that at room temperature the magnetic anisotropy is dominated by the shape anisotropy .
fmr experiments are performed with the microwave pumping field @xmath9 has a frequency of 9.4 ghz with a dc bias field @xmath10 oriented arbitrarily making an angle @xmath11 with the nanowire axis . despite the fact
several studies have considered reversal modes by domain nucleation and propagation ( see for instance henry _ et al . _
@xcite for an extensive discussion of the statistical determination of reversal processes and distribution functions of domain nucleation and propagation fields ; see also ferr _ et al . _
@xcite and hertel @xcite who showed the existence of domains with micromagnetic simulations ) , we do not consider domain nucleation and propagation in this work and rather concentrate on the single domain case that might be expected due to the smallness of ni magnetocristalline with respect to shape anisotropy .
the angular dependence of @xmath0 in the uniform mode is obtained by considering an ellipsoid with an effective uniaxial anisotropy , despite the fact ni has cubic anisotropy @xcite .
this description is sufficient to describe the fmr spectra we find ( see de la torre medina _ et al . _
@xcite for a description fully based on a cubic anisotropy approach ) .
the energy is therefore comprised of a small second - order uniaxial anisotropy contribution @xmath12 , shape demagnetization energy and a zeeman term : @xmath13 \label{energy}\ ] ] with @xmath14 the angle the mgnetization makes with the nanowire axis .
the resonance frequency @xmath15 is obtained from the smit - beljers @xcite formula that can be derived from the landau - lifshitz equation of motion with a damping term @xmath16 and that requires calculating the angular second derivatives of the total energy : @xmath17 ^ 2=\frac{(1+\alpha ^2)}{\sin ^2\theta } \left [ { \frac{\partial ^2e}{\partial \theta ^2}\frac{\partial ^2e}{\partial \phi ^2}- \left [ { \frac{\partial ^2e}{\partial \theta \,\partial \phi } } \right]^2 } \right ] \label{smit1}\ ] ] this provides a relationship between the effective anisotropy field and the external field at the resonance frequency . the frequency - field dispersion relation obtained from the smit - beljers equation is : @xmath18 [ h_{eff } \cos^2\theta + h \cos(\theta -\theta_h ) ] } \label{smit2}\ ] ] where @xmath19 . at the resonance frequency @xmath20 , @xmath21 and the applied field @xmath22 in the saturated case . in the unsaturated case the magnetization angle @xmath23 and one determines it from energy minimization .
calling at equilibrium , magnetization orientation @xmath24 ( taking @xmath25 ) we determine it by evaluating the derivative @xmath26 .
we get : @xmath27 where@xmath28 .
equations [ min ] and [ smit2 ] are used simultaneously to determine the resonance field @xmath0 versus angle @xmath11 at any temperature . from the measured resonance field @xmath0 versus field angle @xmath11 the @xmath29-factor , saturation magnetization @xmath30 and cubic anisotropy constant @xmath12
are determined with a least - squares fitting method ( see appendix a ) . measured fmr absorption derivative spectra @xcite are similar for ( 50 , 80 and 100 nm ) diameters but differ from the 15 nm case .
angular @xmath0 curves versus @xmath11 show a minimum at @xmath31@xmath32for the large diameters ( 50 , 80 and 100 nm ) and a minimum at @xmath33@xmath32for the 15 nm case ( see fig . [ fit ] ) . from this angular variation , we infer that 15 nm samples behave differently from larger diameter samples with a transition observed about 50 nm in agreement with the hysteresis loop vsm measurements ( as displayed in fig .
[ vsm ] ) .
previously neilsch _
et al . _
@xcite mentioned a change of behaviour in nanowire arrays because of the existence of a coherence diameter in ni to be @xmath34 40 nm .
the coherent diameter @xmath35 separates coherent ( stoner - wohlfarth style or homogeneous ) from inhomogeneous reversal ( reversal by curling ) obtained by equating nucleation fields in both cases @xcite : @xmath36 where @xmath37 is the exchange stiffness constant ( for ni , it is about 1.5@xmath38 erg / cm @xcite ) .
approximating the nanowire by an infinitely long cylinder , the demagnetizing factor along the minor axis @xmath39 , thus : @xmath40 with @xmath41 the first positive zero of the first kind bessel function @xmath42 derivative @xmath43 . consequently neilsch _ et al . _
@xcite estimate is recovered . nevertheless in our case ,
the change in behaviour is probably due to change in the values of anisotropy constant @xmath12 and saturation magnetization @xmath30 with the diameter as seen in table [ fit_table ] . [
cols="^,^ " , ] theoretical and experimental values of the resonance field @xmath0 versus temperature are compared in fig .
[ fig5 ] after accounting for the magneto - elastic anisotropy @xmath44 contribution in the form @xmath45 .
the agreement as a function of temperature for both nanowire diameters and both orientations ( @xmath31@xmath32and @xmath33@xmath32 ) of the field is a strong indication of the presence of magneto - elastic effects .
we have performed angular and temperature dependent fmr on ni nanowire arrays with variable diameter and shown with fmr that the easy axis orientation for the 15 nm diameter sample is perpendicular to the wire axis in sharp contrast with the 50 nm , 80 nm and 100 nm samples .
note that we expect ( from bulk ni ) the easy axis along the wire axis by comparing the value of shape energy with respect to anisotropy energy .
results obtained from the angular behavior of @xmath0 versus @xmath46 show that @xmath0 is minimum at 90@xmath32for the 15 nm sample whereas it is minimum at 0@xmath32for the larger diameter samples agree with hysteresis loops obtained from vsm measurements and confirm presence of the transition of easy axis direction from perpendicular at 15 nm to parallel to nanowire axis at 50 nm diameter .
the transition observed at 50 nm is interesting because of several potential applications in race - track mram devices .
_ et al . _
@xcite predicted that in permalloy nanowires of 50 nm and less , moving zero - mass domain walls may attain a velocity of several 100 m / s beating walker limit obeyed in permalloy strips with same lateral size .
hence , nanowire cylindrical geometry in contrast to prismatic geometry of stripes bears important consequences on current injection in nanowires and its effects on domain wall motion with reduced ohmic losses @xcite . at low temperature ,
we find that @xmath2 increases when temperature is decreased .
this may be attributed to the increase of @xmath47 , as temperature is decreased , affecting all anisotropy fields ( @xmath48 , @xmath49 , and @xmath50 ) that depend on @xmath47 . separating the various contributions to anisotropy by using thermal , frequency , diameter and angular variations
we have been able to pinpoint the main contribution to anisotropy at low temperature as stemming from magneto - elastic effects between the nanowire array and the dm supporting it .
magneto - elastic effects affect directly the thermal variation of @xmath2 that increase at low temperature due to the increasing difference between the thermal expansion of the metallic nanowire and the dm .
thermal effects must be analyzed properly for building extremely high density storage devices .
ordered arrays of nanowires are good candidates for patterned media and may also be used in plasmonic applications such as nanoantenna arrays or nanophotonic waveguides in integrated optics @xcite .
recently , heat assisted magnetic perpendicular recording using plasmonic aperture or nano - antenna has been tested on patterned media in order to process large storage densities starting at 1 tbits / in@xmath51 and scalable up to 100 tbits / in@xmath51 . in thermally - assisted perpendicular magnetic recording
, a waveguide delivers light to a plasmonic nanoantenna placed just above the disk platter surface creating an intense optical pattern in the near - field region .
however , this heats the disk on the nanometer scale ( around 25 nm ) and consequently temperature dependent magneto - elastic effects analysed in this work might affect the physics of high - density information writing in these systems .
+ * acknowledgments * + low temperature measurements were kindly performed by dr .
r. zuberek at the institute of physics of the polish academy of science , warsaw ( poland ) .
we have developed a procedure based on a least squares minimization procedure of the curve @xmath0 versus @xmath11 to the set of @xmath52 experimental measurements @xmath53}_{i=1,n}$ ] where @xmath54 and @xmath55 .
@xmath56 represents the set of parameters to fit @xmath57 .
hence the set of three minimum equations for the data points are : @xmath58^{2 } \mbox { minimum } \nonumber \\
\label{eqfit}\end{aligned}\ ] ] where the values @xmath59 are the experimental values of @xmath0 corresponding to the angles @xmath60 and @xmath61 .
the fitting method is based on the broyden algorithm , a generalization to higher dimension of the one - dimensional secant method @xcite that allows us to determine in a least - squares fashion , the set @xmath57 of unknowns .
broyden method is selected because it can handle over or underdetermined numerical problems and that it works from a singular value decomposition point of view @xcite .
this means it is able to circumvent singularities and deliver a practical solution to the problem at hand as an optimal set@xcite within a minimal distance from the real one .
demagnetization energy is given by @xmath62 @xcite where the demagnetization coefficients are such that@xmath63 and @xmath64 .
since the demagnetization field is given by @xmath65 , component @xmath66 of @xmath67 is @xmath68 .
we have two limits : * single isolated nanowire ( for which @xmath69 and @xmath70 ) , @xmath71 and @xmath72 with @xmath73 the unit vector along @xmath74 . * thin film limit ( for which @xmath75 and @xmath76 ) , @xmath77 and @xmath78 .
when the sample is saturated along @xmath74 ( nanowire axis ) @xmath79 , the demagnetization field in both cases has a single @xmath74 component : @xmath80 in the single wire case and @xmath81 in the thin film case . making a linear interpolation between these two limits ,
we get : @xmath82 .
if we rather consider a 2d square lattice of nanowires with parameter @xmath83 the average nanowire separation , the porosity is given by : @xmath84 and we can calculate directly the dipolar energy as follows .
99 z. z. sun and j. schliemann , phys .
lett . * 104 * , 037206 ( 2010 ) .
m. yan , a. kakay , s. gliga and r. hertel , phys .
104 * , 057201 ( 2010 ) . c. t. boone , j. a. katine , m. carey , j. r. childress , x. cheng , and i. n. krivorotov , phys .
lett . * 104 * , 097203 ( 2010 ) .
n. d. mermin and h. wagner , phys .
* 17 * , 1133 ( 1966 ) .
j. k. w. yang , e. dauler , a. ferri , a. pearlman , a ; verevkin , g. goltsman , b. voronov , r ; sobolewski , w. e. keicher and k. k. berggren , * ieee * trans
superconductivity , * 15 * , 626 ( 2005 ) .
s. y. chou , proceedings of the ieee , * 85 * , 652 ( 1997 ) . v. v. kruglyak , s. o. demokritov and d. grundler , j. phys .
phys . * 43 * , 264001 ( 2010 ) .
y. henry , a. iovan , j .-
george and l. piraux , phys .
b. * 66 * , 184430 ( 2002 ) .
r. ferr , k. ounadjela , j. m. george , l. piraux and s. dubois , phys .
b. * 56 * , 14066 ( 1997 ) .
r. hertel , j. appl .
phys . * 90 * , 5752 ( 2001 ) .
sixth - order cubic anisotropy energy is : @xmath111 where : @xmath112 are the cosines of the angles , the magnetization makes with the @xmath113 axes respectively ( see fig .
[ fig1 ] ) .
the following parameters at room temperature apply to bulk nickel : @xmath114 erg/@xmath115 , @xmath116 erg/@xmath115 whereas @xmath30=485 emu/@xmath115 .
j. de la torre medina , m. darques and l. piraux , j. phys .
* 41 * 032008 ( 2008 ) .
j. smit and h. c. beljers , philips res
10 * , 113 ( 1955 ) u. ebels , j. duvail , p. wigen , l. piraux , l. d. buda , k. ounadjela , phys .
b * 64 * , 144421 ( 2001 ) k. neilsch , r. b. wehrspohn , j. barthel , j. kirschner , u. gsele , s. f. fischer and h. kronmller , appl . phys . lett . *
79 * , 1360 ( 2001 ) .
r. skomski , a. kashyap , k. d. sorge and d. j. sellmyer j. appl . phys . *
95 * , 7022 ( 2005 ) .
g. bertotti , _
`` hysteresis in magnetism '' , _ academic press , new - york ( 1998 ) .
a. encinas , m. demand , l. vila , l. piraux , and i. huynen , appl .
. lett . * 81 * , 2032 ( 2002 ) w. j. carr , jr , phys . rev . * 109 * , 1971 ( 1958 ) e. r. callen , h. b. callen , phys . rev . *
129 * , 578 ( 1963 ) m. d. kuzmin , phys .
lett . * 94 * , 107204 ( 2005 ) .
r. m. bozorth , _ `` ferromagnetism '' _ , *
ieee * press ( 1993 ) . o. a. tretiakov , y. liu , and a. abanov , phys .
lett . * 105 * , 217203 ( 2010 ) .
h. ditlbacher , a. hohenau , d. wagner , u. kreibig , m. rogers , f. hofer , f. r. aussenegg and j. r. krenn , phys .
lett . * 95 * , 257403 ( 2005 ) . w. h. press , w. t. vetterling , s. a. teukolsky and b. p. flannery , _ `` numerical recipes in c : the art of scientific computing '' , _ second edition , cambridge university press ( new - york , 1992 ) . l. d. landau and e. m. lifshitz , _ electrodynamics of continuous media _ , pergamon , oxford ( 1975 ) . | ferromagnetic resonance ( fmr ) measurements are performed as a function of temperature on nickel nanowires with different diameters ( 15 nm , 50 nm , 80 nm and 100 nm ) from room temperature down to liquid helium ( 4.2 k ) .
the resonance field @xmath0 measured as a function of field angle and temperature yields interesting information about the size and temperature variation of various anisotropy effects . with diameter increase from 15 nm to 100
nm we observe a transition at 50 nm in easy axis orientation from perpendicular to parallel to the nanowire array axis .
the temperature variation of @xmath0 is analyzed and explained theoretically with presence of strong magneto - elastic effects in small ( 15 nm ) and large diameter ( 100 nm ) nanowire arrays . |
recently , several vector - tensor ( vt ) theories involving a vector field , @xmath1 , and the metric tensor @xmath2 have been applied to cosmology @xcite ; in these theories , the background energy density , @xmath3 , of the vector field @xmath1 plays the role of the dark energy ( hereafter the subscript @xmath4 stands for background ) ; for example , in @xcite , where the theory of gravitation considered in this paper was proposed , the equation of state is @xmath5 , where @xmath6 is the pressure due to the field @xmath1 and @xmath7 ; hence , the constant energy @xmath8 plays the role of vacuum energy .
however , in the theory studied in @xcite , which might be appropriate to explain the anomalies observed in the angular spectrum of the cosmic microwave background ( cmb ) for small @xmath9 multipoles , the equation of state is @xmath10 , where @xmath11 is negative for any value of the scale factor @xmath12 ; hence , in this theory , we have a sort of dynamical dark energy different from that associated to the cosmological constant ( vacuum energy ) . here
, our attention is focused on the theory proposed by @xcite , which was applied to cosmology in @xcite and @xcite . in this last reference ,
the vt theory under consideration was proved to be viable in the sense that : ( i ) its post - newtonian parametrized limit is identical to that of general relativity ( gr ) , and ( ii ) the theory may simultaneously explain the seven year wmap data about the cmb temperature anisotropy and the measurements of supernova ia luminosities . conclusion ( ii )
was obtained by using the well - known bardeen formalism @xcite to write the evolution equations of the scalar linear perturbations in vt theory , and also to find the initial conditions at high redshift necessary to solve these equations [ see @xcite ] . by using these elements , a modified version of the code cosmomc @xcite based on statistical techniques as the markov chains was designed to fit the vt predictions with the observational data mentioned above . a model involving seven free cosmological parameters
results were encouraging @xcite and the theory deserves attention . before writing any field or cosmological equation , let us fix some notation criteria .
our signature is ( ,+,+,+ ) .
latin ( greek ) indexes run from 1 to 3 ( 0 to 3).the symbol @xmath13 ( @xmath14 ) stands for a covariant ( partial ) derivative .
the antisymmetric tensor @xmath15 is defined by the relation @xmath16 , in terms of the @xmath17 vector field .
quantities @xmath18 , @xmath19 , and @xmath20 are the covariant components of the ricci tensor , the scalar curvature and the determinant of the matrix @xmath21 formed by the covariant components of the metric , respectively . the gravitational constant is denoted @xmath22 and the speed of light @xmath23 .
units are chosen in such a way that @xmath24 ; namely , we use geometrized units .
the dimension of any quantity is @xmath25 , n being an integer number .
length unit is chosen to be the kilometer .
our coordinates are denoted @xmath26 , @xmath27 , @xmath28 , and @xmath29 . whatever the quantity @xmath30 may be , @xmath31 stands for a partial derivative with respect to the radial coordinate @xmath27 .
the two vt theories mentioned above correspond to different choices of the parameters @xmath32 , @xmath33 , @xmath34 , and @xmath35 involved in the action @xcite : @xmath36 where the tensor @xmath15 defined above is not the electromagnetic one . the vt theory studied in this paper
[ see @xcite ] corresponds to the following choice of the free dimensionless parameters involved in action ( [ 1.1 ] ) : @xmath37 and @xmath38 .
in this theory of gravitation , it has been proved that there are no ghosts and unstable modes for @xmath39 . moreover , for a homogeneous and isotropic robertson - walker background universe , the energy density of the vector field @xmath3 has been proved to be @xmath40 where @xmath41 . therefore , constant @xmath42 must be positive to have @xmath43 .
the vector field equations , when applied to robertson - walker cosmology , predict a constant value for @xmath44 and , consequently , @xmath3 is strictly constant as the vacuum energy density .
the viability of vt as a theory of gravitation and its cosmological success require @xmath42 and @xmath45 parameters satisfying the inequalities @xmath46 , but their values can not be fixed .
in this situation , it is worthwhile the design of new applications of the vt theory with the essential aim of fixing @xmath42 and @xmath45 and any other arbitrary quantity related with our limited knowledge of the @xmath1 nature and properties , which are being analyzed .
previous outcomes ( based on linearity ) strongly suggest that the new applications should be nonlinear . on account of these considerations
, we have planed the study of various gravitational physical systems as , for example : ( a ) the black hole horizons of different sizes and their neighborhoods , ( b ) the cosmological evolution of nonlinear structures ( galaxies , clusters , superclusters and so on ) by using either approximations or simulations , and ( c ) binary stellar systems radiating gravitational waves . here
, our attention is focused on the simplest of these problems : the study of the vt horizons of outer ( no matter content ) spherically symmetric stationary space - times , which are fully characterized by the mass @xmath47 ( no electrical charge and rotation ) and the cosmological constant @xmath48 . according to its formulation ,
vt is a theory of pure gravitation .
the field @xmath49 has nothing to do with the potential vector of the electromagnetic field .
it does not couple with electrical currents .
the u(1 ) gauge symmetry of maxwell theory is not required in vt .
in other words , vt is a simple and manageable theory of gravitation .
the electromagnetic interaction must be described in the standard way .
there are many alternative theories which are being currently studied , some of these theories are only concerned with gravitation ; e.g. , the so - called @xmath50 and @xmath51 theories , where @xmath52 is the torsion scalar . in these theories the electromagnetic field
is treated in the standard way ( minimal coupling with the gravitational part of the lagrangian ) .
in other theories the action is designed to describe both the gravitational and the electromagnetic fields ; interesting cases may be found , e.g. , in @xcite , where non - minimal couplings of the electromagnetic field with gravity are proposed . a very promising non - minimal coupling between the electromagnetic field and a @xmath53 function
is applied to cosmology in @xcite , where it is claimed that the theory is viable , and also that inflation and late time acceleration may be simultaneously explained . however , in vt theory , as well as in gr , it must be recognized that inflation is to be produced by additional fields .
it is due to the fact that inflation must lead to an isotropic universe , whereas the inflation due to a vector field is expected to be anisotropic .
only a triplet of orthogonal vector fields or n randomly oriented vector fields might produce an isotropic enough expansion @xcite , but this is not the case of the vt theory . in order to explain inflation we could replace @xmath19 by an appropriated function @xmath50 in the lagrangian of the vt theory ; in this way , the field @xmath49 could explain the accelerated late time expansion , whereas the scalar field , associated to @xmath53 in the einstein frame , could account for the required inflation ; hence , function @xmath50 would be chosen to achieve a good inflation , without producing late time acceleration , which implies less restrictions to be satisfied by f(r ) .
nevertheless , we think that before any generalization , the vt theory must be fully developed as a simple viable and manageable gravitation theory , which explains many observations ( see above ) for arbitrary values of @xmath54 and @xmath42 .
this paper is structured as follows : the vt theory is described in sect .
[ sec-2 ] , the stationary spherically symmetric solutions of the field equations are found in sect .
[ sec-3 ] , the horizons are studied in sect .
[ sec-4 ] and ; finally , section [ sec-5 ] contains a general discussion and a summary of the main conclusions .
variational calculations based on action ( [ 1.1 ] ) , with @xmath37 and @xmath38 , lead to the following field equations @xcite : @xmath55 where @xmath56 is the contribution of matter to the energy - momentum tensor , which have the same form as in gr .
tensor @xmath57 is the contribution due to the @xmath49 field of the vt theory , whose form is @xmath58 \nonumber \\ & & -2\gamma [ \{a^{\alpha}\nabla_{\alpha } ( \nabla \cdot a ) + \frac { 1}{2}(\nabla \cdot a)^{2}\ } g^{\mu \nu}\nonumber \\ & & -a^{\mu}\nabla^{\nu } ( \nabla \cdot a ) - a^{\nu}\nabla^{\mu } ( \nabla \cdot a ) ] \ ,
\label{emtee_vt}\end{aligned}\ ] ] and @xmath59 also have the same form as in gr ; namely , @xmath60 eqs .
( [ fieles_vt ] ) are a generalization of the einstein equations of gr
. variations of the vector field @xmath1 in action ( [ 1.1 ] ) give @xmath61 where @xmath62 plays the role of a fictitious current . from eq .
( [ 1.3_vt ] ) one easily gets the relation @xmath63 which may be seen as the conservation law of the fictitious current @xmath64 defined above . since the parameters @xmath42 and @xmath54 are dimensionless , a dimensional analysis of eqs .
( [ fieles_vt ] ) and ( [ emtee_vt ] ) leads to the important conclusion that the dimension of the @xmath49 components is @xmath65 .
this fact will be important below . by using the chosen units and the relation between the einstein ( @xmath66 ) and ricci ( @xmath67 ) tensors , eqs .
( [ fieles_vt ] ) may be written as follows : @xmath68 where @xmath69 is the kronecker delta , and @xmath70 equation ( [ gt ] ) may be easily rewritten in the form : @xmath71 @xmath52 being the scalar @xmath72 .
we have the basic equations to look for horizons in next sections .
it is well known that , in the stationary spherically symmetric case , the line element may be written as follows [ see e.g. , @xcite ] : @xmath73 and , moreover , the covariant components @xmath74 have the form : @xmath75 \ . \label{vf}\ ] ] accordingly , the nonvanishing @xmath76 components are @xmath77 we have the four unknown functions @xmath78 , @xmath79 , @xmath80 , and @xmath81 to be found from the field equations of sect
. [ sec-2 ] .
hereafter , it is assumed that the matter tensor @xmath82 vanishes and , then , taking into account eqs .
( [ emtee_vt ] ) , ( [ tmnl ] ) , ( [ tmn ] ) , plus eqs .
( [ le])([nvf ] ) , one easily get that , in terms of the new dimensionless parameters @xmath83 and @xmath84 , the nonvanishing @xmath85 components are : @xmath86 @xmath87\nonumber \\ & & -\tilde{\gamma}[2 a^1 ( \nabla \cdot a)^{\prime } + ( \nabla \cdot a)^2 ] - \lambda \ } \ , \label{t00}\end{aligned}\ ] ] @xmath88\nonumber \\ & & -\tilde{\gamma}[-2 a^1 ( \nabla \cdot a)^{\prime } + ( \nabla \cdot a)^2 ] - \lambda \ } \ , \label{t11}\end{aligned}\ ] ] @xmath89\nonumber \\ & & -\tilde{\gamma}[2 a^1 ( \nabla \cdot a)^{\prime } + ( \nabla \cdot a)^2 ] - \lambda \ } \ , \label{t22}\end{aligned}\ ] ] @xmath90\nonumber \\ & & -\tilde{\gamma}[2 a^1 ( \nabla \cdot a)^{\prime } + ( \nabla \cdot a)^2 ] - \lambda \ } \ . \label{t33}\end{aligned}\ ] ] the line element ( [ le ] ) does not depend on time and , consequently , it does not describe a cosmological space - time .
this is also valid in gr , where the same line element leads to various metrics as those of schwarzschild and kottler - schwarzschild - de sitter [ see @xcite ] .
the region where these solutions are physically significant must be determined in each case ; e.g. , regions where @xmath91 must be excluded . in vt , it has been claimed ( see above ) that the cosmological constant is related with the value of @xmath92 in the background universe , but this value is different from that of the same divergence in the stationary spherically symmetric case ; by this reason , in spite of its origin , the cosmological constant is treated as in gr , and it is denoted @xmath93 .
let us now look for the stationary spherically symmetric solutions of the vt field equations following various steps .
the calculation of @xmath92 may be performed by solving the tensor field equation ( [ gt ] ) for @xmath94 and @xmath95 . in this case , since the components @xmath96 and @xmath97 vanish , from eq .
( [ ta01 ] ) one easily obtains : @xmath98 ; hence , for @xmath99 , @xmath100 , @xmath101 , and @xmath102 , it follows that @xmath103 vanishes and , consequently , a trivial integration gives @xmath104 where @xmath105 is an integration constant .
the trace @xmath52 is first calculated by using the @xmath107 components calculated from eqs .
( [ ta01 ] ) ( [ t33 ] ) and the @xmath21 metric components .
the result is @xmath108 - 4 \lambda \ , \ ] ] and , then , taking into account this result , eq . ( [ tg ] ) , and eq . ( [ divae ] ) , one easily get the relation @xmath109 from this equation and the nonvanishing components of the ricci tensor : @xmath110 \ , \label{r00}\ ] ] @xmath111 @xmath112 + 1 \ , \ ] ] @xmath113 the following relation is easily obtained : @xmath114 the same equation is also obtained in gr . after integration , it leads to @xmath115 [ see , e.g. , @xcite ] .
evidently , this relation implies that @xmath116 .
function @xmath118 may be calculated by solving eq .
( [ 1.3_vt ] ) in the stationary spherically symmetric case .
since @xmath119 has been proved to be constant [ see eq .
( [ divae ] ) ] , the vector @xmath120 vanishes and , consequently , taking into account the relation @xmath121 , which must be satisfied ( see sect .
[ sec-1 ] ) , eq ( [ 1.3_vt ] ) reduces to @xmath122 .
moreover , taking into account eqs .
( [ vf])([nvf ] ) , the covariant derivative @xmath123 may be easily calculated to get @xmath124 in terms of the new variable @xmath125 , the last equation reduces to @xmath126 .
the solution of this equation is @xmath127 and , then , a new integration gives @xmath128 where @xmath129 and @xmath130 are integration constants . for @xmath131 and @xmath132 , the tensor field equation ( [ tg ] ) may be easily written in the form @xmath133 where we have taken into account the relation @xmath134 ( see sect .
[ s_2 ] ) , the nonvanishing components of @xmath135 and @xmath136 listed in previous sections , and eq .
( [ divae ] ) . in the same way , for @xmath137 and @xmath138
, one finds @xmath139 \ . \label{tg22}\end{aligned}\ ] ] subtracting equations ( [ tg11 ] ) and ( [ tg22 ] ) and multiplying by the factor @xmath140 , the following second order differential equation is obtained : @xmath141 e^{-2\alpha } - r^{-2 } \ = \ 0 \ . \label{edoa}\ ] ] this equation can be solved by using the new variable @xmath142 . in terms of @xmath143 , eq .
( [ edoa ] ) reads as follows : @xmath144 where @xmath145 $ ] and @xmath146 ( see above in this section ) .
the general solution of eq .
( [ edob ] ) is @xmath147 , where @xmath148 is the general solution of the corresponding homogeneous equation , and @xmath149 is a particular solution of the complete inhomogeneous equation .
the general solution @xmath148 is : @xmath150 @xmath151 and @xmath152 being integration constants . in order to obtain a particular solution , @xmath149 ,
we may apply the method of parameter variations ; according to this method , we must look for a solution of the following form : @xmath153 where @xmath154 and @xmath155 is the wronskian : @xmath156 so , the particular solution @xmath149 takes on the form : @xmath157 \ .
\label{wp}\end{aligned}\ ] ] let us now use the explicit form of @xmath158 ( see above ) to easily find @xmath159 finally , eqs .
( [ wh ] ) and ( [ wpr ] ) allow us to write the general form of function @xmath160 , which directly leads to the metric components @xmath161 \ , \label{g00}\ ] ] and @xmath162 the last step is the integration of eq .
( [ divae ] ) derived in sect .
[ s_2] to get the function @xmath164 .
this equation may be easily rewritten as follows @xmath165 \ = \
\label{div}\ ] ] this is a linear first order differential equation of the form @xmath166 , with @xmath167 and @xmath168 .
the solution of this equation is : @xmath169 where @xmath170 is another integration constant , and @xmath171 \ dr $ ] . after performing these integrals ,
one obtains : @xmath172 eqs .
( [ sola0 ] ) and ( [ sola1 ] ) give the vector field @xmath49 , and eqs .
( [ g00 ] ) and ( [ g11 ] ) define the metric of the vt theory in the stationary spherically symmetric case .
the resulting metric is a generalization of the kottler - schwarzschild - de sitter one , which is obtained for @xmath173 , @xmath174 ( @xmath175 being the schwarzschild radius ) , and @xmath176 . in the vt theory
we have found a new term @xmath177 , which is positive due to the fact that the relation @xmath178 must be satisfied [ see sect . [ sec-1 ] ] .
the metric obtained in the framework of the vt theory is similar to the reissner - nordstrm - de sitter metric ( @xcite ) , which corresponds to a stationary spherically symmetric charged system in gr . the form of this known metric is @xmath179 \ , \ ] ] it involves a positive term proportional to @xmath180 which depends on the electrical charge @xmath181 ; evidently , in eqs .
( [ g00 ] ) and ( [ g11 ] ) , there is also a term of this kind , in which , the role of @xmath182 is played by the constant @xmath183 .
in the stationary spherically symmetric case , outside the matter distribution , and in the absence of electrical charge , the solution of the vt field equations involves the integration constants @xmath184 , @xmath185 , @xmath186 , @xmath187 , @xmath188 , and @xmath189 . in this situation , the physical system under consideration is fully described by the quantities @xmath47 and @xmath48 , whose dimensions in geometrized units are @xmath190 and @xmath191 , respectively .
let us now perform a dimensional analysis to predict the dependence of the parameter @xmath184 involved in the metric components in terms of @xmath47 and @xmath93 .
the constants @xmath188 , and @xmath189 also appear in gr .
since the dimensions of @xmath192 are @xmath65 , the term @xmath193 involved in @xmath194 is dimensionless and , consequently , the dimension of @xmath195 must be @xmath190 ; hence , @xmath189 must be a dimensionless number , @xmath196 , multiplied by @xmath47 ; in this case , we have a well known criterion to conclude that @xmath197 , a number leading to the well known term @xmath198 . in the same way
, the dimensionless character of @xmath199 leads to the conclusion that the dimension of @xmath188 is @xmath191 ; hence , this term must be the product of a dimensionless constant @xmath200 by the factor @xmath48 . in this case , there are also arguments to conclude that @xmath201 , a number which leads to the well known term @xmath202 involved in the kottler - schwarzschild - de sitter metric
. a similar analysis may be performed for the constants @xmath184 and @xmath203 ; in fact , according to eq .
( [ sola0 ] ) , the dimensionless component @xmath204 is the sum of two terms of the form @xmath205 and @xmath206 ; hence , the dimensions of @xmath184 and @xmath185 are @xmath190 and @xmath65 , respectively and , consequently , we conclude that the constant @xmath184 must be the product of a dimensionless constant @xmath207 by @xmath47 ; in this case , we have not any criterion to fix the dimensionless constant @xmath207 , which keeps arbitrary by the moment
. this analysis does not give any information about the dimensionless constant @xmath185 , but this information is not necessary to look for the horizons , which follows from the fact that according to eqs .
( [ g00 ] ) and ( [ g11]) the metric components do not depend on @xmath185 .
the dimensional analysis in not extended to the component @xmath208 , since the metric is also independent of the constants @xmath186 and @xmath187 involved in eq .
( [ sola1 ] ) . after the above dimensional considerations we can write : @xmath209 @xmath210 \ , \label{g002}\ ] ] where @xmath211 plays the role of a dimensionless arbitrary constants , which should be fixed by studying appropriate nonlinear problems in the framework of the vt theory , as , e.g. , the geodesic motion of proof particles close to possible horizons . in terms of the function @xmath212 ,
the horizons are the hypersurfaces @xmath213 defined by the condition @xmath214 . in the regions where the inequality @xmath215 is satisfied ,
our description of the stationary spherically symmetric space - time is physically consistent .
condition @xmath216 is not compatible with the assumed metric signature . in the standard @xmath93cdm cosmological model of gr ,
most current observations are explained for values of the vacuum energy density parameter @xmath217 close to @xmath218 , which corresponds to @xmath219 .
the same value also explains current observations in the framework of the vt theory [ see @xcite ] ; hence , the above value of the cosmological constant is hereafter fixed .
the mass @xmath47 is varied between @xmath220 and @xmath221 ; so , the masses of different types of black holes are considered . from stellar black holes due to supernova explosions , to supermassive ones located in the galactic central regions .
once a mass @xmath47 has been fixed , function @xmath222 only involves the unknown positive parameter @xmath223 . for @xmath224 ,
the metric reduces to the kottler - schwarzschild - de sitter one and , in such a case , there are two horizons , the first ( second ) one is the black hole ( cosmological ) horizon , whose radius is hereafter denoted @xmath225 ( @xmath226 ) . in the region limited by these two horizons , namely , for @xmath227 , function @xmath222 is positive and the kottler - schwarzschild - de sitter metric is physically admissible . @xcite studied the horizons in the reissner - nordstrm - de sitter space - time .
if the outcomes obtained in that paper are rewritten in our case , by replacing @xmath228 by @xmath229 , it is straightforward to conclude that from @xmath230 to a certain @xmath231 value , @xmath232 , which is greater than @xmath233 but very close to it , there are both a black hole horizon and a cosmological one ; however , for @xmath234 , there is an unique horizon which is cosmological .
our calculations have verified all this . for appropriate @xmath47 values , the algebraic equation
@xmath235 has been numerically solved for @xmath236 and for many positive @xmath0 values . for @xmath237 and whatever
@xmath47 may be , we have found only a root at @xmath238 ( there is no black hole horizon ) .
however , for any @xmath239 , apart from the above @xmath226 radius for the cosmological horizon , we have also obtained a black hole horizon with a @xmath225 radius depending on both @xmath47 and @xmath0 .
figure [ figu1 ] corresponds to a mass @xmath240 ( stellar black hole ) .
the left panel shows @xmath225 as a function of @xmath0 inside the interval [ 0 , @xmath241 .
the radius of the black hole horizon decreases as @xmath0 separates from the zero value corresponding to the kottler - schwarzschild - de sitter solution of the gr field equations . in the right panel , the relative deviation @xmath242 } { r_{bh}(\gamma=0)+r_{bh}(\gamma ) } \label{drr}\ ] ] is represented , as a function of @xmath0 , in the same interval as in the left panel .
we see that these deviations reach values close to 30% , which are not very large deviations , but moderate significant ones . in figure ( [ figu2 ] )
, the mass is @xmath243 ( galactic supermassive black hole ) and , consequently , the radius @xmath244 is greater than in the top panels by a factor of @xmath245 ; nevertheless , this proportionality factor is the same for any @xmath231 and , consequently , the form of the curves represented in the left panels of figs .
( [ figu1 ] ) and ( [ figu2 ] ) are identical . moreover , the relative deviations @xmath246 defined in eq .
( [ drr ] ) reach the same values in the right panels of the two figures , which means that these deviations do not depend on @xmath47 .
this paper has been devoted to the development of the vt theory of gravitation proposed by @xcite .
previous applications of this theory to both the solar system and cosmology have given excellent results @xcite . here
, we have solved the field equations of the vt theory , in the absence of matter and electrical charge , by assuming a stationary spherically symmetric space - time .
it has been proved that the resulting solution has the same form as the reissner - nordstrm - de sitter solution of gr , but the role of the electrical charge is played by a quantity proportional to the source mass @xmath47 . after reaching this conclusion , we have focused our attention on the horizons associated to stellar and massive black holes .
@xcite have proved that , in the absence of electrical charge , there are @xmath50 theories of gravitation leading to reissner - nordstrm - de sitter space - times , but the authors recognize that -in these theories the meaning of the quantity playing the role of the electrical charge is not clear . since the cosmological constant , @xmath93 ,
is fixed by comparisons between predictions of the vt theory and current observations , the cosmological horizon is practically constant .
its radius is almost independent of the mass , @xmath47 , for any realistic black hole .
there is always a cosmological horizon whatever the value of the parameter @xmath0 defined in sect .
[ sec-4 ] may be .
in the vt theory , we have proved that , for a given mass @xmath47 , the radius of the black hole horizon is smaller than the radius of the kottler - schwarzschild - de sitter black hole having the same mass .
the relative deviations between these two radius are small but significant , reaching values close to 30 % .
this effect is important since it is similar to the effect due to the black hole rotation in gr , which leads to a horizon radius smaller than that corresponding to @xmath247 .
for @xmath237 there is no any black hole horizon in the vt theory under consideration .
various methods have been designed to estimate the mass @xmath47 and angular momentum @xmath248 of a black hole from observations .
if , in future , the mentioned methods become accurate enough , and the estimated @xmath47 and @xmath248 quantities obey the relation predicted by means of the kerr solution of einstein equations , the contribution of the vector field to the horizon radius will have to be considered negligible ( @xmath249 ) ; however , if the kerr relation is not satisfied by the observed values of @xmath47 and @xmath248 , an appropriate @xmath0 value could solve the problem .
let us finally mention two interesting extensions of this paper : first of all , the motion of test particles in the neighborhood of the above vt black hole deserves attention ; so , accretion disks and other phenomena might be studied .
afterward , the stationary axially symmetry line element , plus an appropriate vector field @xmath1 , should be considered to study rotating black holes in the framework of the vt theory ; in this way , a relation between @xmath47 and @xmath248 could be found , which might be satisfied by accurate future observed values of these quantities . | a certain vector - tensor ( vt ) theory of gravitation was tested in previous papers . in the background universe
, the vector field of the theory has a certain energy density , which is appropriate to play the role of vacuum energy ( cosmological constant ) .
moreover , this background and its perturbations may explain the temperature angular power spectrum of the cosmic microwave background ( cmb ) obtained with wmap ( wilkinson map anisotropy probe ) , and other observations , as e.g. , the ia supernova luminosities . the parametrized post - newtonian limit of the vt theory has been proved to be identical to that of general relativity ( gr ) , and there are no quantum ghosts and classical instabilities . here , the stationary spherically symmetric solution , in the absence of any matter content , is derived and studied .
the metric of this solution is formally identical to that of the reissner - nordstrm - de sitter solution of gr , but the role of the electrical charge is played by a certain quantity @xmath0 depending on both the vector field and the parameters of the vt theory . the black hole and cosmological horizons are discussed .
the radius of the vt black hole horizon deviates with respect to that of the kottler - schwarzschild - de sitter radius .
realistic relative deviations depend on @xmath0 and reach maximum values close to 30 per cent . for large enough @xmath0 values
, there is no any black hole horizon , but only a cosmological horizon .
the radius of this last horizon is almost independent of the mass source , the vector field components , and the vt parameters .
it essentially depends on the cosmological constant value , which has been fixed by using cosmological observational data ( cmb anisotropy , galaxy correlations and so on ) . |
the full sky has not been surveyed in space ( imaging ) and time ( variability ) at hard x - ray energies . yet
the hard x - ray ( hx ) band , defined here as 10 - 600 kev , is key to some of the most fundamental phenomena and objects in astrophysics : the nature and ubiquity of active galactic nuclei ( agn ) , most of which are likely to be heavily obscured ; the nature and number of black holes ; the central engines in gamma - ray bursts ( grbs ) and the study of grbs as probes of massive star formation in the early universe ; and the temporal measurement of extremes : from khz qpos to sgrs for neutron stars , and microquasars to blazars for black holes .
a concept study was conducted for the energetic x - ray imaging survey telescope ( exist ) as one of the new mission concepts selected in 1994 ( grindlay et al 1995 ) .
however the rapid pace of discovery in the hx domain in the past 2 years , coupled with the promise of a likely 2 - 10 kev imaging sky survey abrixas2 ( see http://www.aip.de/cgi-bin/w3-msql/groups/xray/abrixas/index.html ) in c.2002 - 2004 and the recent selection of swift ( see http://swift.gsfc.nasa.gov/ ) which will include a 10 - 100 kev partial sky survey ( to 1 mcrab ) in c.2003 - 2006 , have prompted a much more ambitious plan .
a dedicated hx survey mission is needed with full sky coverage each orbit and 0.05 mcrab all - sky sensitivity in the 10 - 100 kev band ( comparable to abrixas2 ) and extending into the 100 - 600 kev band with 0.5 mcrab sensitivity .
such a mission would require very large total detector area and large telescope field of view .
these needs could be met very effectively by a very large coded aperture telescope array fixed ( zenith pointing ) on the international space station ( iss ) , and so exist - iss was recommended by the nasa gamma - ray program working group ( grapwg ) as a high priority mission for the coming decade .
this mission concept has now been included in the nasa strategic plan formulated in galveston as a post-2007 candidate mission . in this paper
we summarize the science goals and briefly present the mission concept of exist - iss .
details will be presented in forthcoming papers , and are partially available on the exist website ( http://hea-www.harvard.edu/exist/exist.html ) .
exist would pursue two key scientific goals : a very deep hx imaging and spectral survey , and a very sensitive hx all - sky variability survey and grb spectroscopy mission .
these survey ( * s * ) and variability ( * v * ) goals can be achieved by carrying out several primary objectives : + + it is becoming increasingly clear that most of the accretion luminosity of the universe is due to obscured agn , and that these objects are very likely the dominant sources for the cosmic x - ray ( and hx ) diffuse background ( e.g. fabian 1999 ) . no sky survey has yet been carried out to measure the distribution of these objects in luminosity , redshift , and broad - band spectra in the hx band where , as is becoming increasingly clear from bepposax ( e.g. vignati et al 1999 ) , they are brightest .
exist would detect at least 3000 seyfert 2s and conduct a sensitive search for type 2 qsos .
spectra and variability would be measured , and detailed followup could then be carried out with the narrow - field focussing hx telescope , hxt , on constellation - x ( harrison et al 1999 ) as well as ir studies .
+ + the study of black holes , from x - ray binaries to agn , in the hx band allows their ubiquitous comptonizing coronae to be measured .
the relative contributions of non - thermal jets at high requires broad band coverage to 511 kev , as does the transition to adafs at lower values .
hx spectral variations vs. broad - band flux can test the underlying similarities in accretion onto bhs in binaries vs. agn .
+ + x - ray novae ( xn ) appear to be predominantly bh systems , so their unbiased detection and sub - arcmin locations , which allow optical / ir identifications , can provide a direct measure of the bh binary content ( and xn recurrence time ) of the galaxy .
xn containing neutron stars can be isolated by their usual bursting activity ( thermonuclear flashes ) , and since they may solve the birth rate problem for millisecond pulsars ( yi and grindlay 1998 ) , their statistics must be established .
a deep hx survey of the galactic plane can also measure the population of galactic bhs not in binaries , since they could be detected as highly cutoff hard sources projected onto giant molecular clouds . compared to ism accretion onto isolated nss , for which a few candidates have been found
, bhs should be much more readily detectable due to their intrinsically harder spectra and ( much ) lower expected space velocities , v , and larger mass m ( bondi accretion depending on m@xmath0/v@xmath6 ) .
+ + type ii sne are expected to disperse 10@xmath7 of 44 , with the total a sensitive probe of the mass cut and ns formation . with a 87y mean - life for decay into @xmath5sc which produces narrow lines at 68 and 78 kev ,
obscured sne can be detected throughout the entire galaxy for 300y given the 10@xmath8 photons cm2sec line sensitivity and 2 kev energy resolution ( at 70 kev ) possible for exist .
thus the likely detection of cas a ( iyudin et al 1994 ) can be extended to more distant but similarly ( or greater ) obscured sn to constrain the sn rate in the galaxy . the all - sky imaging of exist would extend the central - radian galactic survey planned for integral to the entire galaxy .
+ + since at least the `` long '' grbs located with bepposax are at cosmological redshifts , and have apparent luminosities spanning at least a factor of 100 , it is clear that even the apparently lower luminosity grbs currently detected by bepposax could be detected with batse out to z 4 and that the factor 5 increase in sensitivity with swift will push this back to z 5 - 15 ( lamb and reichart 1999 ) .
the additional factor of 4 increase in sensitivity for exist would allow grb detection and sub - arcmin locations for z 15 - 20 and thus allow the likely epoch of pop iii star formation to be probed if indeed grbs are associated with collapsars ( e.g. woosley 1993 ) produced by the collapse of massive stars .
the high throughput and spectral resolution for exist would enable high time resolution spectra which can test internal shock models for grbs .
+ + only 3 sgr sources are known in the galaxy and 1 in the lmc . since a typical 0.1sec sgr burst spike
can be imaged ( 5@xmath9 ) by exist for a peak flux of 200 mcrab in the 10 - 30 kev band , the typical bursts from the newly discovered sgr1627 - 41 ( woods et al 1999 ) with peak flux 2 10@xmath8 ergcm2sec would be detected out to 200 kpc .
hence the brightest `` normal '' sgr bursts are detectable out to 3 mpc and the rare giant outbursts ( e.g. march 5 , 1979 event ) out to 40 mpc . thus the population and physics of sgrs , and thus their association with magnetars and young snr ,
can be studied throughout the local group and the rare super - outbursts beyond virgo .
+ + the cosmic ir background ( cirb ) over 1 - 100@xmath10 is poorly measured ( if at all ) and yet can constrain galaxy formation and the luminosity evolution of the universe ( complementing * s1 * above ) . as reviewed by catanese and weekes ( 1999 ) ,
observing spectral breaks ( from @xmath11 absorption ) for blazars in the band 0.01 - 100 tev can measure the cirb out to z 1 _ if _ the intrinsic spectrum is known .
since the @xmath12-ray spectra of the detected ( low z ) blazars are well described by synchrotron - self compton ( ssc ) models , for which the hard x - ray ( 100kev ) synchrotron peak is scattered to the tev range , the hx spectra can provide both the required underlying spectra and time - dependent light curves for all objects ( variable ! ) to be observed with glast and high - sensitivity ground - based tev telescopes ( e.g. veritas ) . + + the success of batse as a hx monitor of bright accreting pulsars in the galaxy ( cf .
bildsten et al 1997 ) , in which spin histories and accretion torques were derived for a significant sample , can be greatly extended with exist : the very much larger reservoir of be systems can be explored , and wind vs. disk - fed accretion studied in detail .
the wide - field hx imaging and monitoring capability will also allow a new survey for pulsars and axps in highly obscured regions of the disk , complementing * s4 * above .
+ + the rms variability generally and qpo phenomena appear more pronounced above 10kev for x - ray binaries containing both bh and ns accretors , suggesting the comptonizing corona is directly involved .
thus qpos and hx spectral variations can allow study of the poorly - understood accretion disk coronae , with extension to the agn case .
although the wide - field increases backgrounds , and thus effective modulation , the very large area ( 1m@xmath0 ) of hx imaging area on any given source means that multiple 100 mcrab lmxbs could be simultaneously measured for qpos with 10% rms amplitude in the poorly explored 10 - 30 kev band .
to achieve the desired 0.05 mcrab sensitivity full sky up to 100 kev ( and beyond ) requires a very large area array of wide - field coded aperture ( or other modulation ) telescopes .
the very small field of view ( 10 ) of true focussing ( e.g. multi - layer ) hx telescopes precludes their use for all sky imaging and monitoring surveys .
exist - iss would take the coded aperture concept to a practical limit , with 8 telescopes each with 1m@xmath0 in effective detector area and 4040 in field of view ( fov ) .
the individual fovs are offset by 20 for a combined fov of 60x40deg , or 2sr . by orienting the 160axis perpendicular to the orbit vector
, the full sky can be imaged each orbit if the telescope array is fixed - pointed at the local zenith .
this gravity - gradient type orientation , and the large spatial area of the telescope array , are ideally matched for the iss , which provides a long mounting structure ( main truss ) conveniently oriented perpendicular to the motion , as depicted on the exist website .
the sensitivity would yield 10@xmath13 agns full sky , thus setting a confusion limit resolution requirement ( 1/40 `` beam '' ) of 5 . with this coded mask pixel size ,
high energy occulting masks ( 5 mm , w ) can be constructed with 2.5 mm pixel size for minimal collimation .
the mask shadow is then recorded by tiled arrays of cdznte ( czt ) detectors with effective pixel sizes of 1.3 mm , yielding a compact ( 1.3 m ) mask - detector spacing .
the czt detectors would likely be 20 mm square thick ( for 20% efficiency at 500 kev ) and read out by flip - chip bonded asics ( e.g. bloser et al 1999 , harrison et al 1999 ) .
the 8-telescope array is continuously scanning ( sources on the orbital plane drift across the 40fov in 10min ; correspondingly longer exposures / orbit near the poles ) , with each photon time - tagged and aspect corrected ( 10 ) so that iss pointing errors or flexure are inconsequential over the large fov .
source positions are centroided to 1for 5@xmath9 detections .
the resulting sky coverage is remarkably uniform with 25% variation in exposure full sky over the 2mo precession period of the iss orbit .
more details of the current mission concept are given in grindlay et al ( 2000 ) , and will be further developed in the implementation study being conducted by the exist science working group ( exswg ) .
bildsten , l. et al , _ apjs _ , * 113*,367 ( 1997 ) .
+ bloser , p. , grindlay , j. , narita , t. and jenkins , j. , _ proc .
spie _ , * 3765 * , 388 ( 1999 ) .
+ catanese , m. and weekes , t. , _ pasp _ , * 111 * , 1193 ( 1999 ) .
+ fabian , a. _ mnras _ , * 308 * , l39 ( 1999 ) .
+ grindlay , j.e .
et al , _ proc .
spie _ , * 2518 * , 202 ( 1995 ) .
+ grindlay , j.e .
et al , _ proc .
staif-2000 _ , in preparation ( 2000 ) .
+ harrison , f.a .
et al , _ proc .
, * 3765 * , 104 ( 1999 ) .
+ iyudin , a.f .
et al , _ a&a _ , * 284 * , l1 ( 1994 ) .
+ lamb , d.q . and reichart , d.e .
, _ apj _ , submitted ( astro - ph/9909002 ) ( 1999 ) .
+ vignati , p. et al , _ a&a _ , * 349 * , 57l ( 1999 ) .
+ woods , p.m. et al , _ apj _ , * 519 * , l139 ( 1999 ) .
+ woosley , s.e .
_ apj _ , * 405 * , 273 ( 1993 ) .
+ yi , i. and grindlay , j.e . , _ apj _ , * 505 * , 828 ( 1998 ) . | a deep all - sky imaging hard x - ray survey and wide - field monitor is needed to extend soft ( rosat ) and medium ( abrixas2 ) x - ray surveys into the 10 - 100 kev band ( and beyond ) at comparable sensitivity ( 0.05 mcrab ) .
this would enable discovery and study of 3000 obscured agn , which probably dominate the hard x - ray background ; detailed study of spectra and variability of accreting black holes and a census of bhs in the galaxy ; gamma - ray bursts and associated massive star formation ( popiii ) at very high redshift and soft gamma - ray repeaters throughout the local group ; and a full galactic survey for obscured supernova remnants .
the energetic x - ray imaging survey telescope ( exist ) is a proposed array of 8 1m@xmath0 coded aperture telescopes fixed on the international space station ( iss ) with 60x40deg field of view which images the full sky each 90 min orbit .
exist has been included in the most recent nasa strategic plan as a candidate mission for the next decade .
an overview of the science goals and mission concept is presented . /cm2secergs
@xmath1 s@xmath2 2ergs @xmath1 2cm@xmath0 cm2sec @xmath1 s@xmath2 60x40deg160@xmath4 40 0x6deg20@xmath4 6 44@xmath5ti |
in our first paper on this subject matter @xcite we studied a model of @xmath0 quantum oscillators ( qo ) located at different fixed positions in space , which do not interact with each other directly but only through weak couplings to a common scalar massless quantum field .
after coarse - graining the quantum field we derived the late - time covariance matrix of the open system of the @xmath0 oscillators with field - induced coupling and analyzed the structure of their quantum entanglement behavior for the case of @xmath1 .
we show that in the weak coupling limit the correlations adopt a simple pairwise structure , in that the correlation between any two qo is ignorant of the presence of the other qo .
the effects of field - induced interaction on quantum entanglement between two harmonic oscillators have been studied earlier in @xcite .
one interesting new feature shown there is the distance dependence of quantum entanglement , since the coupling is mediated by a quantum field whose influence varies with time and space .
the way how one oscillator s presence alters the entanglement of the other two was used in @xcite to capture a novel notion called ` entanglement domain ' . in this paper
we study the entanglement structure of the @xmath0 quantum oscillator system but in addition to each oscillator s weak coupling with a common quantum scalar field we allow also direct pairwise couplings with varying strength .
the salient features based on detailed studies of the @xmath1 case turn out to be quite different from the case of @xmath0 oscillators with only field - induced coupling .
the first entanglement feature , namely , that the correlation between any two qo are ignorant of the presence of the other qo , is no longer true . the second feature , namely , distance dependence of entanglement due to each oscillator s coupling to a quantum field , is now no longer prominent , being overwhelmed by the direct couplings .
we see that each case studied in these two papers captures one aspect of the overall behavior of the system of nos ( with @xmath1 studied in detail here ) : while the first case may depict neutral atoms the latter case depicts ions ( in an open quantum field rather than in a cavity or in a trap ) .
the idealized model - study here , though not a replica of actual experimental setups , can nonetheless provide us with a clear theoretical understanding and quantitative description of the entanglement structure of such commonly encountered continuous - variable quantum systems . quantifying the entanglement structure of a many - body quantum system can facilitate investigations into related issues in several current research areas .
for example , the results reported in this paper for a system of @xmath0 coupled oscillators with direct yet disparate couplings can help shed light on issues in _ macroscopic quantum phenomena _ ( mqp ) ( see , e.g. , @xcite and references therein ) .
entanglement being an exclusively quantum feature @xcite , it is natural to ask how to use entanglement as a signifier or marker of the existence of , or further , a quantitative measure of , the quantum features and behaviors in a macroscopic quantum system . on this issue
it is important to recognize that ( see , e.g. , @xcite ) there exist many _ levels of structure _ in a composite body and only by judicious choice of an appropriate set of _ collective variables _ can one give a good description of the dynamics of a specific level of structure .
these are often characterized by strong coupling between fine - grained elements or constituents in one particular level of structure ( e.g. the level of quarks / gluons or the level of nucleons ) and weaker coupling between the elements in a coarser - grained level of structure ( e.g. , atoms or molecules ) . capturing the quantum features of a macroscopic object is greatly facilitated by the existence and functioning of these collective variables depicting a specific level of structure . under what conditions
can one capture the quantum dynamics of an @xmath0 oscillator system by its center of mass ( com ) , the so - called com axiom " , was investigated earlier by chou et al @xcite who derived a sufficient condition for this to hold .
a recent essay @xcite ( see also cited work therein e.g. , @xcite ) showed the marked difference between entanglement within one specific level of structure and entanglement between different levels of structure , and identified which types would scale with @xmath0 .
these two aspects , namely , the role of the com in a many - body system and how the level - of - structure enter in the manifestation of macroscopic quantum phenomena , can be examined by using the entanglement structure results obtained here for a three - oscillator system with disparate coupling strengths .
the paper is organized as follows : in the next section we derive a set of langevin equations for @xmath0 coupled quantum oscillators interacting with a quantum scalar field based on the influence functional formalism approach .
this method first introduced by feynman and vernon @xcite for the study of quantum brownian motion has been applied to the derivation of markovian and non - markovian master equations for oscillator baths @xcite and for a quantum field environment @xcite .
( for an alternative treatment of a @xmath0 oscillator system , see @xcite and references therein . ) here we take the direct route of seeking solutions to the evolutionary operator of the reduced density matrix given formally by grabert et al @xcite for the derivation of correlation functions .
( for an alternative derivation of the non - markovian master equation and finding their solutions , see @xcite and references therein . ) in sec .
3 these are used to construct the covariance matrix of this system which is at the heart of our analysis of quantum entanglement . in sec .
4 we specialize to @xmath1 with disparate coupling and analyze the symmetric and asymmetric configurations described above . in sec .
5 we summarize our results and briefly describe their implications for some key issues in mqp and applications in other areas .
the appendix contains more details about symmetric gaussian systems .
we consider a system of @xmath0 interacting @xmath2 detectors where each detector s internal degree of freedom @xmath3 is described by a one - dimensional harmonic oscillator .
these detectors are allowed to have direct coupling via their internal degrees of freedom , and the coupling is assumed to be at most quadratic in the internal degrees of freedom .
in addition , each oscillator is coupled to one and the same massless scalar field @xmath4 which acts as an environment to the @xmath0 oscillator system .
as such there will be indirect interactions between the oscillators mediated by the shared bath .
the entanglement features of such a system with only field - induced interaction was studied in our first paper @xcite . for later purpose when we will consider a system of @xmath0 moving detectors ( such as that studied before in @xcite ) we can allow the detectors to be in motion with the trajectory of the @xmath5 detector specified by @xmath6 @xcite .
we assume that all the oscillators have the same mass @xmath7 , the same bare frequency @xmath8 , and are acted upon by a harmonic potential of the same form @xmath9 .
the lagrangian for the afore - specified system is given by @xmath10&=\int^{t}\!ds\sum_{i=1}^{n}\bigl[\frac{m}{2}\,\dot{\chi}^{(i)2}(s)-v(\chi^{(i)})-\sum_{j > i}^{n}v_{ij}(\chi)\bigr]\notag\\ & \qquad\qquad+\int^{t}\!ds\int\!d^{3}\mathbf{x}\sum_{i=1}^{n}j^{(i)}(\mathbf{x},s)\phi(\mathbf{x},s)\notag\\ & \qquad\qquad\qquad+\frac{1}{2}\int^{t}\!ds\int\!d^{3}\mathbf{x}\bigl[\partial_{\mu}\phi(\mathbf{x},s)\bigr]\bigl[\partial^{\mu}\phi(\mathbf{x},s)\bigr]\,,\label{e : ypowpea}\end{aligned}\ ] ] where @xmath11 describes the source that accounts for the interaction between the internal degree of freedom @xmath12 of the @xmath5 detector and the environmental scalar field @xmath13 with coupling strength @xmath14 which we will assume to be weak .
the direct coupling between the detectors are given by the interaction potential @xmath15 of the form @xmath16^{2}\,,\qquad\sigma_{ij}=\sigma_{ji}\,,\ ] ] with the coupling strength @xmath17 which is positive and can be strong .
the partial derivative @xmath18 in the action of the free scalar field represents @xmath19 and we choose the minkowski metric @xmath20 with the convention @xmath21 .
since we are primarily interested in the dynamics of the internal degrees of freedom @xmath12 of the detectors , obtaining the evolution of their reduced density matrix of @xmath22 constructed by integrating over the scalar field environment will serve our purpose .
this enables us to find the corresponding time evolution of the expectation value of an operator @xmath23 associated with the detector s internal degrees of freedom by @xmath24 where @xmath25 means taking the trace over a complete set of @xmath26 states .
for simplicity we use @xmath26 to represent the collection of the variables @xmath27 .
the time evolution of the reduced density operator , on the other hand , is given by @xmath28 here @xmath29 is the full density operator at the initial time @xmath30 , which describes the initial states of the combined systems @xmath26 and @xmath13 , and @xmath31 is the unitary time evolution operator for the complete system governed by the action .
the trace operator @xmath32 acts on the degrees of freedom of the background scalar field .
thus the reduced density operator @xmath22 alone has offered us sufficient information about the time evolution of the reduced system under the influence of the background scalar field .
here we assume at the initial time @xmath30 , the state of the total system can be factorized into a direct product of the initial state @xmath22 of @xmath26 and the counterpart @xmath33 of the scalar field @xmath13 , @xmath34 each of which explicitly takes the form @xmath35\,,\label{e : erjdkjnds}\\ \varrho_{\beta}(\phi_{a},\phi'_{a};t_{a})&=\langle\phi_{a}\vert e^{-\beta\,h_{\phi}[\phi]}\vert\phi'_{a}\rangle\,.\end{aligned}\ ] ] the parameter @xmath36 is the width of the wavepacket associated with @xmath12 , and @xmath37 $ ] is the free scalar field hamiltonian . the variables @xmath38 , @xmath39 represent @xmath40 and @xmath41 respectively .
we will use this shorthand notation for other variables .
the initial state describes the system ( with variables @xmath12 ) prepared as wavepackets of equal width @xmath36 and the environment as that of a thermal scalar field at temperature @xmath42 . at @xmath30 the nos is brought in contact with the field environment .
the evolution of the combined system + environment will be governed by the unitary evolution operator @xmath43 till the final time @xmath44 according to .
although the initial state is navely simplified and a more sophisticated and physically realistic state has been proposed by @xcite , it is amply sufficient to describe to the late - time dynamics of the reduced system .
the possible artifact induced by a factorized initial state is significant only within a very short time from @xmath45 determined by the highest frequency cutoff in the combined system @xcite . with these provisions
the final reduced density matrix for the internal degrees of the detectors can be calculated exactly with the help of the influence functional formalism @xcite .
once we express in terms of path integrals , we have @xmath46-i\,s\bigl[\chi_{-},\phi_{-}\bigr]\biggr\}\,,\label{e : wondkwqooq}\end{aligned}\ ] ] which has a gaussian integrand . after carrying out the multiple integrals , we arrive at @xmath47 with @xmath48 and @xmath49/2 $ ] .
the evolutionary operator @xmath50 essentially describes how the reduced density operator @xmath22 of the detectors evolves with time .
it is given by @xmath51\bigr\}\,,\end{aligned}\ ] ] where the coarse - grained effective action @xmath52 is defined by @xmath53=s_{\chi}\bigl[\chi_{+}\bigr]-s_{\chi}\bigl[\chi_{-}\bigr]-i\,\ln { \cal f}\bigl[\chi_{+},\chi_{-}\bigr]\,,\ ] ] with the action of @xmath26 given by @xmath54=\int^{t}\!ds\sum_{i=1}^{n}\bigl[\frac{m}{2}\,\dot{\chi}^{(i)2}(s)-v(\chi^{(i)})-\sum_{j > i}^{n}v_{ij}(\chi)\bigr]\,,\ ] ] and the influence functional @xmath55 given by @xmath56\\ & = \exp\left[i\int_{t_{a}}^{t_{b}}\!ds\!\int_{t_{a}}^{t_{b}}ds'\sum_{i,\,j}g_{i}g_{j}\left\{q^{(i)}(s)g^{ij}_{r}(s , s')r^{(j)}(s')+\frac{i}{2}\,q^{(i)}(s)g^{ij}_{h}(s , s')q^{(j)}(s')\right\}\right]\,.\notag\end{aligned}\ ] ] we use an overbar above the action or the influence functional to connote that they are evaluated along the classical trajectory of @xmath57 , @xmath58 .
the normalization constant @xmath59 in will be determined by the completeness relation .
the influence functional @xmath55 merits some discussion .
it essentially summarize the effects of the environment scalar field on the internal degrees of freedom @xmath26 after we trace out the scalar field in .
it contains two kernel functions @xmath60 and @xmath61 , which are respectively the dissipation kernel and the fluctuation kernel @xcite , @xmath62\times\varrho_{\beta}\bigr)\,,\\ g^{ij}_{h}(s , s')&=\frac{1}{2}\,\operatorname{tr}_{\phi}\bigl(\bigl\{\phi(\mathbf{z}^{(i)}(s),s),\phi(\mathbf{z}^{(j)}(s'),s')\bigr\}\times\varrho_{\beta}\bigr)\,,\label{e : enrkea}\end{aligned}\ ] ] where @xmath32 is the trace taken over the field variables .
while the noise kernel represents the coarse - grained effect of the environment as stochastic source(s ) and the dissipation kernel captures the backreaction effect of the coarse - grained environment on the detectors dynamics .
ostensibly not to be arbitrarily assigned by hand they are bound by a set of fluctuation - dissipation relations which should be well - known and in the general case also a set of propagation - correlation relation which are lesser known @xcite .
furthermore , they mediate the response of one detector to the others both by direct couplings amongst the detector and even in the case when there is no direct coupling since the trajectory of each detector affects the field which in turn affects the other detectors .
therefore for such a @xmath0-detector system , the dynamics of a nos is highly non - markovian .
note incidentally they should not be mistaken to be tensorial quantities , the superscripts @xmath63 are merely labels for the action of the @xmath64 detector on the @xmath5 detector .
it proves more convenient later if we cast the evolutionary operator @xmath50 in in the matrix notation and write it explicitly in terms of the initial and the final values of @xmath65 , @xmath66 .
let us denote the column matrix @xmath67 by @xmath68 and the coupling matrix @xmath69 as @xmath70 .
the action @xmath71 then assumes a compact form @xmath72=\int^{t_{b}}_{t_{a}}\!ds\;\bigl[\frac{m}{2}\,\dot{\pmb{\chi}}^{t}\cdot\dot{\pmb{\chi}}-\frac{m}{2}\,\pmb{\chi}^{t}\cdot\pmb{\omega}^{2}\cdot\pmb{\chi}\bigr]\,,\ ] ] where @xmath73 in the superscript means the transpose of a matrix .
the @xmath74 interaction matrix @xmath75 , given by @xmath76 -\sigma_{21 } & \omega_{0}^{2}+\displaystyle\sum^{n}_{k=1}{}'\sigma_{2k } & -\sigma_{23 } & \cdots & -\sigma_{2n}\\[8pt ] -\sigma_{31 } & -\sigma_{32 } & \ddots & \cdots & -\sigma_{3n}\\[8pt ] \vdots & \vdots & \vdots & \ddots & \vdots\\[8pt ] -\sigma_{n1 } & -\sigma_{n2 } & -\sigma_{n3 } & \cdots & \omega_{0}^{2}+\displaystyle\sum^{n}_{k=1}{}'\sigma_{nk } \end{pmatrix}\,,\ ] ] essentially indicates the strength of direct coupling between the internal degrees of freedom of the different detectors .
the prime next to the summation , for example , @xmath77 means that the summation index @xmath78 runs from @xmath79 to @xmath0 except for @xmath80 . if we vary the effective action @xmath52 with respect to @xmath81 or @xmath82 , it produces two sets of equations of motion for the time evolution of @xmath82 , @xmath81 , respectively .
in particular , the equation of motion for the real part of @xmath82 is of special importance @xmath83 where for simplicity we have let the initial time @xmath84 and the final time @xmath85 .
this equation is highly coupled due to both off - diagonal elements of the interaction matrix @xmath75 and the non - local integral expression in the equation of motion . in general even if a suitable basis is found to diagonalize the interaction matrix , its non - markovian nature still render this equation very much tangled . later in sec .
[ s : rnfekjna ] we will focus on a special case when the equation of motion can be totally decoupled . in general the solution to this equation
can formally be expressed in terms of the fundamental solutions @xmath86 , @xmath87 , which satisfy the same equation of motion and the conditions @xmath88 the matrix @xmath89 is an @xmath74 identity matrix .
we immediately see from that these fundamental solutions contain information of dissipation effects due to the backreaction of the environment .
the fundamental solutions to @xmath90 can also be used to construct the solutions of @xmath81 , so the general solutions of @xmath90 and @xmath81 are both given by @xmath91\cdot\mathbf{r}_{a}+\mathbf{d}_{2}(s)\cdot\mathbf{d}^{-1}_{2}(t)\cdot\mathbf{r}_{b}\,,\label{e : uiewjs}\\ \mathbf{q}(s)&=\bigl[\mathbf{d}_{1}(t - s)-\mathbf{d}_{2}(t - s)\cdot\mathbf{d}^{-1}_{2}(t)\cdot\mathbf{d}_{1}(t)\bigr]\cdot\mathbf{q}_{b}+\mathbf{d}_{2}(t - s)\cdot\mathbf{d}^{-1}_{2}(t)\cdot\mathbf{q}_{a}\,,\end{aligned}\ ] ] respectively .
hereafter we will suppress the subscript @xmath92 in @xmath90 to simplify notations .
note that the velocities @xmath93 , @xmath94 at the initial and the final moment can also be given in terms of the fundamental solutions and @xmath95 , @xmath96 . with these results ready at hand ,
the evolutionary operator becomes @xmath97\right\}\,.\notag\end{aligned}\ ] ] the matrices @xmath98 , @xmath99 and @xmath100 are defined by @xmath101 they highlight the backreaction effects due to the fluctuations of the scalar field environment , in contrast with the information about the dissipation backreaction effects wrapped up in the fundamental solutions @xmath102 .
the matrix @xmath103 is composed of elements .
the two matrix functions @xmath104
, @xmath105 are constructed by @xmath106 where @xmath107 are the fundamental solutions introduced in . by , and the explicit expression of the evolutionary operator @xmath50 , we can obtain the reduced density matrix for @xmath26 at the final time , from which we are able to compute the expectation values of the operators associated with @xmath26 at that moment according to .
this enables us to investigate the non - equilibrium dynamics of the reduced system @xmath26 under the influence of the scalar field environment . for the stated purpose of this work the evolution of the correlations between components of @xmath108 and their conjugate momenta @xmath109 are of special interest .
we summarize the results here , @xmath110\bigl[\mathbf{r}_{b}^{(m)}+\frac{1}{2}\,\mathbf{q}_{b}^{(m)}\bigr]\notag\\ & \qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\times j(\mathbf{r}_{b},\mathbf{q}_{b},t;\mathbf{r}_{a},\mathbf{q}_{a},0)\,\varrho_{\chi}(\mathbf{r}_{a},\mathbf{q}_{a})\notag\\ & = \frac{1}{m^{2}}\bigl[\mathbf{d}_{2}(t)\cdot\mathbf{a}(t)\cdot\mathbf{d}_{2}(t)\bigr]_{lm}+\left\{\frac{\varsigma^{2}}{2}\bigl[\mathbf{d}^{2}_{1}(t)\bigr]_{lm}+\frac{1}{2m^{2}\varsigma^{2}}\bigl[\mathbf{d}^{2}_{2}(t)\bigr]_{lm}\right\}\,,\label{e : wouwkn}\\
\langle\,\mathbf{p}_{b}^{(l)}\mathbf{p}_{b}^{(m)}\rangle&=-\int\!d\mathbf{r}_{b}d\mathbf{q}_{b}\!\int\!d\mathbf{r}_{a}d\mathbf{q}_{a}\;\delta^{(n)}(\mathbf{q}_{b})\,\bigl[\frac{1}{2}\,\frac{\partial}{\partial\mathbf{r}^{(l)}_{b}}+\frac{\partial}{\partial\mathbf{q}^{(l)}_{b}}\bigr]\bigl[\frac{1}{2}\,\frac{\partial}{\partial\mathbf{r}^{(m)}_{b}}+\frac{\partial}{\partial\mathbf{q}^{(m)}_{b}}\bigr]\notag\\ & \qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\times j(\mathbf{r}_{b},\mathbf{q}_{b},t;\mathbf{r}_{a},\mathbf{q}_{a},0)\,\varrho_{\chi}(\mathbf{r}_{a},\mathbf{q}_{a};0)\notag\\ & = \bigl[\dot{\mathbf{d}}_{2}(t)\cdot\mathbf{a}\cdot\dot{\mathbf{d}}_{2}(t)\bigr]_{lm}+\mathbf{b}_{lm}+\bigl[\dot{\mathbf{d}}_{2}(t)\cdot\mathbf{c}\bigr]_{lm}+\bigl[\mathbf{c}^{t}\cdot\dot{\mathbf{d}}^{t}_{2}(t)\bigr]_{lm}\notag\\ & \qquad\qquad\qquad\quad+\left\{\frac{m^{2}\varsigma^{2}}{2}\,\bigl[\dot{\mathbf{d}}_{1}(t)\dot{\mathbf{d}}^{t}_{1}(t)\bigr]{}_{lm}+\frac{1}{2\varsigma^{2}}\,\bigl[\dot{\mathbf{d}}_{2}(t)\dot{\mathbf{d}}^{t}_{2}(t)\bigr]{}_{lm}\right\}\,,\\ \langle\,\pmb{\chi}_{b}^{(l)}\mathbf{p}_{b}^{(m)}\rangle&=-i\!\int\!d\mathbf{r}_{b}d\mathbf{q}_{b}\int\!d\mathbf{r}_{a}d\mathbf{q}_{a}\;\delta^{(n)}(\mathbf{q}_{b})\,\bigl[\mathbf{r}_{b}^{(l)}+\frac{1}{2}\,\mathbf{q}_{b}^{(l)}\bigr]\bigl[\frac{1}{2}\,\frac{\partial}{\partial\mathbf{r}^{(m)}_{b}}+\frac{\partial}{\partial\mathbf{q}^{(m)}_{b}}\bigr]\notag\\ & \qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\times j(\mathbf{r}_{b},\mathbf{q}_{b},t;\mathbf{r}_{a},\mathbf{q}_{a},0)\,\varrho_{\chi}(\mathbf{r}_{a},\mathbf{q}_{a})\notag\\ & = \frac{i}{2}\,\delta_{lm}+\frac{1}{m}\,\bigl[\mathbf{d}_{2}(t)\cdot\mathbf{a}\cdot\dot{\mathbf{d}}_{2}(t)\bigr]_{lm}+\frac{1}{m}\,\bigl[\mathbf{d}_{2}(t)\cdot\mathbf{c}\bigr]_{lm}\notag\\ & \qquad\qquad\qquad+\left\{\frac{m\varsigma^{2}}{2}\,\bigl[\mathbf{d}_{1}(t)\cdot\dot{\mathbf{d}}_{1}(t)\bigr]_{lm}+\frac{1}{2m\varsigma^{2}}\,\bigl[\mathbf{d}_{2}(t)\cdot\dot{\mathbf{d}}_{2}(t)\bigr]_{lm}\right\}\,,\\ \langle\,\mathbf{p}_{b}^{(m)}\pmb{\chi}_{b}^{(l)}\rangle&=-i\,\delta_{lm}+\langle\,\pmb{\chi}_{b}^{(l)}\mathbf{p}_{b}^{(m)}\rangle\,.\label{e : youwkn}\end{aligned}\ ] ] in addition we have @xmath111 and @xmath112 due to the initial density matrix @xmath113 we choose . note that the commutation relation is preserved .
furthermore , each correlation in to can be decomposed into two subgroups .
the first group , which depends on @xmath36 in this case , is called the intrinsic or active component .
basically it is caused by the intrinsic quantum fluctuations associated with the internal degrees of freedom ; thus it depends on their wavefunctions . on the other hand , the second group , called the induced or passive component ,
is induced from the quantum fluctuations of the environment scalar field , thus independent of the wavefunctions of the internal degrees of freedom . in general
, the active components gradually decay to zero with time due to the dissipation caused by the environment . on the other hand , the passive components increases from zero owing to the quantum fluctuations of the environment .
usually it will grow without bound if we carelessly ignore the contribution of the dissipation backreaction @xcite .
the fluctuation - dissipation relation ensures the energy balance between these two processes such that the induced or passive components will approach a saturated value and remains bounded . at late time
@xmath114 , when the active components vanish , the correlation functions are completely determined by the induced components , @xmath115_{lm}\,,\label{e : nkqqmna } \intertext{and } \langle\,\mathbf{p}_{b}^{(l)}\mathbf{p}_{b}^{(m)}\rangle&=g^{2}\int^{\infty}_{-\infty}\!\frac{d\omega}{2\pi}\;\omega^{2}\,\bigl[\widetilde{\pmb{\mathfrak{d}}}_{2}(\omega)\cdot\widetilde{\mathbf{g}}_{h}(\omega)\cdot\widetilde{\pmb{\mathfrak{d}}}_{2}(-\omega)\bigr]_{lm}\,,\label{e : rnkwia}\\ \frac{1}{2}\langle\,\{\pmb{\chi}_{b}^{(l)},\mathbf{p}_{b}^{(m)}\}\rangle&=\frac{i}{m}\int^{\infty}_{-\infty}\!\frac{d\omega}{2\pi}\;\omega\,\bigl[\widetilde{\pmb{\mathfrak{d}}}_{2}(\omega)\cdot\widetilde{\mathbf{g}}_{h}(\omega)\cdot\widetilde{\pmb{\mathfrak{d}}}_{2}(-\omega)\bigr]_{lm}=0\,,\end{aligned}\ ] ] where we have chosen @xmath116 and defined a new kernel function @xmath117 , and its fourier transform @xmath118 is given by @xmath119^{-1}\,,\ ] ] if we define the fourier transformation @xmath120 of a given function @xmath121 by @xmath122 it can be further shown that the integrand of becomes @xmath123 and the correlation functions in reduce to @xmath124_{lm}\,,\label{e : kwenkjsdb}\\ \langle\,\mathbf{p}_{b}^{(l)}\mathbf{p}_{b}^{(m)}\rangle&=m\,\operatorname{im}\int^{\infty}_{-\infty}\!\frac{d\omega}{2\pi}\;\omega^{2}\coth\frac{\beta\omega}{2}\bigl[\widetilde{\pmb{\mathfrak{d}}}_{2}(\omega)\bigr]_{lm}\,.\end{aligned}\ ] ] once we know the correlations between the detectors , we may construct the covariance matrix out of them .
constitutes the elements of the covariance matrix , which plays a central role in determining the entanglement of the corresponding system . since the green s functions , in particular , the retarded green s function @xmath125 and the hadamard function @xmath126 play a key role in determining the backreactions of the scalar field environment and the dynamics of the reduced system , it is worthy to expound their properties .
we observe that for the thermal states of the massless scalar field , the retarded green s function is given by @xmath127\times\varrho_{\beta}\bigr)\notag\\ & = i\,\theta(t - t')\int\frac{d^{3}\mathbf{k}}{(2\pi)^{3}}\frac{1}{2\omega}\,\bigl[e^{i\mathbf{k}\cdot(\mathbf{x}-\mathbf{x}')-i\omega(t - t')}-e^{-i\mathbf{k}\cdot(\mathbf{x}-\mathbf{x}')+i\omega(t - t')}\bigr]\notag\\ & = \theta(\tau)\,\frac{1}{4\pi r}\bigl[\delta(\tau - r)-\delta(\tau+r)\bigr]=\frac{1}{2\pi } \,\theta(\tau)\,\delta(\tau^{2}-r^{2})\,,\end{aligned}\ ] ] with @xmath128 and @xmath129 .
it is independent of the temperature @xmath42 of the environment . in the limit @xmath130 , we see that @xmath131=\delta'(\tau)\,,\ ] ] so , in this case , the retarded green s function reduces to @xmath132 the fourier transform of @xmath133 can be derived from as @xmath134 if @xmath135 , @xmath136 are not explicit functions of @xmath137 and @xmath138 respectively .
apparently it gives @xmath139 .
care has to be taken for the limit @xmath130 , where @xmath140 has the real divergence .
usually it is interpreted as a form of self - energy and can be absorbed as part of the renormalized mass , @xmath141 eq .
implies that the self - force on @xmath26 , which essentially is the integral expression in , will be effectively reduced to a local term .
the hadamard function of the scalar field in its thermal state is given by @xmath142\notag\\ & = \frac{1}{4\pi^{2}r}\int_{-\infty}^{\infty}d\omega\;\operatorname{sgn}(\omega)\bigl[\frac{e^{-\beta\lvert\omega\rvert}}{1-e^{-\beta\lvert\omega\rvert}}+\frac{1}{2}\bigr]\,\sin\omega r\,e^{-i\omega\tau}\label{e : gjkndsaa}\,,\end{aligned}\ ] ] again with @xmath128 , @xmath129 , and @xmath42 being the temperature of the thermal state of the scalar field .
carrying out directly gives @xmath143\,.\end{aligned}\ ] ] from , we can immediately read out the fourier transform of the hadamard function at finite temperature @xmath42 , @xmath144\,\frac{\sin\omega
r}{2\pi r}=\coth\frac{\beta\omega}{2}\,\frac{\sin\omega r}{4\pi r}\,.\ ] ] in the limit @xmath130 , it reduces to @xmath145=\frac{\omega}{4\pi}\,\coth\frac{\beta\omega}{2}\,.\ ] ] compare with , we see that @xmath146 this is the fluctuation - dissipation relation .
now we can make a connection with the @xmath0-detector case .
there we have the retarded green s function given by @xmath147 , with @xmath148 being the trajectory of the @xmath5 detector .
if we assume that the trajectories of the probes never intersect , then @xmath149 for @xmath150 . in the special case that all the probes are located at fixed positions , @xmath140 in
will be essentially given by with @xmath151 .
this implies that @xmath152^{-1}=\left[-\omega^{2}\mathbf{i}+\pmb{\omega}^{2}-\dfrac{g^{2}}{m}\,\widetilde{\mathbf{g}}_{r}^{*}(\omega)\right]^{-1}=\widetilde{\pmb{\mathfrak{d}}}_{2}^{*}(\omega)\,.\ ] ] on the other hand , we have @xmath153 given by @xmath154 with @xmath151 , if the probes are located at fixed positions .
since the covariance matrix contains the essential information about a many - body quantum system , before we delve into the entanglement structure of a @xmath0 oscillators system ( nos ) it is helpful to spend a little space to describe its main features .
the covariance matrix is a generalization of the variance function for a single variable theory to a multiple - component theory . for illustration purpose with relevance to the subsequent discussions ,
let us consider an @xmath0-mode quantum system @xcite with the canonical variables @xmath155 and @xmath156 , which satisfy the commutation relations @xmath157=i\,\delta_{jk}\,,\qquad\qquad\bigl[\chi^{(j)},\chi^{(k)}\bigr]=\bigl[p^{(j)},p^{(k)}\bigr]=0\,.\ ] ] it proves convenient to define a vector @xmath158 containing @xmath159 canonical variables , @xmath160 the commutation relations corresponding to can then be written as @xmath161&=i\,\mathbf{j}\,,&\mathbf{j}&=\bigoplus_{k=1}^{n}\,\pmb{\omega}\,,&\pmb{\omega}&=\begin{pmatrix}0 & + 1\\-1 & 0\end{pmatrix}\,.\end{aligned}\ ] ] now if we apply a real linear transformation , represented by a @xmath162 real matrix @xmath163 , upon the canonical variables @xmath26 and @xmath164 , @xmath165 the condition that @xmath166 remain as canonical variables , namely , that @xmath166 obey the same commutation relations as @xmath158 , demands that the transformation matrix @xmath163 must satisfy @xmath167=\mathbf{s}\bigl[\mathbf{r},\mathbf{r}^{t}\bigr]\mathbf{s}^{t}=i\,\mathbf{s}\cdot\mathbf{j}\cdot\mathbf{s}^{t}=i\,\mathbf{j}\,,\qquad\qquad\rightarrow\qquad\qquad \mathbf{s}\cdot\mathbf{j}\cdot\mathbf{s}^{t}=\mathbf{j}\,.\ ] ] this turns out to be the definition for an element in the real symplectic group @xmath168 .
thus the transformation represented by @xmath163 is called the symplectic transformation . since the symplectic transformation maps one set of canonical variables in the phase space to another , it enjoys particular importance in physics .
let the expectation value of an observable @xmath169 be given by @xmath170 , where @xmath171 is the density matrix operator of a given state . without loss of generality , we may assume that the expectation value of @xmath158 in the state @xmath171 is zero , i.e. , @xmath172 , because even if the mean value does not vanish , we can still make it happen by shifting the variable @xmath173 .
next consider the matrix @xmath174 .
if we decompose its expectation value into @xmath175\rangle=\mathbf{v}+\frac{i}{2}\,\mathbf{j}\,,\ ] ] the matrix @xmath176 on the right hand side is called the covariance matrix : @xmath177 sometimes it proves handy to express the covariance matrix @xmath176 by @xmath178 block matrices @xmath179 \pmb{\sigma}_{12}^{t } & \pmb{\sigma}_{22 } & \cdots & \pmb{\sigma}_{2n}\\[6pt ] \vdots & \vdots & \ddots & \vdots\\[6pt ] \pmb{\sigma}_{1n}^{t } & \pmb{\sigma}_{2n}^{t } & \cdots & \pmb{\sigma}_{nn } \end{pmatrix}\,,\ ] ] where @xmath180 \dfrac{1}{2}\langle\bigl\{p^{(j)},\chi^{(k)}\bigr\}\rangle & \dfrac{1}{2}\langle\bigl\{p^{(j)},p^{(k)}\bigr\}\rangle \end{pmatrix}\,.\ ] ] note that @xmath181 is symmetric .
it should be apparent from the construction that the covariance matrix is indeed a generalization of the variance of single variable functions . according to the williamson theorem @xcite
, we can always find some symplectic matrix @xmath163 that will transform the symmetric covariance matrix @xmath176 into a diagonal form @xmath182 this implies that we can always find a new set of canonical bases such that there is no correlation between them .
the parameters @xmath183 are the symplectic eigenvalues of the covariance matrix @xmath176 . alternatively they can be found by solving the symplectic eigenvalue problem , @xmath184 with the corresponding eigenvector @xmath185 , or by the conventional eigenvalue problem @xmath186 for the matrix @xmath187 . obviously the symplectic eigenvalues , as well as the determinant of the covariance matrix , are invariant under the symplectic transformations .
in fact , from eq . , we see that the uncertainty principle requires @xmath188 for all @xmath78 because in the basis @xmath189 that diagonalizes the covariance matrix @xmath176 into the williamson normal form @xmath190 , each @xmath178 block matrix along the diagonal has the form @xmath191 and , for each mode @xmath78 , the uncertainty principle rules that @xmath192 . hence it implies that if the covariance matrix @xmath176 is physically realizable , then its symplectic eigenvalues must be greater than or equal to @xmath193 .
this further shows that @xmath194-\dfrac{i}{2 } & \eta_{k}\end{pmatrix}\geq0\,.\ ] ] since the covariance matrix can always be put into the diagonal form by a symplectic transformation , eq .
asserts that by virtue of the uncertainty principle a physically realizable covariance matrix must satisfy @xmath195 that is , it must be positive - definite . for the benefit of the novice and for the sake of completeness we preface the discussion of entanglement measure by a short summary of what quantum entanglement is .
in the pure state cases , if a state @xmath196 can be written as a product of pure states such as @xmath197 for @xmath198 and @xmath199 in a bipartite system , then it is called a product state or a separable state .
if the state @xmath200 can not be written as , then @xmath200 is an entangled state .
a famous example of an entangled state for a bipartite system is the bell state @xmath201 where @xmath202 , @xmath203 are two possible outcomes of a qubit state .
here comes the natural question : if we are given a pure state @xmath204 , @xmath205 with @xmath206 and @xmath207 , how do we know whether the state @xmath208 can be written into a product state like with an appropriate choice of bases ?
the answer is provided by the singular value decomposition .
it says that for every complex matrix @xmath209 , there always exist unitary transformations @xmath210 and @xmath211 such that @xmath212 is diagonal with real , non - negative diagonal elements .
thus for each state @xmath208 , we can always find suitable bases @xmath213 and @xmath214 in terms of which reduces to @xmath215 where the @xmath216 are known as schmidt coefficients , and the upper limit of the sum is determined by the dimension of the smaller subsystem .
it is then straightforward to see that , compared with , if there is only one nonvanishing schmidt coefficient , then the state @xmath208 is a separable state , and vice versa .
since the singular values @xmath216 are uniquely defined , the number of schmidt coefficients provides a convenient criterion of separability of a state . in realistic situations we more often have to deal with mixed states , such as , when we consider an open quantum system , which interacts with an environment .
due to its huge number of degrees of freedom we can not track all the details of the motion of the environment . if what we need is the overall effect of the environment on the system we can trace out the environment in a fashion similar to and obtain a reduced description for the state of the open system which is usually mixed .
for the mixed state cases , the density matrix of a quantum system takes on the generic form @xmath217 where @xmath218 is a set of complete bases for the system .
the weight @xmath219 is the probability of findng the system in the state @xmath218 .
unitarity requires that @xmath220 a well - known simple criterion for discerning a mixed state from a pure one is : given the density matrix of the state @xmath171 , if the state is pure , the trace of @xmath221 is equal to 1 .
one the other hand , if it is a mixed state , then the trace is less than 1 , that is , @xmath222 the simplest mixed state for a bipartite system is described by the so - called mixed product state @xmath223 where @xmath224 and @xmath225 are the density matrices of the respective subsystems . more generally , the joint system can be described by a convex sum of different product states @xmath226 with @xmath227 and @xmath228 .
the corresponding state is called the separable mixed state . from this definition ,
if a mixed state of a bipartite system can not be expressed as , then the state is entangled .
in other words , a mixed state @xmath171 is entangled if there are no states @xmath229 , @xmath230 , and non - negative value of @xmath231 , such that @xmath171 can be expressed as a convex sum like .
it has bee shown @xcite that if a mixed state is separable , then the partial transpose of the density matrix remains positive @xmath232 where @xmath73 in the superscript represents the transpose operation . here , as an example , we perform transposition on only @xmath224 .
this is the positive - partial - transpose ( ppt ) separability criterion .
thus , in principle , if we find a negative eigenvalue for the partially transposed density matrix , then the corresponding state is entangled .
the sufficient condition of the separability criterion , on the other hand , is valid only for low - dimensional bipartite systems @xcite .
this powerful criterion is further extended to an @xmath233 bi - partitions of the @xmath234-mode bi - symmetric gaussian system @xcite .
both sufficient and necessary conditions apply to such a gaussian system .
when we deal with a system described by continuous variables , the criterion based on the density matrix of the system becomes inconvenient to work with because the dimension of the density matrix for continuous variables is usually infinite .
this is where the covariance matrix finds its use . when translated into the covariance matrix , the partial transpose of the density matrix is equivalent to the mirror reflection of the canonical momentum in the corresponding subsystem
let the mirror reflection operator be @xmath235 .
the peres - horodecki positive partial transpose criterion @xcite then says @xmath236 because when @xmath237 is a bona fide density matrix , @xmath238 will be a bona fide covariance matrix , and then eq
. implies eq . .
conversely , the entanglement criterion @xcite then says that if the state is entangled , the partial transpose of its density matrix is negative , or equivalently the symplectic values @xmath239 of the corresponding partially transposed covariance matrix can be smaller than @xmath193 . a simple and computable entanglement measure known as ` negativity '
is defined based on the idea that if one of the eigenvalues @xmath240 of @xmath237 is negative , then @xmath171 is entangled @xcite .
thus the negativity @xmath241 is given by @xcite @xmath242 if @xmath237 is positive semi - definite , then the negativity @xmath241 vanishes . on the other hand it will take on positive values if @xmath237 has at least one negative eigenvalue .
however , due to the limitation of the positive - partial - transpose ( ppt ) criterion @xcite , the negativity is fully valid only for low - dimensional systems such as a @xmath178 or a @xmath243 system .
an alternative entanglement measure , which is often used for the continuous - variable system , is a variant of the logarithmic negativity @xmath244
it basically collects the symplectic eigenvalues of the partially - transposed covariance matrix , which are smaller than @xmath193 , @xmath245 the equal sign can be satisfied by the ppt entangled state @xcite because in higher dimensions there are entangled states which have a positive partial - transposed density matrix operator .
now we are in a position to discuss entanglement between three strongly coupled detectors .
let these detectors be labeled by @xmath246 , @xmath247 , @xmath248 .
assume that the coupling strength @xmath249 between the @xmath250 pair is the same as that between the @xmath251 pair , but different from the coupling strength @xmath252 between @xmath247 and @xmath248 .
this leads to an interaction matrix @xmath75 of the form @xmath253
-\lambda & \omega_{0}^{2}+\lambda+\sigma & -\sigma\\[6pt ] -\lambda & -\sigma & \omega_{0}^{2}+\lambda+\sigma \end{pmatrix}\,.\ ] ] it describes the case of disparate coupling between three detectors , that is , @xmath254 but @xmath255 in .
these coupling strengths are not assumed weak in comparison with the parameter @xmath256 , which is the oscillator frequency of the internal degrees of freedom ( or the renormalized oscillating frequency after absorbing the divergence in the retarded green s function ) in the absence of the inter - detector coupling .
nonetheless , we still assume the coupling @xmath257 between the detectors and the environmental scalar field is weak . in general , the motion of such a system is highly non - markovian due to the multi - time correlations generated in the sharing of a common bath .
this makes it difficult to find the normal modes of the motion .
however , for certain configurations , this can be done . to accomplish this we first discuss how to do the diagonalization of the interaction matrix .
the eigenvalues @xmath258 and the normalized eigenvectors @xmath259 for the interaction matrix @xmath75 are given by @xmath260 if we construct the matrix @xmath261 in terms of the normalized eigenvectors , it will transform the interaction matrix @xmath75 into the diagonal form @xmath262 by @xmath263 , @xmath264\dfrac{1}{\sqrt{3 } } & -\dfrac{1}{\sqrt{6 } } & \dfrac{1}{\sqrt{2}}\\[10pt ] \dfrac{1}{\sqrt{3 } } & -\dfrac{1}{\sqrt{6 } } & -\dfrac{1}{\sqrt{2 } } \end{pmatrix}\,,\qquad\qquad\qquad\pmb{\lambda}^{2}=\begin{pmatrix } \omega_{0}^{2 } & 0 & 0\\[16pt]0 & \omega_{0}^{2}+3\lambda & 0\\[16pt]0 & 0 & \omega_{0}^{2}+\lambda+2\sigma\end{pmatrix}\,.\ ] ] however , the elements of the @xmath75 matrix has the property that @xmath265_{ij}=\sum_{j}\left[\pmb{\omega}^{2}\right]_{ij}=\text{const.}\,,\ ] ] which is a consequence of the way we construct the lagrangian for inter - detector coupling .
it has an interesting implication .
suppose we try to solve the eigenvalue problem for such a system in general .
let the eigenvector @xmath185 be given by @xmath266 , we need to solve a simultaneous set of homogeneous equations @xmath267_{11}-\nu\bigr)\,a_{1}+\left[\pmb{\omega}^{2}\right]_{12}\,a_{2}+\left[\pmb{\omega}^{2}\right]_{13}\,a_{3}=0\,,\\
\left[\pmb{\omega}^{2}\right]_{21}\,a_{1}+\bigl(\left[\pmb{\omega}^{2}\right]_{22}-\nu\bigr)a_{2}+\left[\pmb{\omega}^{2}\right]_{23}\,a_{3}=0\,,\\ \left[\pmb{\omega}^{2}\right]_{31}\,a_{1}+\left[\pmb{\omega}^{2}\right]_{32}\,a_{2}+\bigl(\left[\pmb{\omega}^{2}\right]_{33}-\nu\bigr)a_{3}=0\ , .
\end{cases}\ ] ] adding these three equations together leads to @xmath268_{ij}-\nu\bigr)\bigl(a_{1}+a_{2}+a_{3}\bigr)=0\,.\ ] ] thus we have either @xmath269_{ij}-\nu=0 $ ] or @xmath270 .
this implies that i ) one of the eigenvalues must be equal to the sum of elements in one of the rows or columns of the matrix @xmath75 , and that ii ) the elements of the eigenvectors for the rest of the eigenvalues must sum to zero . from the condition
i ) , we see from the first equation of that the elements of the corresponding eigenvector must be such that @xmath271 , so we have @xmath272_{ij}\,,\qquad\qquad\mathbf{v}_{1}^{t}=\frac{1}{\sqrt{3}}\,(1,1,1)\,,\ ] ] after proper normalization .
the second condition states that the rest of the eigenvector must be normal to the plane @xmath270 .
therefore as long as the matrix @xmath75 satisfies , the matrix @xmath261 that diagonalizes @xmath75 will be the same up to re - ordering of the rows or columns .
if we define the normal modes @xmath273 by @xmath274 , then it implies one of the normal modes , say @xmath275 , must be such that @xmath276 which highlights the special role of the center - of - mass coordinate of the original variables of motion .
extension to the case of @xmath0 detectors is straightforward .
this seemingly intuitive yet rarely proven fact will play a key role in our analysis of one important facet of macroscopic quantum phenomena later . now in terms of normal mode variables
, the two - point function matrix @xmath277 , shown in for example , will be transformed into @xmath278 and it applies to both the retarded green s function @xmath279 or the hadamard function @xmath103 .
let us examine the structure of the transformed green s function .
generically , we have @xmath280 due to the fact that @xmath281 .
if the coupling constant between the oscillator and the environment is the same for all three detectors , and if the distances between any two of the detectors are the same , then we will have @xmath282 , @xmath150 all the same .
that is , if the detectors sit at the vertices of an equilateral triangle , the green function matrix , which describes the backreaction effects from the environment , will take on only two distinct values .
let @xmath283 and @xmath284 for @xmath150 .
for future reference , we write down the explicit expressions of @xmath285 and @xmath286 , @xmath287 where @xmath288 and @xmath289 , for the retarded green s functions .
furthermore , we immediately see the consequence of this particular configuration leads to @xmath290 for @xmath150 , but @xmath291 in the condensed notation , we will write them as @xmath292 , where @xmath293 , @xmath294 .
the transformed green s function matrix @xmath69 becomes diagonal , too .
it implies that we will have three decoupled equations for the normal modes @xmath273 , @xmath295 as long as the coupling between the detectors and the environment is sufficiently weak and the distance between detectors is not too short , the time - delay term , which describes the mutual indirect influence between oscillators , only introduces a small correction for the late - time dynamics @xcite .
hence the fundamental solutions to the above equations look like those to the damped oscillators , and are given by @xmath296\,,\qquad\mathfrak{d}_{2}^{(i)}(s)=e^{-\gamma_{i}s}\,\frac{1}{\omega_{i}}\,\sin\omega_{i}s\,,\ ] ] where the damping constant @xmath297 and the `` resonance '' frequency , which is ignored here by the assumption that @xmath298 .
however we should keep in mind that this assumption is not necessary . ]
@xmath299 are given by @xmath300\,,\qquad\qquad\omega_{i}=\omega_{i}\left[1-\alpha_{i}\,\frac{\gamma}{\omega_{i}^{2}d}\,\cos\omega_{i}d\right]\,.\ ] ] in terms of the ( renormalized ) oscillator frequency @xmath301 @xmath302 and the parameter @xmath303 . with the choice of the initial density matrix for @xmath26 given in
, the density matrix still takes the same form even in terms of the normal mode variables owing to the fact that the transformation matrix @xmath261 is orthogonal .
let the normal mode variables that correspond to @xmath26 be denoted by @xmath304 .
then it can be shown that the correlations between @xmath26 and their conjugate variables can be expressed in terms of the counterparts of the normal modes @xmath304 , that is , @xmath305 where @xmath306 is the canonical momentum conjugate to @xmath304 .
more importantly the correlation functions for the normal modes have the same form as those for @xmath26 in except that all the relevant matrices are replaced by their counterparts associated with the normal modes .
this correspondence greatly simplifies the calculation when the normal modes can be completely decoupled , as shown in the configuration introduced previously .
hence , let us turn our attention to the evolution of the normal mode variables .
since each normal mode @xmath307 behaves like a damped oscillators driven by the quantum fluctuations of the environment scalar field , it can be shown that the active component of each correlation function , say , @xmath308 , decays to zero as time evolves due to dissipation as a form of backreaction , but the passive component , on the other hand , grows from zero , due to the environment noise , gradually to a saturated value .
this , as we remarked earlier , is a consequence of the energy balance between dissipation and fluctuations @xcite .
this value is also independent of the initial state of the normal mode .
owing to the fact that the motion of the normal modes are totally decoupled , the matrices @xmath309 and @xmath310 are diagonalized .
thus at late - time @xmath114 , the only nonzero terms left are the passive components like @xmath311 and @xmath312 . as seen from , they are @xmath313 where @xmath314 is the fourier transformation of the fundamental solution to each normal mode , that is @xmath315 and @xmath316^{-1}=\left[-\omega^{2}+\omega_{i}^{2}-i\,2\gamma\,\omega-2\gamma\ , \frac{\alpha_{i}}{d}\,e^{i\,\omega d}\right]^{-1}\,,\ ] ] and we have used @xmath317 the divergence in @xmath318 can be absorbed into @xmath301 . here
we have assumed that we have a zero - temperature bath , so that @xmath319 and @xmath320 .
carrying out the integrals in and gives @xmath321\notag\\ & = \frac{1}{2m\omega_{i}}\left\{1-\frac{2}{\pi}\frac{\gamma_{i}}{\omega_{i}}+\frac{\gamma\,\alpha_{i}}{\omega_{i}^{2}d}\biggl[\cos\omega_{i}d-\frac{2}{\pi}\,\sin\omega_{i}d\biggr]+\cdots\right\}\,,\\ \langle\,\pmb{\pi}^{(i)\,2}_{b}\rangle&=\frac{m\omega_{i}}{2}\left[1+\frac{2}{\pi}\frac{\gamma_{i}}{\omega_{i}}\left(\ln\frac{\pi^{2}}{\omega_{i}^{2}}-1\right)+\cdots\right]\notag\\ & = \frac{m\omega_{i}}{2}\left\{1+\frac{2}{\pi}\frac{\gamma_{i}}{\omega_{i}}\,\beta_{i}-\frac{\gamma\,\alpha_{i}}{\omega_{i}^{2}d}\biggl[\cos\omega_{i}d-\frac{2}{\pi}\,\beta_{i}\,\sin\omega_{i}d\biggr]+\cdots\right\}\,,\end{aligned}\ ] ] with @xmath322 where @xmath323 is the cutoff frequency associate with the scalar field , and assumed to take the same value for all oscillators .
note that the correction due to interaction with the environment is linear in @xmath324 .
the damping constant @xmath297 and the resonance frequency @xmath299 are given by @xmath325\,,\qquad\qquad\omega_{i}=\omega_{i}\left[1-\alpha_{i}\,\frac{\gamma}{\omega_{i}^{2}d}\,\cos\omega_{i}d\right]\,.\ ] ] with the parameter @xmath303 .
in the weak coupling limit @xmath326 , the resonance frequency @xmath299 also differs from the normalized oscillating frequency @xmath301 by the order of @xmath327 .
thus to leading order we have @xmath328 \dfrac{1}{2m\omega_{i}}\,,&t\to\infty\ , .
\end{cases } \intertext{and } \langle\,\pmb{\pi}^{(i)\,2}_{b}\rangle&=\begin{cases } \dfrac{1}{2\varsigma^{2}}\,,&t\to0\,,\\[12pt ] \dfrac{m\omega_{i}}{2}\,,&t\to\infty\ , .
\end{cases}\end{aligned}\ ] ] next we can transform the results for the normal modes back into the counterparts for the original variables @xmath67 by @xmath329 explicitly , we have @xmath330 given by @xmath331\,,\label{e : oeowjx}\\ \langle\,\pmb{\chi}_{b}^{(2)\,2}\,\rangle=\langle\,\pmb{\chi}_{b}^{(3)\,2}\,\rangle&=\frac{1}{6}\bigl[2\langle\,\pmb{\zeta}_{b}^{(1)\,2}\,\rangle+\langle\,\pmb{\zeta}_{b}^{(2)\,2}\,\rangle+3\langle\,\pmb{\zeta}_{b}^{(3)\,2}\,\rangle\bigr]\,,\\ \langle\,\pmb{\chi}_{b}^{(1)}\pmb{\chi}_{b}^{(2)}\,\rangle=\langle\,\pmb{\chi}_{b}^{(1)}\pmb{\chi}_{b}^{(3)}\,\rangle&=\frac{1}{3}\bigl[\langle\,\pmb{\zeta}_{b}^{(1)\,2}\,\rangle-\langle\,\pmb{\zeta}_{b}^{(2)\,2}\,\rangle\bigr]\,,\label{e : krnktiw}\\ \langle\,\pmb{\chi}_{b}^{(2)}\pmb{\chi}_{b}^{(3)}\,\rangle&=\frac{1}{6}\bigl[2\langle\,\pmb{\zeta}_{b}^{(1)\,2}\,\rangle+\langle\,\pmb{\zeta}_{b}^{(2)\,2}\,\rangle-3\langle\,\pmb{\zeta}_{b}^{(3)\,2}\,\rangle\bigr]\,,\label{e : neowjx}\end{aligned}\ ] ] and similar structures for @xmath332 .
also note that in these cases we have @xmath333 , and thus @xmath334 now it is instructive to make comparison with the case that there is no inter - detector coupling @xcite . in the latter case
, the late - time correlation between the variables @xmath335 solely results from their interaction with the environment , thus it is typically of the order @xmath324 .
in addition , as a consequence of the intervention of the background field , this correlation depends on the spatial separation between the detectors . thus the detectors tend to have the pairwise correlation due to the facts that they have stronger correlation if they get closer to one another , and that additional correlation mediated by the third parties will be at least of the order @xmath336 . on the other hand , in the presence of inter - detector coupling ,
the later - time correlation comes from the superposition of their counterparts from the normal modes ; it is of the same order as the uncertainty of the corresponding variables .
thus the correlation due to inter - detector coupling tends to be much stronger than the induced correlation by the shared bath , and it does not depend on the distance between two detectors .
these are the fundamental differences between the two cases studied in the present two papers .
in the subsequent discussion , since the contributions from the environment are much smaller than those from inter - detector coupling , we will only keep terms which are zeroth order in @xmath324 .
the covariance matrix at late time is then explicitly given by @xmath337 with @xmath338 given by and
@xmath339 some information of entanglement for this tripartite system can be revealed in the symplectic eigenvalues of a covariance matrix which corresponds to the partially transposed density matrix of this system . for a system which involves more than two parties , there is more than one way to take the partial transpose of the density matrix .
denote detector 1 as @xmath246 , and the remaining two as @xmath247 and @xmath248 .
we see two distinct cases , one where swapping @xmath247 and @xmath248 leads to identical results , because q is coupled with equal strength @xmath249 to @xmath247 and @xmath248 .
this is the symmetric case we called case [ sym ] before . in the other case we called case [ asm ] before the entanglement between @xmath247 and @xmath246 is different from that between @xmath247 and @xmath248 because the coupling strengths of these pairs are different , being @xmath249 in the former and @xmath252 in the latter .
thus there are two ways to perform the partial transpose .
we can either do it to @xmath246 , which will be called the case @xmath246 versus @xmath340 .
or we partially transpose variables associated with @xmath247 , which will be called @xmath247 versus @xmath251 .
the consequence of taking partial transpose in the density matrix is equivalent to changing the sign of the conjugate momentum of the corresponding subsystem .
take the case @xmath246 versus @xmath340 for example , the covariance matrix associated with the partial transpose of the subsystem @xmath246 becomes @xmath341 and likewise for the case @xmath247 versus @xmath251 , it is @xmath342 the symplectic eigenvalue @xmath343 of the covariance matrix can be found by solving the eigenvalue problem of the form @xmath344 where the matrix @xmath345 is the fundamental symplectic matrix @xmath346 and in our case @xmath1 .
the vector @xmath185 is the corresponding eigenvector .
the eigenvalues always appear in pairs and this is most transparent to see in the williamson normal form of the covariance matrix @xmath347 which is the diagonal form of a symmetric matrix under suitable symplectic transformations .
the eigenvalues in our case are then given by the roots to the polynomial @xmath348 since it is a third - degree polynomial in @xmath349 , its roots can always be found exactly . for the @xmath246 versus @xmath340 or what we called the symmetric [ sym ] case
, the symplectic eigenvalues are @xmath350 they do not depend on @xmath351 , thus independent of the coupling constant @xmath252 between @xmath247 and @xmath248 . among these three eigenvalues ,
we observe that @xmath352 is always greater than @xmath193 , but @xmath353 is always smaller than @xmath193 , so it signals the presence of entanglement between @xmath246 and the pair @xmath340 according to .
hence we only focus on the symplectic eigenvalue @xmath353 . in terms of the coupling constant @xmath249 between @xmath246 and @xmath247 or @xmath246 and @xmath248
, we can show @xmath354 \dfrac{3^{3/4}}{4\sqrt{2}}\,\left(\dfrac{\omega_{0}^{2}}{\lambda}\right)^{1/4}+\cdots\ , , & \lambda\gg\omega_{0}^{2}\ , , \end{cases}\ ] ] so it ranges from @xmath193 to 0 .
the fact that @xmath353 does not depend on the interaction between @xmath247 and @xmath248 may not be that surprising because the interaction between the @xmath250 pair is of the same strength as the @xmath251 pair , whatever happens between @xmath247 and @xmath248 will be `` equally '' distributed to @xmath246 over two channels through the @xmath250 , @xmath251 couplings .
this observation is supported by the fact that the correlation functions between the canonical variables associated with @xmath246 and @xmath247 , say @xmath355 , taking on the same value as the corresponding correlation between @xmath246 and @xmath248 , as seen in .
the value does not depend on the coupling constant between @xmath247 and @xmath248 .
this balance or symmetry is built - in in the parity between @xmath247 and @xmath248 since they start from the same initial configurations , have the same initial oscillating frequency , and are coupled to @xmath246 with equal strength .
it can be easily disrupted if their motion is out of phase by changing any of the above - mentioned factors or by skewing their relative positions , such that the dependence on @xmath252 will re - appear .
when we put back the influence of the environment , the conclusion that @xmath353 does not depend on @xmath252 still holds because both @xmath247 and @xmath248 experience the same self - force , and the same non - markovian effects mediated by the environment . in comparison ,
the correlations , say @xmath356 , between @xmath247 and @xmath248 , do depend on the coupling constant @xmath252 between @xmath247 and @xmath248 . following this line of reasoning
, we see that when the coupling between @xmath246 and @xmath247 ( or @xmath248 ) is vanishing small , the entanglement of @xmath246 with @xmath340 gradually disappears .
this is consistent with our intuition because the correlation between @xmath246 and @xmath247 ( or @xmath248 ) vanishes , as seen in . in this limit
, the only connection between @xmath246 and @xmath247 ( or @xmath248 ) comes from their coupling to the environment which is weak .
if we even ignore that part , then @xmath246 is essentially isolated from @xmath247 and @xmath248 .
thus if @xmath246 gets disentangled with @xmath340 , it will remain disentangled throughout its evolution . on the other hand ,
if @xmath246 interacts strongly with @xmath247 , their momenta tend to be strongly anti - correlated while their positions remain positively correlated . at the same time the symplectic eigenvalue @xmath353 deviates further away from @xmath193 from below , signaling stronger entanglement between @xmath246 and @xmath340 in the sense of negativity .
in contrast to the case @xmath246 versus @xmath340 , we now turn to the case @xmath247 versus @xmath251 .
the corresponding symplectic eigenvalues share several identical features as in the case @xmath246 versus @xmath340 . among the three distinct symplectic eigenvalues ,
one of them is always greater than @xmath193 , the second is equal to @xmath193 , and the third one , denoted by @xmath353 , is always smaller than @xmath193 , regardless of the values of @xmath256 , @xmath249 and @xmath252 , @xmath357^{2}}\biggr\}\end{aligned}\ ] ] nonetheless , in this case @xmath353 does depend on the coupling constant @xmath252 between @xmath247 and @xmath248 . generally speaking , for fixed values of @xmath256 and @xmath249 , the symplectic eigenvalue @xmath353 monotonically decreases with larger values of @xmath252 .
therefore , it implies that when the coupling between @xmath247 and @xmath248 is stronger than the interaction between @xmath246 and @xmath247 , the entanglement for @xmath247 versus @xmath251 is stronger than the counterpart for @xmath246 with @xmath340 , and vice versa .
thus the role inter - detector coupling plays on determining the entanglement structure is more transparent from this comparison . on the other hand , for fixed values of @xmath256 and @xmath252 , the symplectic eigenvalue @xmath353 is not always a monotonically decreasing function of @xmath249 . for small @xmath249 ,
the value of @xmath353 can increase with @xmath249 to a maxima and then monotonically decreases .
it is particularly significant for larger values of @xmath252 . to better understand the behavior of @xmath353 ,
let us take a look at some limiting cases : first , suppose @xmath249 is vanishingly small and @xmath252 is fixed in value , and then the only channel for possible entanglement of @xmath247 with @xmath251 comes from the interaction between @xmath247 and @xmath248 .
@xmath246 is out of reach of @xmath247 . in this case
, we can ( wait for the system reaches a steady state and ) examine the cross - correlation between them .
we see @xmath358 but @xmath359 due to the fact that @xmath360 in this limit .
then we let @xmath249 increase from zero , there comes an extra channel between @xmath247 and @xmath251 owing to the interaction between @xmath247 and @xmath246 . in addition
, @xmath248 can also be related to @xmath247 via @xmath246 , thus improving the connection between @xmath247 and @xmath251 . for @xmath361
, we have @xmath362 if @xmath252 is still greater than @xmath249 , so it implies that the cross - correlations then become : @xmath363 while @xmath364>0\,,\\ \langle\,\mathbf{p}_{b}^{(a)}\mathbf{p}_{b}^{(b)}\,\rangle&=\frac{m}{12}\left[2\left(\omega_{1}-\omega_{3}\right)+\left(\omega_{2}-\omega_{3}\right)\right]<0\,.\end{aligned}\ ] ] as @xmath249 increases beyond @xmath252 , even we have @xmath365 , the cross - correlation between @xmath247 and @xmath248 does not change qualitatively .
however , we find both @xmath366 and @xmath367 reach their extremal values at @xmath368 .
that is , @xmath366 decreases from some positive value as @xmath249 moves away from zero , until it reaches its minimum at @xmath368 . after that point
the correlation between @xmath369 and @xmath370 monotonically increases with @xmath249 , approaching the value @xmath371 .
the minimum is @xmath372 on the other hand , the correlation @xmath367 increases from some negative values with @xmath249 until it reaches its maximum value @xmath373 at @xmath368 . beyond that point
, it decreases monotonically without bound .
this partially explains the non - monotonic behavior of @xmath353 for a fixed @xmath252 because some other factors which enter in determining the covariance matrix , such as @xmath374 , do not depend on @xmath252 . for the case with a fixed @xmath249 , if we let @xmath375 , that is , no direct coupling between @xmath247 and @xmath248 , the only connection between @xmath247 and @xmath251 results from interaction between @xmath247 and @xmath246 .
in addition , in this case @xmath248 is not entirely out of contact with @xmath247 .
their correlation can be mediated by @xmath246 .
now we examine their cross - correlation . for @xmath375
, we have @xmath365 , and @xmath376 these are in contrast with and . as for the cross - correlation between @xmath247 and @xmath248
, we have @xmath377>0\,,\\ \langle\,\mathbf{p}_{b}^{(a)}\mathbf{p}_{b}^{(b)}\,\rangle&=\frac{m}{12}\left[2\left(\omega_{1}-\omega_{3}\right)+\left(\omega_{2}-\omega_{3}\right)\right]<0\,.\end{aligned}\ ] ] these qualitative features do not change with increasing @xmath252 . however , there is a major distinction in this case .
there is no extremum for the cross - correlation between @xmath247 and @xmath248 when we vary @xmath252 .
this is consistent with the behavior of @xmath353 . since we have known that the entanglement @xmath246 with @xmath340 does not depend on @xmath252 , it is equivalent to the special case @xmath378
. therefore we can draw the conclusion that when the coupling @xmath252 is greater than @xmath249 , the entanglement for the partition @xmath247 versus @xmath251 is stronger than the counterpart for @xmath246 versus @xmath340 . on the other hand , when the coupling @xmath252 is weaker than @xmath249 , the entanglement for the partition @xmath246 versus @xmath340 is stronger than the counterpart for @xmath247 versus @xmath251 ( see fig .
[ fi : qab ] ) .
here we outline some general features of cross - correlations .
it has been shown that with stronger inter - oscillator couplings , the cross - correlation @xmath379 tends to be more strongly correlated with the exception that @xmath380 is not sensitive to the coupling strength @xmath252 between @xmath247 and @xmath248 .
on the other hand , @xmath381 tends to be more strongly anti - correlated with stronger inter - oscillator couplings .
we also observe that since @xmath379 is bounded , they do not change appreciably with @xmath249 and @xmath252 .
contrary to @xmath379 , @xmath381 is not bounded below , it can vary more dramatically with the coupling constants @xmath249 and @xmath252 .
in particular , @xmath382 varies more significantly with @xmath252 but less with @xmath249 , while @xmath383 changes more significantly with @xmath249 but is independent of @xmath252 . in conclusion
, we see from this detailed analysis how the entanglement structure of coupled quantum oscillators depends on the coupling strength , from a knowledge of the cross - correlations between the oscillators .
this quantitative description is valuable in aiding our understanding of entanglement behavior because it is the simplest continuous variable many - body system which is amenable to such a detailed analysis . for larger systems with more components and less symmetry
this rapidly becomes impossible .
we shall explore the implications of our results for some key issues in macroscopic quantum phenomena in the next section .
the key findings are given in the form of an executive summary below ( as a retribution for the length of prior sections ) : * please extract from the text for the main results on the influence functional , the reduced density matrix evolutionary operator , the langevin equations , the green s functions , the correlation functions and the covariance matrix . * comparison with the case studied in @xcite where there is no direct coupling amongst the @xmath0 oscillators but only field - induced coupling : with direct inter - oscillator coupling studied here , the effect of the system - environment coupling becomes of secondary importance . *
as a measure of entanglement for the @xmath0 oscillator system with direct coupling studied here , it is decided by whether any of the symplectic eigenvalues of the partially transposed covariance matrix is smaller than 1/2 .
* entanglement is enhanced with stronger coupling between any two oscillators , which confirms intuitive reasoning . *
two cases are studied in details where three oscillators are placed at the vertices of an equilateral triangle .
main features : + case [ sym ] q vs ab , where q is coupled to a and b with equal strength : entanglement of q with ab as a group is independent of the coupling between a and b. it is as if q is entangled with a ` center - of - mass ' variable of the two oscillators ab as a group .
+ case [ asy ] a vs qb , where the coupling strength between a and q is different from between a and b. entanglement for a vs qb does depends on qa coupling and ab coupling .
a. for a fixed qa coupling , the entanglement measure is a monotonically decreasing function of ab coupling ; its values ranges from 1/2 to 0 .
( measure decreases means entanglement increases . )
b. for a fixed ab coupling , the entanglement measure does not always monotonically decrease with increasing qa coupling .
it increases first to reach the maximum and then monotonically decreases .
this is because part of it results from the behavior of cross - correlations . *
comparison between the two cases : a. for qa and qb coupling greater than ab coupling , entanglement for q vs ab is stronger than entanglement a vs qb .
b. for ab coupling greater than qa coupling , entanglement for a vs qb is stronger than that for q vs ab as mentioned in the introduction , one emergent area of research where results from this analysis can be applied to is macroscopic quantum phenomena .
we identified two key issues to understand , one is the role of the com in the quantum behavior of a many - body system ; the second is how the level - of - structure enter in the manifestation of macroscopic quantum phenomena , can be examined using the entanglement structure results obtained here for a three oscillator system with disparate coupling strengths .
call one of the oscillators q which is distinguished from the other two a and b by different couplings .
consider the two cases : case [ sym ] when q is equally coupled to a and b ( symmetric configuration ) versus case [ asy ] when a is coupled to q differently from its coupling to b ( asymmetric configuration ) . in the sym case we show that as far as quantum entanglement is concerned q is effectively coupled to the com of a and b. this shows that as far as quantum behavior of this nature is concerned the com variable functions similar to its role in classical mechanics . now view q as the center of mass variable representing a macroscopic object 1 , and a , b as representative constituents of a similar macroscopic object 2 .
one interesting result we found here of relevance to the level - of - structure aspect of mqp is that ( at least for quadratic inter - particle coupling ) the entanglement between q and ab is equivalent to that between q and the com of 2 , _ independent of the coupling between a and b _ provided a and b are interchangeably identically .
this is of interest because it explains why as one reports on the quantum dynamics of say a cantilever ( e.g. , @xcite ) one does not need to also provide information of how say , the nucleons in the metal bar , are coupled .
comparison of the results in this case [ sym ] with the other case [ asy ] is also informative about why levels of structure , or a hierarchical ordering , of matter is a natural way of organization for a better understanding of the manifestation of quantum behavior of macroscopic objects .
the open system of @xmath0 coupled quantum oscillator system is a generic model for investigating continuous - variables many - body systems under different environmental influence . continuing the theme of mqp
, one can also examine the size and the ` correlation ' aspects in addition to the entanglement aspects explored here for interacting quantum systems .
( these three aspects of mqp are explained in @xcite see references therein . )
we mention two additional directions currently under study .
it can easily be generalized to a finite temperature field whereby one can investigate issues in quantum thermodynamics such as those related to thermal entanglement in @xcite .
one can also partition the @xmath0 oscillators into two parts : one consisting of @xmath7 oscillators representing the system and the second part consisting of @xmath384 oscillators as the bath . by varying @xmath7 versus @xmath385 one can use this model to investigate mesoscopic phenomena @xcite addressing new and different facets such as those mentioned above for mqp .
.5 cm * acknowledgment * we are obliged to prof . s .-
lin for helpful discussions on the covariance matrix structure , prof .
chou , c. fleming , and y. subai for discussions on multi - partite entanglements in relation to macroscopic quantum phenomena .
blh thanks professors y. s. wu and j. q. you for their kind hospitality at fudan university in the spring of 2013 when this work was finished .
in this appendix , we would like to illustrate two interesting observations in the course of investigating entanglement of the bi - partition system .
first , we highlight the role the com coordinate in such a system .
we then explicitly point out the connection between an orthogonal matrix in the configuration space and the corresponding symplectic matrix in the phase space . in general ,
the three - oscillator system , discussed in the previous section , is not a fully symmetric gaussian system due to disparate inter - oscillator coupling .
however , when we make a particular partition such that it is a bipartite @xmath246 versus @xmath340 system , then we have a @xmath386 bi - symmetric gaussian system .
we may apply some well - known facts @xcite of the @xmath233 gaussian system to this case . in the case of disparate coupling we considered ,
if we ignore the influence of the background field , then the lagrangian corresponding to for @xmath1 can be expressed in terms of the center - of - mass coordinate and the relative coordinate .
following , we write the lagrangian as @xmath387 from this construction we see that the degree of freedom @xmath388 is decoupled from the remaining degrees of freedom . in addition , only the center - of - mass coordinate @xmath389 for the subsystem 2 and 3 couple with the coordinate @xmath390 of the system 1 .
thus during the course of evolution , the degree of freedom @xmath388 is always separate from the others , and may not be involved in creation of entanglement of the whole system .
for example , in the case of @xmath246 versus @xmath340 , the entanglement is obtained by partial transpose the density matrix associated with the subsystem 1 , that is @xmath246 , so the degree of freedom @xmath388 remains uninvolved .
hence the entanglement for @xmath246 versus @xmath340 is completely determined and quantified by the degrees of freedom @xmath391 and @xmath389 . from
, we can see that the entanglement measure should not depend on the coupling constant @xmath252 between subsystem 2 and 3 because it does not appear in the first line of eq . . on the other hand , when we make a partition like @xmath247 versus @xmath251 , the coordinate transformation of @xmath391 , @xmath392 and @xmath393 into @xmath391 , @xmath389 and @xmath388 does not have any advantageous convenience because it does not conform to any symmetry this particular partition renders .
in addition , when we make partial transpose on either subsystem 2 ( that is @xmath247 ) , or subsystem 3 ( that is @xmath248 ) , both degrees of freedom @xmath389 , @xmath388 will be involved .
thus in this case , the entanglement measure for @xmath247 versus @xmath251 will inevitably depend on @xmath252 .
we would like to generalize the system introduced in to a @xmath394 fully symmetric system .
let the corresponding lagrangian be given by @xmath395 it describe a specially partitioned @xmath0-component system which consists of two groups , subsystem 1 and the rest of the components .
the subsystem 1 has the same coupling strength @xmath249 with the the rest of the subsystems , meanwhile the remaining subsystems interact with one another with the same coupling constant @xmath252 .
apparently this is a fully symmetric @xmath394 system because it does not change when we permute @xmath396 for @xmath397 .
it is a generalization of @xmath246 versus @xmath340 configuration to @xmath246 versus @xmath398 .
recall that we can always find the normal modes for a multi - component system described by the lagrangian .
we first rewrite the lagrangian as @xmath399\chi^{(1)\,2}\bigr\}+\frac{m}{2}\,\lambda\,\chi^{(1)}\sum_{i=2}^{n}\chi^{(i)}\notag\\ & \qquad\qquad\qquad+\sum_{i=2}^{n}\bigl\{\frac{m}{2}\,\dot{\chi}^{(i)\,2}-\frac{m}{2}\bigl[\omega_{0}^{2}+\lambda+(n-2)\sigma\bigr]\chi^{(i)\,2}\bigr\}+\sum_{j\neq i=2}^{n}\frac{m}{2}\,\sigma\,\chi^{(i)}\chi^{(j)}\notag\\ & = \bigl\{\frac{m}{2}\,\dot{\chi}^{(1)\,2}-\frac{m}{2}\bigl[\omega_{0}^{2}+(n-1)\lambda\bigr]\chi^{(1)\,2}\bigr\}\notag\\ & \qquad\qquad\qquad+\frac{m}{2}\,\lambda\,\chi^{(1)}\sum_{i=2}^{n}\chi^{(i)}+\frac{m}{2}\mathbf{y}^{t}\cdot\mathbf{y}-\frac{m}{2}\,\mathbf{y}^{t}\cdot\pmb{\omega}^{2}\cdot\mathbf{y}\,,\label{e : yuwkwz}\end{aligned}\ ] ] where @xmath400 , and @xmath401 this @xmath402 interaction matrix has an interesting feature that when we sum elements along either one of the rows or columns , we obtain the same result .
this feature , as discussed before , implies that one of the normal modes must be of the form @xmath403 which apparently is the center of mass coordinate for the original variables @xmath392 , ... , @xmath404 .
if we let @xmath405 , for @xmath406 be the normal mode coordinates , then the lagrangian becomes @xmath399\chi^{(1)\,2}\bigr\}+\frac{m}{2}\,\sqrt{n-1}\lambda\,\chi^{(1)}\xi^{(2)}\notag\\ & \qquad\qquad\qquad+\frac{m}{2}\,\dot{\xi}^{(2)\,2}-\frac{m}{2}\,\omega_{2}^{2}\,\xi^{(2)\,2}+\sum_{j=3}^{n}\bigl\{\frac{m}{2}\,\dot{\xi}^{(j)\,2}-\frac{m}{2}\,\omega_{j}^{2}\,\xi^{(j)\,2}\bigr\}\,,\end{aligned}\ ] ] where @xmath407 is the eigenfrequency associated with the normal mode @xmath405 , or in other words @xmath4080 & \omega_{3}^{2 } & 0 & \cdots & 0\\[4pt ] 0 & 0 & \ddots & & \vdots\\[4pt ] \vdots & \vdots & & \ddots & \vdots\\[4pt ] 0 & 0 & \cdots & 0 & \omega_{n}^{2}\end{pmatrix}\,.\ ] ] for the later reference , @xmath409 is given by @xmath410_{ij}=\omega_{0}^{2}+\lambda$ ] and we have @xmath411 degenerate eigenfrequencies @xmath412 .
therefore following the previous arguments , if we compute entanglement for the partition @xmath246 versus @xmath398 , then only these two degrees of freedom are involved in creation of entanglement and the corresponding entanglement measure will not depend on @xmath252 .
in fact , the remaining @xmath411 normal modes @xmath413 , for @xmath414 , remain decoupled from one another and evolve with time by themselves .
essentially only @xmath391 and @xmath415 interact together and evolve into an entangled state .
furthermore their evolution does not depend on the details of evolution of the degrees of freedom @xmath413 .
that is , @xmath413 will not contribute to entanglement between @xmath391 and @xmath415 .
the covariance matrix of the whole system in terms of @xmath391 and the normal modes @xmath405 , @xmath416 , will look like @xmath417 \pmb{\gamma}^{t } & \pmb{\beta}_{2 } & \mathbf{0 } & \cdots & \mathbf{0}\\[4pt ] \mathbf{0 } & \mathbf{0 } & \pmb{\beta}_{3 } & & \mathbf{0}\\[4pt ] \vdots & \vdots & & \ddots & \vdots\\[4pt ] \mathbf{0 } & \mathbf{0 } & \mathbf{0 } & \cdots & \pmb{\beta}_{n } \end{pmatrix}\ ] ] not that this is not the williamson form of the covariance matrix .
the @xmath178 matrix @xmath418 account for cross - correlation between @xmath391 and @xmath415 .
the covariance matrix happens to take the `` localized '' form as demonstrated in @xcite , that is , the center - of - mass coordinate @xmath415 is the representative mode of the @xmath419 fully symmetric subsystems , responsible for setting up correlation with the other party , represented by @xmath391 .
then how does this observation go along with the procedures of diagonalizing the covariance matrix for a @xmath233 bi - symmetric system , discussed in @xcite ?
we again use the example of the @xmath386 system with disparate coupling for illustration .
we will closely follow the procedures outlined in @xcite . generically speaking
the covariance matrix for our tripartite system takes the form @xmath4200 & e & 0 & f & 0 & f\\[4pt]b & 0 & c & 0 & d & 0\\[4pt]0 & f & 0 & g & 0 & h\\[4pt]b & 0 & d & 0 & c & 0\\[4pt]0 & f & 0 & h & 0 & g\end{pmatrix}\,,\ ] ] where @xmath421 as shown in and .
next we would like to write the covariance matrix by a series symplectic transformations into the standard form @xmath422 where @xmath423 , @xmath424 , @xmath425 , @xmath418 are @xmath178 block matrices , and take the form @xmath4260 & \alpha\end{pmatrix}\,,&\pmb{\beta}&=\begin{pmatrix}\beta & 0\\[4pt]0 & \beta\end{pmatrix}\ , , & \pmb{\epsilon}=\begin{pmatrix}\epsilon_{1 } & 0\\[4pt]0 & \epsilon_{2}\end{pmatrix}\,.\end{aligned}\ ] ] since for our simple example , is already closed to standard form , what we will do is to apply a symplectic transformation @xmath427 , which is equivalent to rescaling or change of the units , upon @xmath428 such that the @xmath178 block matrices along the diagonal of @xmath428 are proportional the identity matrix . for lower dimensional
matrix , this transformation matrix can be found without much effort .
let us take the upper left block of @xmath429 for example .
let the @xmath178 matrix @xmath98 denote @xmath430 and then the corresponding @xmath178 transformation matrix @xmath163 @xmath431 can be obtained by requiring that @xmath432 and that @xmath433 should be proportional to a @xmath178 identity matrix .
we then find @xmath434{\dfrac{e}{a } } & 0\\[10pt]0 & \sqrt[4]{\dfrac{a}{e}}\end{pmatrix}\,.\ ] ] we can do the same thing to the remain @xmath178 block matrices along the diagonal .
thus the symplectic matrix @xmath427 , which we use to diagonalize @xmath429 , will take the form @xmath435{\dfrac{e}{a } } & 0 & 0 & 0 & 0 & 0\\[10pt ] 0 & \sqrt[4]{\dfrac{a}{e } } & 0 & 0 & 0 & 0\\[10pt ] 0 & 0 & \sqrt[4]{\dfrac{g}{c } } & 0 & 0 & 0\\[10pt ] 0 & 0 & 0 & \sqrt[4]{\dfrac{c}{g } } & 0 & 0\\[10pt ] 0 & 0 & 0 & 0 & \sqrt[4]{\dfrac{g}{c } } & 0\\[10pt ] 0 & 0 & 0 & 0 & 0 & \sqrt[4]{\dfrac{c}{g } } \end{pmatrix}\,,\ ] ] and then @xmath436{\dfrac{eg}{ac } } & 0 & b\sqrt[4]{\dfrac{eg}{ac } } & 0\\[10pt ] 0 & \sqrt{ae } & 0 & f\sqrt[4]{\dfrac{ac}{eg } } & 0 & f\sqrt[4]{\dfrac{ac}{eg}}\\[10pt ] b\sqrt[4]{\dfrac{eg}{ac } } & 0 & \sqrt{cg } & 0 & \sqrt{\dfrac{g}{c}}\,d & 0\\[10pt ] 0 & f\sqrt[4]{\dfrac{ac}{eg } } & 0 & \sqrt{cg } & 0 & \sqrt{\dfrac{c}{g}}\,h\\[10pt ] b\sqrt[4]{\dfrac{eg}{ac } } & 0 & \sqrt{\dfrac{g}{c}}\,d & 0 & \sqrt{cg } & 0\\[10pt ] 0 & f\sqrt[4]{\dfrac{ac}{eg } } & 0 & \sqrt{\dfrac{c}{g}}\,h & 0 & \sqrt{cg}\end{pmatrix}\,.\ ] ] since the upper left @xmath178 block matrix of @xmath437 is already diagonalized and is in its williamson form , next we will diagonalize the lower right @xmath438 block matrix @xmath4390 & \sqrt{cg } & 0 & \sqrt{\dfrac{c}{g}}\,h\\[10pt]\sqrt{\dfrac{g}{c}}\,d & 0 & \sqrt{cg } & 0\\[10pt]0 & \sqrt{\dfrac{c}{g}}\,h & 0 & \sqrt{cg}\end{pmatrix}\ ] ] into the williamson normal form .
we may follow the procedures outlined in @xcite to find the williamson form ; however it becomes a little impractical in reality due to the ambiguity that we can always introduce an additional orthogonal matrix when we construct a symplectic matrix that transforms a given symmetric matrix into its williamson form .
instead , we make an observation that a symplectic transformation in the phase space has its corresponding rotation transformation in the configuration space .
in addition , in quantum mechanics the canonical momentum operator is defined in such a way that it transforms in the same manners as the position operator does .
therefore , to avoid cluttering of the square root notations , we write the matrix @xmath440 as @xmath4410 & a_{1 } & 0 & c_{1}\\[4pt]b_{1 } & 0 & a_{1 } & 0\\[4pt]0 & c_{1 } & 0 & a_{1}\end{pmatrix}\ ] ] and suppose that the position column vector @xmath442 is rotated by an orthogonal matrix @xmath443-\sin\theta & \cos\theta\end{pmatrix}\,,\ ] ] so that the corresponding symplectic transformation matrix @xmath4440 & \cos\theta & 0 & \sin\theta\\[4pt]-\sin\theta & 0 & \cos\theta & 0\\[4pt]0 & -\sin\theta & 0 & \cos\theta\end{pmatrix}\ ] ] makes the matrix @xmath440 into the diagonal form . of course
this does not guarantee that the resulting diagonal matrix is in the williamson form .
the requirement of @xmath445 being a diagonal matrix gives @xmath446 , and the resulting matrix is @xmath4470 & a_{1}+c_{1 } & 0 & 0\\[4pt]0 & 0 & a_{1}-b_{1 } & 0\\[4pt]0 & 0 & 0 & a_{1}-c_{1}\end{pmatrix}\,,\ ] ] if we choose @xmath448 . to make this into the williamson form
, we again apply another transformation @xmath449 to rescale the diagonal elements , @xmath450{\dfrac{a_{1}+c_{1}}{a_{1}+b_{1 } } } & 0 & 0 & 0\\[8pt]0 & \sqrt[4]{\dfrac{a_{1}+b_{1}}{a_{1}+c_{1 } } } & 0 & 0\\[8pt]0 & 0 & \sqrt[4]{\dfrac{a_{1}-c_{1}}{a_{1}-b_{1 } } } & 0\\[8pt]0 & 0 & 0 & \sqrt[4]{\dfrac{a_{1}-b_{1}}{a_{1}-c_{1}}}\end{pmatrix}\ ] ] so that @xmath4510 & \eta_{3 } & 0 & 0\\[8pt]0 & 0 & \eta_{2 } & 0\\[8pt]0 & 0 & 0 & \eta_{2}\end{pmatrix}\,.\ ] ] where @xmath452 and @xmath453 .
now we have arrived at the williamson form of the matrix @xmath440 , .
the diagonal elements are its symplectic eigenvalues , and can be verified by solving the secular equation of the corresponding symplectic eigenvalue problem @xmath454 now putting these results together , we have the symplectic matrix @xmath455 @xmath456{\dfrac{a_{1}+c_{1}}{a_{1}+b_{1 } } } & 0 & \dfrac{1}{\sqrt{2}}\sqrt[4]{\dfrac{a_{1}+c_{1}}{a_{1}+b_{1 } } } & 0\\[8pt]0 & 0 & 0 & \dfrac{1}{\sqrt{2}}\sqrt[4]{\dfrac{a_{1}+b_{1}}{a_{1}+c_{1 } } } & 0 & \dfrac{1}{\sqrt{2}}\sqrt[4]{\dfrac{a_{1}+b_{1}}{a_{1}+c_{1}}}\\[8pt ] 0 & 0 & -\dfrac{1}{\sqrt{2}}\sqrt[4]{\dfrac{a_{1}-c_{1}}{a_{1}-b_{1 } } } & 0 & \dfrac{1}{\sqrt{2}}\sqrt[4]{\dfrac{a_{1}-c_{1}}{a_{1}-b_{1 } } } & 0\\[8pt]0 & 0 & 0 & -\dfrac{1}{\sqrt{2}}\sqrt[4]{\dfrac{a_{1}-b_{1}}{a_{1}-c_{1 } } } & 0 & \dfrac{1}{\sqrt{2}}\sqrt[4]{\dfrac{a_{1}-b_{1}}{a_{1}-c_{1}}}\end{pmatrix}\,,\ ] ] with @xmath457 it will transform the lower right @xmath438 block matrix of @xmath437 into its williamson form @xmath4580 & \eta_{\circ } & 0 & \zeta & 0 & 0\\[8pt]\xi & 0 & \eta_{- } & 0 & 0 & 0\\[8pt]0 & \zeta & 0 & \eta_{- } & 0 & 0\\[8pt]0 & 0 & 0 & 0 & \eta_{+ } & 0\\[8pt]0 & 0 & 0 & 0 & 0 & \eta_{+}\end{pmatrix}\,.\ ] ] with @xmath459{\dfrac{e(g+h)}{a(c+d)}}\ , , & \zeta&=\sqrt{2}f\,\sqrt[4]{\dfrac{a(c+d)}{e(g+h)}}\,.\end{aligned}\ ] ] note that @xmath460 , @xmath461 shown above are not the symplectic eigenvalues of the @xmath462 matrix @xmath429 .
instead @xmath460 is the symplectic eigenvalue of the upper left @xmath178 matrix but @xmath461 are the symplectic eigenvalues of the lower right @xmath438 matrix .
the true symplectic eigenvalues of the @xmath462 matrix are in fact given by @xmath463^{2}-\bigl[2b^{2}-a(c+d)\bigr]\bigl[2f^{2}-e(g+h)\bigr]}\biggr\}^{1/2}\,,\notag\end{aligned}\ ] ] so only @xmath353 coincides with one of the symplectic eigenvalues of the covariance matrix .
the result in shows that the cross - correlation between the subsystem 1 ( @xmath246 ) and the union of the subsystems 2 & 3 ( @xmath340 ) is linked up by one representative degree of freedom from each partition .
this correlation is described by the block matrix @xmath464 it will be interesting to see what degrees of freedom are involved in cross - correlation between two partitions .
to do so , we first combine @xmath455 and @xmath427 together @xmath465{\dfrac{e}{a } } & 0 & 0 & 0 & 0 & 0\\[10pt ] 0 & \sqrt[4]{\dfrac{a}{e } } & 0 & 0 & 0 & 0\\[10pt ] 0 & 0 & \dfrac{1}{\sqrt{2}}\sqrt[4]{\dfrac{g+h}{c+d } } & 0 & \dfrac{1}{\sqrt{2}}\sqrt[4]{\dfrac{g+h}{c+d } } & 0\\[10pt ] 0 & 0 & 0 & \dfrac{1}{\sqrt{2}}\sqrt[4]{\dfrac{c+d}{g+h } } & 0 & \dfrac{1}{\sqrt{2}}\sqrt[4]{\dfrac{c+d}{g+h}}\\[10pt ] 0 & 0 & -\dfrac{1}{\sqrt{2}}\sqrt[4]{\dfrac{g - h}{c - d } } & 0 & \dfrac{1}{\sqrt{2}}\sqrt[4]{\dfrac{g - h}{c - d } } & 0\\[10pt ] 0 & 0 & 0 & -\dfrac{1}{\sqrt{2}}\sqrt[4]{\dfrac{c - d}{g - h } } & 0 & \dfrac{1}{\sqrt{2}}\sqrt[4]{\dfrac{c - d}{g - h}}\end{pmatrix}\ ] ] and apply the transformation matrix @xmath466 to the column vector @xmath467 , which describes the state of the tripartite system .
it gives @xmath468{\dfrac{e}{a } } & 0 & 0 & 0 & 0 & 0\\[10pt ] 0 & \sqrt[4]{\dfrac{a}{e } } & 0 & 0 & 0 & 0\\[10pt ] 0 & 0 & \dfrac{1}{\sqrt{2}}\sqrt[4]{\dfrac{g+h}{c+d } } & 0 & \dfrac{1}{\sqrt{2}}\sqrt[4]{\dfrac{g+h}{c+d } } & 0\\[10pt ] 0 & 0 & 0 & \dfrac{1}{\sqrt{2}}\sqrt[4]{\dfrac{c+d}{g+h } } & 0 & \dfrac{1}{\sqrt{2}}\sqrt[4]{\dfrac{c+d}{g+h}}\\[10pt ] 0 & 0 & -\dfrac{1}{\sqrt{2}}\sqrt[4]{\dfrac{g - h}{c - d } } & 0 & \dfrac{1}{\sqrt{2}}\sqrt[4]{\dfrac{g - h}{c - d } } & 0\\[10pt ] 0 & 0 & 0 & -\dfrac{1}{\sqrt{2}}\sqrt[4]{\dfrac{c - d}{g - h } } & 0 & \dfrac{1}{\sqrt{2}}\sqrt[4]{\dfrac{c - d}{g - h}}\end{pmatrix}\cdot\begin{pmatrix}\chi^{(1)}\vphantom{\bigg|}\\[12pt]p^{(1)}\vphantom{\bigg|}\\[12pt]\chi^{(2)}\vphantom{\bigg|}\\[12pt]p^{(2)}\vphantom{\bigg|}\\[12pt]\chi^{(3)}\vphantom{\bigg|}\\[12pt]p^{(3)}\vphantom{\bigg|}\end{pmatrix}\notag\\ & = \begin{pmatrix}\sqrt[4]{\dfrac{e}{a}}\;\chi^{(1)}\vphantom{\bigg|}\\[12pt]\sqrt[4]{\dfrac{a}{e}}\;p^{(1)}\vphantom{\bigg|}\\[12pt]\sqrt[4]{\dfrac{g+h}{c+d}}\;\dfrac{\chi^{(2)}+\chi^{(3)}}{\sqrt{2}}\vphantom{\bigg|}\\[12pt]\sqrt[4]{\dfrac{c+d}{g+h}}\;\dfrac{p^{(2)}+p^{(3)}}{\sqrt{2}}\vphantom{\bigg|}\\[12pt]-\sqrt[4]{\dfrac{g - h}{c - d}}\;\dfrac{\chi^{(2)}-\chi^{(3)}}{\sqrt{2}}\vphantom{\bigg|}\\[12pt]-\sqrt[4]{\dfrac{c - d}{g - h}}\;\dfrac{p^{(2)}-p^{(3)}}{\sqrt{2}}\vphantom{\bigg|}\end{pmatrix}\,.\label{e : ernkwiq}\end{aligned}\ ] ] again from this and , we explicitly show that the cross - correlation between @xmath246 and @xmath340 are set up by the degrees of freedom @xmath391 and @xmath469/2 $ ] . here
we emphasize that this decomposition is suitable only for entanglement of @xmath246 with @xmath340 .
we may want to perform further transformations to get rid of the factors before @xmath391 , @xmath470 , @xmath471 , and etc . in .
however once we do so the resulting covariance matrix no longer takes the williamson form .
in addition , we may observe that those factors play the role of rescaling or changes of the unit in such a way that all elements in the column vector have the same footing or unit .
a comment is ready .
we may use the idea of normal modes to help find the williamson form of a given covariance matrix .
suppose the @xmath74 matrix @xmath261 is composed of the orthonormal eigenvectors of the interaction matrix @xmath75 , as shown in .
we know that the matrix @xmath261 will diagonalize @xmath75 into @xmath472 and the normal mode coordinates will be given by @xmath473 , where @xmath474 .
we may construct a @xmath475 symplectic transformation matrix @xmath163 based on @xmath261 such that @xmath476 this transformation matrix will put the original covariance matrix into an diagonal form @xmath477 recall that there is not correlation between the position and the canonical momentum of a free oscillator .
the diagonal elements of @xmath478 consists of the position and the momentum uncertainty of the normal modes . with appropriate scaling or change of the units
, it becomes the williamson form .
however , this form is not very useful for computing the entanglement measure associated with the partial transpose , because even though the williamson form @xmath478 is related to its original covariance matrix @xmath428 by the symplectic transformation , their partial - transposed counterparts @xmath479 , @xmath480 are not . in other words ,
the partial transpose of the williamson form of the covariance matrix @xmath428 is not necessarily the same as the williamson form of the partial transpose of the same covariance matrix .
thus the entanglement endowed in the original covariance can not be recovered from the partial transpose of its williamson form .
for example , let the matrix @xmath481 account for flipping the sign of @xmath470 , that is @xmath482 we note that the transformation represented by @xmath481 is not a symplectic transformation because @xmath483 this implies that the covariance matrix corresponding to the partial transpose with respect to subsystem 1 is given by @xmath484 where @xmath479 is the partial transpose of the williamson form of the original covariance matrix @xmath428 .
let @xmath485 , and we can check that it is not a symplectic matrix , but an orthogonal matrix . hence @xmath480 is not related to @xmath479 by a symplectic transformation , and their symplectic eigenvalues do not coincide , not necessarily describing compatible entanglement information . instead , to find the entanglement measure based on the symplectic eigenvalues , we should find a symplectic matrix @xmath486 , which will diagonalize @xmath480 into its own williamson form @xmath487 @xmath488 because @xmath489 in general is not a symplectic matrix .
next we turn to the @xmath0 fully symmetric gaussian systems ; in particular , we stress the connection between the orthogonal matrix in the configuration space and the corresponding symplectic matrix in the phase space .
suppose that the lagrangian is given by @xmath490 where @xmath491 , and the @xmath74 interaction matrix is @xmath492 it implies that one of the normal modes must be of the form @xmath493 corresponding to the eigenfrequency @xmath494 and the eigenvector @xmath495 the remaining @xmath496 normal modes @xmath405 for @xmath406 are associated with the eigenfrequency of @xmath496-fold degeneracy @xmath497 , and @xmath419 orthonormal eigenvectors @xmath498 .
the normal modes @xmath499 are in fact related to the original variables @xmath500 by a orthogonal transformation @xmath261 such that @xmath501 .
this @xmath74 transformation matrix is constructed by the eigenvectors of the interaction matrix @xmath75 , @xmath502 and can be used to diagonalized @xmath75 , @xmath503 let the covariance matrix in terms of the normal modes is given by @xmath504 we may find the corresponding covariance matrix @xmath428 in terms of the original variables by an appropriate @xmath162 symplectic transformation matrix @xmath163 , @xmath505 so that @xmath506 if @xmath507 and @xmath508 with @xmath164 , @xmath306 being the conjugated momenta to @xmath26 , @xmath509 , respectively
. it is just the compact form for the statements @xmath510 it then implies that @xmath511 because there is no correlation between the conjugated pair of the normal mode @xmath509 ; on the other hand , the cross - correlation between @xmath12 , @xmath155 is @xmath512 where @xmath513 is the position uncertainty of the normal mode @xmath413 , @xmath514 it will take two different values , depending whether @xmath515 or not , so we have @xmath516 where @xmath517 and @xmath518 the orthogonal matrix @xmath261 has the property that @xmath519 therefore we arrive at @xmath520 since @xmath521 .
similarly we have @xmath522 with @xmath523 , @xmath524 this tells us that generically the covariance matrix @xmath428 takes the form @xmath5250 & e\end{pmatrix}\,,&\pmb{\epsilon}&=\begin{pmatrix}b & 0 \\[4pt]0 & f\end{pmatrix}\,,\end{aligned}\ ] ] and now @xmath526 we convert it to the standard form @xmath527 with @xmath5280 & \lambda\end{pmatrix}\,,&\pmb{\epsilon}'&=\mathbf{k}\cdot\pmb{\epsilon}\cdot\mathbf{k}^{t}=\begin{pmatrix}\zeta_{1 } & 0 \\[4pt]0 & \zeta_{2}\end{pmatrix}\,,\end{aligned}\ ] ] by the symplectic matrix @xmath176 @xmath529{\frac{e}{a}},\sqrt[4]{\frac{a}{e}})\,.\end{aligned}\ ] ] therefore we have @xmath530 since it has been shown in @xcite that for such a fully symmetric system , there are only two distinct symplectic eigenvalues for the covariance matrix @xmath531 .
they are @xmath532\bigl[\lambda+\bigl(n-1\bigr)\zeta_{2}\bigr]}\,,\ ] ] and the @xmath496-fold degenerate @xmath533 from we have derived so far , we obtain @xmath534=\sqrt{\frac{e}{a}}\,\delta_{<}\,,\\ \lambda+\bigl(n-1\bigr)\zeta_{2}&=\sqrt{\frac{a}{e}}\bigl[e+(n-1)f\bigr]=\sqrt{\frac{a}{e}}\,\theta_{<}\,,\\ \lambda-\zeta_{1}&=\sqrt{\frac{e}{a}}\bigl[a - b\bigr]=\sqrt{\frac{e}{a}}\,\delta_{>}\,,\\ \lambda-\zeta_{2}&=\sqrt{\frac{a}{e}}\bigl[e - f\bigr]=\sqrt{\frac{a}{e}}\,\theta_{>}\,.\end{aligned}\ ] ] hence these two symplectic eigenvalues take rather simple form @xmath535 and both of them are exactly @xmath193 .
this is the consequence of the uncertainty principle for the ground state .
if the states for the normal modes are not their ground states , then both @xmath353 , @xmath352 should be greater than @xmath193 . | in this paper we study a system of @xmath0 coupled quantum oscillators interacting with each other directly with varying coupling strengths and indirectly through linear couplings to a scalar massless quantum field as its environment .
the influence of the quantum field on the system is calculated with the use of the influence functional formalism .
we take the direct route of seeking solutions to the evolutionary operator of the reduced density matrix for the derivation of the correlation functions .
they are then used to construct the covariance matrix which we use to perform an analysis of the structure of quantum entanglement in the open system at a stationary state . to see the physical features more explicitly we specialize to a system of three quantum coupled oscillators placed at the vertices of a equilateral triangle and allowed to have disparate pairwise couplings .
we analyze the entanglement between one oscillator and the other two with equal ( symmetric ) and unequal ( asymmetric ) coupling strengths . as an illustration we use the results for these two different configurations to address two representative issues in macroscopic quantum phenomena .
we also mention possible extensions of our work and applications of our analysis and results to issues in some current areas of research in quantum thermodynamics and mesoscopic quantum systems . quantum entanglement , open quantum systems , macroscopic quantum phenomena , influence functional formalism , quantum field theory , quantum brownian motion , non - markovian dynamics |
non abelian gauge theories are the most important ingredient in our present understanding of elementary particles and their interactions .
in particular , quantum chromodynamics ( qcd ) is now universally believed to be the correct theory of the strong interactions . however , while perturbation theory has been used successfully in describing the scattering of particles by partons , the perturbative series does not converge at hadronic energy scales .
moreover , the phenomena of confinement and the hadronic spectrum are fundamentally beyond the reach of perturbation theory . therefore , non - perturbative monte carlo simulations of lattice - regularised qcd are crucial in order to obtain a full description and understanding of qcd phenomena .
the lattice regularisation with a lattice spacing @xmath0 does , however , introduce a sharp momentum cutoff at the momentum scale @xmath1 . connecting lattice measurements to their continuum counterparts therefore requires renormalisation factors accounting for the excluded high - frequency modes .
in particular , renormalisation is needed for qcd matrix elements , and for fixing the bare quark masses to be used in the lattice lagrangian .
renormalisation is also necessary to determine the running of the strong coupling constant @xmath2 and to relate the lattice regularisation scale @xmath3 to @xmath4 .
since the lattice regularisation also introduces discretisation errors , renormalisations are also used to `` improve '' the lattice action in order to reduce these discretisation errors at a given lattice spacing . while in some cases the renormalisation constants can be determined non - perturbatively @xcite
, results at finite lattice spacing can depend upon the method used ( cf .
e.g. @xcite ) , and non - perturbative methods do not cope well with operators that mix under renormalisation . for these reasons , lattice perturbation theory plays an important role in determining the renormalisation constants needed to extract continuum predictions from lattice qcd .
given the breakdown of perturbation theory in low energy qcd , one might doubt whether it could work on the lattice .
an argument in favour of its use is given in @xcite : since the renormalisation factors may be thought of as compensating for the ultraviolet modes excluded by the lattice regulator , and since for typical lattice spacings @xmath5 , the excluded modes have momenta in excess of 5 gev , the running qcd coupling @xmath6 is small enough that perturbation theory should rapidly converge .
the wide range of results reviewed in @xcite show that perturbation theory is useful for a large range of lattice qcd processes .
the assumption that non - perturbative effects do not contribute on these short length scales , can be tested directly in some cases by comparing higher order perturbative calculations with monte carlo simulations performed at a range of weak couplings @xcite ) , or by using so - called stochastic techniques @xcite . in all these cases , the non - perturbative contributions turned out to be small .
other comparisons , such as @xcite , can not distinguish non - perturbative effects from higher - order perturbative corrections .
lattice perturbation theory therefore provides a reliable , and the only systematically improvable , method for determining the full range of renormalisation constants @xcite . as in the continuum ,
the calculation of lattice feynman diagrams is a two stage process .
first , the lattice action and any operator insertions are taylor expanded in the ( bare ) coupling constant to give the propagators and vertices that form the feynman rules ( the `` vertex expansion '' stage ) .
secondly , these rules are then used to construct and numerically evaluate feynman diagrams , possibly after some algebraic simplification ( the `` feynman diagram evaluation '' stage ) .
the latter task is more complicated than in the continuum due to the presence of lorentz symmetry violating terms at finite lattice spacing , and on a finite lattice volume also by the more complicated nature of discrete momentum sums as compared to momentum integrals .
diagrams are therefore usually evaluated using numerical routines like vegas@xcite , or proprietary mathematical packages , possibly after manipulation using other computer algebra packages like form @xcite .
the greater difficulty is posed , however , by the task of vertex expansion .
deriving the feynman results on the lattice is far more complicated than in the continuum for a number of reasons .
firstly , lattice gauge fields are elements of the gauge lie group itself rather than of its lie algebra . to obtain the perturbative expansion of the action
, we must therefore expand exponentials of non commuting objects . as a consequence ,
the feynman rules even for the simplest lattice action are already much more complicated than their continuum counterparts .
secondly , modern lattice theories generally contain a large number of additional ( renormalisation group irrelevant ) terms chosen to improve specific aspects of the monte carlo simulation , such as the rate of approach to the continuum or chiral limits of qcd .
since there is no unique prescription for these terms , and the best choice depends on which quantities we are most interested in simulating , a large number of different lattice actions and operators are in use .
subtle though the differences between the lattice formulations may be , each choice provides a completely separate regularisation of qcd with its own set of renormalisation constants and in particular its own lattice feynman rules .
for a long time , the complications of the perturbative expansion have led to the calculations of renormalisation factors lagging far behind new developments in the improvement of lattice theories . in many cases
this has restricted the physical predictions that could be obtained from the simulations . an automated method for deriving lattice feynman rules for as wide a range of different theories as possible
is therefore highly desirable .
the vertex expansion should be fast enough not to impose undue constraints on the choice of action . to avoid human error
, the user should be able to specify the action in a compact and intuitive manner . since the evaluation of the feynman diagrams can be computationally intensive , and
will often need to be carried out in parallel on costly supercomputing facilities , parsimony dictates that the feynman rules should be calculated in advance on a different system , and rendered as machine readable files that can be copied to the supercomputer for the feynman diagram evaluation stage . in this paper
we describe a pair of software packages for deriving the feynman rules for arbitrary lattice actions and for evaluating the resulting vertices in a numerical feynman diagram calculation .
our algorithm is based through our older algorithm @xcite on the seminal work of lscher and weisz @xcite .
a different implementation of the latter has been used in @xcite , and a similar method is employed in @xcite .
the new feature of the algorithm presented here is that it is capable of expanding not only gluonic actions like the algorithm of @xcite , and fermionic actions like our algorithm from @xcite , but also far more complicated multiply - smeared fermionic actions with reunitarisation such as hisq @xcite , and that it supports taking advantage of the factorisation inherent in some lattice actions , such as improved lattice formulations of nrqcd @xcite . as in @xcite and @xcite ,
the vertex expansion is performed completely independently of any boundary conditions , allowing for instance , the use of twisted periodic boundary conditions as a gauge invariant infrared regulator @xcite or for changing the discrete momentum spectrum in other ways @xcite .
we have used the software packages described for calculations of the renormalised anisotropy in gauge theories @xcite , to study the mean link in landau gauge for tadpole improvement @xcite , to measure the electromagnetic decays of heavy quark systems using nrqcd @xcite , to calculate the radiative corrections to the gluonic action due to highly improved staggered ( hisq ) sea quarks @xcite and the renormalisation of the self energy of heavy quarks using moving nrqcd ( mnrqcd ) @xcite .
the code is flexible and can be easily extended , as has already been done for lattice regularised chiral perturbation theory @xcite , perturbative calculations in the schrdinger functional @xcite and anisotropy calculations @xcite .
the structure of this paper is as follows . in the next section ,
we review the basic expansion algorithm described in ref .
@xcite , outlining some improvements in , for instance , the handling of automatic derivatives and spin matrices . in sec .
[ sec_new_stuff ] , we present novel extensions to the algorithm that are needed to describe complicated fermion actions , including hisq , nrqcd and mnrqcd . sec .
[ sec_software ] provides details of the implementation of the algorithm in the hippy and hpsrc codes and of their installation , testing and use .
we make some concluding remarks in sec .
[ sec_conc ] .
technical details are relegated to the appendices .
the hippy and hpsrc codes are released under the second version of the gnu general public licence ( gpl v2 ) .
therefore anyone is free to use or modify the code for their own calculations . as part of the licensing
, we ask that any publications including results from the use of this code or of modifications of it cite refs .
@xcite as well as this paper .
finally , we also ask that details of these publications , as well as of any bugs or required or useful improvements of this core code , would be communicated to us .
in this section we describe the algorithms used to give the most efficient , yet generic , implementation of the feynman rules for lattice actions .
these extend the original work of lscher and weisz @xcite and developments described in ref .
@xcite .
we consider a @xmath7-dimensional hypercubic spacetime lattice with lattice spacing @xmath0 and extent @xmath8 in the @xmath9-direction : @xmath10 where lattice sites are labelled by a vector @xmath11 . in the following
, we will usually set @xmath12 ( a lattice anisotropy can be introduced by rescaling the coupling constants in the action @xcite ) .
let @xmath13 be a right - handed orthonormal basis , and @xmath14 .
a lattice path consisting of @xmath15 links starting at site @xmath16 can be specified by an ordered set of directions given by integers , @xmath17 : @xmath18 \ } \ ; , \label{eqn_path_def}\ ] ] with the @xmath19 point on the path being @xmath20 and @xmath21
. for periodic boundary conditions , the momentum vectors are @xmath22 and the fourier expansion of a function @xmath23 is@xmath24 where @xmath25 is the lattice volume .
twisted boundary conditions @xcite provide a useful gauge - invariant infrared regulator in perturbative calculations @xcite .
these change the momentum spectrum , converting colour factors into `` twist matrices '' associated with momenta interstitial to the reciprocal lattice ( with the introduction of an additional quantum number , `` smell '' , for fermions @xcite ) .
the hpsrc code fully supports such boundary conditions but for simplicity we only discuss periodic boundary conditions in this paper . following ref .
@xcite , we denote the gauge field associated with a link as @xmath26 , and define @xmath27 .
the gauge potential @xmath28 associated with the midpoint of the link is defined through @xmath29 where @xmath30 is the bare coupling constant . in terms of the anti - hermitian generators of @xmath31 ,
@xmath32 = -f_{abc } t_c , \qquad { \mathop{\mathrm{tr}}}\left ( t_a t_b \right ) = -{\frac{1}{2}}\ , \delta_{ab } \;.\ ] ] quark fields @xmath33 transform according to the representation chosen for the generators @xmath34 , which we take to be the fundamental representation ( other choices will affect the colour factors , but not the structure of our algorithm ) .
the wilson line @xmath35 associated with the lattice path @xmath36 is a product of links @xmath37 \ ; .\ ] ] as all actions and operators can be written as sums of wilson lines ( possibly terminated by fermion fields ) , our goal is to efficiently expand @xmath38 in terms of the gauge potential in momentum space : @xmath39 the vertex functions @xmath40 factorise as @xmath41 with a momentum- and path - independent clebsch
gordan ( colour ) factor @xmath42 @xmath43 it is therefore more efficient to calculate just the expansion of the reduced vertex functions , @xmath44 ( with an appropriate description of the colour trace structure where ambiguous see appendix b of ref .
@xcite for further details ) . the reduced vertex function can be written as a sum of terms , each of which contains an exponential . for convenience
, we will call each term a `` monomial '' : @xmath45 where for each combination of @xmath46 lorentz indices we have @xmath47 terms , each with an amplitude @xmath48 and locations @xmath49 of the @xmath46 factors of the gauge potential , which are drawn from the locations of the midpoints of the links in the path @xmath50 . to avoid floating point ambiguities
, we express the components of all position @xmath7-vectors as integer multiples of @xmath51 ( accounting for the factor of @xmath52 in the exponent ) . in the hpsrc code , we use the convention that all momenta flow into the vertex , so @xmath53 .
( [ eqn_y ] ) makes clear that the number of monomials depends not just on the number of gluons @xmath46 , but also on the choice of lorentz indices @xmath54 , and that each monomial has a different amplitude and set of @xmath46 positions . for clarity of presentation
, we will , however , suppress these additional arguments in later expressions ( notably eqns .
( [ eqn_sy ] , [ eqn_gy ] , [ eqn_diff_y ] , [ eqn_zy ] , [ eqn_xy ] ) ) .
the generation of the feynman rules for generic momenta thus reduces to a calculation of the amplitudes @xmath48 and locations @xmath49 of the monomials that build up the various reduced vertices @xmath55 .
this is all carried out in the hippy code , a description of which can be found in sec . 4 of ref .
the amplitudes and locations defining each monomial are encoded as an instance of the class ` entity ` , and the collection of monomials that make up the reduced vertices is encoded as an instance of class ` field ` .
the data structures have been chosen to ensure that equivalent monomials are combined to minimise the size of the reduced vertex description .
once expanded , the monomials required for the reduced vertices at each order are written to disk as a text file .
the hpsrc code reads these ( previously generated ) vertex files at runtime . for given momenta @xmath56 , lorentz indices @xmath57 and colour indices , the @xmath55 are constructed as given in eqn . ( [ eqn_y ] ) , which is then multiplied by the appropriate clebsch - gordan colour factor(s ) to form the ( euclidean ) feynman rule , @xmath58 .
realistic lattice fermion and gauge actions require some refinements to this generic description .
we begin with the fermion sector .
the most general gauge- and translation - invariant action can be written as @xmath59 and consists of wilson lines @xmath60 defined by open paths @xmath61 , each carrying an associated coupling constant @xmath62 and a spin matrix @xmath63 ( possibly the identity ) . using the convention that all momenta flow into the vertex ,
the perturbative expansion is @xmath64 the euclidean feynman rule for the @xmath46-point gluon
anti - fermion vertex is @xmath65 , where the symmetrised vertex is : @xmath66 where @xmath67 is the permutation group of @xmath46 objects and @xmath68 is applied to the gluonic variables , @xmath56 , @xmath57 and @xmath69 .
the normalisation factor of @xmath70 for this is additional to the @xmath70 factor arising from the taylor expansion of the exponential in eqn .
( [ eqn_ss ] ) .
the reduced vertex @xmath71 is the sum of contributions from paths @xmath72 . in most cases
the clebsch - gordan colour factor is the matrix element : @xmath73 and the reduced vertex function has the structure : @xmath74 where we understand @xmath75 .
cases with more complicated colour structures do arise , for example the use of traceless field strengths in qcd .
such structures are accommodated in the codes ; monomials with different colour structures are distinguished using `` pattern lists '' ( discussed in appendix b of ref .
@xcite ) and appropriate colour factors are applied to each when the feynman rule is constructed .
as there are no permutation symmetries in @xmath76 , there is no advantage to carrying out any symmetrisation in the hippy expansion code . in the hpsrc code , symmetrisation of the feynman rule shown in eqn .
( [ eqn_ferm_symm ] ) carries a potentially significant computational overhead : the reduced vertices must be calculated afresh for each permutation .
not all such permutations may be needed because symmetries of a feynman diagram can reduce the number of distinct contributions to its value from the terms in eqn .
( [ eqn_ferm_symm ] ) . for this reason
, symmetrisation is not carried automatically in the hpsrc code and the user must therefore explicitly construct all permutations requiring a different calculation from eqn .
( [ eqn_ferm_symm ] ) , applying the appropriate normalisation factor .
a typical gluonic action is @xmath77 \ ; , \label{eqn_gen_glue}\ ] ] built of wilson loops @xmath78 defined by closed paths @xmath79 , each with coupling constant @xmath80 .
the perturbative action is @xmath81 the euclidean feynman rule for the @xmath46-point gluon vertex function is @xmath82 , and the vertex @xmath83 is @xcite @xmath84 the reduced vertex @xmath85 is the sum of contributions from paths @xmath86 . as before , the @xmath87 factor normalises the symmetrisation .
@xmath88 can be expanded as @xmath89 in most cases we expect the lattice action to be real .
thus , for every monomial @xmath90 in eqn .
( [ eqn_gy ] ) , there must be a corresponding term @xmath91 . we can therefore speed up the evaluation of the feynman rules by removing the latter term , and replacing the exponentiation in eq .
( [ eqn_gy ] ) with `` @xmath92 '' for @xmath46 even , and with `` @xmath93 '' for @xmath46 odd .
this can either be done by recognising conjugate contours in the action ( e.g. @xmath94 $ ] ) and expanding only one , or by attaching a flag to each monomial to signal whether its complex conjugate has already been removed . if in addition the action has the form eq .
( [ eqn_gen_glue ] ) with a single trace in the fundamental representation , the colour factors are @xmath95 \ ; . \label{cg_notwist}\ ] ] which has a number of permutation symmetries : @xmath96 there is thus a great advantage in carrying out some of the symmetrisation in eqn .
( [ eqn_symmetrise ] ) at the expansion stage in the hippy code .
many of the extra monomials generated by symmetrising over subgroup @xmath97 ( generated by cyclic permutations and inversion ) are equivalent and can be combined in the hippy code , significantly reducing the number of exponentiation operations required to construct the partially - symmetrised @xmath98 : @xmath99 @xmath100 the @xmath101 factors go into the amplitudes of the new monomials coming from the partial symmetrisation .
the number of symmetrisation steps remaining to be carried out in the hpsrc code is the number of cosets in @xmath102 ( one for @xmath103 , three for @xmath104 , twelve for @xmath105 etc . ) .
these symmetrisation steps ( and the normalisation ) are carried out automatically in the hpsrc gluon vertex modules . in many cases , such as when computing wavefunction renormalisation constants , one needs to calculate the derivative of a feynman diagram with respect to one or more momenta . whilst derivatives can be computed numerically using an appropriate local difference operator , such differencing schemes are frequently numerically unstable and require computing the feynman diagram multiple times .
automatic differentiation methods @xcite are a stable and cost - saving alternative
. we can easily construct the differentiated feynman vertex using eqn .
( [ eqn_y ] ) .
if we want to differentiate with repsect to momentum component @xmath106 , we first construct a rank @xmath46 object @xmath107 $ ] which represents the proportion of momentum @xmath108 in each leg of the feynman diagram .
momentum conservation dictates @xmath109 . for instance , for a gluon 3-point function with incoming momenta @xmath110 , we would have @xmath111 $ ] .
the differentiated vertex is @xmath112 and so on for higher derivatives .
we may therefore simultaneously calculate as many differentials as we need for the cost of just one exponentiation . if this momentum expansion is placed into an appropriate data structure for which appropriately overloaded operations have been defined , it is straightforward to create the taylor series for a feynman diagram by simply multiplying the vertex factors together .
we use a slightly modified version of the taylur package @xcite to do this , which encodes the taylor series expansion in the hpsrcfortran code as a derived type ` taylor ` , for which all arithmetic operations and elementary functions have been overloaded so as to respect leibniz s and fa di bruno s rules for higher derivatives of products and functions . in calculations that require only certain higher - order derivatives ,
the taylor multiplication can be significantly sped up by defining a mask that only propagates certain terms in the taylor expansion .
this has not , however , been implemented in the distributed version of the code .
the feynman rules for fermions contain spin matrices ( which may be pauli matrices , e.g. for nrqcd , or dirac matrices for relativistic fermions ) .
we can expand a generic spin matrix @xmath60 using a basis @xmath113 : @xmath114 the product of any two spin basis matrices @xmath115 and @xmath116 is another basis matrix ( with label @xmath117 ) times a phase : @xmath118 we choose the basis so that these phases @xmath119 are real .
where another convention is desired , the amplitudes @xmath120 need to be adjusted appropriately .
for pauli matrices , we use a basis @xmath121 and @xmath122 , giving : @xmath123 for the ( euclidean ) dirac matrices , we choose @xmath124 and @xmath125 .
we define @xmath126 and note the definition here @xmath127 $ ] .
the multiplication for the basic @xmath128 matrices is thus @xmath129 we can thus write the product of two general spin matrices as @xmath130 where @xmath131 .
we use this implicit representation of the spin matrices .
matrix operations on explicit dirac matrices take @xmath132 operations and introduce additional rounding errors .
( [ eqn_spin_multiply ] ) , by contrast , needs @xmath133 multiplications , depending on how many basis matrices are needed to describe @xmath60 and @xmath134 . if @xmath135 , the latter method is more efficient and this is almost always the case .
there are greater gains when inverting spin matrices of the form @xmath136 as we might do when obtaining the propagator of a relativistic quark ( for nrqcd , the propagator is spin diagonal in the pauli matrices and trivial to invert ) .
the inverse is @xmath137 which is far more efficient than inverting a @xmath138 matrix .
inversion of a general spin matrix ( not of the form eqn .
( [ eq_specialspinor ] ) ) is less efficient with an implicit representation , but this is irrelevant in most perturbative calculations .
since all basis matrices except the identity are traceless , taking the trace of a spin matrix is a free operation in the implicit representation . in the hippy code ,
spin basis matrices are associated with monomials using an appropriate integer @xmath139 , which is part of the ` entity ` data structure .
when terms are multiplied together , the factors @xmath140 are absorbed into the amplitude @xmath48 of the resulting monomial . in the hpsrc code , we represent a spin matrix @xmath60 as a defined type , ` spinor ` , which is encoded as a double array @xmath141 it turns out to be significantly more efficient for the order of terms in this array to not necessarily match a standard order of basis elements @xmath113 , hence the use of the index array @xmath142 . in particular , this allows us to omit basis elements with zero coefficient .
arithmetic operations have been overloaded to act appropriately on objects of this type , including implemention of the multiplication table . during inversion ,
an additional function argument , ` short_spinor ` , is used to employ the more efficient expression in eqn .
( [ eq_specialspinor_invert ] ) .
sec . [ sec_old_stuff ]
summarised the general method that was described in ref . @xcite . in general , fermionic actions
are much more complicated than gluonic actions and several algorithmic improvements are needed to efficiently calculate the associated feynman rules .
we stress that by efficient we mean speed - ups of at least an order of magnitude .
the algorithms described in this section can be used independently or together , with the choice configurable at runtime of the code .
all of these features have also been implemented using taylor - valued variables ( as per sec .
[ sec_taylor ] ) to provide automatic differentiation of feynman rules . in many cases
an action has a block - like structure that we can exploit to make the evaluation of the reduced vertices more efficient .
this is particularly useful in the case of nrqcd and mnrqcd actions , which can be heuristically written as @xmath143 .
such an action can be expanded directly in the hippy code but the extremely large number of monomials makes this is inefficient ( or impossibly slow and memory - hungry ) .
it is clear why : there is often little scope for monomial compression between blocks . for instance , in ( m)nrqcd , blocks @xmath144 and @xmath145 live on different timeslices of the lattice , and no compression is possible when combining them . instead , we recognise that the blocks are combined in a gauge covariant manner , so that in @xmath144 , for instance , each contour in @xmath146 starts where a contour in @xmath147 finishes .
summing over the start / end location of each block we obtain a convolution of terms and can use the convolution theorem to construct the overall reduced vertex ( i.e. the momentum - space fourier transform ) from those of the individual blocks .
we refer to the action as being a sum over terms that we call `` summands '' . in the above example , there are two .
each summand is the convolution of a number of `` factors '' , with one factor in the first summand and four in the second .
the overall reduced vertex @xmath148 is the sum of the reduced vertices for each summand . for each summand ,
@xmath148 is calculated by combining the reduced vertices @xmath149 for each of @xmath150 factors that make up that summand , @xmath151 .
we generate these by expanding each factor of the action separately in the hippy code , with the convolution then carried out in the hpsrc code . here
we give the method for constructing @xmath148 for a summand with @xmath150 factors for general @xmath46 .
expressions for specific @xmath152 are given in appendix [ sec_concrete_factors ] .
in giving the general expression , we first establish some useful notation .
consider the ordered set of the first @xmath46 integers : @xmath153 .
we can form an ordered partition of this set : @xmath154 , where the cardinality @xmath155 is the number of elements in the partition .
the set of all such partitions we denote @xmath156 .
for instance , the partitions @xmath156 for @xmath157 , @xmath158 , @xmath159 are shown in table [ tab_partitions ] .
note that we do not consider unordered partitions ( e.g. @xmath160 ) because the gauge fields are explicitly ordered in the paths ( wilson lines ) making up the action .
.[tab_partitions ] the elements @xmath78 in the set of ordered partitions , @xmath156 , of the ordered set of the first @xmath46 integers , @xmath153 , for @xmath157 , @xmath158 and @xmath159 . [ cols="^,^,^",options="header " , ] the general reduced vertex for a summand with @xmath150 factors is then : @xmath161 \\ \left
. \times \left ( \prod_{k = n_{|p|}+1}^{n } y_{f,0}^{(k)}(-{\bm{q}},{\bm{q } } ) \right ) \right\ } \label{eqn_factors}\end{gathered}\ ] ] where @xmath162 is the position of element @xmath163 in the ordered set @xmath78 and @xmath164 .
momentum is conserved in the diagram , so @xmath165 and subsequent @xmath166 ( with @xmath139 defined as above ) . here
@xmath167 refers to the summed momenta for the set @xmath163 ( e.g. @xmath168 ) , implying @xmath169 .
fermion actions often use fattened links to reduce discretisation errors in numerical simulations . by fattening or smearing a link
, we will , in general , end up with an @xmath170 complex colour matrix @xmath171 , which can be expanded in powers of the gauge field @xmath172 .
it is often the case that this matrix is reunitarised by projecting it back onto the gauge group @xmath31 or , more usually , simply back onto the related group @xmath173 .
fattened links can then be further fattened , in an iterative procedure .
an example of this is the hisq action . to complement these numerical simulations
, we need to do perturbative calculations using the same actions .
we confine our attention here to the hisq action ( and simpler variants of the same form , for testing ) , with an iterated , two - level smearing procedure with an intermediate reunitarisation : @xmath174\ ] ] where @xmath175 is the unsmeared gauge field , @xmath176 denotes the polar projection onto @xmath177 ( as used in simulations , and _ not _ @xmath178
@xcite ) , and the fat7 and asq ( a slightly modified version of asq ) smearings are defined in ref .
straightforward application of the expansion algorithm above is theoretically possible but practically unfeasible : the number of monomials is huge and the memory requirements of the hippy code quickly become excessive .
we get around this by taking advantage of the two - level structure inherent in the definition of the action and that the intermediate reunitarisation .
this allows us to express the partially - smeared gauge field as a member of the associated gauge lie algebra , allowing us to split the derivation and subsequent application of the feynman rules into two steps .
in the first step , the feynman rules for the outer ( or `` top '' ) layer , the asq action , are derived in the same way as before .
representing the reunitarised ( and so far uncalculated ) fat7r smeared links by a new , lie - algebra valued gauge potential , @xmath179 : @xmath180 = e^{b_\mu(x+\tfrac{1}{2}\hat\mu ) } \;,\ ] ] we can use the hippy code to expand the asq action in terms of @xmath179 as before .
we obtain similar position - space contributions @xmath181 that fourier transform to give monomials of the usual form .
to complete the derivation of the hisq feynman rules , we also need to know the expansion of @xmath179 in terms of the original gauge potential @xmath172 . to obtain this , we write inner ( or `` bottom '' ) layer , the fat7-smeared link , as @xmath182 = m = hw$ ] , where @xmath183 and @xmath184 ( we suppress lorentz and lattice site indices in the following ) .
we again use the hippy code to obtain an expansion @xmath185 \label{eqn_m_reunit}\ ] ] where @xmath186 is a normalisation constant which , for simplicity in the text , we assume has been rescaled to unity .
the notation here is , for instance , @xmath187 then unitarity of @xmath60 implies that @xmath188 and hence @xmath189 using the expansion @xmath190 rearranging the result as @xmath191 , i.e. @xmath192 finally yields the desired expansion of @xmath146 . these formul
are implemented in this form in the hippy code .
given this , we can now numerically reconstruct the hisq feynman rules for any given set of momenta from eqn .
( [ eqn : asqfr ] ) by a convolution of the asq feynman rules of eqn .
( [ eqn : asqfr ] ) with the expansion of @xmath179 in terms of @xmath172 , summing up all the different ways in which the gluons @xmath172 going into the vertex could have come from the fields @xmath179 appearing in eqn .
( [ eqn : asqfr ] ) . in more detail
, we now separately have the expansions of the asq action in terms of the fattened gauge fields @xmath146 , and of @xmath146 in terms of the unfattened gauge field @xmath147 . to obtain the correct reduced vertex , we must find all the ways that we can get unfattened gluons of the correct lorentz polarisations ( `` directions '' ) .
in doing this , we bear in mind that @xmath179 contains , in principle , gauge fields @xmath193 in all directions and not just @xmath194 .
below we give an expression for the reduced vertex for general @xmath46 .
explicit formulfor @xmath103 , as implemented in the hpsrc code , are given in appendix [ sec_concrete_algebra ] .
using the partitions @xmath156 as before , the reduced vertex for a two - level action is : : @xmath195 where @xmath162 is again the position of element @xmath163 in the ordered set @xmath78 . in the algebra
reduced vertex @xmath196 , the momenta @xmath197 are each one of the arguments to the overall reduced vertex @xmath198 .
as before , in the field reduced vertex @xmath199 , @xmath200 refers to the summed momenta for the partition @xmath201 . in fig .
[ fig_upper_lower ] we represent this expansion graphically for @xmath202 .
solid circles represent the @xmath148 , whilst crossed and open circles represent @xmath203 and @xmath204 respectively .
top - level , fattened gluons @xmath179 are represented by dashed ( brown ) lines , whilst bottom - level , unfattened gluons @xmath172 are shown as wavy ( blue ) lines .
the two sub - diagrams represent the two partition contributions listed in eqn .
( [ eqn_y_f2 ] ) .
as we shall discuss later , this partitioning translates naturally into blocks of code . for certain calculations
, symmetries of the feynman diagram will lead to the contributions of some of these blocks being zero .
we can then improve the performance of the code by commenting them out in these circumstances .
for instance , in the `` tadpole '' diagram in the one - loop fermion self energy , we can remove the term in @xmath205 that is proportional to @xmath206 @xcite .
similarly , in the calculation of the radiative corrections to the gauge action @xcite , we can suppress the term proportional to @xmath207 in the `` octopus '' diagram .
we can further optimise the calculation of the reduced vertices @xmath148 by reusing terms @xmath208 that appear multiple times in the expressions .
for instance , in eqn .
( [ eqn_y_f3 ] ) in appendix [ sec_concrete_algebra ] we can reuse @xmath209 and @xmath210 . for testing it
is useful to define a simpler variant of the smearings described on page 3 of ref .
@xcite , which we call `` fat3 '' .
it is composed just of a central link and adjacent 3-staples with weights @xmath211 , @xmath212 .
the field reduced vertex essentially calculates the feynman rule for obtaining @xmath46 _ fattened _ gluons in particular directions @xmath213 .
the effect of fattening is treated separately in the @xmath204 .
@xmath203 is composed of a sum of @xmath47 monomials , each representing a different way of obtaining the required gluons from the contours comprising the action : @xmath214 where @xmath215 is a spin matrix . note that if there is no fattening of the underlying gluons , this is precisely the reduced vertex @xmath148 described in eqn .
( [ eqn_sy ] ) , if we incorporate the @xmath216 factor that was in eqn .
( [ eqn_ss ] ) .
if the gluons are fattened , the effect of this is contained in the reduced vertices @xmath204 .
these represent how we would choose @xmath46 unfattened gluons in directions @xmath217 from a fattened link in the @xmath218 direction : @xmath219 note that there is no spin matrix in @xmath196 , so the order of the multiplication in the expression for @xmath198 does not matter .
the position vectors @xmath220 refer to the position of the unfattened link relative to the start of the fattened .
as such , we need to remove the half - link shift that was implicit in the definition of @xmath221 as the _ centre _ of the link in the basic expansion algorithm
. the start and end sites of the contour @xmath16 and @xmath222 do not appear in @xmath196 , making @xmath223 for @xmath224 .
of course , the actual values of @xmath47 , @xmath225 and @xmath226 will be different in eqns .
( [ eqn_zy ] ) and ( [ eqn_xy ] ) , as they come from different smearing prescriptions .
we described above how splitting a fermion action such as hisq into two parts simplifies the generation of the feynman rules .
this allowed us to exploit the reunitarisation of the inner fat7r smeared links allows them to be expressed in terms of a new gauge potential @xmath146 ( suppressing the lorentz index ) . nonetheless , the feynman rules for @xmath146 ( as calculated in @xmath204 ) are still too complicated to derive for @xmath227 .
the reason is clear if we look a tthe formula for the reunitarisation .
using eqn .
( [ eqn_m_reunit ] ) , the reunitarised field is @xmath228 , which we can express as an element of the algebra @xmath229 : @xmath230 + \left[a_{\mu \nu } + a^\dagger_{\mu \nu}\right ] a_\sigma \right ) + \frac{1}{3 } a_\mu a_\nu a_\sigma \label{eqn_reunit}\end{aligned}\ ] ] if there are @xmath231 monomials in @xmath232 , there will be at least @xmath233 in @xmath234 arising from the term @xmath235 .
there will be some compression , but this will at best only reduce this number by a factor of around 2 . for the simpler smearing fat3 , @xmath236 and already the hippy code struggles to produce @xmath237 for fat3r .
for fat7 , @xmath238 and direct reunitarisation to produce fat7r links using the hippy code is out of the question .
the alternative is to use the hippy code to produce the @xmath69 and to do the reunitarisation in the hpsrc code by hardwiring the formul in eqn .
( [ eqn_reunit ] ) . as before , the @xmath204 algebra reduced vertices are based on expansion coefficients @xmath239 .
the inputs from the hippy code , however , implement the coefficients @xmath69 .
we use these to build a set of algebra reduced vertices @xmath240 and eqn .
( [ eqn_reunit ] ) gives the relation between the reunitarised @xmath204 and the @xmath240 : @xmath241 \right .
\\ \left .
+ \left [ w^\nu_{f,2}({\bm{k}}_1,\mu_1;{\bm{k}}_2,\mu_2 ) + w^\nu_{f,2}({\bm{k}}_2,\mu_2;{\bm{k}}_1,\mu_1 ) \right ] w^\nu_{f,1}({\bm{k}}_3,\mu_3 ) \right ) \\
+ \frac{1}{3 } w^\nu_{f,1}({\bm{k}}_1,\mu_1 ) w^\nu_{f,1}({\bm{k}}_2,\mu_2 ) w^\nu_{f,1}({\bm{k}}_3,\mu_3 ) \ ; . \end{split } \label{eqn_wy}\end{gathered}\ ] ]
the hippy code is used to taylor expand lattice actions , producing output files encoding the monomials making up the reduced vertices that can later be used by the hpsrc code to evaluate feynman diagrams and integrals .
the code is written in python , for which interpreters are freely available for a wide range of computational platforms @xcite .
the hippy code can be run on any machine which has python version 2.5.x installed on it .
it is also expected to work with any version 2.x , but is not ( yet ) compatible with version 3.x .
only the packages supplied in a standard installation of python are required .
installation consists of unpacking the source files in the chosen location . a set of unit test programs
are contained in subdirectory ` tests ` .
they are run from the command line with no options and the code is expected to pass all of these .
no input files are needed for the tests . whilst the unit testing does not cover each code feature separately , those in ` test_class_field.py ` do cover most features combined .
the hippy code consists of the following main programs that make use of the core routines described below .
a full list of options for each of the main programs can be obtained by running them from the command line with option ` help ` .
[ [ sample_link.py ] ] ` sample_link.py ` + + + + + + + + + + + + + + + + used to generate monomials from the algebra of the group for a single fattened link in the specified direction .
these are used to evaluate @xmath204 in eqn .
( [ eqn_xy ] ) or @xmath240 in eqn .
( [ eqn_wy ] ) in the hpsrc code .
the fattening style is specified using ` bottom_style ` .
the output filenames have the form ` algebra_%s%i_qq*.in ` , where ` % s ` is replaced by the name of the fattening style and ` % i ` by the direction of the link ( with value ` parameters.d ` mapped to zero ) .
wildcard ` * ` is @xmath46 copies of the character `` ` g ` '' ( up to maximum ` parameters.r ` ) . for monomials that will be used in eqn .
( [ eqn_wy ] ) ( with hardwired reunitarisation carried out in the hpsrc code ) , the option ` fld_as_alg ` should be specified . in this case ` %
s ` is a concatenation of character `` ` f ` '' and the name of the fattening .
if an undefined fattening style is specified with ` --bottom_style ` , a list of acceptable styles is printed .
[ [ sample_naive.py ] ] ` sample_naive.py ` + + + + + + + + + + + + + + + + + used to generate monomials for the naive ( or , equivalently , staggered ) lattice dirac action using fattened links .
these are used to evaluate @xmath149 in eqn .
( [ eqn_factors ] ) in the hpsrccode .
the fattening style(s ) are specified using ` top_style ` and ` bottom_style ` , and should not include fattening already carried out at the algebra stage above .
the output filenames have the form ` vertex_%s_qq*.in ` , where ` % s ` is replaced by the name of the top fattening style and ` * ` is @xmath46 copies of the character `` ` g ` '' . for complicated fattening prescriptions ,
there is likely to be some filename clashes , so care should be exercised .
if an undefined fattening style is specified with ` --top_style ` , a list of acceptable styles is printed .
[ [ sample_glue.py ] ] ` sample_glue.py ` + + + + + + + + + + + + + + + + used to generate monomials for a range of simple gauge actions , specified by a command line argument .
as described above , these vertices have been ( partially ) symmetrised . as an example , here are the commands one would use to generate the four - dimensional hisq reduced vertex files using a two - level action with hardwired reunitarisation and mass @xmath242 : .... python sample_link.py --bottom_style fat7 --fld_as_alg 1 python sample_link.py --bottom_style fat7 --fld_as_alg 2 python sample_link.py --bottom_style fat7 --fld_as_alg 3 python sample_link.py --bottom_style fat7 --fld_as_alg 4 python sample_naive.py --top_style asq_for_hisq 0.2 .... which creates files ` algebra_ffat7%i_qq*.in ` and ` vertex_asq_for_hisq_qq*.in ` .
the same feynman rules can be generated using ....
python sample_naive.py --top_style asq_for_hisq \ --bottom_style fat7r 0.2 .... but this generates _ far _ more monomials , which soon becomes prohibitive as @xmath46 increases , as discussed above . for hisq fermions ,
there is the option to set the mass dependent parameter @xmath243 to zero for small @xmath244 .
this is done using a command line option ` --hisq_eps_zero ` for ` sample_naive.py ` . for main programs to calculate feynman rules for other actions , e.g. nrqcd
, interested readers should contact the authors .
these main programs make use of the following core routines : [ [ parameters.py ] ] ` parameters.py ` + + + + + + + + + + + + + + + defines physical and numerical parameters associated with the expansion .
if the monomials require complex - valued amplitudes , e.g. for mnrqcd actions , this is enabled here .
[ [ template.py ] ] ` template.py `
+ + + + + + + + + + + + + defines some link fattening prescriptions that can be used to define actions .
each fattening is a set of paths which do not need to be gauge covariant , so three - link naik terms can be included , for instance . instructions for adding new fattenings
are included in this file .
[ [ spin_multiplication.py ] ] ` spin_multiplication.py ` + + + + + + + + + + + + + + + + + + + + + + + + implements the implicit representations of the spin algebras described in sec .
[ sec_spinor ] . running this from the command line displays the multiplication table from eqn .
( [ eqn_spin_mult ] ) for either pauli or dirac spin matrices .
[ [ wilson.py ] ] ` wilson.py ` + + + + + + + + + + + the main expansion engine that converts a collection of wilson lines into a taylor expansion in the form of reduced vertices built from monomials .
the collection of wilson lines is defined by ....
thepath = [ [ c1,c2,\dots],[p1,p2,\dots ] ] .... where ` c1 ` , ` c2 ` etc .
are the amplitude of each contour , the path of which is defined by a set of signed integers .
for instance , a plaquette would be ` p1=[1,2,-1,-2 ] ` .
as an example of a complete path , a two - dimensional naive fermion action would be defined by a path : .... thepath = [ [ 0.5,-0.5,0.5,-0.5,m ] , [ [ 1],[-1],[2],[-2 ] , [ ] ] ] .... the expansion is not symbolic , so the mass ` m ` must take a specific numerical value .
it is assumed that all links are of the same , fattened style .
the fattening can contain any number of iterated styles defined in ` template.py ` , given as a list ` top_style ` interpreted from right to left .
at the lowest level of this iteration , a further single level of fattening can be specified using ` bottom_style ` .
uniquely , a reunitarisation by projection onto @xmath173 can be imposed after this fattening by appending character `` r '' to the end of the name of the style ( again , chosen from the definitions in ` template.py ` ) .
[ [ class_entity.py ] ] ` class_entity.py ` + + + + + + + + + + + + + + + + +
each monomial in the reduced vertex is encoded as an instance of this class , a description of which can be found in sec . 4 of ref .
translation invariance is exploited so that all paths are assumed to start at the origin , @xmath245 .
if this is not the case ( e.g. for open boundary conditions in the temporal direction for schrdinger functional calculations @xcite ) , then an attribute ` start_site ` would need to be introduced to this class and appropriate changes made to ensure gauge covariance in the multiplication and in other places .
[ [ class_field.py ] ] ` class_field.py ` + + + + + + + + + + + + + + + + the collection of monomials defining the reduced vertices returned by ` wilson.path ( ) ` are stored as a dictionary in an instance of this class .
a fuller description can be found in sec . 4 of ref .
@xcite .
[ [ class_algebra.py ] ] ` class_algebra.py `
+ + + + + + + + + + + + + + + + + + very similar in form to ` class_field.py ` , this class contains the reduced vertices associated with the algebra in eqn .
( [ eqn_algebra2 ] ) .
[ [ mathfns.py ] ] ` mathfns.py ` + + + + + + + + + + + + contains a collection of methods not tied to any particular class .
the hpsrc code uses the input files from the hippy code to create feynman rules , and combines these to evaluate feynman integrals by exact mode summation or statistical estimation using the vegas algorithm @xcite .
the hpsrc code requires a standards - compliant fortran95 compiler with support for minimal cpp directives ( chiefly ` # ifdef ` and ` # ifndef ` ) .
the command ` make ` is optional but useful .
an appropriate mpi wrapper for the compiler will be needed for creating a parallel version of the executable .
no additional libraries are required .
the code is known to compile and execute correctly with the following compilers : intel ifort , portland pgf90 , nag f95 and gnu gfortran on unix - based systems ( the first two also with mpi on parallel machines ) and silverfrost / salford ftn95 and lahey - fujitsu lf95 on microsoft windows systems using cygwin . to install the code ,
the source should be unpacked in a suitable directory and the file ` makefile_details ` ( which contains machine - dependent information ) edited to point to the correct compiler .
it is a good idea to execute ` make clean ` before compiling the code . a given target ` this_executable `
can be compiled using the command .... make this_executable .... use the command ` make ` to see a list of possible executables , or examine ` makefile ` .
main programs are named ` main_.f90 ` and compile to form executables with the same filestem and extension ` .$(exe ) ` as specified by ` makefile_details ` . typically `
$ ( exe ) = x ` , but certain compilers expect ` $ ( exe ) = exe ` . a fortran module ` this ` is typically contained in a file named ` mod_this.f90 ` .
input files containing information that should be specified at compile time are named with extension ` .i ` .
the command ` make clean ` must be executed before recompiling if any of these files are changed .
these files typically contain unavoidable cpp directives and the number of these has been kept to a minimum .
input files that are read at runtime are have names with extension ` .in ` .
most input files have a filename that reflects that of the module that reads them .
the code has been rigorously tested as a whole in a number of scientific calculations , being compared with identical calculations carried out using different programs .
in addition , various manual and automatic tests are routinely carried out to verify the continued correctness of the code .
the manual tests are located in subdirectory ` tests/ ` .
[ [ automatic - differentiation ] ] automatic differentiation + + + + + + + + + + + + + + + + + + + + + + + + + without a templating option in fortran , each vertex is effectively coded twice : once as a complex - valued object and once as a taylor - valued object that includes the automatic derivatives .
the code exploits this by performing a finite difference of the first to compare with the derivatives included in the second . to avoid multiplication of errors in higher - order difference operators ,
the tests are carried out as follows .
first , the non - taylor version is compared with the zero - order derivative from the taylor version . if this agrees , a first - order , symmetric difference is made of the zero - order derivative in the taylor in direction @xmath246 .
we check that , for all @xmath246 , this agrees ( within a specified tolerance ) with the analytic first order derivative in the taylor object .
if so , we form first - order differences of the analytic first - order derivatives and compare these with the analytic second order derivatives and so on .
these comparisons are done at runtime when the vertex modules are first initialised .
this test checks not only that the automatic differentiation ( and associated overloading of operators ) is working correctly , but also that there are no coding inconsistencies between the taylor and non - taylor versions of the vertices .
it can not , of course , detect algorithmic errors that have been consistently coded in both versions .
an additional , lower level test of the taylur package is provided by ` test_taylors.x ` . [
[ test_compare_vertices.x ] ] ` test_compare_vertices.x
` + + + + + + + + + + + + + + + + + + + + + + + + + can be used to check that the feynman rules are the same for two different implementations of the same quark action .
this provides a rigorous check of the algorithms in ` mod_vertex_qq*.f90 ` . having compiled ` test_compare_vertices.x ` , a python script ` test_compare_vertices.py `
can be executed to complete various independent checks of the two level action decomposition , the hardwired reunitarisation and the summand / factor division of the action .
the last is based on an nrqcd action and also provides a strong test of the implementation of the pauli spin algebra .
[ [ tests - of - vegas ] ] tests of vegas + + + + + + + + + + + + + + the vegas code is automatically tested at runtime , evaluating the area under a two - dimensional , normalised gaussian as well as a contour integral .
these tests can also be run using ` test_vegas.x ` .
[ [ lower - level - hpsrctests ] ] lower level hpsrc tests + + + + + + + + + + + + + + + + + + + + + + + ` test_spinors.x ` tests type ` spinor ` defined in sec .
[ sec_spinor ] .
in addition , ` test_pauli_dirac_spinors.x ` tests the embedding of pauli two - spinors into dirac 4-spinors . `
test_print_vertices.x ` prints the values of the fermionic vertices for a random set of momenta .
the same random number seed is used each time , so running the code for different ` vertex_qq_composite.in ` on the same machine will yield outputs that can be directly compared . to illustrate the use of the hpsrc code , we provide three sample programs with makefile targets : [ [ quark_sigma.x ] ] ` quark_sigma.x ` + + + + + + + + + + + + + + + calculates the one - loop renormalisation of asqtad improved staggered fermions . to do this
, it evaluates the self - energy feynman diagrams and their derivatives .
[ [ nrqcd_sigma.x ] ] ` nrqcd_sigma.x ` + + + + + + + + + + + + + + + performs a similar calculation for heavy nrqcd fermions .
[ [ nflwimp.x ] ] ` nflwimp.x ` + + + + + + + + + + + calculates the hisq fermion contribution to the one - loop radiative corrections to the lscher - weisz gauge action , as per refs .
@xcite .
this subsection describes modules that are needed to use the vertex modules .
if mpi is to be used , the file ` use_mpi.i ` must be changed prior to compilation .
similarly , if monomials require complex amplitudes , e.g. for mnrqcd calculations , this is enabled in ` use_complex_amplitudes.i ` .
[ [ mod_num_parm.f90 ] ] ` mod_num_parm.f90 ` + + + + + + + + + + + + + + + + + + numerical parameters that are used by most of the other modules .
most parameters in this module will not need to be changed .
it also optionally reads from file ` paths.in ` the pathnames of the directories holding the vertex and algebra files generated by the hippy code .
[ [ mod_phys_parm.f90 ] ] ` mod_phys_parm.f90 ` + + + + + + + + + + + + + + + + + + + physical parameters used in the calculation .
the code has only been tested for number of dimensions ` ndir ` @xmath247 and colours ` ncol ` @xmath248 .
other choices will require careful testing before use .
values for many parameters can be changed at runtime by editing file ` phys_parm.in ` twisted boundary conditions are selected by setting the number of twisted directions ` ntwisted_dirs ` to 2 , 3 or 4 .
value 0 gives periodic boundary conditions .
for finite lattices , the lattice size is specified in ` latt_size(0:ndir-1 ) ` , with the first component the temporal one .
if momenta are to be squashed @xmath249 @xcite , the squash factors @xmath250 in each direction are specified in ` mom_squash(0:ndir-1 ) ` .
the anisotropy metric ` anis ` is only supported currently for gluonic vertices . [ [ mod_momenta.f90 ] ] ` mod_momenta.f90 ` + + + + + + + + + + + + + + + + + defines type ` mom ` which encodes momenta .
each instance contains the momentum components , the twist vector ( only relevant for twisted boundary conditions ) and a route variable that is used in automatic differentiation .
[ [ mod_taylors.f90 ] ] ` mod_taylors.f90 ` + + + + + + + + + + + + + + + + + defines type ` taylor ` , as described in sec .
[ sec_taylor ] .
[ [ mod_spinors.f90 ] ] ` mod_spinors.f90 ` + + + + + + + + + + + + + + + + + defines types ` spinor ` and ` tayl_spinor ` , as described in + sec .
[ sec_spinor ] .
[ [ mod_matrices.f90 ] ] ` mod_matrices.f90 ` + + + + + + + + + + + + + + + + + + defines types ` mat4x4 ` , used to explicitly represent objects such as the gluon propagator in lorentz index space . a corresponding taylor - valued type ` tayl_mat4x4 ` is also defined .
[ [ mod_colour.f90 ] ] ` mod_colour.f90 ` + + + + + + + + + + + + + + + + calculates the clebsch - gordan factors for both gluonic and fermionic vertices , for periodic and twisted boundary conditions .
colour factors for periodic boundary condition are pre - computed and stored in the compile - time files ` tr_cols_n.i ` and ` mat_cols_n.i ` .
for twisted boundary conditions , the clebsch - gordan factors are computed as needed .
[ [ mod_mod_mpi.f90 ] ] ` mod_mod_mpi.f90 ` + + + + + + + + + + + + + + + + + the code can either be run as a scalar code on a single processor or as a parallel code using mpi . to achieve this with the minimum of compile - time code modifications ( either by hand or using cpp directives ) , ` mod_mod_mpi.f90 `
contains a set of dummy mpi subroutines . whether these dummy routines are used or not is controlled by the single cpp directive in file ` use_mpi.i ` .
note that ` mod_mod_mpi.f90 ` is the only file that reads ` use_mpi.i ` .
only the mpi commands we use are included in ` mod_mod_mpi.f90 ` .
we do not know of any externally maintained library of dummy mpi commands for use in such situations .
if this were to exist , it could easily be used to replace this file . throughout the code
we are ( or try to be ) careful that input / output files are only opened by one processor , with information then shared using mpi commands . [ [ mod_vegasrun.f90 ] ] ` mod_vegasrun.f90 ` + + + + + + + + + + + + + + + + + + provides an implementation of the vegas algorithm , based on the original code of peter lepage @xcite , who also helped convert the code to a parallel version using mpi . on parallel machines
, the processors are divided into farms , each of which does an independent vegas estimate of the integral . within the farm ,
all processors carry out the same calculation , interacting only to share the work of evaluating the integrand at each of the monte carlo points .
the number of processors per farm is controlled in file ` vegas_parm.in ` , and must be a factor of the number of processors .
farming has the advantage of avoiding potentially slow global communications on large clusters .
an improved parallel implementation is currently being prepared as part of a deisa deci - funded project @xcite .
feynman rules @xmath83 for the gluonic vertices are implemented in + ` mod_vertex_*.f90 ` , where ` * ` is @xmath46 ( the number of gluons ) copies of character `` ` g ` '' .
currently vertices for @xmath251 have been implemented .
the modules for @xmath252 are called ` vertex _ * ` and have a very similar structure . the top level routine is ` vert_(k , lorentz , colour ) ` , which takes as arguments the momenta , lorentz and colour indices and returns the complete _ symmetrised _ feynman rule for that vertex as a complex number .
the colour array is ignored if we are using twisted boundary conditions .
the module assumes that the monomials written by the hippy code have been partially symmetrised as described above .
the remaining symmetrisation over the distinct cosets is carried out automatically in the top level routine .
the reduced vertex is calculated from the monomials in ` yvertex _ * ( ) ` .
this code structure is repeated using instances of type ` taylor ` instead of complex numbers , providing analytic derivatives of the feynman rules .
the top level routine in this case is called ` taylor_vert _ ( ) ` . as detailed above , coding the algorithm twice in this way is exploited in the runtime testing of the code . the remainder of the ` mod_vertex_qq*.f90 ` files is taken up with initialisation code ( which reads in all the vertex files at runtime ) and testing routines . only `
mod_vertex_gg.f90 ` differs from this overall structure .
this calculates the gluon propagator . in this case , the top - level routine is + ` gluon_prop(k , colour ) ` .
the propagator is simultaneously calculated for all lorentz index combinations , packaged as a @xmath253 matrix implemented as derived type ` mat4x4 ` .
the choice of gauge is controlled by variable ` galpha ` , specified in runtime input file ` vertex_gg_parm.in ` .
value ` 1 ` corresponds to feynman gauge and ` 0 ` to landau gauge . as detailed in ref .
@xcite , the inverse propagator does not exist in landau gauge , so an intermediate gauge parameter ` gbeta ` should be used . `
gauge_family = ' landau ' ` is the usual choice , where the gauge correction is based on the four - vector @xmath254 .
changing this to ` ' coulomb ' ` uses just the spacelike components instead , but this option is not fully tested . as evaluation of the gluon propagator
can be an expensive part of a perturbative calculation , the propagators for various gluon actions are hardwired in ` mod_vertex_gg.f90 ` .
their use is specified by changing variable ` action ` in ` vertex_gg_parm.in ` from ` ' python ' ` to appropriate other character strings which are listed in subroutine ` vertex_gg_params ( ) ` .
a gluon mass can be added to the propagator .
its square is ` gluon_mass2 ` , specified in ` vertex_gg_parm.in ` .
again , this code is repeated using automatic derivatives packaged instead in a ` taylor`-valued matrix of type ` tayl_mat4x4 ` .
the remainder of the module contains runtime tests and initialisation routines .
associated with the gluon modules are modules calculating the feynman rules associated with the fadeev - popov ghost fields and the haar measure .
these are implemented as hardwired formul in ` mod_vertex_hhstar.f90 ` ( for @xmath255 ) and in ` mod_vertex_meas_gg.f90 ` ( only for @xmath202 ) .
feynman rules @xmath256 for the fermionic vertices ( the _ unsymmetrised _ version of eqn .
( [ eqn_ferm_symm ] ) ) are calculated in the files ` mod_vertex_qq*.f90 ` , where ` * ` is @xmath46 ( the number of gluons ) copies of character `` ` g ` '' . currently vertices for @xmath103 have been implemented .
the modules for @xmath252 are called ` vertex_qq * ` and have a very similar structure .
the top level routine is ` vert_qq*(k , lorentz , colour ) ` , which takes as arguments the momenta , lorentz and colour indices and returns the complete feynman rule for that vertex as an instance of type ` spinor ` .
the convention is that ` k(1:r+2 ) ` is @xmath257 .
the colour array has the same ordering , but is ignored if we are using twisted boundary conditions .
internally , this function calculates the reduced vertex using ` yvertex_qq * ( ) ` and multiplies on the appropriate clebsch - gordon factor , calculated using + ` mod_colour.f90 ` .
[ [ yvertex_qq ] ] ` yvertex_qq * ( ) ` + + + + + + + + + + + + + + + implements the summands and factors decomposition described in sec .
[ sec_summands_factors ] .
the reduced vertices for each factor are calculated in + ` yvertex_qq*_partial ( ) ` .
the sum over partitions @xmath258 naturally translates into a block structure for this routine , the order of which follows that of the expressions in appendix [ sec_concrete_factors ] .
[ [ yvertex_qq_partial ] ] ` yvertex_qq*_partial ( ) ` + + + + + + + + + + + + + + + + + + + + + + + implements the two - level field / algebra algorithm as per section [ sec_two_level ] .
the `` upper - level '' , fattened , field vertices @xmath203 are calculated in ` yvertex_qq*_partial_noalg ( ) ` , combining monomials that have been read into the module at runtime .
the `` lower - level '' , algebra vertices @xmath204 are calculated in ` yvertex_qq*_algebra ( ) ` .
once more , the sum over partitions translates into a block structure for the code , and the order reflects the expressions in appendix [ sec_concrete_algebra ] . as discussed earlier , for certain feynman diagrams we can comment out some blocks on symmetry grounds . [ [ yvertex_qq_algebra ] ] ` yvertex_qq*_algebra ( ) ` + + + + + + + + + + + + + + + + + + + + + + + calculates @xmath204 in terms of the monomial - calculated @xmath240 from ` y_qq*_alg_basic ( ) ` , as detailed in sec .
[ sec_reunit ] .
there is the option to apply no reunitarisation ( setting @xmath259 ) , or the hardwired projection described in the text .
additional projection methods , such as `` stouting '' @xcite , could also be implemented here .
this code structure is repeated using instances of type ` tayl_spinor ` instead of type ` spinor ` , providing analytic derivatives of the feynman rules .
the top level routine in this case is called ` taylor_vert_qqg * ( ) ` . as detailed above , coding the algorithm twice in this way is exploited in the runtime testing of the code . the remainder of the ` mod_vertex_qq*.f90 ` files is taken up with initialisation code ( which reads in all the vertex files at runtime ) and testing routines . only `
mod_vertex_qq.f90 ` differs from this overall structure .
this calculates the fermion propagator . in this case , the top - level routine is + ` quark_prop(k , colour ) ` where the momentum specified is @xmath260 ( which flows in the direction of the fermion arrow ) . `
inv_quark_prop ( ) ` calculates the two - point function for the fermions and implements the summands and factors decomposition described in sec .
[ sec_summands_factors ] .
the reduced vertex for each factor is calculated in ` inv_quark_prop_partial ( ) ` .
this is the lowest - level routine , as there is no algebra contribution for @xmath224 .
these routines are followed in the code by type ` tayl_spinor ` versions and by initialisation and tests .
the hpsrc code allows for the use of multiple fermion types , each with their own feynman rules constructed in their own way .
the details of all of this are specified at runtime by the file ` vertex_qq_composite.in ` , which is read by all the ` mod_vertex_qq*.f90 ` files .
the input takes the form of a namelist .
a typical example is given in appendix [ app_composite ] .
the accuracy of simulations of lattice qcd ( and other field theories ) has been markedly improved by the use of symanzik- and radiatively - improved actions and measurement operators .
in addition to the calculation of the radiative corrections , this improvement programme also requires a concomitant effort in calculating associated renormalisation parameters that are vital to the continuum and chiral extrapolations .
lattice perturbation theory is an essential tool in these calculations .
for some quantitities we can use other techniques , in particular non - perturbative renormalisation ( npr ) @xcite , high-@xmath261 simulations @xcite and stochastic techniques @xcite . in cases where these can be used , lattice perturbation theory provides a valuable check of their results and , in the case of high-@xmath261 simulations , important constraints on the coefficients of the low powers of @xmath2 .
when there is complicated mixing of operators , which happens in a lot of physically relevant measurements , statistical errors lead to the failure of the npr and high-@xmath261 techniques and lattice perturbation theory is the only alternative .
stochastic perturbation theory is also very expensive for theories with dynamical fermions .
lattice perturbation theory for improved actions is complicated : not only does it include many vertices not present in the continuum , the feynman rules also contain a very large number of terms .
these complications vanish , of course , in the continuum limit as the lattice spacing @xmath0 tends to zero .
the point is , however , that lattice perturbation theory is used to correct for the missing momentum modes on a discrete lattice , so the presence of these complications are central to its utility .
many of the complicating terms violate lorentz symmetry and are mathematically complicated .
this makes analytic evaluation of feynman integrals impossible in almost all cases .
it is therefore crucial that we develop a robust computational method for deriving and then using these feynman rules in perturbation theory .
this method must be efficient as well as accurate . in this paper
we have described such a method .
we have used a two - pronged approach , with separate programs to expand the action and derive the feynman rules , and then to use these rules to calculate feynman integrals . in both cases ,
we have developed efficient algorithms to minimise the computational expense .
indeed , without these efficient algorithms we find the calculations to be impossible for many realistic action choices .
we have also presented a full , working implementation of the algorithm in the form of the hippy and hpsrc codes .
the programming languages used for the implementations are python and fortran95 , respectively .
python offers good list - handling and object - orientation features that are suited to the expansion of the action to derive the feynman rules .
fortran95 is numerically efficient , an over - riding concern for the evaluation ( or estimation ) of expensive feynman integrals .
we stress that our method is generically applicable . in the text ,
we have focused on the features of our method that make possible calculations using realistic nrqcd and hisq fermion actions .
this bias merely reflects the problems for which we have used the method most so far .
we have considered other actions , both by expanding them using the hippy code and by hardwiring the hand - derived feynman rules in the hpsrc vertex modules .
particular strengths of our approach are the ability to deal with complicated colour and spin structures in actions and the breaking down of actions into simpler sub - structures for faster evaluation .
we have also provided a particularly simple way of describing actions that reduces the scope for transcription errors .
the comprehensive suite of tests in the code also give confidence in the accuracy of the basic components of both the hippy and hpsrc code suites . as well as extending the basic functionality of the codes , e.g. to describe actions with multiply nested levels of link fattening ,
the codes are easily extended to a wide range of related problems in lattice simulation , as has already been done for lattice chiral perturbation theory @xcite and schrdinger functional calculations @xcite .
another area in which the code might be used is in calculations where the lattice gauge field is split into a fluctuating , quantum piece and an external background field @xcite .
the authors are happy to provide further details on the implementation of this , as well as additional advice on the use of the hippy and hpsrc codes .
a.h . thanks the u.k . royal society for financial support .
the university of edinburgh is supported in part by the scottish universities physics alliance ( supa ) .
is supported by the deutsche forschungsgemeinschaft in the sfb / tr 09 .
this work has made use of the resources provided by the edinburgh compute and data facility ( ecdf ; http://www.ecdf.ed.ac.uk ) .
the ecdf is partially supported by the edikt initiative ( http://www.edikt.org.uk ) .
we also acknowledge support from the deisa extreme computing initiative ( http://www.deisa.eu/science/deci ) .
using table [ tab_partitions ] , we here give explicit expressions for eqn .
( [ eqn_factors ] ) for @xmath148 for a summand in the action with @xmath150 factors , for @xmath262 ( the range of @xmath46 implemented in the hpsrc code ) : @xmath263 @xmath264 @xmath265 @xmath266
the reduced vertex from the @xmath267 is denoted @xmath268 .
the formula for @xmath269 is illustrated graphically in fig .
[ fig_summands ] .
to make things more concrete , we give explicit formul for the @xmath157 , @xmath202 and @xmath269 reduced vertices for a two - level action as defined in eqn .
( [ eqn_algebra ] ) .
this is the range of @xmath46 implemented in the hpsrc code . using the partitions in table [ tab_partitions ]
we obtain : @xmath270 @xmath271 @xmath272
here we provide an explicit example of the runtime file used to specify the feynman rules used by the hpsrc code .
the calculation will use two types of quarks : relativistic hisq and heavy nrqcd .
it is assumed that the appropriate vertex and , where relevant , algebra files have been precomputed using the hippy code and stored in the locations specified in the hpsrc file ` paths.in ` . .... & vertex_qq_composite no_quark_types = 2 ! first quark type : hisq !
all rh indices set to 1 to denote first quark type summands(1 ) = 1 factors(1:1,1 ) = 1 summand_amps(1:1,1 ) = 1.d0 opname(1:1,1,1 ) = " asq_for_hisq _ " algname(1 ) = " ffat7 " reunit_to_apply_to_alg(1 ) = " project " ! second quark type : nrqcd !
all rh indices set to 2 to denote second quark type summands(2 ) = 2 factors(1:2,2 ) = 0,4 summand_amps(1:2,2 ) = 1.d0,-1.d0 opname(1:4,2,2 ) = " nrqcd_a","nrqcd_b","nrqcd_c","nrqcd_a " algname(2 ) = " simple " reunit_to_apply_to_alg(2 ) = " none " !
no opname(:,1,2 ) definition because first summand has no ! hippy inputs because it is a constant / .... that there are two quark types is specified in the namelist using + ` no_quark_types ` . within the hpsrc code
we control which set of feynman rules we use by setting variable ` quark_type ` to 1 or 2 prior to calling + ` quark_prop ( ) ` or ` vert_qqg * ( ) ` .
the first fermions are relativistic hisq quarks .
there is no splitting of the action into summands and factors , so ` summands(qt)=1 ` and + ` factors(1:summands(qt),qt)=1 ` .
similarly , the only summand has amplitude ` summand_amps(1:summands(qt),qt)=1.0d0 ` .
the files used to construct @xmath203 for each factor for each summand are assumed to have been generated by the hippy code and stored in files named ` vertex_%sqq*.in ` where ` % s ` is replaced by appropriate text for each factor for each summand as specified in ` opname(ft , sm , qt ) ` , where ` ft ` runs from 1 to ` factors(sm , qt ) ` and ` sm ` runs from 1 to ` summands(qt ) ` . in this case , the only vertex files are those with ` % s ` replaced by ` asq_for hisq _ ` .
the name of the algebra file is given by ` algname(qt ) ` ( note there is no trailing underscore in this case ) .
we specify that hardwired projection is required to relate @xmath240 in eqn .
( [ eqn_wy ] ) to @xmath204 in eqn .
( [ eqn_xy ] ) using ` reunit_to_apply_to_alg(qt ) ` .
the hpsrc code currently assumes that the same algebra files are used for all factors and summands of a particular quark type .
again , it is assumed that the files defining these have been pre - produced using the hippy code and stored in files named ` algebra_%s%i_qq*.in ` .
` % s ` is replaced by ` algname(qt ) ` and ` % i ` runs from 0 to @xmath273 , where @xmath7 is the number of dimensions . the second block of the namelist defines an nrqcd action of the heuristic form @xmath274 .
there are two summands , the first with 0 factors ( which gives default answer 1 ) and the second with 4
. the minus sign in front of the second summand is specified in ` summand_amps ` .
the first summand has no factors , so ` opname ` is not specified .
the second summand has 4 factors , and the filenames are specified .
note that the first and last filenames are the same .
the algebra is specified as before , and no hardwired reunitarisation is required .
g. martinelli , c. pittori , c. t. sachrajda , m. testa , a. vladikas , a general method for nonperturbative renormalization of lattice operators , nucl .
b445 ( 1995 ) 81108
. http://arxiv.org/abs/hep-lat/9411010 [ ] , http://dx.doi.org/10.1016/0550-3213(95)00126-d [ ] .
r. sommer , non - perturbative qcd : renormalization , o(a)-improvement and matching to heavy quark effective theory , lectures given at workshop on perspectives in lattice qcd , nara , japan ( 2006 ) . http://arxiv.org/abs/hep-lat/0611020 [ ] .
t. bhattacharya , r. gupta , w .- j .
lee , s. r. sharpe , order @xmath0 improved renormalization constants , phys .
d63 ( 2001 ) 074505 . http://arxiv.org/abs/hep-lat/0009038 [ ] , http://dx.doi.org/10.1103/physrevd.63.074505 [ ] .
h. d. trottier , higher - order perturbation theory for highly - improved actions , nucl .
suppl . 129
( 2004 ) 142148 .
http://arxiv.org/abs/hep-lat/0310044 [ ] , http://dx.doi.org/10.1016/s0920-5632(03)02515-5 [ ] .
g. p. lepage , p. b. mackenzie , n. h. shakespeare , h. d. trottier , perturbative two- and three - loop coefficients from large beta monte carlo , nucl .
phys . proc .
83 ( 2000 ) 866871 . http://arxiv.org/abs/hep-lat/9910018 [ ] .
r. horsley , p. e. l. rakow , g. schierholz , separating perturbative and non - perturbative contributions to the plaquette , nucl .
106 ( 2002 ) 870872 . http://arxiv.org/abs/hep-lat/0110210 [ ] , http://dx.doi.org/10.1016/s0920-5632(01)01870-9 [ ] .
h. d. trottier , n. h. shakespeare , g. p. lepage , p. b. mackenzie , perturbative expansions from monte carlo simulations at weak coupling : wilson loops and the static - quark self- energy , phys .
d65 ( 2002 ) 094502 . http://arxiv.org/abs/hep-lat/0111028 [ ] , http://dx.doi.org/10.1103/physrevd.65.094502 [ ] .
a. hart , r. r. horgan , l. c. storoni , perturbation theory vs. simulation for tadpole improvement factors in pure gauge theories , phys .
d70 ( 2004 ) 034501 . http://arxiv.org/abs/hep-lat/0402033 [ ] , http://dx.doi.org/10.1103/physrevd.70.034501 [ ] .
a. hart , g. m. von hippel , r. r. horgan , l. c. storoni , automatically generating feynman rules for improved lattice field theories , j. comput .
( 2005 ) 340353 .
http://arxiv.org/abs/hep-lat/0411026 [ ] , http://dx.doi.org/10.1016/j.jcp.2005.03.010 [ ] .
m. a. nobes , h. d. trottier , g. p. lepage , q. mason , second order perturbation theory for improved gluon and staggered quark actions , nucl .
106 ( 2002 ) 838840 .
http://arxiv.org/abs/hep-lat/0110051 [ ] , http://dx.doi.org/10.1016/s0920-5632(01)01860-6 [ ] .
m. a. nobes , h. trottier , one loop renormalization of fermilab fermions , nucl .
suppl . 119
( 2003 ) 461463 .
http://arxiv.org/abs/hep-lat/0209017 [ ] , http://dx.doi.org/10.1016/s0920-5632(03)01586-x [ ] .
m. a. nobes , h. d. trottier , progress in automated perturbation theory for heavy quark physics , nucl .
suppl . 129 ( 2004 ) 355357 .
http://arxiv.org/abs/hep-lat/0309086 [ ] , http://dx.doi.org/10.1016/s0920-5632(03)02580-5 [ ] .
b. alles , m. campostrini , a. feo , h. panagopoulos , lattice perturbation theory by computer algebra : a three loop result for the topological susceptibility , nucl .
b413 ( 1994 ) 553566 . http://arxiv.org/abs/hep-lat/9301012 [ ] , http://dx.doi.org/10.1016/0550-3213(94)90632-7 [ ] .
e. follana , et al .
, highly improved staggered quarks on the lattice , with applications to charm physics , phys . rev .
d75 ( 2007 ) 054502 .
http://arxiv.org/abs/hep-lat/0610092 [ ] , http://dx.doi.org/10.1103/physrevd.75.054502 [ ] .
g. p. lepage , l. magnea , c. nakhleh , u. magnea , k. hornbostel , improved nonrelativistic qcd for heavy quark physics , phys .
d46 ( 1992 ) 40524067 . http://arxiv.org/abs/hep-lat/9205007 [ ] , http://dx.doi.org/10.1103/physrevd.46.4052 [ ]
. g. m. de divitiis , r. petronzio , n. tantalo , on the discretization of physical momenta in lattice qcd , phys .
b595 ( 2004 ) 408413 . http://arxiv.org/abs/hep-lat/0405002 [ ] , http://dx.doi.org/10.1016/j.physletb.2004.06.035 [ ] . i. t. drummond , a. hart , r. r. horgan , l. c. storoni , one loop calculation of the renormalised anisotropy for improved anisotropic gluon actions on a lattice , phys .
d66 ( 2002 ) 094509 . http://arxiv.org/abs/hep-lat/0208010 [ ] , http://dx.doi.org/10.1103/physrevd.66.094509 [ ] .
i. t. drummond , a. hart , r. r. horgan , l. c. storoni , the contribution of @xmath276 radiative corrections to the renormalised anisotropy and application to general tadpole improvement schemes , phys .
d68 ( 2003 ) 057501
. http://arxiv.org/abs/hep-lat/0307010 [ ] , http://dx.doi.org/10.1103/physrevd.68.057501 [ ] .
i. t. drummond , a. hart , r. r. horgan , l. c. storoni , lattice perturbation theory for gluonic and fermionic actions , nucl .
suppl . 119
( 2003 ) 470475 .
[ ] , http://dx.doi.org/10.1016/s0920-5632(03)01589-5 [ ] .
a. hart , g. m. von hippel , r. r. horgan , leptonic widths of heavy quarkonia : s - wave qcd / nrqcd matching coefficients for the electromagnetic vector annihilation current at o(@xmath277 ) , phys .
d75 ( 2007 ) 014008 .
http://arxiv.org/abs/hep-lat/0605007 [ ] , http://dx.doi.org/10.1103/physrevd.75.014008 [ ] .
a. hart , g. m. von hippel , r. r. horgan , leptonic widths of heavy quarkonia : qcd / nrqcd matching for the electromagnetic current at o(@xmath277 ) , pos lat2006 ( 2006 ) 098 . http://arxiv.org/abs/hep-lat/0609002 [ ] .
a. hart , g. m. von hippel , r. r. horgan , heavy quarkonium decays on and off the lattice , in : g. grindhammer , k. sachs ( eds . ) , proceedings of 15th international workshop on deep - inelastic scattering and related subjects ( dis2007 ) , 2007 , p. 162 .
a. hart , g. m. von hippel , r. r. horgan , radiative corrections to the lattice gluon action for hisq improved staggered quarks and the effect of such corrections on the static potential , phys .
d ( in press ) http://arxiv.org/abs/0812.0503 [ ] . g. m. von hippel , taylur , an arbitrary - order automatic differentiation package for fortran 95 , comput .
174 ( 2006 ) 569576 . http://arxiv.org/abs/physics/0506222 [ ] , http://dx.doi.org/10.1016/j.cpc.2005.12.016 [ ] . k. orginos , d. toussaint , r. l. sugar , variants of fattening and flavor symmetry restoration , phys . rev .
d60 ( 1999 ) 054503 . http://arxiv.org/abs/hep-lat/9903032 [ ] , http://dx.doi.org/10.1103/physrevd.60.054503 [ ] . c. morningstar , m. j. peardon , analytic smearing of su(3 ) link variables in lattice qcd , phys .
d69 ( 2004 ) 054501 . http://arxiv.org/abs/hep-lat/0311018 [ ] , http://dx.doi.org/10.1103/physrevd.69.054501 [ ] .
m. lscher , p. weisz , background field technique and renormalization in lattice gauge theory , nucl . phys . b452 ( 1995 ) 213233 .
http://arxiv.org/abs/hep-lat/9504006 [ ] , http://dx.doi.org/10.1016/0550-3213(95)00346-t [ ] .
m. lscher , p. weisz , computation of the relation between the bare lattice coupling and the ms coupling in su(n ) gauge theories to two loops , nucl .
b452 ( 1995 ) 234260 . http://arxiv.org/abs/hep-lat/9505011 [ ] , http://dx.doi.org/10.1016/0550-3213(95)00338-s [ ] . | the derivation of the feynman rules for lattice perturbation theory from actions and operators is complicated , especially for highly improved actions such as hisq .
this task is , however , both important and particularly suitable for automation .
we describe a suite of software to generate and evaluate feynman rules for a wide range of lattice field theories with gluons and ( relativistic and/or heavy ) quarks . our programs are capable of dealing with actions as complicated as ( m)nrqcd and hisq .
automated differentiation methods are used to calculate also the derivatives of feynman diagrams .
quantum chromodynamics , qcd , lattice qcd , perturbation theory 11.15.ha , 12.38.gc 81 - 04 , 81t13 , 81t15 , 81t18 , 81t25 , 81v05 , 65s05 ,
41a58 * program summary * _ manuscript title : _ automated generation of lattice qcd feynman rules + _ authors : _ a. hart , g.m .
von hippel , r.r .
horgan , e.h .
mller .
+ _ program title : _ hippy , hpsrc + _ journal reference : _
+ _ catalogue identifier : _ + _ licensing provisions : _ gplv2 ( see note in sec . [ sec_licence ] . )
+ _ programming languages : _ python , fortran95 + _ ram : _ problem specific , typically less than 1 gb for either code . + _ keywords : _ quantum chromodynamics , qcd , lattice qcd , perturbation theory + _ pacs : _ 11.15.ha ; 12.38.gc + _ classification : _ 4.4 feynman diagrams ; 11.5 quantum chromodynamics , lattice gauge theory + _ nature of problem : _
+ derivation and use of perturbative feynman rules for complicated lattice qcd actions .
+ _ solution method : _ + an automated expansion method implemented in python ( hippy ) and code to use expansions to generate feynman rules in fortran95 ( hpsrc ) .
+ _ restrictions : _ + no general restrictions .
specific restrictions are discussed in the text . + _ running time : _
+ very problem specific , depending on the complexity of the feynman rules and the number of integration points . typically between a few minutes and several weeks . |
the standard paradigm for long - duration gamma - ray bursts ( grbs ) involves a broad - lined type ic supernova ( sn ic - bl ) with @xmath6 mag @xcite , such as those predicted by the collapsar model @xcite .
this is based on spectroscopic evidence in sne from low - luminosity grbs @xcite , such as sn 1998bw accompanying grb 980425 @xcite , as well as relatively - higher - luminosity grbs , such as sn 2003dh accompanying grb 030329 @xcite .
interestingly , for the two low - redshift cases of grb 060505 and grb 060614 no associated sn was found to deep limits @xcite , but since then no similar events have been reported .
grb 130427a @xcite is remarkable as it is both extremely energetic and located at a moderately low redshift of @xmath2 ( @xcite and this work ) . using the spectral parameters for the prompt emission given by @xcite
, we derive an isotropic @xmath0-ray energycdm cosmology with @xmath7 km s@xmath8 mpc@xmath8 , @xmath9 , and @xmath10 @xcite .
] of @xmath11 erg in the 110000 kev rest - frame energy band .
the optical afterglow peaked at @xmath12 mag during the prompt emission phase @xcite .
only @xmath13 of all grbs with measured redshifts are at such a low distance ( see e.g. @xcite ; figure [ eisoz ] ) , and those are usually low - luminosity events .
by contrast , grb 130427a was an extremely energetic burst , and hence it has allowed detailed studies of a system similar to those usually only found at higher redshifts @xcite .
due to the dearth of such extremely luminous grbs at low redshift and their faintness at high redshift , we so far have had no spectroscopic evidence for accompanying sne in very energetic grbs with @xmath14 erg ( for photometric evidence of a sn associated with very energetic grb 080319b , see @xcite ) . in figure [ eisoz ]
we show the @xmath0-ray energy release of grbs as a function of redshift .
overplotted are systems with spectroscopic evidence for a sn .
grb 130427a stands out as an exceptional system .
it is unclear if progenitor models involving a sn can power such energetic grbs ( see , e.g. , @xcite ) , and observationally it has proven hard to test the sn properties of these events . in this work
we focus on the search for and discovery of a sn accompanying this remarkable burst .
our first follow - up photometry was carried out at the 1-m optical telescope located in weihai , shandong province , china .
the bright optical counterpart of the grb was well detected in the cousins @xmath15 filter with @xmath16 mag at 4.178 hr post - burst .
initially , sdss filters were not available at the telescope , but 2 days into our monitoring campaign , sdss @xmath4 and @xmath17 filters were installed and have been available since then .
our photometric follow - up observations were mainly obtained at the 2.5-m nordic optical telescope ( not ) equipped with the alfosc instrument .
photometry was primarily obtained in the @xmath4-band , complemented by @xmath18 data useful to monitor the spectral evolution of the counterpart . as shown in figure [ field ] , there is indication from the not images that the grb counterpart lies in the nw part of an extended host galaxy . with our spatial resolution ,
the afterglow is blended with the host , and we thus use a relatively big aperture ( @xmath19 diameter ) to measure the magnitudes of the counterpart plus host . using a smaller aperture
would provide lower fluxes , but with a host contribution hard to quantify , especially at late times ( i.e. , since @xmath20 days after the burst ) . in this way
, we consistently include most of the host light .
additionally , follow - up observations were obtained at the 2.16-m telescope in xinglong , hebei province , china .
photometry was done in the cousins @xmath15 and @xmath21 filters and then transformed to sdss @xmath4 and @xmath17 magnitudes based on our afterglow spectra .
all optical data were reduced in a standard way using iraf v2.15 in the scisoft 7.7 package .
magnitudes were calibrated with two nearby bright sdss field stars , sdss j113230.55 + 274420.3 and sdss j113220.11 + 274133.5 , whose zero - point errors are 0.01 mag and propagated into the final magnitude measurements .
the sdss @xmath4-band lightcurve is presented in figure [ rlc ] and a log of the observations is shown in table [ log ] .
we fit the not multi - color photometry taken in the first night after the trigger ( @xmath22 mag , @xmath23 mag , @xmath24 mag , @xmath25 mag ) and the simultaneous xrt spectrum . using synchrotron models and extinction laws from the local group ( see * ? ? ?
* for details ) , we estimate the reddening of the host to be @xmath26 mag for a mw - type extinction law . within the errors ,
this value is consistent with what derived assuming an smc or lmc extinction law , because of the small amount of reddening and wavelength range probed by our observations .
our first spectrum was obtained using not / alfosc .
the total exposure was 1800 s with a mean time of 0.44 days post - burst .
the spectrum covers the range 32009100 with a resolving power of @xmath27 .
we identify prominent absorption lines of 2796 & 2803 , 2852 and 3934 & 3968 , as well as weak emission lines of [ ] 3727 and h@xmath28 , all at a common redshift of @xmath29 .
a second spectrum with intermediate resolution was obtained shortly afterwards using the very large telescope ( vlt ) equipped with the x - shooter spectrograph . in the spectra
the continuum was well detected in the full range 300024800 .
a number of absorption features are visible , including 2344 , 2577 , 2796 & 2803 , 2852 , 3074 , 3934 & 3968 , 5890 & 5896 , and emission lines such as [ ] 3727 , h@xmath28 , [ ] 5007 , and h@xmath30 , all at a common redshift of @xmath2 . in the x - shooter spectrum
d 5890 & 5896 absorption was detected at the redshift of the host .
we measure @xmath31 and @xmath32 for the equivalent widths of the d1 and d2 components . using the relations in @xcite
, we obtain an estimate for @xmath33 mag , but remark that there exists substantial dispersion of @xmath34 mag in this relation . considering different calibrations / systematics involved in above @xmath35 measurements , we adopt @xmath36mag for the host extinction .
given the relatively low redshift , we planned a third spectroscopic observation with the aim of detecting the sn signatures .
based on the light curve evolution , we obtained a spectrum of the optical counterpart and host galaxy with the 10.4-m gran telescopio canarias ( gtc ) 16.7 days after the grb .
this corresponds to 12.5 days in the grb rest frame .
observations consisted of @xmath37 s , covering the range of 480010000 with a resolving power of @xmath38 .
the slit was oriented to cover both the afterglow position and the host galaxy nucleus .
all spectroscopic data were reduced in a standard way using dedicated iraf pipelines or eso pipelines .
the resulting spectra are presented in figure [ spectrum ] .
we use the catalogued pre - explosion imaging from the sdss data release @xcite to estimate physical properties of the grb host galaxy and to build a physical model of the stellar emission .
the sdss @xmath39 photometry was fitted within lephare @xcite using stellar population synthesis models from @xcite as detailed in @xcite .
based on the model fit to the data ( figure [ spectrum ] ) , we derive a luminosity @xmath40 mag , a stellar mass of @xmath41 , a star - formation rate @xmath42 yr@xmath8 and an age of the starburst of @xmath43 myr for the host of grb 130427a .
host galaxy emission lines are detected above the sn continuum , including [ ] 6584 in the gtc spectrum , albeit at low significance .
this allows us to place constraints on the metallicity of the explosion host environment by using the calibrations in @xcite .
we measure @xmath44 and @xmath45 by using the o3n2 and the n2 methods , respectively ( statistical errors only ) , which is at the top - right of the grb - sn range in the metallicity@xmath46 plane and similar to the cases of sne ic - bl without observed grbs and sne ib+iib ( see fig . 2 in @xcite ) .
we note that the host galaxy also nicely lies within the 1@xmath47 dispersion of the mass - metallicity relation for normal field galaxies @xcite . after including the systematic dispersion of 0.14 and 0.18 dex @xcite for the two methods , these results translate to a metallicity of @xmath48 and @xmath49 @xmath50 , respectively @xcite .
we scaled the gtc spectrum with our simultaneous photometry from the same night , dereddened it for @xmath51mag @xcite in the milky way galaxy , and then subtracted the model host galaxy spectrum , after bringing it to the same resolution , in order to obtain a `` clean '' spectrum of the transient .
afterwards , the spectrum is dereddened by @xmath52mag for the host extinction at @xmath53 .
both the original and the final gtc spectra are shown in figure [ spectrum ] .
although the resulting spectrum is noisy , it shows clear sn features , with the most prominent being a strong bump peaking at @xmath54 ( observer - frame ; @xmath55 rest - frame ) .
the features are broad ( and no h or he can be seen ) , justifying the classification of sn 2013cq as a sn ic - bl @xcite .
sn 1998bw @xcite , associated with grb 980425 , does not provide a good spectral match to sn 2013cq , mainly because its main peak is located more to the red ( rest - frame @xmath56 at similar phases ) . the same is true for sn 2006aj @xcite , associated with grb 060218 .
instead , we find a better match with sn 2010bh , associated with grb 100316d , which is known to have high expansion velocities , up to 10000 km s@xmath8 higher than other previous sne ic - bl associated with grbs at all phases @xcite .
in particular , the best match is obtained with the spectrum of sn 2010bh at a rest - frame time of 12.7 days , very close to the rest - frame 12.5 days for sn 2013cq here ( shown in figure [ spectrum ] ) .
the similarity to sn2010bh is striking although there may be small colour differences ( e.g. , difference redward of @xmath577500 ) , which may reflect the diversity in grb - sn spectra and/or the uncertain extinction correction towards sn2010bh . considering that the p - cygni feature on the left of the strong bump can be primarily attributed to 5169 , we measure an expansion velocity of @xmath58 km s@xmath8 from the absorption minimum .
this is very similar to the peak photospheric velocity , @xmath59
km s@xmath8 , of sn 2010bh @xcite . the @xmath4-band lightcurve presented in figure [ rlc ] has been corrected for the foreground extinction ( @xmath51 mag ) . from the foreground - extinction - corrected flux we subtract the contribution of the host galaxy ( @xmath60 , as determined via photometry on pre - explosion sdss imaging ) , so that the flux powering the lightcurve can be attributed solely to the afterglow and accompanying sn .
visual inspection of the light curve reveals a deviation away from a power - law decay a few days post - burst , followed by a plateau that lasts for about 10 days , before decaying further . at the gtc spectrum time
the galactic - extinction - corrected flux of the sn / afterglow is 1.85 times that of the host galaxy .
next , we fit a smoothly broken power - law model @xcite to the lightcurve up to the first four days ( some @xmath4-band data points @xmath61 days published in gcns are used for the fitting ; @xcite ) , and derive the following best - fit parameters : @xmath62 , @xmath63 , and @xmath64 days ( the smoothness factor of the break is fixed to be unity and @xmath65/dof = 1.47 for 7 dof ) .
this afterglow model is then subtracted from the already host - subtracted lightcurve , and the resultant flux resembles that of a supernova both in brightness and shape ( e.g. , @xcite ) .
the peak brightness of sn 2013cq is found to be @xmath66 mag at @xmath67 days observer - frame ( @xmath68 days rest - frame ) . at @xmath29 ,
the observer - frame @xmath4 band corresponds approximately to the rest - frame @xmath69 band , and we find that sn 2013cq has an absolute @xmath69-band magnitude of @xmath70 , which is about 0.2 mag fainter than sn 1998bw .
we then compare the optical properties of sn 2013cq relative to two other grb - associated sne : sn 1998bw and sn 2010bh .
we create synthetic @xmath4-band light curves of sn 1998bw and sn 2010bh as they would appear at @xmath29 .
next , using equations 1 and 2 from @xcite , we determine the stretch ( @xmath71 ) and luminosity factor ( @xmath72 ) of sn 2013cq relative to sn 1998bw to be @xmath73 and @xmath74 .
we further estimate the bolometric properties of sn 2013cq .
the method , developed by @xcite , uses the relative stretch and luminosity factors of a sn relative to a template ( in this case sn 1998bw ) , and makes the assumption that the relative shape of a given sn in a given filter is a good proxy for the relative shape of the sn bolometric light curve relative to the template .
the bolometric light curve of the template filters . ]
is then altered by @xmath71 and @xmath72 , and then fit with an analytical model derived from @xcite .
the arnett model depends on knowing the photospheric velocity at peak bolometric light to determine the ejecta mass and explosion energy , which for many sne is determined from the velocity of different species in the sn spectra at peak light .
we have used the photospheric velocity value determined in section [ subsec : gtc ] ( i.e. @xmath58 km s@xmath8 ) . on comparison with sn 1998bw , for sn 2013cq ( in @xmath75 )
we derived its ni mass , ejecta mass , and kinetic energy @xmath76 , @xmath77 , and @xmath78 erg , respectively .
the quoted errors are statistical only .
our photometric and spectroscopic campaign has led to the unambiguous discovery of a sn ic - bl , sn 2013cq @xcite , associated with grb 130427a at @xmath29 .
grb 130427a is one of the most energetic bursts ever detected with @xmath79 , comparable to that of high - redshift grbs and much larger than local events .
the fact that a supernova progenitor model accounts for even very energetic bursts suggests a common progenitor model , such as the collapsar model , may account for the majority of all long - duration grbs . to overcome the challenge of providing enough energy to power the grb , it is likely that the large observed energy is due
to beaming ( for a strong beaming of grb 130427a , see @xcite ) , making the true energy much lower ( typically two orders of magnitude ) .
our discovery now suggests that not only core - collapse sne , but specifically stripped envelope , high velocity sne are almost an inevitable consequence of the deaths of stars that form all grbs .
a common mechanism is therefore at play to power grbs with very different high - energy properties .
it is worth noting that the comparable peak @xmath69-band luminosity of sn 2013cq and of sn 1998bw is consistent with the suggestion of @xcite that there is an upper envelope to the brightness of grb - sne which drops slightly with isotropic luminosity .
the origin of such an upper envelope is intriguing , but currently not clear .
as shown in @xcite , among previous grb - sn events , the highest oxygen abundance @xmath80 at the sn position was less than 8.3 , and when the oxygen abundance rises towards a seemingly critical value of 8.5 , grb - sn events tend to locate in dwarf galaxies with @xmath81 mag .
note that the abundance value of 8.5 is a typical one for sne ic - bl without observed grbs which occur in both luminous and dwarf galaxies and also it is typical for sne ib+iib which happen in relatively luminous galaxies with @xmath82 mag @xcite . with an abundance of @xmath83 for the sn position and @xmath40 mag for the host , grb 130427a / sn
2013cq is consistent with subclasses of core - collapse sne such as sne ic - bl without observed grbs and sne ib+iib in the metallicity@xmath46 plane , implying that grbs do not exclusively explode in low metallicity dwarf galaxies .
we thank the referee for valuable comments and suggestions .
we are grateful to jens jessen - hansen , tiina liimets , dagmara oszkiewicz , yeisson martinez - osorio , olesja smirnova , jari kajava , tuomas kangas , emanuel gafton , anthony macchiavello , tore mansson , and joachim wiegert for taking images at the not .
we are also grateful to jinyi yang for carrying out observations at the xinglong 2.16-m telescope .
dx thanks adam s. trotter for helpful discussion .
dx , jpuf and bmj acknowledge support from the erc - stg grant eggs-278202 .
adup acknowledges support by the european commission under the marie curie career integration grant programme ( fp7-people-2012-cig 322307 ) .
adup , ct and jg are supported by the spanish research project aya2012 - 39362-c02 - 02 .
gl is supported by the swedish research council through grant no .
623 - 2011 - 7117 .
zc and pj acknowledge support from the icelandic research fund .
ss acknowledges financial support from the iniciativa cientifica milenio grant p10 - 064-f ( millennium center for supernova science ) , with input from `` fondo de innovacin para la competitividad , del ministerio de economa , fomento y turismo de chile '' , and basal - cata ( pfb-06/2007 ) .
lpx , tmz , and ycz acknowledge the support of the national science foundation of china grants 11103036 , 11203034 & 11178003 , and u1231101 , respectively .
the dark cosmology centre is funded by the danish national research foundation . based on observations made with the nordic optical telescope , operated by the nordic optical telescope scientific association at the observatorio del roque de los muchachos , la palma , spain , of the instituto de astrofsica de canarias . based on observations made with eso telescopes at the la silla paranal observatory under programme
i d 091.c-0934 based on observations made with the gran telescopio canarias ( gtc ) , at the spanish observatorio del roque de los muchachos ( la palma ) .
this work made use of data supplied by the uk _ swift _ science data centre at the university of leicester . | long - duration @xmath0-ray bursts ( grbs ) at @xmath1 are in most cases found to be accompanied by bright , broad - lined type ic supernovae ( sne ic - bl ) .
the highest - energy grbs are mostly located at higher redshifts , where the associated sne are hard to detect observationally . here
we present early and late observations of the optical counterpart of the very energetic grb 130427a . despite its moderate redshift @xmath2 , grb 130427a
is at the high end of the grb energy distribution , with an isotropic - equivalent energy release of @xmath3 erg , more than an order of magnitude more energetic than other grbs with spectroscopically confirmed sne . in our dense photometric monitoring ,
we detect excess flux in the host - subtracted @xmath4-band light curve , consistent with that expected from an emerging sn , @xmath5 mag fainter than the prototypical sn 1998bw .
a spectrum obtained around the time of the sn peak ( 16.7 days after the grb ) reveals broad undulations typical of sne ic - bl , confirming the presence of a sn , designated sn 2013cq .
the spectral shape and early peak time are similar to those of the high expansion velocity sn 2010bh associated with grb 100316d .
our findings demonstrate that high - energy long - duration grbs , commonly detected at high redshift , can also be associated with sne ic - bl , pointing to a common progenitor mechanism . |
several observations made over the recent years , related to a large extension to large scale structures ( lss ) and anisotropies of the cosmic microwave background ( cmb ) as well as the magnitude redshift relation for type ia supernovae have given us a convincing picture of the energy and matter density in the universe @xcite .
baryonic matter accounts for no more than 30@xmath4 of the mass in galaxy clusters while the existence of a large clustered component of dark matter ( dm ) seems now firmly established , although its nature is still unknown . however , they contribute to the total energy density of the universe with only a few percent and about 25@xmath4 respectively .
no more than another few percent could be accounted for by massive neutrinos , but only in the most favorable , but unlikely case . according to @xcite
( see also @xcite ) the total mass of neutrinos can not exceed the limit of 1.43 @xmath5 ( see , however , @xcite for a recent analysis on neutrino mass limits in coupled dark energy models ) . a very small part @xmath6 of the total energy density is due to massless neutrinos and cmb radiation .
the model suggested by observations is only viable if the remaining 75@xmath4 is ascribed to the so called dark energy ( de ) responsible for the present day cosmic acceleration .
although strongly indicated by the observations , the existence of de is even more puzzling than dm .
it can be identified with a cosmological constant @xmath0 or with a yet unknown dynamical component with negative pressure . on the other hand
, its manifestation can be interpreted as a geometrical property of the gravity on large scales resulting from a failure of general relativity ( gr ) on those scales ( see @xcite for a review ) . within the context of gr , as an alternative to the cosmological constant
, de is usually described as a scalar field @xmath7 , self interacting through a suitable potential @xmath8 , or a cosmic fluid with negative pressure @xcite ( see @xcite for alternatives and references therein ) .
scalar fields naturally arise in particle physics .
furthermore , if they are _ tracker _ fields @xcite , _ fine tunings _ associated to the small value of the present de energy density can be significantly alleviated unlike the cosmological constant case .
in addition to self interaction , a scalar field can in principle be coupled to any other field present in nature . however , in order to drive the cosmic acceleration , its present time mass is expected to be , at least on large scales , @xmath9 ( @xmath10 being the present hubble parameter ) .
such a tiny mass gives rise to long range interactions which could be tested with fifth force type experiments .
couplings to ordinary particles are strongly constrained by such a kind of experiments but limits on the dm coupling are looser ( constraints on coupling for specific models were obtained in @xcite from cmb , n - body simulations and matter power spectrum analysis ) . a possible common origin of dm and de and/or a their direct coupling @xcite .
would ease one of the most critical problems in modern cosmology , the so called _ coincidence problem _ : why expansion started to accelerate just at the eve of our cosmic epoch , after decelerating during all epochs after inflation ? why de and dm have similar densities just now ? because of the coupling dm and de densities keep similar values during a long period and the only peculiar feature of the present epoch is the recent overtaking of dm density by de density .
if present , dm de coupling could have a relevant role in the cosmological evolution affecting not only the overall cosmic expansion but also modifying the dm particles dynamics with relevant consequences on the growth of the matter density perturbations in both linear and nonlinear regime ( e.g. , on halo density profiles , cluster mass function and its evolution , see @xcite .
lss is then a powerful probe of de nature which permit to put significant constraints on de parameters . constraints often become even more stringent when data from other probes are simultaneously taken into account .
cmb is another powerful probe of de nature . in principle , by joining anisotropy and polarization data , de parameters can be significantly constrained .
cmb and lss probe the universe at different epochs and are therefore complementary to each other
. future data from high resolution cmb experiment like planck and lss surveys ( euclid , lsst , des , jdem , etc ) will allow to constrain de up to an unprecedented accuracy .
in this paper we will focus on the integrated sachs wolfe ( isw ) effect @xcite .
isw effect is a secondary anisotropy of the cmb and a direct signature of de .
the effect arises when a photon from the last scattering surface passes through a time
dependent gravitational potential changing its energy so that additional temperature anisotropies are generated .
decay of gravitational potentials may occur through cosmic curvature , in the presence of de or in alternative gravity models .
assuming general relativity is the correct theory of gravity and that the universe is spatially flat , large scale gravitational potentials do not evolve significantly in the matter era .
cosmic acceleration , however , causes the gravitational potentials to decay making the isw effect highly sensitive to the presence of de . though difficult to detect directly in the cmb , isw signal can be measured by cross correlating the cmb with tracers of lss and has recently been detected using wmap data of cmb in combination with several lss surveys at the the @xmath11 confidence level providing independent evidence for the existence of the de ( see @xcite and references therein ) .
cross - correlation then provides a powerful method to discriminate among different de models and , in particular , to detect a possible interaction between de and dm other than investigate the clustering properties of de on large scales . if present , dm de coupling changes both the scaling of the dm energy density and the growth rate of matter perturbations yielding a significant evolution of the metric potentials even in the matter era .
the aim of this paper is to provide an exhaustive analysis of the isw effect in the contest of the _ so called _ coupled de cosmologies @xcite mainly focusing on the effects of the dm de coupling on the cross
correlation between galaxy / quasar distributions and isw effect .
such models can be motivated in the contest of scalar
tensor theories of gravity or describe the low energy limit of a more fundamental theory beyond the standard model of particle physics , e.g. string theory . the models which we aim to investigate differ from the standard @xmath0cdm in three different aspects : ( i ) de is a self interacting scalar field @xmath7 rather than a cosmological constant @xmath0 . we shall consider a class of self interaction potentials @xmath8 admitting _
tracker solutions_. ( ii ) a linear dm de coupling is allowed .
( iii ) we allow neutrinos to be massive . the effects of massive neutrinos in cosmology have been studied thoroughly for many years ( for a review see @xcite ) .
cosmological observations are mostly sensitive to the sum of neutrino masses , @xmath12 .
currently , the strongest upper limit on neutrino mass scale comes from cosmology .
one of the effects of massive neutrinos is to induce a small decay of the gravitational potentials during both matter and de domination so that , in principle , isw effect would also provide information on their mass .
furthermore , as recently outlined in @xcite the effects that massive neutrinos have on the angular power spectrum of the cmb anisotropies , @xmath13 , and matter power spectrum , @xmath14 , are almost opposite to those of the dm de coupling , resulting in a strong degeneracy between the coupling strength @xmath15 and @xmath12 .
a recent analysis by means of monte carlo markov chain method has shown that a cosmology with significant @xmath12 and @xmath15 is statistically preferred to one with no coupling and almost massless neutrinos .
further , when priors on the neutrino mass from earth - based neutrino mass experiments ( heidelberg - moscow neutrinoless double @xmath15-decay , katrin tritium @xmath15-decay ) are added to the analysis , a @xmath16 detection of a dm - de coupling is found .
the plan of the paper is as follows : in section 2 we describe our model while the isw effect theory is reviewed in section 3 where we also discuss how the isw signal depends on the main parameters of the model . in section 4
we discuss galaxy bias and magnification bias .
comparison between theoretical prediction and observation data is presented in section 5 while a tomographic analysis is performed in section 6 .
section 7 is devoted to the conclusions .
we assume a spatially flat friedmann robertson walker ( frw ) background with metric @xmath17 ( @xmath18 is the conformal time ) filled with baryons , photons , neutrinos , dm and a component of de which will be ascribed to a scalar field @xmath7 self interacting through a potential @xmath8 . in the following , the indexes @xmath19 , @xmath20 , @xmath21 and @xmath7 will denote baryons , cold dm , massive neutrinos and de .
photons and massless neutrinos will be referred as radiation and denoted by @xmath22 .
in addition to self interaction we also consider a possible interaction between the scalar field and dm . here
we give only the equations for baryons , dm and de being the equations for the other components the usual ones ( see , e.g @xcite ) . the friedmann equation for the scale factor @xmath23 , the continuity equations for baryons and dm and the evolution equation for the scalar field read : @xmath24 where @xmath25 denotes the derivative with respect to @xmath18 , @xmath26 , @xmath27 is the energy density of the component @xmath28 and the constant @xmath29 parametrizes the dm de coupling strength . working in the conformal newtonian gauge
the metric of a perturbed flat frw universe takes the form : @xmath30 \label{metric}\ ] ] where @xmath31 plays the role of the newtonian potential , @xmath32 is the newtonian spatial curvature , @xmath33 and : @xmath34 \\
\psi=\phi-\sum_i{9 \over 2 } { { \cal h}^2\over k^2}(1+w_i ) \omega_i \sigma_i\end{aligned}\ ] ] where @xmath35 , @xmath36 , @xmath37 , @xmath38 and @xmath39 are the density parameter , density contrast , four velocity divergence , shear and state parameter of the component @xmath40 ( @xmath41 and @xmath42 being the pressure of the component @xmath40 and the critical density of the universe ) .
linear perturbation equations for dm and baryons read : @xmath43 where @xmath44 is the baryon sound speed , @xmath45 is the standard term describing momentum exchange with photons due to thomson scattering ( see , e.g ma & bertschinger 1995 ) and @xmath46 is the perturbation to the scalar field which evolves according to : @xmath47 as we are interested in the cross
correlation between isw effect and galaxy distributions , above equations can be simplified . at late time
radiation and massive neutrinos can be neglected so that no shear stresses are present and @xmath48 .
furthermore , the cross
correlation signal comes from scales well within the horizon , @xmath49 mpc , so that the second term in ( 6 ) can be neglected .
the above equations then reduce to the usual poisson equation for the gravitation potential : @xmath50,\ ] ] a modified jeans equation for dm and the usual one for baryons : @xmath51 \nonumber \\ \ddot\delta_b+ { \cal h } \dot\delta_b & = & { 3\over 2}{\cal h}^2\left[\omega_c\delta_c + \omega_b \delta_b + \omega_\nu\delta_\nu\right]=0,\end{aligned}\ ] ] and a poisson like equation for the scalar field perturbation : @xmath52\ ] ] where we have defined the dimensionless coupling parameter : @xmath53 ( @xmath54 is the planck mass ) . as clearly visible from the above equations and
widely discussed in @xcite , coupling affects the dynamics of dm particles . as a consequence baryons and
dm develop a bias @xmath55 , i.e. @xmath56 .
notice that , this bias , which origin is to ascribe to the coupling , is something completely different from the galaxy bias due to hydrodynamical effects , discussed in the subsequent sections .
it is also worth mentioning that , unlike the uncoupled case , in the presence of coupling , universe goes through an evolutionary phase named @xmath7matter dominated era ( @xmath7mde ) just after matter radiation equivalence . in this period
the scalar field @xmath7 behaves as @xmath57 ( @xmath58 ) having a non
negligible kinetic energy which dominates over the potential one . after this stage ,
the usual matter era follows before entering in the accelerated regime with a final de sitter attractor .
notice also that , because of the @xmath7mde and the non
usual scaling of the dm energy density , i.e. @xmath59 , after equivalence the background expansion law will differ from the usual @xmath60 .
[ f1 ] we shall consider the ratra peebles ( rp ) @xcite and sugra self interaction potentials
@xcite , reading @xmath61 and @xmath62 respectively ; they allow tracker solutions for any @xmath63 . for both potentials , once @xmath64 and @xmath0 are assigned , the present time de density parameter @xmath65 is uniquely defined . limits on these models without coupling between de and cdm have been studied in la vacca & kristiansen 2009 .
they find , that only quite small @xmath66 are allowed . in the sugra case in particular
only @xmath67 is allowed .
such small values are well below the range motivated by particle physics .
therefore the physical appeal of the sugra potential is spoiled .
let us however outline that , when the @xmath15 degree of freedom is opened , @xmath0 values as large as 30@xmath68gev become allowed , at the 1@xmath69 level , while , at the 2@xmath69 level , no significant constraint on the energy scale @xmath0 remains . even for the rp potential , for which a limit @xmath70 held , in the absence of coupling , values @xmath71 become allowed ( la vacca et al 2009 )
. in the absence of dm de coupling , rp yields
quite a slowly varying @xmath72 state parameter . on the contrary
, sugra yields a fast varying @xmath73 .
although coupling causes a @xmath73 behavior significantly different from the uncoupled case , one could again consider these potentials as examples of rapidly or slowly varying @xmath73 .
the results shown in the next two sections are qualitatively the same for rp and sugra models .
we will show them only for sugra cosmologies while the results for rp models will be shown only when comparing the theoretical predictions with observational data and dealing with redshift tomography .
isw effect arises when cmb photons from the last scattering surface pass through a time
dependent gravitational potential changing its energy so that additional temperature anisotropies are generated .
the isw temperature fluctuation , @xmath74 , is given by : @xmath75 where @xmath76 is the cmb temperature and @xmath77 is the visibility function of the photons .
as outlined in previous section , we will deal with scales within the horizon and redshifts such that radiation and any anisotropic stress can be neglected ( @xmath48 ) . in the matter era and in the absence of dm de coupling ( @xmath78 ) ,
poisson equation reads : @xmath79 ( @xmath80 and @xmath81 are the total matter density parameter and density contrast respectively ) from which one can appreciate that the gravitational potential stays constant , @xmath82 , and no isw effect arises .
however , when de starts to dominate the cosmic expansion , @xmath31 is no longer constant and isw effect generates secondary anisotropies in the cmb . on the other hand ,
as explained in the previous section , dm - de coupling affects both the background and density perturbation evolution , resulting in a variation of @xmath31 even during the matter domination .
figure 1 shows how @xmath15 and @xmath83 affect the time evolution of the sum @xmath84 which time derivative forms the source of the isw effect .
evolution of the gravitational potentials is obtained by a modified version of the public code cmbfast integrating the fully relativistic equations and taking into account the contributions of all of the components , i.e , photons , dm , baryons , neutrinos and de .
notice how @xmath15 and @xmath83 affect @xmath84 in an opposite fashion .
performing a measurement of the isw effect is , however , a difficult task because of its small signal compared with that of primary anisotropies ( @xmath85 times larger ) .
furthermore , while on small scales the small differences in temperature tend to cancel out , the large scales , from which the most isw effect contributes come from , are strongly affected by the cosmic variance
. the problem can be overcome by cross correlating the isw anisotropies with some tracers of the matter density , e.g. astrophysical objects like galaxies .
the observed galaxy density contrast in the direction @xmath86 is : @xmath87 where @xmath88 is the normalized selection function of the galaxy survey and @xmath2 is the galaxy bias , which will be discussed in the next section , relating the galaxy density contrast to the inhomogeneities in the mass distribution , @xmath89 .
since @xmath90 is related to the gravitational potential through the poisson equation , the observed galaxy density will be correlated with the isw anisotropies in the nearby direction @xmath91 : @xmath92 the 2point angular cross correlation function ( ccf ) and auto correlation function ( acf ) in the harmonic space are then defined as : @xmath93 \nonumber \\ c^{gal - gal}(\theta ) & \equiv & \left\langle\delta_{gal}(\hat{\bf n}_1)\delta_{gal}(\hat{\bf n}_2)\right \rangle \nonumber \\ & = & \sum^{\infty}_{l=2}\frac{2l+1}{4\pi}c^{gal - gal}_lp_l[\cos(\theta ) ] \label{ctheta}\end{aligned}\ ] ] where @xmath94 , @xmath95 s are the legendre polynomials and the cross and auto correlation power spectra are given by : @xmath96 ^ 2 } \label{cl}\end{aligned}\ ] ] where @xmath97 is the primordial power spectrum of scalar perturbations and the integrands @xmath98 and @xmath99 are : @xmath100dz}~,\\ i^{gal}_l(k)=\int{b(z)\frac{dn}{dz}(z)\delta_{\rm m}(k , z)j_l[k\chi(z)]dz}\end{aligned}\ ] ] ( here @xmath101 are the spherical bessel functions , and @xmath102 is the comoving distance ) . in
the following we use our modified cmbfast code to calculate the theoretical ccf and acf . in order to better understand the effects of @xmath0 , @xmath15 and @xmath83 on them
, we start to compute the isw - matter ccf and acf , @xmath103 and @xmath104 , and power spectra , @xmath105 and @xmath106 , the values of which are obtained similar to ( [ ctheta ] ) and ( [ cl ] ) by replacing @xmath107 and @xmath108 with : @xmath109dz}\end{aligned}\ ] ] where the only difference from @xmath110 and @xmath111 is the bias factor . here , we model @xmath112 as a narrow gaussian centered at two different redshifts , z=0.3 and z=3 .
this will permit to obtain some information about the time evolution of the correlations .
realistic selection functions will be considered in the next sections .
cosmological parameters are assumed to be the same as in the wmap 5 years best fit @xmath0cdm model @xcite .
let us consider first the uncoupled case ( @xmath78 ) and no massive neutrinos . figure [ f2 ] shows the redshift evolution of the gravitational potentials ( left panel ) and the source of isw effect @xmath113 ( right panel ) for different values of @xmath0 .
the dependence on @xmath0 of the isw - matter correlations @xmath114 and @xmath103 is then displayed in fig .
[ f3 ] at two different redshifts , @xmath115 and @xmath116 .
when increasing @xmath0 , both @xmath114 and @xmath103 show an opposite behavior at low and high redshifts .
this reflects the behavior of @xmath117 . in the presence of coupling one
can distinguish between two different behaviors for small and large @xmath0 s .
this is shown in figs .
[ f4 ] and [ f5 ] for @xmath118 . in the first case , the evolution of the gravitational potentials and the cross correlation signal are almost independent from @xmath0 .
it can be understood noticing that for small @xmath0 , @xmath7mde is very long and the usual tracker solution is ( almost ) never reached . in this phase , coupling terms in the field equations dominate so that the tracker solution is almost independent from @xmath0 .
when increasing @xmath0 , @xmath7mde becomes shorter and the behavior resemble that of the uncoupled case .
the transition between the two above regimes occurs around @xmath119 gev .
dependence on @xmath15 is shown in figs [ f6 ] and [ f7 ] .
again , the behavior of the cross
correlation reflects that of the isw source .
however , while coupling can have opposite effects on the cross - correlation at different redshifts , i.e. it can increase or decrease the signal , massive neutrinos always decrease the signal .
it is shown in fig .
[ fnu ] which displays the behavior of @xmath114 as a function of @xmath83 at two different redshifts .
cross - correlation signal always decreases at the increasing of the neutrino mass .
the galaxy bias @xmath19 can , in general , evolve both in redshift or as a function of the scale .
however , on the large scales of interest for the isw effect , the bias is usually assumed to be linear , spatially constant and only redshift dependent , i.e. @xmath120 .
this assumption is fully consistent with results from numerical simulations , redshift surveys and semi analytic calculation in the contest of the so called halo model ( see @xcite ) .
however , given a galaxy selection function @xmath112 picked at certain redshift @xmath121 , the bias can be approximated with an appropriated constant . in this case
it will be @xmath122 within the above approximation , given a particular survey and assumed a cosmological model , the bias is usually estimated by fitting the theoretical matter
matter correlation function , @xmath123 , for the assumed cosmology , to the observed galaxy galaxy correlation function , @xmath124 .
biases have been estimated for different surveys by several authors assuming the wmap best fit @xmath0cdm cosmology ( see @xcite ) .
since we are considering cosmological models different from a @xmath0cdm , we need to appropriately rescale those biases to each of the our models .
note , however , how the estimation of @xmath19 in ( [ bias ] ) depends on the normalization of the power spectrum in @xmath123 ( see ( [ cl ] ) ) . for a fixed normalization ,
taking into account ( [ bias ] ) , biases will be rescaled according to : @xmath125 where @xmath126 indicates the average on the angular scales @xmath127 of interest .
however , ( [ rescal ] ) should be generalized when magnification bias effect due to gravitational lensing is non negligible .
gravitational lensing by intervening matter changes the observed galaxy number density @xmath128 , leading a correction term @xmath129 being added to the intrinsic galaxy fluctuation @xmath110 @xmath130 with this correction , the observed acf becomes : @xmath131 where @xmath132 and the rescaled bias will then be the solution of : @xmath133 auto and cross
correlations corrected for magnification bias are obtained considering in ( 25 ) the function @xcite : @xmath134 where @xmath135 is the slope of the number counts of galaxy number density as a function of the flux , @xmath136 .
it depends on the choice of galaxy sample and is redshift dependent . the lensing window function ( in a flat universe )
is : @xmath137 magnification bias increases with redshift and could be important when dealing with deep survey , e.g. quasars .
this is shown in fig .
[ mag ] which compares the effect of the magnification bias on the isw - gal correlation at @xmath138 and @xmath116 . a detailed analysis on how magnification bias affects
acf and ccf can be found in @xcite .
.best fit values and @xmath139 error bars for rp and sugra models . only upper limits on @xmath140 , @xmath15 and @xmath0 are shown . [ cols="<,^,^",options="header " , ] further , we assume @xmath141 as expected from future experiments , and consider three different splitting schemes , each with @xmath142 bins in the redshift range from @xmath143 to @xmath144 : \(i ) bins equally spaced in @xmath1 , with @xmath145 ; \(ii ) same number of galaxies , @xmath146 , in each bin ; \(iii ) bin sizes increasing proportionally to the photometric error , @xmath147 .
the three splitting schemes are shown in fig .
the thick line is the overall quasar distribution while the other curves are the true ( spectroscopic ) distributions that correspond to the divisions ( vertical lines ) in photo - z space .
we also take into account magnification bias effect which can be important at higher @xmath1 . in a recent analysis by @xcite , the slope of the quasar counts , @xmath64 , entering in ( [ magb ] ) ,
was found to be redshift dependent .
they found @xmath148 in the photometric redshift range @xmath149 and @xmath150 for @xmath151 .
for simplicity , we assume a constant slope @xmath150 as in @xcite . before comparing the different models we discuss the effects of ignoring magnification bias and quasar bias evolution .
this is done for the @xmath0cdm cosmology .
results , however , are valid for rp and sugra as well .
notice that implications of galaxy bias evolution on isw measurements and parameter estimation has been considered in a pioneer work of schaefer et al 2009 . for the quasar bias evolution in @xmath0cdm
, we use the empirical formula derived by @xcite : @xmath152 which provides a good fit of the recent observational findings by @xcite .
we first consider the overall quasar distribution and calculate the expected signal to noise ratio , _ snr _ , of the cross correlation in the following cases : constant quasar bias @xmath153 ( where @xmath121 is the mean redshift of the survey ) and @xmath154 as given by ( [ biasfit ] ) .
for each of them , the _ snr _ is obtain by neglecting or considering the magnification bias correction .
results are summarized in table [ mb ] . for gaussian fileds , the expected @xmath3
is given by ( see e.g. @xcite ) : @xmath155 where @xmath156 is the noise contribution to the isw ( @xmath157 and @xmath158 being the total anisotropy contribution and any detector noise contribution , negligible on the scales we are interested in ) while @xmath159 , is the shot noise associated with the galaxy / quasar catalog ( @xmath160 is the surface density of galaxies / quasars per steradians ) .
@xmath161 is the fraction of sky common to cmb and galaxy / quasar survey maps .
note that ( [ stn ] ) is strictly true only in the case of gaussian fields and full sky coverage , @xmath162 . in this section ,
we assume that to be the case other than a negligible @xmath163 .
these assumptions will be relaxed in the following .
anyway , for partial sky coverage , one can , in first approximation , multiply for @xmath161 the values of @xmath3 here presented . the value of the lowest multipole , @xmath164 , can be approximately set to @xmath165 in order to account for the loss of low multipole modes .
anyway , although a significant part of the isw signal comes from lower multipoles , we set @xmath166 in order to avoid effects of gauge correction on very large scales recently discussed by @xcite even if this implies a reduction of @xmath3 .
the maximum multipole , @xmath167 , is set to @xmath168 .
however , we will show later that the contribution to @xmath3 from @xmath169 multipoles is negligible . from table [ mb ]
, it follows that cross correlation measurements are more affected by errors when ignoring the quasar bias evolution rather than the magnification bias correction .
in fact , in this last case , an error of @xmath170 on _ snr _ is obtained while in this last case while the error raises to @xmath171 if the quasar bias is approximated by a constat value , @xmath172 ( @xmath121 being the mean redshift of the quasar distribution ) and it is of @xmath173 if both quasar bias evolution and magnification bias are neglected .
we now turn to tomography . according to ( [ stn ] ) , the signal to noise ratios , @xmath174 , in the _
i_th tomographic bin are calculated for each of the cases listed in table [ mb2 ] .
[ snrbm ] shows the errors @xmath175 relative to the case @xmath176 for each bin of the three splitting schemes described above . unlike the overall quasar distribution , in each bin , distributions are very narrow and the quasar bias can be approximated by a constant , @xmath177 ( @xmath178 being the mean redshift of the @xmath40th bin ) , leading to only a minor error @xmath179 on @xmath174 ( case @xmath180 ) . on the other hand , ignoring magnification bias correction might be critical at high @xmath1 ( @xmath181 , case @xmath182 ) and if a constant bias , @xmath153 , is used for all bins , the error can reach @xmath183 at higher redshifts
. however , since photometric redshift errors cause the bins to overlap and magnification of the galaxies at @xmath184 probes structures at @xmath185 , cross correlation measurements at high @xmath1 are quite correlated with those at low redshifts .
taking into account such correlations , the net accumulated @xmath3 for measurements from all the bins up to @xmath186 is given by : @xmath187_{ij}c_{l}^{isw - gal_j } \label{snrcov}\ ] ] + @xmath188_{ij } = { { c_{l}^{isw - gal_i}c_{l}^{isw - gal_j}+\left(c_l^{isw}+c_l^{n_{isw}}\right ) \left(c_l^{gal_i}c_l^{gal_j}+\delta_{ij}c_l^{n_{gal_i}}\right ) \over \left[f_{sky } \left(2l+1\right)\right]^{-1 } } } \label{cov}\ ] ] + the relative error on @xmath189 is shown in fig .
[ snrbmc ] for the same cases as in fig .
[ snrbm ] . in all cases , despite the significant differences in @xmath174 in high @xmath1 bins , errors are always @xmath190 .
this is understood because most of the cumulative _ snr _ comes from low redshifts .
however , in the case @xmath176 , errors considerably increase if the shot noise associated to the catalog is non negligible . fig . [ snrbmnoise ] compares the relative errors on @xmath174 ( top panel ) and @xmath189 ( bottom panel ) in the case of negligible shot noise and assuming a sdss dr6like survey with a number density of quasars , @xmath160 , of @xmath191 . in this last case
, the error on the cumulative @xmath3 ranges from @xmath192 , to @xmath193 , depending on the number of bins considered while the error on @xmath174 can also reach the @xmath194 in bins at higher redshift . in the previous section we have shown that when dealing with narrow tomographic bins , we can approximate the galaxy bias in each bin with a constant committing an error of no more than a few percent . then , after calculating the cross correlation power spectra for the best fit @xmath0cdm , sugra and rp models we apply ( [ rescalmag ] ) to each bin of the three splitting schemes .
the rescaled biases @xmath195 and @xmath196 are then fitted with an expression similar to ( [ biasfit ] ) obtaining : @xmath197 for rp and @xmath198 for sugra .
these expressions , which will be used in the rest of the paper , are valid up to @xmath144 leading a galaxy bias evolution only slightly different from that of @xmath0cdm . in fig .
[ f12 ] , we show the cross
correlation power spectra for the splitting ( i ) .
lower frames of each panel display the ratio between the sugra and rp spectra and the @xmath0cdm spectrum .
mean redshifts of the true bin distributions are also indicated .
a similar redshift evolution could be obtained by considering ( ii ) or ( iii ) . as clearly visible from the figure , a better discrimination among the models
is expected at higher redshifts .
the expected @xmath174 and @xmath189 for the different splitting schemes are shown in figs .
[ snr ] and [ snrc ] respectively , for the cosmologies considered . despite the same qualitative behavior for all the models , higher @xmath3 are expected in the sugra case ; rp being in the between of sugra and @xmath0cdm .
horizontal lines in fig .
[ snrc ] indicate the @xmath3 obtained by using the full distribution . for each splitting scheme , in figs .
[ snr2 ] , we plot the ratio between the rp and sugra @xmath174 and @xmath189 values and those expected in the @xmath0cdm case .
a first thing to note is the overlapping , for both rp and sugra , of the three curves corresponding to the different splittings indicating that the three schemes considered perform equally well in discriminating among the models .
a better discrimination is however achieved looking at high redshifts . in fig .
[ snr3 ] , we investigate how @xmath174 changes when increasing the number of the bins ( left panel ) and whether a greater number of bins could permit to better discriminate between models ( right panel ) .
results are shown for the splitting scheme ( iii ) with 2,3,5,10 bins in the rp model .
similar results are obtained in the other cases .
@xmath3 in the case of no splitting is also shown .
more bins , in principle , would permit to have a more detailed description of the redshift evolution of isw effect .
however , as clearly visible in the left panel , the @xmath3 decreases at the increasing of the bin number . in the right panel , rp model is compared to @xmath0cdm . even though , at high redshifts , tomography permits to distinguish among the models better than using the full distribution , the figure shows that increasing the number of the bins from 2 to 10 would permit only a minor improvement in discriminating between the models . up to now
, we have considered the ideal case of gaussian fields , full sky coverage and negligible shot noise .
however , after cutting out our galaxy from the analysis , future cmb and galaxy maps are expected to cover , at best , a sky fraction @xmath199 . in this case
different multipoles are no longer independent and ( [ stn ] ) , ( [ snrcov ] ) and ( [ cov ] ) only provide approximated estimations and a more rigorous analysis taking into account for the effective survey geometry is needed ( @xcite , @xcite , @xcite ) .
it has been however shown that , under the above approximations , a better estimation can be obtained by binning the power spectra data in bins of appropriate size @xmath200 making the bins independent . in this case , ( [ stn ] ) and ( [ snrcov ] ) are increased by a multiplicative factor of @xmath200 .
@xcite found that @xmath201 works well for @xmath202 . in figs .
[ snrl ] and [ snrcontour ] we show some results for the cumulative @xmath3 from all the bins of the splitting ( iii ) .
results are however the same for the other schemes . the left panel of fig .
[ snrl ] shows the contribution to the cumulative @xmath3 from each multipole while the cumulative @xmath3 up to @xmath203 is displayed in the right panel .
the dependence on @xmath161 and @xmath204 has been removed . as clearly visible ,
most of the cross - correlation signal comes from lower multipoles and contributions from @xmath169 are negligible . in fig .
[ snrcontour ] cumulative @xmath3 contour levels are plotted in the plane @xmath205 where @xmath206 is the quasar number density for a sdss dr6-like survey .
given that such a survey cover @xmath207 of the sky , future experiments covering a sky fraction @xmath208 will increase the cumulative @xmath3 of a factor @xmath209 if the shot noise is reduced by @xmath210 and a factor @xmath211 in the case @xmath212 .
no significant improvement is obtained by further reducing the shot noise . for
@xmath208 and negligible shot noise , the increasing of @xmath3 in the @xmath40th tomographic bin can range from a factor of @xmath213 ( low @xmath1 bins ) up to @xmath214 ( high @xmath1 bins ) .
this is shown in fig .
in this work we have investigated isw - lss cross - correlation in coupled dark energy models with massive neutrinos .
the presence of a coupling between dm and de as well as massive neutrinos change both the background and matter perturbation evolutions yielding , unlike the @xmath0cdm case , a time variation of the gravitational potentials even during the matter domination .
a significant isw signal is thus expected also at high redshifts .
firstly , we have investigated the dependence on the energetic scale , @xmath0 , of the de potential , the coupling strength @xmath15 and the neutrino mass @xmath83 .
we considered first the uncoupled case ( @xmath78 ) and massless neutrinos .
we found that , when increasing @xmath0 , both @xmath114 and @xmath103 show an opposite behavior at low and high redshifts .
this in fact reflects the behavior of @xmath117 . in the presence of coupling one
can distinguish between two different behaviors for small and large @xmath0 s .
in the first case , the evolution of the gravitational potentials and the cross correlation signal are almost independent from @xmath0 .
it can be understood noticing that for small @xmath0 , coupling terms in the de field equations dominate so that its solution is almost independent from @xmath0 .
when increasing @xmath0 , the behavior resemble that of the uncoupled case .
dependence on @xmath15 was also investigated and , again , the behavior of the cross correlation at different redshifts reflects that of the isw source
. however , while coupling can affect @xmath114 ( and @xmath103 ) in an opposite fashion at high and low redshifts , massive neutrinos always decrease the cross
correlation signal .
secondly , we have provided a simple expression , eq .
( [ rescal ] ) , which permits to appropriately rescale the galaxy bias when comparing different cosmologies once the bias of a particular model , e.g. @xmath0cdm , is known and the normalization of the power spectrum in each model is fixed .
we also give , a generalized version of ( [ rescal ] ) to the case when the magnification bias effect due to gravitational lensing is non negligible ( see eq .
( [ rescalmag ] ) ) .
then , we compare the theoretical prediction on the cross
correlation function for our models with the observational data obtained for six different galaxy catalogues by @xcite .
we found that , while it is not possible to distinguish among the models at low redshifts , discrepancies between coupled models and @xmath0cdm increase with @xmath1 even though rp and sugra models remain indistinguishable . in spite of this , however , current data alone seem not able to discriminate between coupled models and @xmath0cdm .
finally , we studied the redshift tomography .
upcoming galaxy surveys will cover a large redshift range also providing photometric redshifts of the galaxies with high accuracy .
this will permit to split a survey into multiple photometric redshift bins allowing for tomographic analysis . here , we were interested to study how a tomographic analysis of the isw
lss cross
correlation is affected when considering different splitting schemes and assuming photometric redshift errors as expected from future experiments . as we were interested in high redshifts , where our models , unlike the @xmath0cdm case
, are expected to provide a significant isw effect signal , isw effect was cross correlated with quasars .
the quasar distribution was thus split in tomographic bins according to three different schemes : ( i ) bins equally spaced in @xmath1 ; ( ii ) same number of galaxies , in each bin ; ( iii ) bin sizes increasing proportionally to the photometric error .
cross - correlation were then calculated in each bin .
our tomographic study was based on a signal to noise analysis .
we started our discussion investigating the effect on @xmath3 of ignoring the quasar bias evolution and magnification bias correction for an ideal survey .
we found that , if the overall quasar distribution is used , cross correlation measurements are more affected by errors when ignoring the quasar bias evolution ( @xmath215 ) rather than the magnification bias correction ( @xmath216 ) however , when dealing with tomography the error on @xmath174 ( @xmath40 indicating the @xmath40th bin ) never overcome @xmath217 if the quasar bias in each bin is approximated with an appropriated constant , but it can reach @xmath218 , at high redshifts , when magnification bias is ignored .
errors on the cumulative @xmath3 , however , always stay below the @xmath219 . on the other hand errors
can increase up to @xmath220 if the shot noise associated to the quasar survey is set to the current values .
we then used the tomographic analysis in order to compare different cosmologies .
we found that the above splitting schemes , perform equally well in discriminating among the models .
a better discrimination is however achieved looking at high redshifts .
we also investigated how the expected signal to noise ratio , @xmath3 , of the cross
correlation changes when increasing the number of the bins and whether a greater number of bins could permit to better discriminate between models .
even though more bins would allow to have more information on the redshift evolution of the isw effect , the @xmath3 decreases at the increasing of the bin number . as a consequence , although tomography , at high redshifts , would permit to distinguish among the models better than using the full distribution , when comparing our models to @xmath0cdm it was shown that increasing the number of the bins from 2 to 10 would permit only a minor improvement in the discrimination .
finally , we showed that future wide field surveys ( @xmath221 ) can increase the cumulative @xmath3 of the cross correlation of a factor @xmath222 ( @xmath223 ) if the current shot noise is reduced by @xmath210 ( @xmath224 ) while the @xmath3 of the single bins can increase up to a factor @xmath214 at high redshift .
our @xmath3 analysis suggest a discrimination power of future isw
lss cross - correlation measurements able to distinguish among different cosmologies .
however , in order to assess the discrimination , more rigorous analysis in terms of fisher matrix and monte carlo markov chain are needed .
they are currently under investigation and left for future works .
tommaso giannantonio is gratefully thanked for providing observational data on cross correlation and useful hints .
dfm thanks the research council of norway frinat grant 197251/v30 and the abel extraordinary chair ucm - eea - abel-03 - 2010 .
dfm is also partially supported by the projects cern / fp/109381/2009 and ptdc / fis/102742/2008 . | we provide an exhaustive analysis of the integrated sach wolfe effect ( isw ) in the context of coupled dark energy cosmologies where a component of massive neutrinos is also present .
we focus on the effects of both the coupling between dark matter and dark energy and of the neutrino mass on the cross - correlation between galaxy / quasar distributions and isw effect .
we provide a simple expression to appropriately rescale the galaxy bias when comparing different cosmologies .
theoretical predictions of the cross
correlation function are then compared with observational data .
we find that , while it is not possible to distinguish among the models at low redshifts , discrepancies between coupled models and @xmath0cdm increase with @xmath1 . in spite of this
, current data alone seems not able to distinguish between coupled models and @xmath0cdm .
however , we show that upcoming galaxy surveys will permit tomographic analysis which allow to better discriminate among the models .
we discuss the effects on cross - correlation measurements of ignoring galaxy bias evolution , @xmath2 , and magnification bias correction and provide fitting formulae for @xmath2 for the cosmologies considered .
we compare three different tomographic schemes and investigate how the expected signal to noise ratio , @xmath3 , of the isw
lss cross correlation changes when increasing the number of tomographic bins .
the dependence of @xmath3 on the area of the survey and the survey shot noise is also discussed . |
the evolution of galaxy clustering is a prediction of hierarchical models of galaxy and structure formation ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
hierarchical models for a concordance cosmology , @xmath11 , @xmath12 , and comoving coordinates . ] predict little or no evolution of the clustering of @xmath13 red galaxies at @xmath14 .
precise measurements of galaxy clustering at @xmath4 can therefore test the predictions of these models .
extremely red objects ( eros ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) could be the progenitors of local ellipticals ( e.g. , * ? ? ?
roughly 80% of @xmath15 eros have spectra with the absorption features of old stellar populations @xcite and @xmath16 of @xmath17 eros have early - type morphologies @xcite .
some eros contain super - massive black holes , as @xmath18 of eros contain an obscured active galactic nucleus which can be detected by deep x - ray surveys @xcite .
a direct test of the relationship between @xmath4 eros and the most massive local ellipticals is to compare the spatial clustering of the two populations .
previous constraints on the spatial correlation function of eros , summarized in table [ table : prev ] , are provided by pencil - beam surveys with @xmath19 areal coverage each .
individual structures comprised of eros can have sizes comparable to the field of view of these surveys ( e.g. , * ? ? ?
* ) , and small surveys do not sample representative volumes of the universe for highly clustered objects ( e.g. , * ? ? ?
* ) . at @xmath4 ,
the transverse comoving distance spanned by previous ero studies is @xmath20 , which is much smaller than the size of individual structures observed in the present - day universe .
spatial clustering measurements derived from the angular correlation function depend on ero redshift distribution models .
previous angular clustering studies were unable to verify their model redshift distributions , as complete spectroscopic samples of eros were unavailable .
previous ero spatial clustering measurements have large uncertainties and possibly large ( and sometimes unaccounted for ) systematic errors . in this paper
, we present a measurement of the clustering of eros using @xmath3 imaging of a @xmath2 subset of the noao deep wide - field survey ( ndwfs ) . the large area of our study provides a more representative volume than previous studies .
the @xmath3 passbands of the ndwfs allow us to constrain the ero redshift distribution with photometric redshifts and their uncertainties .
we also use photometric redshifts to select eros as a function of luminosity and redshift .
we use ero spectroscopic redshifts to verify the accuracy of our photometric redshifts and we compare our estimate of the ero redshift distribution with spectroscopic redshift distributions from the literature . the structure of the paper is as follows . in [ sec : ndwfs ] we provide a brief description of the ndwfs imaging and catalogs from which the @xmath0 ero sample was selected .
we discuss our estimates of ero photometric redshifts , and provide a comparison of ero photometric and spectroscopic redshifts in [ sec : photoz ] .
the selection of the ero sample and ero number counts are discussed in [ sec : ero ] . in [ sec : cor ] , we describe the techniques used to measure the angular and spatial correlation functions . the angular and spatial clustering of eros , as a function of apparent magnitude , apparent color , absolute magnitude , and redshift are discussed in [ sec : clu ] .
we discuss the implications of our results in [ sec : dis ] and summarize the paper in [ sec : sum ] .
the ndwfs is a multiband ( @xmath21 ) survey of two @xmath22 high galactic latitude fields with the ctio @xmath23 , kpno @xmath24 , and kpno @xmath25 telescopes @xcite .
a thorough description of the optical and @xmath26-band observing strategy and data reduction will be provided by jannuzi et al . and dey et al .
( both in preparation ) .
this paper utilizes @xmath2 of @xmath3 data in the botes field .
@xmath27 imaging and catalogs for the entire ndwfs botes field became available from the noao science archive on 22 october 2004 .
@xmath26-band imaging and catalogs for approximately half of the botes field are also available from this archive .
we generated object catalogs using sextractor @xmath28 @xcite , run in single - image mode in a similar manner to @xcite .
at faint magnitudes , detections in the different bands were matched if the centroids were within @xmath29 of each other . at bright magnitudes , detections in the different bands
were matched if the centroids were within an ellipse defined using the second order moments of the light distribution of the object , @xmath30 , and @xmath31 . ] . throughout this paper
we use sextractor mag_auto magnitudes ( which are similar to kron total magnitudes ; * ? ? ?
* ) , due to their small uncertainties and systematic errors at faint magnitudes .
our clustering measurements are not particularly sensitive to how we measure ero photometry , and the clustering of eros selected with @xmath32 diameter aperture photometry is only marginally weaker than the clustering of eros selected with mag_auto photometry .
we determined the completeness as a function of magnitude by adding artificial objects to copies of the data and recovering them with sextractor . to approximate @xmath4 galaxies , the artificial objects have an intrinsic profile with a full width at half maximum of @xmath33 , which was then convolved with a moffat profile model of the seeing .
the @xmath34 completeness limits vary within the sample area in the ranges of @xmath35 , @xmath36 , @xmath37 and @xmath38 .
regions surrounding saturated stars were removed from the catalog to exclude ( clustered ) spurious objects detected in the wings of the point spread function .
we excluded regions where the root - mean - square of the sky noise in the @xmath26-band data was @xmath39 higher than the mean , as the depth of these regions is significantly less than the mean depth across the field . while it is plausible that smaller variations in the sky noise could alter the measured clustering of the faintest eros , our main conclusions remain unchanged if we exclude @xmath40 eros from the sample .
we used sextractor s star - galaxy classifier to remove objects from the galaxy catalog which had a stellarity of @xmath41 in 2 or more bands brighter than @xmath42 , @xmath43 , and @xmath44 . at fainter magnitudes
we do not use the star - galaxy classification and correct the angular correlation function for the estimated stellar contamination of the sample .
we do not use the @xmath26-band for star - galaxy classification as there are image quality variations across the @xmath26-band image stacks .
we estimated stellar contamination of the galaxy sample using the same technique as @xcite , where the stellar number counts were assumed to be a power - law and the distribution of stellar colors does not change with magnitude at @xmath45 . the contamination of the ero sample ( [ sec : ero ] ) by stars is estimated to be @xmath46 , and the conclusions of this paper remain unaltered unless stellar contamination is higher than @xmath47 .
photometric redshifts were determined for all objects with @xmath48 and @xmath26-band detections .
we provide a brief overview of the photometric redshifts here and refer the reader to our earlier study of @xmath49 red galaxy clustering in the ndwfs @xcite for a more detailed description of the photometric redshift code . to model galaxy spectral energy distributions ( seds ) , we used pegase2 evolutionary synthesis models @xcite with exponentially declining star formation rates ( @xmath50 models ) and @xmath51 ages of @xmath52 ( formation @xmath53 ) .
the effect of @xmath54 dust reddening with @xmath55 , comparable to estimates for @xmath56 early - type galaxies @xcite , was included in the @xmath50 models . in @xcite , we used models with solar metallicity at @xmath51 , which resulted in small systematic underestimates of galaxy redshifts .
simple solar metallicity @xmath50 models underestimate the uv luminosity of galaxies ( e.g. , * ? ? ?
* ) , so in this work we let the metallicity of the models be a function of @xmath50 .
this has the effect of slightly increasing the uv flux of the model seds .
we verified the accuracy of the photometric redshifts at @xmath8 with 89 @xmath57 galaxies with rest - frame @xmath58 and spectroscopic redshifts .
after decreasing the metallicity of the models , the photometric redshifts of these red galaxies did not have significant systematic errors .
we note , however , that the uv flux in galaxies can also be increased by the presence of young stars or by altering the properties of the dust extinction ; our approach is merely a proxy for correcting any systematic effects in our photometric redshifts and is not meant to be interpreted as justifying sub - solar metallicities in the red galaxy population .
we use these solar and sub - solar @xmath50 models throughout the remainder of the paper .
color - tracks for 2 of the models are shown in figure [ fig : colmod ] .
for comparison , we also show two ultra - luminous infrared galaxy ( ulirg ) templates from @xcite , which have bluer @xmath59 colors than the @xmath50 models at @xmath4 .
photometric redshifts were estimated by finding the minimum value of @xmath60 as a function of redshift , spectral type ( @xmath50 ) , and luminosity . for objects not detected in the @xmath61 or @xmath62-bands , we estimated the probability of a non - detection using the completeness estimates discussed in [ sec : ndwfs ] . as the model seds do not account for the observed width of the galaxy locus
, we increased the photometric uncertainties for the galaxies by @xmath63 magnitudes ( added in quadrature ) . to improve the accuracy of the photometric redshifts , the estimated redshift distribution of galaxies as a function of spectral type and apparent magnitude was introduced as a prior . the 2df galaxy redshift survey ( 2dfgrs ) luminosity functions for different spectral types @xcite , with spectral evolution given by the @xmath50-models , were used to estimate the redshift distributions .
we tested the reliability of the photometric redshifts with simulated galaxies and real galaxies with spectroscopic redshifts .
simulated galaxies were generated using the pegase2 @xmath50 models .
the simulated data consisted of @xmath0 galaxies with @xmath64 in the redshift range @xmath65 and luminosity range @xmath66 .
the simulated object photometry was scattered using the estimated uncertainties , thus mimicking what would be present in the real catalogs .
we tested the accuracy of the photometric redshifts with a few spectroscopic redshifts and @xmath3 photometry for eros in the ndwfs botes field .
a comparison of our photometric and spectroscopic redshifts for eros is shown in figure [ fig : photoz ] .
we discuss the selection criteria for the eros in [ sec : ero ] .
the simulated galaxies in the left panel of figure [ fig : photoz ] have @xmath67 uncertainties of @xmath68 . for galaxies with seds similar to the pegase2 @xmath50 models
, our procedure should yield accurate photometric redshifts .
there are 4 ndwfs @xmath0 eros with spectroscopic redshifts .
as shown in the right hand panel of figure [ fig : photoz ] , real eros in the ndwfs exhibit a @xmath67 scatter of @xmath69 between the photometric and spectroscopic redshifts .
there are more outliers than would be expected if the @xmath50 models reproduced the variety of ero seds . for comparison , goods obtains ero photometric redshifts with accuracies of @xmath70 @xcite , as they have photometry and upper limits in more bands ( @xmath71 ) .
the accuracy of photometric redshifts is a complex function of redshift , sed and apparent magnitude .
the accuracy of the ero photometric redshifts could not be extrapolated from @xmath8 red galaxy photometric redshifts , which can have uncertainties of @xmath72 ( e.g. , * ? ? ?
the accuracy of ero photometric redshifts can not , and should not , be extrapolated from other samples of galaxies , such as samples selected by apparent magnitude only or from the hubble deep fields ( hdfs ) . though our ero photometric redshifts can only be considered approximations , they provide a good estimate of the ero redshift distribution ( see [ sec : cor ] ) .
we selected eros with the @xmath1 criterion ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , though redder color cuts are sometimes used in the literature ( e.g. , * ? ? ?
* ; * ? ? ?
we have limited the sample to @xmath0 eros , to reduce the effects of completeness variations across the survey area on the measured clustering .
as shown in figure [ fig : colmag ] , the percentage of eros increases from @xmath73 of the total galaxy counts at @xmath74 to @xmath68 at @xmath75 .
contamination of the ero sample by other galaxies could significantly alter the measured correlation function . at the magnitude limit of our sample ,
the uncertainty in the @xmath76 color is @xmath77 magnitudes .
for the distribution of galaxy colors shown in figure [ fig : colmag ] , and assuming gaussian photometric uncertainties , approximately 6% of the @xmath0 ero sample is contamination by @xmath78 galaxies .
even if @xmath78 galaxies were completely ( and implausibly ) unclustered , the amplitude of the @xmath1 angular correlation function would only be decreased by 12% .
contamination by @xmath79 galaxies could be as high as 22% in the @xmath0 ero sample .
this would significantly alter our results if the clustering of galaxies is a very strong function of color at @xmath80 .
however , as discussed in
[ sec : color ] , we do not see evidence of this within our dataset .
@xcite bias does increase the observed number of eros .
if we assume the ero number counts in table [ table : ang ] are a good approximation of the true ero number counts , then the contribution of malmquist bias to the ndwfs counts is @xmath81 .
this would alter the measured clustering if ero angular clustering is an extremely strong function of apparent magnitude .
we assume the bulk of our sample consists of galaxies with red stellar populations .
the colors of dusty starbursts are predicted to differ significantly from galaxies with red stellar populations .
as shown in figure [ fig : colcol ] , 77% of the ndwfs ero sample has @xmath82 , which is redder than the @xcite non - evolving ulirg templates shown in figure [ fig : colmod ] .
our assumption that most @xmath0 eros have red stellar populations is also consistent with the conclusions of @xcite , who find 86% of @xmath83 eros have the absorption features of old stellar populations .
the final sample consists of 671 objects , of which 318 are detected in the @xmath62-band and 635 are detected in the @xmath61-band .
the @xmath84 eros have photometric redshifts in the range @xmath85 , with the median of the distribution at @xmath86 . only 5 of the 671 @xmath84 eros have photometric redshifts of @xmath87 .
ero number counts as a function of @xmath26-band limiting magnitude are provided in table [ table : ang ] and figure [ fig : counts ] , along with results from previous surveys .
we evaluated the uncertainties of the sky surface density , for the ndwfs and previous work , using the method discussed by @xcite , which includes the contribution of large - scale structure .
the contribution of clustering to the uncertainties is typically several times larger than uncertainties determined by poisson statistics . for our @xmath84 ero sample ,
accounting for the clustering increases the @xmath67 uncertainty from 5% to 20% !
we note that the uncertainties quoted by some studies do not include this contribution ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the distribution of the ero sample on the plane of the sky is shown in figure [ fig : dist ] .
ero surveys of @xmath88 often have individual structures with sizes comparable to the field of view ( e.g. , * ? ? ?
while clustering and voids are evident in figure [ fig : dist ] , there are no obvious @xmath89 structures or gradients in the distribution of eros in our sample .
we determined the angular correlation function using the @xcite estimator : @xmath90 where @xmath91 , @xmath92 , and @xmath93 are the number of galaxy - galaxy , galaxy - random and random - random pairs at angular separation @xmath94 .
the pair counts were determined in logarithmically spaced bins between @xmath95 and @xmath96 .
we employed the same methodology as @xcite to generate random object catalogs , correct for the integral constraint @xcite , and estimate the covariance of the @xmath97 bins ( using the technique of * ? ? ? * ) .
the random object catalog contains 100 times the number of objects as the ero catalog , so @xmath92 and @xmath93 are renormalized accordingly .
the angular correlation function was assumed to be a power - law given by @xmath98 where @xmath99 is a constant .
this is a good approximation of the observed galaxy spatial correlation function from the 2dfgrs and sloan digital sky survey ( sdss ) on scales of @xmath100 @xcite . throughout this paper
we assume @xmath101 , the approximate value of @xmath99 for @xmath102 red galaxies from the 2dfgrs and sdss surveys . for a @xmath101 power - law ,
the integral constraint for this study was approximately @xmath103 of the amplitude of the correlation function at @xmath104 .
pair counts and the estimate of the angular correlation function ( including the integral constraint correction ) for @xmath105 and @xmath106 eros are presented in table [ table : pairs ] .
the spatial correlation function was obtained using the @xcite equation ; @xmath107 dz \left/ \left ( \int^\infty_0 \frac{dn}{dz } dz \right)^2 \right.\ ] ] where @xmath108 is the redshift distribution without clustering , @xmath109 is the spatial correlation function and @xmath110 is the comoving distance between two objects at redshifts @xmath111 and @xmath112 separated by angle @xmath113 on the sky .
the spatial correlation function was assumed to be a power law given by @xmath114^{-\gamma}. \label{eq : spa}\ ] ] we estimated the redshift distribution for the sample by summing the redshift likelihood distributions of the individual galaxies in each subsample .
model redshift distributions for subsamples selected by apparent magnitude and photometric redshift are shown in figure [ fig : dndz ] . while the individual photometric redshifts are not especially accurate , they do include information provided by the observed ero photometry and are likely to provide a fair approximation of the ero redshift distribution .
redshift distribution models which only reproduce the apparent ero number counts and local galaxy luminosity functions ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) have fewer constraints and may have larger systematic errors .
the estimated median redshift of the @xmath0 eros is 1.18 , which is almost identical to the spectroscopic median redshift of 24 @xmath15 eros from @xcite .
the median redshift is also similar to @xmath115 eros in the k20 spectroscopic sample ( * ? ? ?
* a. cimatti 2003 , private communication ) .
we measured the angular and spatial correlation functions for a series of apparent magnitude , apparent color , absolute magnitude , and redshift bins .
a power - law of the form @xmath116 was fitted to the data with @xmath99 fixed to @xmath117 .
much larger imaging surveys , including the completed ndwfs botes and cetus fields , will have sufficient area to accurately measure @xmath99 .
when parameterizing the power - law fits , we use @xmath118 instead of @xmath119 as it depends less on the assumed value of @xmath99 . using @xmath120 instead of @xmath101 increases @xmath118 by @xmath121 and @xmath119 by @xmath122 .
the best - fit values of @xmath123 do depend on the assumed value of @xmath99 , but for the ndwfs ero sample , changing @xmath99 from @xmath117 to @xmath124 increases @xmath123 by only @xmath121 .
measurements of @xmath118 for eros as a function of @xmath26-band limiting magnitude from our study and the literature are summarized in table [ table : ang ] .
angular correlation functions for apparent magnitude limited samples are also plotted in figure [ fig : ang ] .
estimates of @xmath118 and @xmath123 for each of the ndwfs subsamples are presented in table [ table : r0 ] and discussed in [ sec : app ] to [ sec : foz ] .
the amplitude of the angular correlation function for a series of apparent magnitude limited samples is presented in figure [ fig : omegak ] and table [ table : ang ] , along with estimates from the literature . while our @xmath84 sample has a larger volume and more objects than previous studies , our uncertainties are comparable to the published uncertainties of many previous studies .
this is due to our inclusion of the covariance when fitting a power - law to the data .
our estimates of the amplitude of the angular correlation function are @xmath125 lower than the smaller @xmath126 ero samples from @xcite . within our sample
we do not see a significant change in the angular clustering amplitude with apparent magnitude , but this is not unexpected as we span a small range of apparent magnitudes .
the first section of table [ table : r0 ] provides an estimate of @xmath123 for eros as a function of apparent limiting magnitude . while the ndwfs @xmath123 values are accurate to @xmath18 , each apparent magnitude bin spans a large range of redshift and absolute magnitude . as @xmath123 is correlated with luminosity in other galaxy samples ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , a correlation between @xmath123 and apparent magnitude might be expected .
we do not observe a significant correlation within the ndwfs , but our uncertainties are too large to rule out such a correlation . combining published ero samples provides spatial clustering measurements over a broad magnitude range .
however , it is not possible to directly compare the published @xmath123 measurements of different ero samples ( table [ table : prev ] ) , as different authors use different models of the ero redshift distribution .
different studies also estimate the uncertainties of the angular correlation function and @xmath123 using different techniques . in table
[ table : pairs ] , we present the ndwfs pair counts for the @xmath105 and @xmath106 @xmath0 ero angular correlation functions , so other researchers can apply their techniques for estimating the correlation function to our data .
poisson statistics underestimate the uncertainties of the correlation function on large - scales , where object pair counts are high and the uncertainties of the correlation function are dominated by large - scale structure .
the uncertainties of clustering measurements from deep pencil - beam surveys should be larger than those of the ndwfs , and the large scatter of @xmath127 ero clustering measurements shown in figure [ fig : omegak ] may reflect this .
if we assume the published best - fit values of the amplitude of the angular correlation are correct , then the angular clustering of eros does decrease with increasing limiting magnitude .
we find @xmath5 for @xmath0 eros while @xcite find @xmath128 for @xmath129 eros .
unless the redshift distribution of faint eros is very broad , the spatial clustering of eros is decreasing with increasing apparent magnitude . several studies to measure @xmath130 for faint eros @xcite , but their model redshift distributions contain more high redshift objects than the goods @xmath131 ero photometric redshift distribution @xcite .
if @xmath132 for faint eros and the goods photometric redshifts are accurate , then @xmath123 decreases from @xmath133 for @xmath0 to @xmath10 for @xmath134 eros .
red galaxies at @xmath8 have a comparable range of @xmath123 values ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) and their spatial clustering is correlated with absolute magnitude .
the current measurements of ero clustering are consistent with eros being the progenitors of local red galaxies .
we present the clustering of @xmath105 and @xmath106 eros in figure [ fig : omegak ] and table [ table : r0 ] .
we find the angular and spatial clustering of @xmath106 galaxies does not differ significantly from the remainder of the sample .
low redshift galaxies may have a bimodal distribution of clustering properties as a function of color @xcite .
this could be due to the bimodal distribution of galaxy colors at low redshift , or a bimodality of the clustering properties of galaxies as a function of star formation rate .
if the clustering is bimodal at all redshifts , we would not expect a correlation between clustering and color within a red galaxy sample .
we do not see a correlation of clustering with color but a larger sample with improved photometric redshifts is required so accurate spatial clustering measurements can be performed as a function of rest - frame color .
the clustering of eros as a function of absolute magnitude is presented in table [ table : r0 ] .
we have determined the absolute magnitudes ( without evolution corrections ) of the eros using the best - fit @xmath50-model sed . as shown in figure [ fig : photoz ] , ero photometric redshifts can have large uncertainties and our ero absolute magnitudes are , at best , approximations .
the two absolute magnitude bins are approximately volume limited samples with the same photometric redshift range .
both absolute magnitude bins are extremely luminous , and contain eros approximately 4 times brighter than the local value of @xmath135 ( @xmath136 ; * ? ? ?
we do not see a significant correlation between luminosity and @xmath123 within the sample .
the correlation between galaxy luminosity and clustering is only seen unambiguously in @xmath8 samples ( e.g. , 2dfgrs and sdss ) which contain a factor of @xmath137 more galaxies than the ndwfs ero sample .
a strong correlation between ero luminosity and spatial clustering remains plausible , and may be detected with an analysis of the complete ndwfs .
we measured the clustering of eros within the sample with two photometric redshift bins , @xmath138 and @xmath139 .
we exclude eros beyond these redshift ranges , as they contribute less than @xmath140 of the total @xmath0 ero number counts .
the results are presented in table [ table : r0 ] and figure [ fig : r0z ] .
we do not observe significant evolution of @xmath123 with redshift within the ero sample .
however , our uncertainties are large and the redshift distributions of the two samples overlap , so they are not entirely independent .
the clustering of @xmath141 , @xmath0 eros is well approximated by a power law with @xmath142 and @xmath99 fixed at @xmath117 .
as eros are thought to be the progenitors of local ellipticals , it is useful to compare the clustering measurements of these populations . in the 2dfgrs ,
the spatial correlation function increases from @xmath143 for @xmath144 red galaxies to @xmath145 for @xmath146 red galaxies @xcite .
other @xmath8 surveys , including the ndwfs , measure comparable spatial clustering for red galaxies ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
it is not unreasonable to assume the brightest eros are the progenitors of the @xmath146 red galaxies in the local universe , as the comoving spatial clustering of the two populations is comparable .
however , this assumes models predicting little or no evolution of the @xmath13 galaxy correlation function ( e.g. , * ? ? ?
* ; * ? ? ?
* ) are valid . if @xmath0 eros are the progenitors of the most luminous local red galaxies , fainter eros could the progenitors of @xmath147 red galaxies .
several previous studies of fainter eros find they are very strongly clustered , with @xmath148 @xcite .
this is much stronger than the clustering of local @xmath135 red galaxies , where @xmath149 @xcite .
however , the ero spatial clustering measurements could be subject to large , and possibly systematic , errors .
several of these measurements use model redshift distributions which are primarily constrained by the local galaxy luminosity function and faint galaxy number counts @xcite .
other model redshift distributions use unverified photometric redshifts @xcite or redshifts which could only be verified with galaxies other than eros @xcite . as shown in figure [ fig : omegak ] , the angular clustering of eros does decrease with increasing apparent magnitude . unless the redshift distribution of @xmath9 eros is very broad ,
the spatial clustering of @xmath9 eros is weaker than the spatial clustering of @xmath150 eros .
while @xmath0 eros may be the progenitors of most luminous local ellipticals , our spatial clustering measurement should be treated with some caution .
our ero sample spans broad ranges of redshift and absolute magnitude ( @xmath151 ) .
the luminosity and density evolution of eros and red galaxies at @xmath8 has not been accurately determined .
the pegase @xmath50 models predict @xmath152 magnitudes of luminosity evolution at @xmath8 , so eros would be the progenitors to @xmath144 red galaxies .
if this were the case , the spatial correlation function would be decreasing with decreasing redshift , which is unphysical .
the uncertain luminosity and density evolution of eros limits the use of @xmath153 eros to measure the evolution of the galaxy spatial correlation function .
current ero spatial clustering measurements , including our study , have large uncertainties and may be subject to systematic errors .
we will significantly reduce the random uncertainties of ero clustering measurements when we analyze the entire ndwfs botes field . the uncertainties and systematic errors of our photometric redshifts
will be accurately determined as we obtain more spectroscopic redshifts .
we will also improve the photometric redshifts for eros by using ndwfs , flamingos @xcite , and _
space telescope data
. we will then be able to accurately measure the spatial clustering of eros as a function of luminosity , color , and redshift .
we have measured the clustering of 671 @xmath0 eros with @xmath2 subset of the ndwfs .
this study covers an area nearly 5 times larger and has twice the sample size of any previous ero clustering study .
the angular clustering of @xmath0 , @xmath105 eros is well described by a power - law with @xmath5 and @xmath101 . using a model of the ero redshift distribution derived from photometric redshifts
, we find the spatial clustering of @xmath0 eros is given by @xmath154 comoving .
within our study , we detect no significant correlations between ero clustering and apparent magnitude , apparent color , absolute magnitude , or redshift . however , our uncertainties are large and such correlations may exist . when combined with data from other studies , there is evidence of the angular clustering of eros decreasing with increasing apparent magnitude .
unless the redshift distribution of @xmath9 eros is very broad , the spatial clustering of eros decreases with increasing apparent magnitude .
as the uncertainties and systematic errors of current ero spatial clustering measurements are large , they do not yet provide strong tests of models of structure evolution and galaxy formation .
this research was supported by the national optical astronomy observatory which is operated by the association of universities for research in astronomy ( aura ) , inc . under a cooperative agreement with the national science foundation .
we thank our colleagues on the ndwfs team and the kpno and ctio observing support staff .
we are grateful to frank valdes , lindsey davis and the iraf team for the packages used to reduce the imaging data .
we thank alyson ford , chris greer , heather gross , lissa miller , and erin ryan , for helping reduce kpno mosaic-1 and onis imaging data used for our work . while working on this paper , the authors had many useful discussions about galaxy clustering with tod lauer and daniel eisenstein .
jannuzi , b. t. , & dey , a. , 1999 , in asp conf .
191 , photometric redshifts and high redshift galaxies , ed .
r. j. weymann , l. j. storrie - lombardi , m. sawicki , & r. j. brunner ( san francisco : asp ) , 111 ccccccccccc ndwfs & 3529 & 671 & @xmath156 & @xmath1 & & @xmath157 & photz & @xmath158 & 1.87 + + k20 & 52 & 18 & @xmath159 & @xmath1 & dusty sf sed & @xmath160 & spectra & @xmath161 & 1.8 + k20 & 52 & 15 & @xmath159 & @xmath1 & old stellar sed & @xmath162 & spectra & 5.5 to 16 & 1.8 + ntt - wht & 701 & 400 & @xmath159 & @xmath1 & & @xmath163 & pe & @xmath164 & 1.8 + lcirs & 744 & 337 & @xmath165 & @xmath166 & & @xmath167 & photz & @xmath168 & 1.8 + lcirs & 407 & 312 & @xmath169 & @xmath166 & & @xmath167 & photz & @xmath170 & 1.8 + subaru & 114 & 134 & @xmath171 & @xmath172 & dusty sf sed & @xmath173 & photz & @xmath174 & 1.8 + subaru & 114 & 143 & @xmath171 & @xmath172 & old stellar sed & @xmath173 & photz & @xmath175 & 1.8 + elais n2 & 81.5 & 158 & @xmath176 & @xmath1 & & @xmath177 & m - de & @xmath178 & 1.8 + elais n2 & 81.5 & 158 & @xmath176 & @xmath1 & & @xmath177 & ne & @xmath179 & 1.8 + cdf - s & 50.4 & 198 & @xmath180 & @xmath181 & & @xmath177 & m - de & @xmath182 & 1.8 + hdf - s & 4 & 18 & @xmath183 & @xmath184 & & & photz & @xmath185 & 1.8 + hdf - s & 4 & 39 & @xmath183 & @xmath186 & & & photz & @xmath187 & 1.8 + hdf - s & 4 & 23 & @xmath183 & @xmath186 & @xmath188 & @xmath163 & photz & @xmath189 & 1.8 + cccccccc ndwfs & 3529 & 256 & @xmath190 & @xmath191 & @xmath105 & @xmath192 & 1.87 + ndwfs & 3529 & 421 & @xmath193 & @xmath194 & @xmath105 & @xmath195 & 1.87 + ndwfs & 3529 & 671 & @xmath196 & @xmath0 & @xmath105 & @xmath197 & 1.87 + ntt - wht & 701 & 58 & @xmath198 & @xmath199 & @xmath200 & @xmath201 & 1.8 + ntt - wht & 701 & 106 & @xmath202 & @xmath203 & @xmath200 & @xmath204 & 1.8 + ntt - wht & 701 & 158 & @xmath205 & @xmath206 & @xmath200 & @xmath207 & 1.8 + ntt - wht & 701 & 279 & @xmath208 & @xmath209 & @xmath200 & @xmath210 & 1.8 + lcirs & 744 & 337 & @xmath211 & @xmath212 & @xmath213 & @xmath214 & 1.8 + lcirs & 744 & 201 & @xmath215 & @xmath212 & @xmath216 & @xmath217 & 1.8 + ntt - wht & 447.5 & 281 & @xmath218 & @xmath219 & @xmath200 & @xmath220 & 1.8 + subaru & 114 & 111 & @xmath221 & @xmath222 & @xmath200 & @xmath223 & 1.8 + lcirs & 407 & 312 & @xmath224 & @xmath225 & @xmath213 & @xmath226 & 1.8 + lcirs & 407 & 170 & @xmath227 & @xmath225 & @xmath216 & @xmath228 & 1.8 + elais n2 & 81.5 & 73 & @xmath229 & @xmath230 & @xmath105 & @xmath231 & 1.8 + elais n2 & 81.5 & 93 & @xmath232 & @xmath233 & @xmath105 & @xmath234 & 1.8 + elais n2 & 81.5 & 112 & @xmath235 & @xmath236 & @xmath105 & @xmath237 & 1.8 + elais n2 & 38.7 & 63 & @xmath238 & @xmath239 & @xmath105 & @xmath240 & 1.8 + cdf - s & 50.4 & 137 & @xmath241 & @xmath242 & @xmath181 & @xmath243 & 1.8 + cdf - s & 50.4 & 179 & @xmath244 & @xmath245 & @xmath181 & @xmath246 & 1.8 + cdf - s & 50.4 & 198 & @xmath247 & @xmath248 & @xmath181 & @xmath249 & 1.8 + hdf - s & 4 & 18 & @xmath250 & @xmath251 & @xmath184 & @xmath252 & 1.8 + cccccc @xmath105 & @xmath253 to @xmath254 & @xmath255 & 88 & 5515 & 558532 + @xmath105 & @xmath254 to @xmath256 & @xmath257 & 424 & 33602 & 3346550 + @xmath105 & @xmath256 to @xmath258 & @xmath259 & 2274 & 192715 & 19207198 + @xmath105 & @xmath258 to @xmath260 & @xmath261 & 12084 & 1115798 & 107760586 + @xmath105 & @xmath260 to @xmath262 & @xmath263 & 60806 & 5811770 & 555035500 + @xmath105 & @xmath262 to @xmath264 & @xmath265 & 244670 & 23592613 & 2273794800 + + @xmath106 & @xmath253 to @xmath254 & @xmath266 & 20 & 1193 & 122608 + @xmath106 & @xmath254 to @xmath256 & @xmath267 & 80 & 7502 & 733626 + @xmath106 & @xmath256 to @xmath258 & @xmath268 & 510 & 41951 & 4199066 + @xmath106 & @xmath258 to @xmath260 & @xmath269 & 2546 & 242766 & 23547906 + @xmath106 & @xmath260 to @xmath262 & @xmath270 & 13240 & 1269134 & 121503332 + @xmath106 & @xmath262 to @xmath264 & @xmath271 & 53148 & 5137427 & 497077474 + + cccccccc @xmath105 & 0.80 - 3.00 & @xmath272 & @xmath273 & 256 & @xmath274 & 1.17 & @xmath275 + @xmath105 & 0.80 - 3.00 & @xmath276 & @xmath277 & 421 & @xmath278 & 1.17 & @xmath279 + @xmath105 & 0.80 - 3.00 & @xmath280 & @xmath281 & 671 & @xmath282 & 1.18 & @xmath283 + @xmath106 & 0.80 - 3.00 & @xmath284 & @xmath273 & 96 & @xmath285 & 1.22 & @xmath286 + @xmath106 & 0.80 - 3.00 & @xmath287 & @xmath277 & 180 & @xmath288 & 1.22 & @xmath289 + @xmath106 & 0.80 - 3.00 & @xmath290 & @xmath281 & 314 & @xmath291 & 1.24 & @xmath292 + @xmath105 & 0.80 - 1.25 & @xmath293 & @xmath294 & 108 & @xmath295 & 1.16 & @xmath296 + @xmath105 & 0.80 - 1.25 & @xmath297 & @xmath298 & 208 & @xmath299 & 1.11 & @xmath300 + @xmath105 & 0.80 - 1.15 & @xmath301 & @xmath302 & 318 & @xmath303 & 1.04 & @xmath304 + @xmath105 & 1.15 - 1.40 & @xmath305 & @xmath306 & 292 & @xmath307 & 1.28 & @xmath308 + @xmath105 & 0.80 - 1.40 & @xmath309 & @xmath306 & 610 & @xmath310 & 1.15 & @xmath311 + | we have measured the angular and spatial clustering of 671 @xmath0 , @xmath1 extremely red objects ( eros ) from a @xmath2 sub - region of the noao deep wide - field survey ( ndwfs ) .
our study covers nearly 5 times the area and has twice the sample size of any previous ero clustering study .
the wide field of view and @xmath3 passbands of the ndwfs allow us to place improved constraints on the clustering of @xmath4 eros .
we find the angular clustering of eros is slightly weaker than in previous measurements , and @xmath5 for @xmath0 eros .
we find no significant correlation of ero spatial clustering with redshift , apparent color or absolute magnitude , although given the uncertainties , such correlations remain plausible .
we find the spatial clustering of @xmath0 , @xmath1 eros is well approximated by a power - law , with @xmath6 in comoving coordinates .
this is comparable to the clustering of @xmath7 early - type galaxies at @xmath8 , and is consistent with the brightest eros being the progenitors of the most massive ellipticals .
there is evidence of the angular clustering of eros decreasing with increasing apparent magnitude , when ndwfs measurements of ero clustering are combined with those from the literature .
unless the redshift distribution of @xmath9 eros is very broad , the spatial clustering of eros decreases from @xmath6 for @xmath0 to @xmath10 for @xmath9 eros . |
in the fireball model , afterglows of gamma - ray bursts are described by synchrotron radiation from a relativistic blast wave ( piran 1999 ) .
radio observations of the afterglow of gamma - ray bursts at millimeter wavelengths are very important , since they can trace time evolution of the synchrotron spectrum directly and make it possible to derive physical parameters of gamma - ray bursts .
therefore , many observations of radio afterglows of gamma - ray bursts have been made so far ( frail et al . 2003 ) .
grb 030329 is one of the closest events .
its redshift is @xmath0 .
the afterglow of grb 030329 was so bright that the light curves at various frequencies , from x - ray to radio , have been obtained ( e.g. , smith et al .
2003 ; uemura et al . 2003
; price et al . 2003 ; sheth et al . 2003 ; berger et al .
furthermore , a connection between the grb and a supernova was confirmed ( kawabata et al .
2003 ) . in this paper
, we present the results of the multi - frequency observations of the radio afterglow of grb 030329 with the nobeyama 45-m telescope .
radio observations at 23.5 ghz , 43 ghz and 90 ghz were made using the 45-m telescope at nobeyama radio observatory ( nro ) from 2003 april 3 to may 21 .
the receiver front ends were cooled hemt amplifiers for 23.5 ghz and sis receivers for 43 ghz and 90 ghz .
the 43 ghz and 90 ghz observations were performed simultaneously .
we selected the observing frequency according to the weather condition .
the typical atmospheric opacities at zenith were 0.06 at 23.5 ghz , 0.07 at 43 ghz and 90 ghz .
the beam sizes measured in 2001 were 72.4@xmath1 at 23 ghz , 38.4@xmath1 at 43 ghz , and 18.2@xmath1 at 86 ghz .
the observations were performed by on - off scan with beam switching of 15 hz .
the integration time of on and off points was 10 sec for a scan .
we continued the on - off scan for 30 min - 2 hour to average the data .
antenna pointing was checked before and after each observation using a radio bright qso , 1055 + 018 .
the deviations after the observations were less than 13@xmath1 for 23.5 ghz and 8@xmath1 for 43/90 ghz . for 90 ghz
, we used the data when the deviation was less than 5@xmath1 , since the influence of the pointing error is more serious for 90 ghz .
we have to note that these pointing errors yield an uncertainty of the flux .
the uncertainties due to the pointing error are estimated to be less than 9% for 23.5 ghz , less than 11% for 43 ghz , and less than 19% for 90 ghz , assuming that the beam patterns at all frequencies are gaussian .
the chopper - wheel method was used to get antenna temperature corrected for both atmospheric and antenna ohmic loss . to check the aperture efficiency of the 45-m telescope and to convert the antenna temperature into the flux density , some qsos
( 1055 + 018 for all frequencies , 3c286 for 23.5 ghz and 43 ghz , 3c345 for 90 ghz ) were observed by the same scan mode .
since the variation of the antenna temperature of 1055 + 018 during the observing period was small at all frequencies ( less than 10% ) , we used it as a flux calibrator .
the flux densities of 1055 + 018 at 23.5 ghz , 43 ghz , and 90 ghz were 3.64 jy , 2.66 jy , and 2.5 jy , respectively , where the flux at 23.5 ghz was measured by comparing with ngc 7027 whose flux was assumed to be 5.36 jy ( ott et al .
1994 ) , that at 43 ghz by using saturn , and that at 90 ghz by the nobeyama millimeter array ( nma ) on 2003 april 10 and 13 .
we have to note that the fluxes at 43 ghz and 90 ghz in this paper are lower than the values reported in kuno et al .
this is because the 43-ghz flux in kuno et al .
( 2003 ) was derived using 3c286 as a flux calibrator assuming that its flux was 2.375 jy ( ott et al .
1994 ) , and 90 ghz flux was derived using 3c345 assuming that its flux was 5.5 jy that was measured by nma on 2003 apr 1 . however , since the time variation of the antenna temperature of 3c286 during the observing period was larger than that of 1055 + 018 , and 1055 + 018 was stronger than 3c286 , we recalibrated the 43-ghz data using 1055 + 018 .
it was found later that the flux of 3c345 had varied before our observing period , so that we recalibrated the 90-ghz data also using 1055 + 018 .
figure 1 shows the light curves at 23.5 ghz , 43 ghz , and 90 ghz and the data are listed in table 1 .
the light curve at 23.5 ghz is almost flat until 11 days after the burst , then , shows steep decline . the light curve at 43 ghz is flat until 8 days after the burst , although the sampling is sparse .
on the other hand , the decay at 90 ghz began earlier than 8 days after the burst .
the difference in start time of the decline is consistent with that expected from the fireball model in which afterglows are described by synchrotron emission from a relativistic blast wave . in that case ,
when @xmath2 , the typical synchrotron frequency of the minimal electron in the power law , becomes lower than the observed frequency , the decay begins .
therefore , the decay begins earlier at higher frequency ( piran 1999 ) .
berger et al . ( 2003 ) and sheth et al . ( 2003 ) have pointed out that there are two jet breaks at t=0.55 days ( price et al .
2003 ) and 9.8 days .
the first break is due to a narrow - angle jet , and the second one is due to a wide - angle jet .
( 80mm,80mm)fig1.eps .summary of the radio observations of grb030329 [ cols="^,^,^,^",options="header " , ] ( 86mm,)fig2.eps the light curves lower than 15 ghz shows different behavior from those at higher frequencies . rising of the light curves is consistent with that these are lower than the self - absorption frequency , @xmath3 ( berger et al .
2003 ) , since an increase of flux density below @xmath3 is expected from both the ism model and the wind model . from the power - law fitting of the decline
, we can derive _ p _ , the power - law index of the electron energy distribution assuming the power - law listed in table 2 .
however , if we use the powers in figure 1 , _
p _ may be underestimated , since the decline becomes steep gradually .
sheth et al . ( 2003 ) and berger et al . (
2003 ) have shown steeper power - laws using more data at late stage than ours .
figure 3 shows that the time evolution of the spectra .
we made the spectra using the data during apr 3.0 - 4.0 ut , apr 6.16 - 7.0 ut , and apr 16.2 - 20.8 , respectively .
the peak frequency , which corresponds to @xmath2 , is about 100 ghz on apr 3.0 - 4.0 ut , and between 43 ghz and 90 ghz on apr 6.16 - 7.0 ut .
for apr 16.2 - 20.8 ut , the peak is at about 15 ghz .
we compare the time evolution of the peak position of the spectrum with the predictions from the above models roughly .
we assumed that the peak frequency and peak flux density on apr 3.0 - 4.0 ut are 100 ghz and 57.5 mjy , the average flux density of 90 ghz and 100 ghz data , respectively .
then , we derived the time evolution of the peak frequency and peak flux density for each model assuming the power - law ; @xmath4 and @xmath5 constant for the ism model , @xmath4 and @xmath6 for the wind model , and @xmath7 and @xmath8 for the jet model , where @xmath9 is the flux density at @xmath2 .
the results are shown in figure 3 .
the ism model seems to be not consistent with the data , since the peak flux is expected to be constant in the model . from both of the wind model and the jet model , the decreases of the peak flux density and frequency
are expected .
the observed results are consistent with these models qualitatively .
we made radio observations of the afterglow of grb 030329 at 23.5 ghz , 43 ghz , and 90 ghz .
the light curves show steep decline after constant phase .
the difference in the start time of the decline was seen .
namely , the decay started earlier at higher frequency .
the time evolution of the spectrum shows a decay of the peak frequency and peak flux .
these results are consistent with the fireball model qualitatively .
we compared the radio data with three models ( ism model , wind model , and jet model ) . among the three models , the wind model and the jet model are relatively consistent with the observed data .
berger , e. , et al .
2003 , nature 426 , 154 chevalier , r.a .
, & li , z .- y .
1999 , 520 , l29 frail , d.a . , kulkarni , s.r . , berger , e. , & wieringa , m.h . 2003 , 125 , 2299 galama , t.j . ,
1998 , 500 , l101 galama , t.j . , et al .
2000 , 541 , l45 galama , t.j . ,
frail , d.a . , sari , r. , taylor , g.b . , & kulkarni , s.a .
2003 , 585 , 899 gallant , y.a . , achterberg , a. , & kirk , j.g .
1999 , 138 , 549 granot , j. , & sari , r. 2002 , 568 , 820 hoge , j. c. , meijerink , r. , tilanus , r. p. j. , & smith , i. a. 2003 , gcn2088 kawabata , k. s. , et al .
2003 , 593 , l19 kohno , k. , et al .
2003 , in preparation kuno , n. , sato , n. , & nakanishi , h. 2003 , gcn2089 ott , m. , witzel , a. , quirrenbach , a. , krichbaum , t. p. , standke , k. j. , schalinski , c. j. , & hummel , c. a. 1994 , 284 , 331 piran , t. 1999 , 314 , 575 price , p.a . ,
2003 , nature 423 , 844 sari , r. , piran , t. , & halpern , j. p. 1999 , 519 , l17 sheth , k. , frail , d. a. , white , s. , das , m. , bertoldi , f. , walter , f. , kulkarni , s. r. , & berger , e. 2003 , 595 , l33 smith , d.a . ,
2003 , 596 , l151 uemura , m. , et al .
2003 , nature 423 , 843 waxman , e. 1997 , 485 , l5 | we present the results of the radio observations of the afterglow of grb 030329 with the nobeyama 45-m telescope .
the observations were made at 23.5 ghz , 43 ghz , and 90 ghz .
the light curves show steep decline after constant phase .
the start time of the decline depends on frequency .
namely , the decline started earlier at higher frequency .
the spectrum has a peak at mm wavelength range . the peak frequency and
the peak flux decreased with time .
these results are consistent with the expectation from the fireball model . |
sparsity encodes key structural information about data and permits estimating unknown , high - dimensional vectors robustly .
no wonder , sparsity has been intensively studied in signal processing , machine learning , and statistics , and widely applied to many tasks therein . but the associated literature has grown too large to be summarized here ; so we refer the reader to @xcite as starting points .
sparsity constrained problems are often cast as instances of the following high - level optimization problem @xmath0 where @xmath1 is a differentiable loss - function , @xmath2 is a convex ( nonsmooth ) regularizer , and @xmath3 is a scalar .
alternatively , one may prefer the constrained formulation @xmath4 both formulations and continue to be actively researched , the former perhaps more than the latter .
we focus on the latter , primarily because it often admits simple but effective first - order optimization algorithms
. additional benefits that make this constrained formulation attractive include : * even when the loss @xmath1 is nonconvex , gradient - projection remains applicable ; * if the loss is separable , it is easy to derive highly scalable incremental or stochastic - gradient based optimization algorithms ; * if only inexact projections onto @xmath5 are possible ( a realistic case ) , convergence analysis of gradient - projection - type methods remains relatively simple . in this paper
, we study a particular subclass of that has recently become important , namely , _
groupwise sparse regression_. two leading examples are multitask learning @xcite and group - lasso @xcite .
a key component of these regression problems is the regularizer @xmath6 , which is designed to enforce ` groupwise variable selection'for example , with @xmath6 chosen to be a _
mixed - norm_. [ def.mn ] let @xmath7 be _ partitioned _ into subvectors @xmath8 , for @xmath9$]$ ] as a shorthand for the set @xmath10 . ] . the @xmath11-_mixed - norm _ for @xmath12 , @xmath13 ,
is then defined as @xmath14 the most practical instances of are @xmath15-norms , especially for @xmath16 .
the choice @xmath17 yields the ordinary @xmath18-norm penalty ; @xmath19 is used in group - lasso @xcite , while @xmath20 arises in compressed sensing @xcite and multitask lasso @xcite .
less common , though potentially useful versions allow interpolating between these extremes by letting @xmath21 ; see also @xcite .
definition [ def.mn ] can be substantially generalized : we may allow the subvectors @xmath22 to overlap ; or to even be normed differently @xcite .
but unless the overlapping has special structure @xcite , it leads to somewhat impractical mixed - norms , as the corresponding optimization problem becomes much harder .
since our chief aim is to develop fast , scalable algorithms for , we limit our discussion to @xmath15-norms this choice is widely applicable , hence important @xcite .
before moving onto the technical part , we briefly list the paper s main contents : * batch and online ( stochastic - gradient based ) algorithms for solving ; * theory of and algorithms for fast projection onto @xmath15-norm balls ; * application to @xmath15-norm based multitask lasso ; both batch and online versions ; * application to computing projections for matrix mixed - norms ; * a set of open problems .
we begin by developing some basic theory . our aim is to efficiently implement a generic ` first - order ' algorithm : generate a sequence @xmath23 by iterating @xmath24 where @xmath25 is a stepsize , @xmath26 is an estimate of the gradient , and @xmath27 is the projection operator that enforces the constraint @xmath28 .
below we expand on the most challenging component of iteration when applied to mixed - norm regression , namely efficient computation of the projection operator @xmath27 .
formally , the ( orthogonal ) _ projection operator _
@xmath29 is defined as @xmath30 closely tied to projection is the _ proximity operator _
@xmath32 where @xmath33 is a convex function on @xmath34 .
operator generalizes projections : if in the function @xmath33 is chosen to be the indicator function for the set @xmath35 , then the operator @xmath36 reduces to the projection operator @xmath27 .
alternatively , for convex @xmath2 and @xmath33 , operators @xmath27 and @xmath36 are also intimately connected by duality .
indeed , this connection proves key to computing a projection efficiently whenever its corresponding proximity operator is ` easier ' .
the idea is simple ( see e.g. , @xcite ) , but exploiting it effectively requires some care ; let us see how .
let @xmath37 be the lagrangian for ; and let the optimal dual solution be denoted by @xmath38 .
assuming strong - duality , the optimal primal solution is given by @xmath39 but to compute , we require the optimal @xmath38the key insight on obtaining @xmath38 is that it can be computed by solving a single nonlinear equation .
here is how .
first , observe that if @xmath40 , then @xmath41 , and there is nothing to compute .
thus , assume that @xmath42 ; then , the optimal point @xmath43 satisfies @xmath44 next , observe from that for a fixed @xmath45 , the point @xmath46 equals the operator @xmath47 .
consider , therefore , the nonlinear function ( residual ) @xmath48 which measures how accurately equation is satisfied . the optimal @xmath38 can be then obtained by solving @xmath49 , for which the following lemma proves very useful .
[ lemm : mono ] let @xmath6 be a gauge is nonnegative , positively homogeneous , and disappears at the origin @xcite ] , and let @xmath50 be as defined in .
then , there exists an interval @xmath51 $ ] , on which @xmath50 is monotonically decreasing , and differs in sign at the endpoints . by assumption on @xmath52
, it holds that @xmath53 .
we claim that for @xmath54 , where @xmath55 denotes the _ polar _ of @xmath2 , the optimal point @xmath56 . to see why , suppose that @xmath57 , but @xmath58 .
then , @xmath59 .
but since @xmath60 is strictly convex , the inequality @xmath61 also holds for any @xmath62 .
thus , it follows that @xmath63 , whereby , for @xmath64 , the optimal @xmath46 must equal @xmath65 .
hence , we may select @xmath66 .
monotonicity of @xmath67 follows easily , as it is the derivative of the concave ( dual ) function @xmath68 .
finally , @xmath69 , so it differs in sign . since @xmath50 is continuous , changes sign , and is monotonic in the interval @xmath70 $ ] , it has a unique root therein .
this root can be computed to @xmath71-accuracy using bisection in @xmath72 iterations .
we recommend not to use mere bisection , but rather to invoke a more powerful root - finder that combines bisection , inverse quadratic interpolation , and the secant method ( e.g. , matlab s ` fzero ` function ) .
pseudocode encapsulating these ideas is given in algorithm [ algo2 ] .
after the generic approach above , let us specialize to projections for the case of central interest to us , namely , @xmath27 with @xmath73 .
algorithm [ algo2 ] requires computing the upper bound @xmath66 . to that end
lemma [ lem.dual ] , which actually proves much more , proves useful .
[ lem.dual ] let @xmath74 ; and let @xmath75 be `` conjugate '' scalars , i.e. , @xmath76 and @xmath77 .
the polar ( dual - norm ) of @xmath78 is @xmath79 . by definition , the norm dual to an arbitrary norm @xmath80 is given by @xmath81 to prove the lemma , we prove two items : ( i ) for any two ( conformally partitioned ) vectors @xmath62 and @xmath82 , we have @xmath83 ; and ( ii ) for each @xmath82 , there exists an @xmath62 for which @xmath84 .
let @xmath62 be a vector partitioned conformally to @xmath82 , and consider the inequality @xmath85 which follows from hlder s inequality .
define @xmath86 $ ] and @xmath87 $ ] , and invoke hlder s inequality again to obtain @xmath88 .
thus , from definition we conclude that @xmath89 . to prove that the dual norm actually equals @xmath90
, we show that for each @xmath82 , we can find an @xmath62 that satisfies @xmath91 , for which the inner - product @xmath92 .
define therefore @xmath93some juggling with indices suggests that we should set @xmath94 where @xmath95 denotes the @xmath96-the element of the subvector @xmath22 ( similarly @xmath97 ) . to see that works ,
first consider the inner - product @xmath98 next , we check that @xmath99 .
consider thus , the term @xmath100 .
using we have @xmath101 thus , it follows that @xmath102 where the last equality holds because @xmath76 .
finally , from it follows that @xmath103 since by definition @xmath104 .
this concludes the proof .
the next key component for algorithm [ algo2 ] is the proximity operator @xmath105 . for @xmath106 ,
this operator requires solving @xmath107 fortunately , problem separates into a sum of @xmath108 _ independent _ , @xmath109-norm proximity operators .
it suffices , therefore , to only consider a subproblem of the form @xmath110 for @xmath17 , the solution to is given by the _ soft - thresholding _
operation @xcite : @xmath111 where operator @xmath112 performs elementwise multiplication . for @xmath19 ,
we get @xmath113 while the case @xmath20 is slightly more involved .
it can be solved via the _ moreau decomposition _
@xcite , which , for a norm @xmath114 implies that @xmath115 for @xmath116 , the dual - norm ( polar ) is @xmath117 ; but projection onto @xmath18-balls has been extremely well - studied
see e.g. , @xcite . for @xmath118 ( different from @xmath119 and @xmath120 ) ,
problem ( [ eq.17 ] ) is much harder .
fortunately , this problem was recently solved in @xcite , using nested root - finding subroutines .
but unlike the cases @xmath121 , the proximity operator for general @xmath122 can be computed only approximately ( i.e. , in ( [ eq:18 ] ) , each iteration generates only approximate @xmath46 ) .
we now make a brief digression , which is afforded to us by the above results .
our digression concerns mixed - norms for matrices , as well as their associated projection , proximity operators , which ultimately depend on the results of the previous section .
our discussion is motivated by applications in @xcite , where the authors used mixed - norms on matrices to simultaneously .
we define mixed - norms on matrices by building upon the classic _ schatten-@xmath122 matrix norms _
@xcite , defined as : @xmath123 where @xmath124 is an arbitrary complex matrix , and @xmath125 is its @xmath126th singular value .
now , let @xmath127 be an arbitrary set of matrices , and let @xmath128 .
we define the matrix @xmath129-norm by the formula @xmath130 as for the vector case , we have a similar lemma about norms dual to . [ lem.hold ]
let @xmath124 and @xmath131 be matrices such that @xmath132 is well - defined .
then , for @xmath133 , such that @xmath134 , it holds that @xmath135 from the well - known _ von neumann _ trace inequality @xcite we know that @xmath136 now invoke the classical hlder inequality and use definition of matrix mixed - norms to conclude .
[ lem.mtxdual ] let @xmath128 ; and let @xmath137 be their conjugate exponents .
the norm dual to @xmath138 is @xmath139 . by the triangle - inequality and lemma [ lem.hold ]
we have @xmath140 applying hlder s inequality to the latter term we obtain @xmath141 now , we must show that for any @xmath142 , we can find an @xmath143 such that holds with equality . to that end , let @xmath144 be the svd of matrix @xmath145 . setting @xmath146
, we see that @xmath147 ; since both @xmath148 and @xmath149 are diagonal , this reduces to the vector case ( [ eq:35 ] ) , completing the proof . * projections onto @xmath150-norm balls : * + as for vectors , we now consider the matrix @xmath151-norm projection @xmath152 algorithm [ algo2 ] can be used to solve .
the upper bound @xmath153 can be obtained via lemma [ lem.mtxdual ] .
it only remains to solve proximity subproblems of the form @xmath154 since both @xmath155 and @xmath156 are unitarily invariant , from corollary 2.5 of @xcite it follows that if @xmath145 has the singular value decomposition @xmath157 , then is solved by @xmath158 , where the vector @xmath159 is obtained by solving @xmath160 we note in passing that operator generalizes the popular singular value thresholding operator @xcite , which corresponds to @xmath17 ( trace norm ) .
we describe two realizations of the generic iteration that can be particularly effective : ( i ) spectral projected gradients ; and ( ii ) stochastic - gradient descent .
the simplest method to solve ( [ eq.9 ] ) is perhaps _ gradient - projection _
@xcite , where starting with a suitable initial point @xmath161 , one iterates @xmath162 we have already discussed @xmath27 ; the other two important parts of are the stepsize @xmath163 , and the gradient @xmath164 . even when the loss @xmath1 is not convex , under fairly mild condition , we may still iterate to obtain convergence to a stationary point see ( * ? ? ?
* chapter 1 ) for a detailed discussion , including various strategies for computing stepsizes .
if , however , @xmath1 is convex , we may invoke a method that typically converges much faster : _ spectral projected gradient _
( spg ) @xcite .
spg extends ordinary gradient - projection by using the famous ( nonmonotonic ) _ spectral stepsizes _ of barzilai and borwein @xcite ( bb ) .
formally , these stepsizes are @xmath165 where @xmath166 , and @xmath167 .
spg substitutes stepsizes in ( using safeguards to ensure bounded steps ) .
thereby , it leverages the strong empirical performance enjoyed by bb stepsizes @xcite ; to ensure global convergence , spg invokes a nonmontone line search strategy that allows the objective value to occasionally increase , while maintaining some information that allows extraction of a descending subsequence .
[ [ inexact - projections ] ] inexact projections : + + + + + + + + + + + + + + + + + + + + theoretically , the convergence analysis of spg @xcite depends on access to a subroutine that computes @xmath27 _ exactly_. obviously , in general , this operator can not be computed exactly ( including for many of the mixed - norms ) . to be correct , we must rely on an _ inexact _ spg method such as @xcite .
in fact , due to roundoff error , even the so - called exact methods run inexactly .
so , to be fully correct , we must treat the entire iteration as being inexact .
such analysis can be done ( see e.g. , @xcite ) ; but it is not one of the main aims of this paper , so we omit it .
suppose the loss - function @xmath1 in is separable , that is , @xmath168 for some large number @xmath169 of components ( say @xmath170 ) .
in such a case , computing the entire gradient @xmath164 at each iteration ( [ eq:9 ] ) may be too expensive , and it might be more preferable to use stochastic - gradient descent ( sgd ) instead . in its simplest realization , at iteration @xmath171 , sgd picks a random index @xmath172 $ ] , and replaces @xmath173 by a stochastic estimate @xmath174 .
this results in the iteration @xmath175 where @xmath163 are suitable ( e.g. , @xmath176 ) stepsizes . again
, some additional analysis is also needed for to account for the potential inexactness of the projections .
we present below numerical results that illustrate the computational performance of our methods .
in particular , we show the following main experiments : 1 . running time behavior of our root - finding projection methods , including * comparisons against the method of @xcite for @xmath177 projections * some results on @xmath15 projections for a few different values of @xmath122 .
2 . application to the @xmath177-norm multitask lasso @xcite , for which we show * running time behavior of spg , both with our projection and that of @xcite ; * derivation of and numerical results with a sgd based method for mtl . for ease of comparison , we use the notation of @xcite , who seem to be the first to consider efficient projections onto the @xmath178-norm ball .
the task is to solve @xmath179 where @xmath180 is a @xmath181 matrix , and @xmath182 denotes its @xmath126th row . in our comparisons
, we refer to the algorithm of @xcite ( c implementation)aquattoni / codetoshare/ _ ] , as ` qp , ' and to our method as ` fp ' ( also c implementation ) .
the experiments were run on a single core of a quad - core amd opteron ( 2.6ghz ) , 64bit linux machine with 16 gb ram .
we compute the optimal @xmath183 , as @xmath184 varies from @xmath185 ( more sparse ) to @xmath186 ( less sparse ) settings .
tables [ tab.1][tab.3 ] present running times , objective function values , and errors ( as measured by the constraint violation : @xmath187 , for an estimated @xmath183 ) .
the tables also show the absolute difference in objective value between qp and fp . while for small problems , qp is very competitive , for larger ones , fp consistently outperforms it .
although on average fp is only about twice as fast as qp , it is noteworthy that despite fp being an `` inexact '' method ( and qp an `` exact '' one ) , fp obtains solutions of accuracy many magnitudes of order better than qp .
.runtime and accuracy for qp and fp on a @xmath188 matrix @xmath189 . [ cols=">,>,>,>,>,>",options="header " , ] the results in table [ tab : data ] indicate that for large - scale problems , the savings accrued upon using our faster projections ( in combination with spg ) can be substantial .
we now show a running comparison between three methods : ( i ) spg@xmath190 , ( ii ) spg@xmath191 , and ( iii ) sgd ( with projection step computed using fp ) . for our comparison , we solve mtlon a subset of the cmu newsgroups datasettextlearning/ _ ; we use the reduced version of @xcite . ] .
the dataset corresponds to 5 feature selection tasks based on data taken from the following newsgroups : computer , politics , science , recreation , and religion .
the feature selection tasks are spread over the matrices @xmath192 , each of size @xmath193 , while the dependent variables @xmath194 correspond to class labels .
figure [ fig.cmu ] reports running time results obtained by the three methods in question ( all methods were initialized by the same @xmath195 ) .
as expected , the stochastic - gradient based method rapidly achieves a low - accuracy solution , but start slowing down as time proceeds , and eventually gets overtaken by the spg based methods .
interestingly , in the first experiment , spg@xmath190takes much longer than spg@xmath191to convergence , while in the second experiment , it lags behind substantially before accelerating towards the end .
we attribute this difference to the difficulty of the projection subproblem : in the beginning , the sparsity pattern has not yet emerged , which drives spg@xmath190to take more time . in general , however , from the figure it seems that either sgd or spg@xmath191yield an approximate solution more rapidly
so for problems of increasingly larger size , we might prefer them .
we described mixed - norms for vectors , which we then naturally extended also to matrices .
we presented some duality theory , which enabled us to derive root - finding algorithms for efficiently computing projections onto mixed - norm balls , especially for the special class of @xmath15-mixed norms . for solving an overall regression problem involving mixed - norms we suggested two main algorithms , spectral projected gradient and stochastic - gradient ( for separable losses ) .
we presented a small but indicative set of experiments to illustrate the computational benefits of our ideas , in particular for the multitask lasso problem . * designing fast projection methods for certain classes of non - separable mixed norms .
some algorithms already exist for particular classes @xcite . * studying norm projections with additional simple constraints ( e.g. , bounds ) . * extending the fast methods of this paper to non - euclidean proximity operators . * exploring applications of matrix mixed - norm regularizers .
tomioka , r. , suzuki , t. , sugiyama , m. : augmented lagrangian methods for learning , selecting , and combining features . in : s. sra , s. nowozin , s.j .
wright ( eds . )
optimization for machine learning .
mit press ( 2011 ) | joint sparsity offers powerful structural cues for feature selection , especially for variables that are expected to demonstrate a `` grouped '' behavior .
such behavior is commonly modeled via group - lasso , multitask lasso , and related methods where feature selection is effected via mixed - norms . several mixed - norm based sparse models have received substantial attention , and for some cases efficient algorithms are also available .
surprisingly , several _ constrained _ sparse models seem to be lacking scalable algorithms .
we address this deficiency by presenting batch and online ( stochastic - gradient ) optimization methods , both of which rely on efficient projections onto mixed - norm balls .
we illustrate our methods by applying them to the multitask lasso .
we conclude by mentioning some open problems . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * keywords : * mixed - norm , group sparsity , fast projection , multitask learning , matrix norms , stochastic gradient _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ |
strong bursts of decametric radio emission below 40 mhz were identified by @xcite as originating from jupiter .
the emission in the decametric range ( referred to as dam range hereafter ) occurs from 3 mhz to 39.5 mhz , although specific io - related dam arcs have been seen at frequencies below 2 mhz @xcite .
jovian dam has a peak of the spectral power at a frequency of about 10 mhz , and average fluxes of @xmath1 mjy watt m@xmath2 hz@xmath3 , as well as s - bursts that can be 100 times stronger @xcite .
the lower - limit on observations from ground - based observatories is generally above 5 to 10 mhz due to attenuation by the earth s ionosphere . from ground - based and space - based observations
, it has been long known that much of the jupiter dam emission is influenced by the satellite io @xcite , and that some emissions not related to io have been correlated with solar wind parameters @xcite . for a review see @xcite .
these coherent emissions are thought to be the result of instabilities near the local electron cyclotron frequency , called the cyclotron maser instability ( cmi ) mechanism @xcite .
the cmi emission is beamed in a hollow cone with thickness of a few degrees @xcite which has a large half - angle of 60@xmath0 to 90@xmath0 relative to the local magnetic field direction @xcite .
the dam emission is observed at earth when the edge of the emission cone sweeps over the observer .
general properties of the dam emission include 1 ) high brightness temperatures , @xmath4 k , 2 ) emission at frequencies near the x - mode cutoff which is close to the local electron cyclotron frequency from both northern and southern hemisphere emission regions , and 3 ) 100% circular or elliptical polarization @xcite .
the io - controlled jovian emission regions are designated io - a and io - b for the primarily right hand circular ( rhc ) emission sources in the northern hemisphere while the predominantly left hand circular ( lhc ) sources in the southern hemisphere are designated io - c and io - d .
the handedness of the polarized emission is set by the direction of propagation of the emission relative to the local magnetic field direction as seen by the observer @xcite .
these io - related emission regions are defined by their occurrence probability in the central meridian longitude ( ) and io - phase ( @xmath5 ) plane @xcite .
[ cols="^,^,^,^,^,^ " , ] [ tbl : obs_time ] the lwa1 has not yet been characterized for polarization response thus observations at large zenith angles are expected to have significant polarization leakage across the stokes parameters .
determining the polarized response of a dipole array is considerably more difficult than for a single dish telescope .
the reasons for this are : ( 1 ) dipole arrays produce beams with higher side lobe levels , therefore a significant contribution to the beam power comes from beyond the primary beam ; ( 2 ) the dipole array beam shape and sensitivity vary significantly across the sky . the first factor violates the usual assumption that the source polarization properties are constant in time ( or change only as a function of parallactic angle ) , and the second factor violates the assumption that the instrumental polarization is constant in time .
we have minimized the impact of the lack of polarization calibration by observing jovian bursts at high elevations near transit . to characterize the impact of this leakage on the measured polarization ,
we have compared the fractional circular polarization measured for jovian bursts with measurements from the literature . using io - b and io - c bursts
we have measured the fractional circular polarization of the emission over zenith angles from @xmath6 20@xmath0 to over 50@xmath0 and compared those measurements to statistical measurements of fractional polarization of the same type of bursts observed with the nda as presented in @xcite .
we find very good agreement between the lwa1 measured fractional polarization and the nda statistical results up to zenith angles of 40@xmath0 . at larger zenith angles
we see an elevation dependent fractional polarization which is due to the lack of polarization calibration .
a similar effect is seen observing cygnus a over a variety of zenith angles where the fractional circular polarization shows a slow increase with increasing zenith angle beyond roughly 50@xmath0 ( j. hartman , 2012 , private communication ) .
we would not expect that the lack of polarization calibration would change the handedness of the emission observed and , in fact , we find that the handedness of the emission observed in all bursts recorded with the lwa1 matches that expected from the literature . due to the lack of polarization calibration
, we do not measure polarization fraction within this paper and report mainly the handedness of the emission as well as source structure .
rfi is present in the data in both persistent form with known emitters spread throughout the spectrum as well as intermittent rfi events @xcite .
observations were preferentially scheduled at night when rfi levels are lower but many of the observations had to be taken at least partially during the daytime to target high probability events in the observing cycle . for the majority of the jovian observations , we found that the rfi only impacted a small amount of the spectrum and thus we have not excised these regions . during quiescent rfi times
we were easily able to study jovian emission down to frequencies of 12 mhz or lower ( see for example figure [ fig:262 ] ) .
lwa1 observations of jupiter were targeted during times of io - controlled emission . in this section
we present analysis of bursts from all four of the io - controlled emissions regions . in table
[ tbl : obs ] we list event details of each burst discussed in this paper while the observing session details are given in table [ tbl : obs_time ] . in figure
[ fig : prob ] we show a probability plot of the io phase vs with the positions of jupiter marked for each of the bursts discussed in this paper .
we show in figure [ fig:262 ] a stokes v spectrogram of an io - a / io - c event from 01-dec-2012 .
the full observation lasts 5 hours and we have combined the data from the two frequency tunings .
the colorbar at the right shows intensity ( labeled as power spectral density or psd ) with negative stokes v ( lhc ) emission in blue and positive stokes v ( rhc ) emission in red on a scale that is linear in power in arbitrary units .
the ordinate axis shows observing frequency while the top abscissa is labeled in time since the start of the observation and the bottom abscissa is labeled in central meridian longitude system iii ( ) .
the io - a event ( 185@xmath0 - 270@xmath0 ) shows classic vertex late arc structures , which are possibly caused by multiple alfvn wave reflections between the io plasma torus and the source regions @xcite .
we see additional ( fainter ) rhc emission between = 270@xmath0 to roughly 350@xmath0 where it appears to drop below our observing frequency .
this second rhc emission component appears distinct from the main io - a emission ; the overall intensity is weaker and the envelope reaches a higher frequency of @xmath635 mhz . @xcite studied similar io - c morphology using nda data from 1986 to 1991 and identified one rhc component ( called a great arc ) and three lhc components ( type i , ii , and iii ) .
they interpret their data as components of io - c and suggest that both rhc and lhc emissions are coming from the same source in the same hemisphere .
recent modulation lane analysis ( section [ sect : modulation ] ) also favors this interpretation ; however , the different spectral structure and different frequency maxima of the rhc and lhc data are difficult to explain .
we also note that bimodal polarizations emitted by the same source are not favored by theory @xcite .
the lhc event begins at about 230@xmath0 and increases in frequency to reach a maximum frequency of @xmath624.5 mhz before decreasing again in frequency to fall below our observing band around = 350@xmath0 .
we interpret this event as an io - c burst but we note that it begins well before the generally accepted beginning value of = 280@xmath0 for io - c events @xcite . although the previously published source boundaries are based on long - term statistics and are frequency dependent , individual io - c events beginning at values prior to 280@xmath0 have been shown before ( e.g. * ? ? ?
* ; * ? ? ?
the cml boundaries were not previous discussed , however , these new lwa1 observations suggest that the range for the io - c source needs refining .
more precise source boundaries will be of value to theorists and modelers to better know the active cml regions and their associated magnetic field lines . furthermore we hope the lwa1 data can help disentangle the io - a from the io - c source(s ) and be used to determine the hemispheric source(s ) for the io - a and/or io - c components . the vertex late arc structures characteristic of io - c events are clearly seen by the lwa1 . in figure [ fig:262_l_zoom ] we show a higher resolution section of the lhc polarization in the lower frequency tuning for the portion of the burst from 10800 s onward .
we can clearly see that the io - c event splits into at least two different structures near the end of the event .
the very subtle , nearly horizontal amplitude modulations seen near 19 mhz in the burst emission are ionospheric faraday lanes induced in the highly elliptically polarized emission as it passes through the earth s ionosphere ( see figure [ fig:288_fl ] for more obvious faraday lanes ) . )
.,width=316 ] we have observed two other io - c events with the lwa1 and find remarkable similarity in the events . in figure [ fig:996 ]
we show the upper frequency tuning ( centered on 26 mhz ) for an io - c event from 10-mar-2012 .
rhc emission is seen from the observation start at = 302@xmath0 until = 323@xmath0 , where the emission drops below an observing frequency of 17 mhz .
the lhc emission is seen from the observation start until it falls below our band at @xmath6 = 342@xmath0 .
similar to the burst described above , we interpret the rhc emission as a great arc and the lhc emission as io - c , most likely coming from the southern hemisphere .
the io - c emission reaches a maximum frequency of roughly 25 mhz and shows the same vertex late arc structures as the previous burst .
the third io - c event we present , figure [ fig:357 ] , was observed on 06-mar-2013 for a total of 3.1 hours .
the early portion of the observation shows the io - a emission in rhc ending around = 268@xmath0 while the lhc emission of the io - c event starts around = 277@xmath0 and runs past the end of our observing window . as with the other two events discussed above
, the io - c emission shows the vertex late structure typical of these events .
we measure a peak frequency of just above 24.6 mhz for the lhc io - c emission .
similar to the other two events , we also see the great arc rhc emission co - incident with the io - c event between of roughly 275@xmath0 and 318@xmath0 before it drops below our upper tuning frequency .
although the lower tuning is significantly impacted by rfi , we are able to continue tracing all rhc and lhc emission to frequencies near 12 mhz . comparing the three events in figures [ fig:262 ] to [ fig:357 ]
we see that the overall picture is one of the io - a event ending before = 270@xmath0 followed by a rhc event which proceeds from higher frequencies ( 35 mhz ) around = 270@xmath0 to lower frequencies reaching up to 350@xmath0 before falling below our lowest frequency observing window .
more surprisingly , the lhc io - c events appear to start at as early as 230@xmath0 at low frequencies ( 11 mhz ) , peaking at roughly 25 mhz near of 300@xmath0 to 315@xmath0 before falling again to frequencies as low as 11 mhz near of 345@xmath0 .
the most recent event reported here ( 06-mar-2013 ) appears to have the io - c emission shifted in by at least 20@xmath0 compared to the event observed one year earlier .
the typical range quoted for io - c events is 280@xmath0 - 60@xmath0 @xcite .
observations of io - a / io - c storms with the nda are suggestive of an early start to the io - c , but there has not been any discussion or publications regarding the boundaries of the io - c source . in panel b of figure 1 in @xcite
we see the rhc and lhc emission from an event on 15-oct 1987 that shows structure remarkably similar to the events presented herein in figures [ fig:262 ] to [ fig:357 ] .
@xcite note the occurrence of the lhc io - c emission in the same range as the rhc but do not trace the early start due to significant rfi present in the data below 20 mhz .
polarization studies of jovian decametric emission using voyager planetary radio astronomy ( pra ) data by @xcite reported a change in the handedness of circular polarization of continuous dam emission below 15 mhz .
@xcite showed that these anomalies and sign reversals in the polarization of the pra below 15 mhz can be explained by the antenna response when the dipole length is close to the observing frequency . using a model for the polarization response for the pra
, @xcite were able to analyze the polarization properties of the hectometric emission and separate the contributions of two main sources of emission .
they found these two sources had opposite handedness and independently changing intensities . due to the complexities of the higher frequency pra band , they were not able to map the polarization details of the dam . at these higher frequencies from voyager
only total intensity analyses of jupiter s dam were made and thus it would have been very difficult to detect the early start of the io - c emission we observe with the lwa1 .
io - b events are generally considered as originating in the northern hemisphere and are thought to represent the leading edge of the northern emission cone .
in figure [ fig:997_full ] we show the upper tuning ( centered at 26 mhz ) of a classic io - b observation from 11-mar-2012 . early in the event
we see emission drifting upward in frequency and peaking about 35 mhz followed by a more narrow - band tail that drifts downward in frequency .
the io - b source shows vertex early arc structure that have been well studied and modeled @xcite . as is common in io - b events ,
s - bursts are seen in the early part of the event .
we also see a faint , and more narrow - band lhc emission ranging from values of about 130@xmath0 to 180@xmath0 .
this emission peaks about 21 mhz and is most likely southern hemisphere io - d emission .
we discuss s - bursts in section [ sect : sbursts ] and the io - d emission in section [ sect : io - d ] .
a very different io - b event showing a series of narrow - banded emission regions is shown in figure [ fig:194_full ] .
this observation covers a total of 2.5 hours .
this entire event consists of narrow - band emission bands .
analysis of the high temporal resolution spectrograms of this burst show a number of fine structure details such as s bursts and tilted v events . these latter events are seen as regions within narrow - band events that are bounded by two s bursts @xcite and are discussed further in section [ sect : sbursts ] .
the final io - b / io - d event we present is shown in figure [ fig:288_full ] .
this event is peculiar because the io - b emission cuts off sharply at 07:43 ut ( 121@xmath0 , io phase 100@xmath0 ) , presumably just as the source cone rotates past the earth . because the lwa1 off - beam data ( not shown here ) show good continuity over this interval , we believe that this sharp cutoff is real and displays the narrow beaming nature of the dam emission cones .
figure [ fig : prob ] shows that this event was a high pass through the io - b probability region on the io phase- plane .
soon after the io - b cone moves out of our line of sight an io - d event begins .
this event is seen as the lhc emission .
unfortunately this observation ended at 09:40 ut ( 150@xmath0 , io phase 107@xmath0 ) just as io - d appeared to be reaching its peak frequency .
the maximum frequency observed before the end of the observations was 22.5 mhz .
this io - d event is also particularly interesting due to the detection of s - bursts within it ( see section [ sect : d_fs ] ) .
the leading edge of the southern emission cone is thought to be associated with io - d events .
these sources are more elusive due to the lower emission intensity , lower emission frequency and the overlap of the source emission region with the same general and io phase ( @xmath7 ) range as the stronger io - b radiation .
the and io phase ranges for the io - d source are = 0@xmath0 to 200@xmath0 and @xmath7 = 80@xmath0 to 130@xmath0 @xcite .
@xcite found that the io - d reaches down to frequencies of 3 mhz and that it appears as an isolated arc .
their model argues for the io - d source in the southern hemisphere along a single io flux tube .
we note that we did not specifically target io - d events during our initial observing program but several events were detected within the targeted io - b event observations . based on these detections ,
we have initiated a follow - up observing program which specifically targets the io - d sources for detailed structure analysis and those results will be presented in a future paper . in figure [ fig:997_full ]
we see evidence of weak , narrow - band lhc polarization extending over = 134@xmath0 to 185@xmath0 and @xmath7 = 93@xmath0 to 104@xmath0 . the emission is slowly drifting in time with the peak frequency moving from a maximum near 21 mhz to around 18 mhz .
the emission characteristics of this io - d event are consistent with results from both nda and voyager data ( see * ? ? ?
* ; * ? ? ?
structures identified in higher time - resolution images of this lhc emission are discussed in section [ sect : d_fs ] .
the second io - d event we present can be seen beginning near the end of the 27-dec-2012 event shown in figure [ fig:288_full ] . shortly after the io - b emission cone passes out of our line - of - sight ,
the lhc io - d emission begins .
the emission is initially at very low frequencies ( @xmath6 13 mhz ) near 121@xmath0 .
the emission frequency increases until the end of our observations at which point the peak frequency is near 22 mhz . due to the remarkably clean radio frequency interference spectrum at these low frequencies we are able to see evidence of both riihimaa modulation lanes as well as nearly horizontal , closely spaced faraday lanes that are clearly seen below 21 mhz across both the io - b and io - d emission regions .
we show these faraday fringes in figure [ fig:288_fl ] and discuss this further in section [ sect : faraday ] .
high time - resolution study of this io - d burst shows clear s - burst structure within the event . in section [
sect : d_fs ] we discuss the rare io - d fine structure and characterize the drift rate of these s - bursts .
there are many examples of fine structures clearly seen in the lwa1 data .
most of these structures have been seen by other ground - based and spacecraft observers @xcite in one or more io - controlled events .
examples include frequency envelopes for different sources , nested arc features , narrow band events , modulation lanes , faraday lanes , and simple and complex s - bursts .
we will highlight many of the excellent lwa1 observations and address these features in the sections below .
jovian emission undergoes a variety of propagation effects as it passes from the emission cone near the planet through the magnetoionic plasma in the io torus , the interplanetary medium and the earth s ionosphere .
@xcite identified a set of amplitude modulation lanes that cut across the l - bursts and s - bursts and which could have either positive or negative slopes .
these modulation lanes are characterized by a strong periodicity in time , with periods ranging from about 1 to 5 sec and an average of about 2 sec . in figure
[ fig:288_mod_lanes ] , we show a small 120 second portion of the io - b event shown in figure [ fig:288_full ] where positive slope modulation lanes are clearly visible .
this image has a temporal resolution of 10 ms and spectral resolution of 10 khz .
the modulation lanes show regular spacing crossing the io - b emission .
slope measurements of the lines in the frequency range of 21 - 23 mhz in early parts of this 27-dec-2012 event show slopes consistent with measurements from @xcite .
high temporal resolution spectrograms of the 10-mar-2012 event show modulation lanes visible through the rhc and lhc emission of this event ( figure [ fig : mod_lanes ] ) .
analysis of these lanes shows that they are continuous across the two different polarization hands @xcite .
we defer a detailed analysis of this modulation lane structure to a future paper but do note that the detection of these modulation lanes may have very important implications for our understanding of the emission mechanism at jupiter . if both the rhc and lhc emission regions are shown to be crossed by continuous modulation lanes then , according to the widely accepted @xcite model , this would mean that the two polarization hands must be coming from sources in the _ same _ hemisphere .
this would then suggest that the emission detected would be both the l - o and r - x modes of the cmi .
one concern is that theoretical studies of the cmi emission mechanism of earth s auroral kilometric radiation ( akr ) , which is analogous to jovian dam emission , predict that the r - x mode growth rates up to two orders of magnitude larger than the l - o mode @xcite .
observational studies of akr support these predictions @xcite .
our data show roughly similar powers for both hands of polarization which may imply that other processes are necessary to explain the l - o emission , like refractions , reflections , or mode conversions .
alternately , if the emissions are coming from opposite hemispheres , the propagation paths of each source emission through an interference screen that is comprised of field - aligned columns of enhanced plasma density from the io torus would need to be remarkably similar to form the continuous modulation lanes across both polarizations ( see * ? ? ?
* ) .
we have measured the slope of the modulations lanes for the above lwa1 events .
we show in figure [ fig : slope_cml ] the measured slopes of several of the lwa1-detected modulation lanes included on the modulation lane plot of @xcite .
each lwa1 slope is the mean of independent slope measurements for 5 consecutive modulation lanes around the plotted .
the figure shows the modulation lane slopes we measure are consistent with earlier papers although we note that we do detect modulation lanes in io - c regions of the plot that are currently sparsely populated .
we defer a detailed modulation lane slope analysis with many more measurements to a future paper .
highly elliptically polarized emission passing through dense plasma media , such as jupiter s io plasma torus plus magnetosphere and the earth s ionosphere , will undergo faraday rotation due to the bi - refringent nature of the magnetized and ionized plasma .
each magneto - ionic medium between the source and observer will introduce a rotation measure ( rm ) in the polarized jovian emission where @xcite show the faraday rotation measure along some path @xmath8 is defined as : @xmath9 where @xmath10 is the electron charge , @xmath11 is the mass of the electron , @xmath12 is the speed of light , @xmath13 is the electron density at each point @xmath14 along the path , and @xmath15 is line of sight component of the magnetic field at point @xmath14 .
there are at least four distinct contributions to the observed faraday rotation of jovian decametric emission .
these are ( 1 ) the jovian magnetosphere , ( 2 ) the io plasma torus , ( 3 ) the interplanetary medium , and ( 4 ) the earth s ionosphere .
each of these components will make a time - variable contribution to the observed faraday rotation .
taking into account the propagation effects from these media , @xcite used io - b events observed from the nda to determine that at least some of the faraday rotation comes from the io plasma torus .
the fraction of faraday rotation from each of the components above is unclear from work in the literature and is discussed in some detail in @xcite .
the majority of the rm signal likely comes from the earth s ionosphere which has typical electron density of @xmath16 @xmath17 , field strength of @xmath18 g , and a path length of 500 km .
this results in a contribution of roughly rm@xmath19 rad / m@xmath20 .
linearly polarized feeds , such as used at lwa1 , will detect the emission with a time - variable visibility in each of the two orthogonal polarizations .
we have investigated several of the bursts using the orthogonal linear polarizations from the lwa1 and see clear evidence of faraday fringes .
we show in figure [ fig:288_fl ] a linear polarization difference spectrogram of a portion of the 27-dec-2012 io - b / io - d event with clear nearly horizontal faraday fringes .
similar faraday fringes have been seen for the io - b event on 11-mar-2012 .
we have used the locations of the nulls in the latter event to determine the rotation measure of rm=2.2 rad m@xmath2 which is similar to the expected contribution from the earth s ionosphere . in this section
we discuss some of the fine structure seen in the lwa1 observations of the jovian bursts .
short bursts of emission , known as s - bursts , are described in detail in @xcite .
these sources consist of a series of very short pulses of emission that range in length from 1 ms to several hundred ms .
the instantaneous emission is very narrow in frequency and typical drift rates are between -5 and -30 mhz / s @xcite .
drift rates have been measured for s - burst seen in io - b and io - c events , and more rarely in io - a events @xcite .
assuming that the emission occurs near the local electron cyclotron frequency , the negative slope is interpreted as resulting from motion of the packet of radiating electrons away from the jovian surface toward lower magnetic field strengths . @xcite
show the fine structure within an s - burst and the coherence of the oscillations inside the microstructure while wavelet analysis of these bursts shows modulations on scales of 6 - 15 @xmath21s @xcite .
n - events are described in @xcite .
these events are relatively rare and consist of narrow frequency bands with trains of emission that can drift up or down with time .
the n - events are often either composed of short s - bursts of multiple split bands of emission .
typical bandwidths are 200 - 600 khz with durations of a few seconds to tens of minutes .
these events are also seen to drift slowly in frequency with time in some cases .
@xcite studied the occurrence of the n - events and found they are often seen over a wide range of .
we have investigated the fine structure of the 11-march-2012 io - b event ( figure [ fig:997_full ] ) at high temporal resolution .
figure [ fig : iob_sn_wide ] shows a short @xmath6 1.8 s long portion of the burst with 0.25 ms temporal resolution and 10 khz spectral resolution for a 9.2 mhz wide frequency window centered around 21.8 mhz .
we see that the portion of the spectrum between 21 and 25 mhz is largely dominated by s - bursts while below 21 mhz there is a mix of s - bursts and n - events .
we have measured the slopes of the s - bursts in several regions of the io - b event and find values of around -15 mhz / s , consistent with typical values for io - b events reported in the literature ( see figure 3 in * ? ? ?
* ) .
more complex s - burst interactions with n - events are seen below 23 mhz in figure [ fig : iob_sn_wide ] where there appears to be multiple instances of n - events terminating at the time when an s - burst passes through the same frequency as the n - event . in figure
[ fig : iob_sn_zoom ] we isolate details of one such interaction over a total time of 0.21 s. this figure shows the full details of one s - burst that starts near 25.6 mhz and drifts at a rate of -17.9 mhz / s to roughly 23.7 mhz where it encounters and apparently quenches an n - event which began @xmath6 15 ms before the s - burst encounter .
multiple s - bursts appear to be generated after the termination of the n - event starting at frequencies of 23.2 mhz and with slopes of -16.3 mhz / s .
analysis of high temporal resolution data on this event appears to show that there are several regions where the s - bursts and short n - events are apparently interacting .
this would suggest that the source regions for each type of event are cospatial .
previous observations have been made that show interactions between n - events and s - bursts , referred to as s - n burst events @xcite .
the s - n burst events typically occur between 20 - 24 mhz and show an inverted `` tilted - v '' shape in the spectrum that connects the s - burst with the n event .
figure [ fig:194_tiltv ] shows a 4.3-second portion of one of the narrow bands of emission during the 24-sept-2012 narrow - band io - b event ( see figure [ fig:194_full ] ) .
the spectrogram in figure [ fig:194_tiltv ] has a bandwidth of 2.3 mhz centered on 22 mhz and clearly shows many inverted - v shapes as narrow band emissions changing rapidly in frequency and in time .
also seen are the `` shadow event '' regions beneath the tilted - v shapes .
the origin of these events and interactions are not clearly understood , but are thought to be bunched electrons accelerated by electric fields upward along the io flux tube ( cf .
* and references therein ) .
thus these s - n burst events are believed to be generated in the same region in jupiter s magnetosphere ; our observations of s- and n - events shown in figures [ fig : iob_sn_wide ] and [ fig : iob_sn_zoom ] are consistent with this interpretation .
more high resolution observations of the io - b source using the lwa instrument may provide some additional clues as to the nature and origin of these peculiar events . very little information is available in the literature regarding the io - d events . in general , the io - d emission seems to be characterized by relatively narrow emission seen over a wide range of and confined to frequencies below @xmath6 18 mhz @xcite .
we have undertaken high temporal resolution studies of the io - d emission from the 27-dec-2012 event and we see clear s - bursts during this lhc emission event
. a short portion of the spectrum ( @xmath6 1.44 s ) is shown in figure [ fig : iod_sbursts ] where we show the region of the spectrum from @xmath6 15 mhz to 18 mhz .
we measure drift rates of the io - d s - bursts of -11 to -15 mhz / s at frequencies of 15 to 16 mhz which are consistent with drift rates from other io sources .
there are few discussions in the literature regarding s - burst detections during known io - d events , and only one definitive event could be found in the literature ( see * ? ? ?
* ) . because both io - b and io - d have significant overlap in cml , it is difficult to distinguish the io - related source of the s - bursts
there is some possible evidence showing io - d related s - bursts in the voyager pra data ( see @xcite ) and cassini radio and plasma wave science ( rpws ) data ( see @xcite ) .
in figure [ fig : iod_997 ] we show the details of the lhc io - d emission from the 11-march-2012 event with 0.25 ms temporal resolution and 10 khz spectral resolution .
the figure shows a window of 1.7 s and a frequency window of 2.6 mhz centered around 19.1 mhz .
the emission is seen to consist of several narrow bands of emission at frequencies near 20 mhz as well as s - bursts at frequencies between 18 and 20 mhz .
we note that the horizontal structures seen in the figure are rfi .
analysis of additional sections of the io - d emission show structures such as tilted - v events and other narrow - band structures .
more observations are currently being conducted in an effort to further characterize the fine structure of the io - d events which can be useful in understanding the electron acceleration method and the source generation mechanism .
the long wavelength array ( lwa1 ) instrument is shown to be an excellent instrument for studies of jupiter decametric radio emission ( dam ) .
the large bandwidth , fine spectral and temporal resolution , and full stokes capabilities of the lwa1 instrument allow for a comprehensive analysis of jupiter s broad and fine dam structures .
the location and low rfi environment are suitable to measure jupiter s emissions down to about 11 mhz , well below most other earth - based radio telescopes .
this overview paper of targeted io - related jovian dam emissions not only highlights the capabilities of the lwa1 instrument , but also provides some new data and insight into the nature of jupiter s decametric emissions .
the jupiter observations from dec-2011 to mar-2013 highlighted in this paper show good io - a , io - b , io - c , and io - d structures , and the broad and fine detail in these data are further enhanced by the excellent polarization capabilities of the lwa1 instrument .
data reduction and analysis show some characteristic io - related spectral signatures but also show some fine structures and details not seen before .
a summary of the findings follows .
the io - a observations show characteristic vertex - late arc structures as well as some narrow band emission events at frequencies below 19 mhz ( see figure [ fig:262 ] ) .
the remarkable observations are the io - c emissions that occur simultaneously with the io - a emissions .
the onset of the lhc io - c emissions at 11 mhz was measured to be 230@xmath0 ; this has not been clearly measured before and allows the extent of the io - c source to be significantly expanded .
the cml range of the io - c source can tentatively be defined as 230@xmath0 @xmath22 60@xmath0 ( * ? ? ?
* previous range was 280@xmath0 @xmath22 60@xmath0 ) .
further observations are warranted , but this can give clues to the nature and geometry of the southern hemisphere emission regions and be of value to new decameter emission models .
in addition , the rhc component emissions from io - c are seen in excellent detail ; these observations , along with modulation lane measurements , give additional evidence that may suggest that both rhc and lhc emissions are coming from the same hemisphere source regions .
observations of io - b and io - d emissions are clearly seen with the dual polarization lwa1 measurements .
the io - b data show characteristic arc and nested band structures , normal envelope and tail structures due to source geometry and beaming , and characteristic s - bursts that occur early in the io - b events .
the io - d events are easily seen and analyzed with the lwa1 instrument .
the io - d events show characteristic lhc polarization and one event ( figure [ fig:997_full ] ) shows narrow band structure .
the lwa1 instrument shows excellent fine structure in jupiter s emission .
modulation lanes are clearly seen in both polarizations and drift rate measurements are consistent with earlier results .
however , modulation lane measurements from one io - a / io - c event where both polarizations are observed simultaneously ( figure [ fig : mod_lanes ] ) might provide good constraints on the emission theory and growth mechanisms .
more detailed analysis of additional io - a / io - c events are underway to further investigate the modulation lane details .
faraday fringes are also clearly seen in the lwa1 data and provide clues about the local conditions at the emission source , in the io plasma torus , and the earth s ionosphere . finally , and perhaps most interestingly , we have made excellent spectral and temporal measurements of s - bursts , n - events , and s - n events .
first we have clearly detected s - bursts in the io - d source , a measurement that is rare and will be useful to help understand the mechanism of s - burst emission .
drift rates of -12 mhz / s were measured for the io - d region and are consistent with known drift rates from other io - related sources .
several n - events are seen in our data in io - a , io - b , and io - d events , but most interesting are the s - n events seen in an io - b observation ( figures [ fig : iob_sn_wide]-[fig : iob_sn_zoom ] ) .
these events display complex interaction where s - bursts appear to quench n - events , and n - events may trigger more s - bursts .
these effects demonstrate that the emission mechanisms for the s - n events are similar and/or interacting , and that they are generated in the same spatial location . as mentioned previously , more observations of io - d
are being made in anticipation of further study of the s - bursts measured with the lwa1 instrument .
we thank the referees for a careful reading of the manuscript and helpful comments .
we also wish to thank the staff of the long wavelength array .
basic research in radio astronomy at the naval research laboratory is supported by 6.1 base funding .
construction of lwa1 was supported by the office of naval research under contract n00014 - 07-c- 0147 .
support for operations and continuing development of lwa1 is provided by the national science foundation under grants ast-1139963 and ast-1139974 of the university radio observatories program .
cah acknowledges support from the tennessee space grant consortium . this research has been supported in part by jsps kakenhi grant number 25400480 .
js ( a student at montgomery blair high school ) was supported by the science and engineering apprenticeship program ( seap ) at the naval research laboratory .
gbt acknowledges partial support for this work from nsf grant ast-1212162 .
data used for this publication are available through request to the authors .
zarka , p. 1988
, planetary radio emissions ii ; proceedings of the second international workshop , graz , austria , sept .
7 - 9 , 1987 ( a89 - 39951 17 - 91 ) . wien , verlag der oesterreichischen akademie der wissenschaften , 1988 , p. 327
- 342 . , 327 | new observations of jupiter s decametric radio emissions have been made with the long wavelength array station 1 ( lwa1 ) which is capable of making high quality observations as low as 11 mhz .
full stokes parameters were determined for bandwidths of 16 mhz . here
we present the first lwa1 results for the study of six io - related events at temporal resolutions as fine as 0.25 ms .
lwa1 data show excellent spectral detail in jovian dam such as simultaneous left hand circular ( lhc ) and right hand circular ( rhc ) polarized io - related arcs and source envelopes , modulation lane features , s - bursts structures , narrow band n - events , and interactions between s - bursts and n - events .
the sensitivity of the lwa1 combined with the low radio frequency interference environment allow us to trace the start of the lhc io - c source region to much earlier than typically found in the literature .
we find the io - c starts as early as = 230@xmath0 at frequencies near 11 mhz .
this early start of the io - c emission may be valuable for refining models of the emission mechanism .
we also detect modulation lane structures that appear continuous across lhc and rhc emissions , suggesting that both polarizations may originate from the same hemisphere of jupiter .
we present a study of rare s - bursts detected during an io - d event and show drift rates are consistent with those from other io - related sources .
finally , s - n burst events are seen in high spectral and temporal resolution and our data strongly support the co - spatial origins of these events . |
the type ia supernovae ( sne ia ) observational data has forced us to discard or change the majority of theoretical cosmological models supposed to be correct until the second half of the last decade @xcite .
the crossing of the sne ia statistics with other observational data , like the anisotropy of the cosmic microwave background radiation ( cmbr ) @xcite , gravitational lensing @xcite , the x - ray gas mass fraction of galaxy clusters @xcite , etc , leads to a scenario where the matter content of the universe is described by an unclustered component of negative pressure , the dark energy , and a clustered component of zero pressure , the cold dark matter .
there are many candidate for dark energy , the most natural seems to be the cosmological constant @xcite , since it can be connected with the vacuum energy in quantum field theory @xcite , but the small value resulting from observations for the energy density of the cosmological constant term yields a discrepancy of about @xmath2 orders of magnitude with the theoretically predicted value @xcite . among many other possibilities , for example
there is the quintessence model with scalar fields @xcite . here
we will focus on the chaplygin gas models ( cgm ) @xcite .
it is based on a string inspired configuration that leads to a specific equation of state where pressure is negative and varies with the inverse of the density @xcite .
this model has been generalized , giving birth to the generalized chaplygin gas model ( gcgm ) , where now the pressure varies with a power of the inverse of the density @xcite .
these proposals have many advantages , among which we can quote the following : in spite of presenting a negative pressure , the sound velocity is positive , what assures stability @xcite ; these models can unify the description of dark energy and dark matter , since the fluid can clusters at small scale , remaining a smooth component at large scales @xcite ; the cgm has an interesting connection with string theory @xcite .
some criticisms have been addressed to the gcgm ( cgm ) mainly connected with its features related to the power spectrum for the agglomerated matter @xcite . however , in our opinion , this specific criticism is not conclusive , since the introduction of baryons may alleviate the objections presented against the cosmological scenarios based on the gcgm ( gcm ) @xcite .
the gcgm ( generalized chaplygin gas model ) is defined as a perfect fluid with an equation of state given by @xmath3 where @xmath4 and @xmath5 are constants . when @xmath6 we re - obtain the equation of state for the cgm ( chaplygin gas model ) , the traditional chaplygin gas model .
@xcite for more detailed definitions of the chaplygin gas models used in the present work .
all free parameters for each model are considered . in the case of the gcgm
there are five free parameter : the hubble constant @xmath7 ; the equation of state parameter @xmath5 ; the `` sound velocity '' related parameter @xmath8 ; the density parameter for the pressureless matter @xmath9 ; the density parameter for the chaplygin gas @xmath10 ( or alternatively the density parameter for the curvature density of the universe @xmath11 ) . for the cgm ,
the number of parameters reduce to four , since @xmath6 .
we also consider the @xmath12cdm , where the number of parameters reduce to three : @xmath7 , @xmath9 and @xmath10 ( alternatively , @xmath13 ) .
one important point is how to perform this statistical analysis : the final conclusions may , in some cases , depend on the statistical framework ( bayesian , frequentist , etc . ) , as well as on the parameters that are allowed to be free , and how these parameters are constrained ( through a joint probability for two parameters , minimizing the error function or through a marginalization of all parameters excepted one , etc . ) . in some cases
, the different procedures adopted may lead to quite different conclusions on the best value for a given set of parameters .
the choice of the observational data sample may of course be important as well .
the present work is intended to : 1 .
announce the first public version of * betocs * ( * * b**ay**e**sian * * t**ools for * * o**bservational * * c**osmology using * * s**ne ia @xcite ) , which is a powerful and high productivity tool aimed to help the theoretical physicist community investigate cosmological models using type ia supernovae ( sne ia ) observational data .
betocs is a freeware and open source tool written in the _ mathematica _
@xcite language .
the _ mathematica _ notebooks of betocs contain documentation , source code and practical examples with textual and graphical outputs
emphasize that fixing @xmath7 is not acceptable , yielding arbitrary parameter estimations and usually bad best - fittings . on the other hand ,
the hst ( hubble space telescope ) prior @xcite for @xmath7 implies minor effects on all best - fittings and parameter estimations ( as shown here comparing the tables with flat and hst priors for @xmath7 ) , so its use is just a matter of choice .
nevertheless , it is recommended to use the hst prior as it is standard when using other observational cosmological data ( x - ray gas mass fraction , etc ) .
3 . continue the work of refs .
@xcite and show a very complete view and analysis of the chaplygin gas models ( generalized and traditional ) using sne ia .
the present work specially fixes an error on ref .
@xcite when calculating the best - fittings and parameter estimations for non - flat universes ( @xmath14 ) , due to an earlier version of betocs with wrong optimization code .
b@xmath15ut the side - effects of this error were not critical , just worsening some positive features
. the best - fitting parameters and the parameter estimations were calculated for each case of gcgm , cgm or @xmath0cdm using a corresponding * betocs * ( * * b**ay**e**sian * * t**ools for * * o**bservational * * c**osmology using * * s**ne ia @xcite ) notebook ( written in _
@xcite language ) containing the : definitions of the theoretical cosmological model , reading of the observational sne ia data , @xmath16 definition , bayesian tools library , calculation of the bayesian pdf , global maximization of pdf , pdf visualization in @xmath17 dimensions ( if available ) , pdf visualization and analysis in @xmath18 dimensions and finally pdf visualization in @xmath19 dimension with parameter estimation .
this article does not include all the graphics and analyses of the betocs notebooks , but some of them are available on the internet site of the betocs project @xcite .
this paper is organized as follows . in section [ sectionbestfitting ]
we detail the best - fitting analysis using betocs , with many results presented in tables .
section [ sectionparameterestimation ] explains how the parameter estimations are made using betocs , such that a detailed bayesian analysis is performed for each independent and dependent parameter , with the results shown in many tables and figures .
the conclusions are discussed in section [ sectionconclusions ] .
in order to compare the theoretical results with the observational data , the first step in this sense is to compute the quality of the fitting through the least squared fitting quantity @xmath20 . in the case of flat priors for all independent parameters of the theoretical cosmological model
, we get @xmath21 in this expression , @xmath22 is the distance moduli observationaly measured for each supernova of the @xmath23 gold sne ia dataset @xcite , @xmath24 is the value calculated through the theoretical cosmological model , @xmath25 is the measurement error and includes the dispersion in the distance modulus due to the dispersion in galaxy redshift due to peculiar velocities , following ref .
it is useful to define @xmath26 : @xmath20 divided by the number of degrees of freedom of the observational data , i.e. , the number of sne ia , here @xmath23 .
as we also want to compare the fitting and estimation of the parameters without priors and with the hst ( hubble space telescope ) prior @xcite for @xmath7 , then the @xmath20 used for the calculations with the hst prior is simply @xmath27 .the best - fitting parameters , i.e. , when @xmath28 is minimum , for each type of spatial section and matter content of the generalized chaplygin gas model .
@xmath7 is given in @xmath29 , @xmath30 in units of @xmath31 , @xmath32 in @xmath33 and @xmath34 in units of @xmath35 . [ cols="^,^,^,^,^,^,^ " , ] note that the marginalized estimations differs substantially from those extracted from the minimization of @xmath36 , which gives a large positive best value @xmath37 , but the dispersion is quite high , so even large positive values are not excluded , at least at @xmath38 level . and
figure [ figsalphaa ] for the joint probabilities for @xmath37 and @xmath39 and figures [ figsalpha ] and [ figsa ] for @xmath37 and @xmath39 , respectively , clearly show that the marginalization process changes the peak values and credible regions depending on the number of dimensions . of course , the cgm is obtained when @xmath37 is fixed to unity . from the analysis of the gcgm
it can be inferred that @xmath40 , i.e. , the cgm is favoured with a probability of @xmath41 . restricting to null curvature or fixing the pressureless matter density increases considerably this value , from @xmath42 to @xmath43 .
analogously , the probability to have @xmath44 ( with more physical meaning ) is @xmath45 and this value is quite increased when one or two parameters are fixed . both @xmath46 and @xmath47
are increased with respect to ref .
@xcite . like ref .
@xcite , the results indicate that the value of @xmath8 is close to unity , but now the dispersion is slightly smaller . in the case of gcgm with no fixed parameters
, the marginalization of the remaining four other parameters leads to @xmath48 .
again , this could suggest the conclusion that @xmath12cdm ( @xmath49 ) model is favoured .
however , the accuracy of the computation , due to the step ( between @xmath50 and @xmath51 ) used in the evaluation of the parameter , does not allow this conclusion .
instead , it means the peak happens for @xmath52 .
in fact , fixing the curvature or the pressureless matter , the preferred value differs slightly from unity , for example the quartessence scenario , @xmath53 , yields @xmath54 .
but , differently from ref .
@xcite , the cgm now predicts a best value for @xmath8 smaller than unity , @xmath55 , and the best value becomes smaller when one or two parameters are fixed . in figure [ figsa ]
the pdf for @xmath8 is displayed , both for the gcgm and the cgm , where the marginalization is made in all other parameters .
note that the probability to have @xmath56 ( meaning how much the @xmath0cdm is rule out ) is zero only for the gcgm ( due to the step used in the evaluation of @xmath8 ) , but this probability varies from about @xmath57 to @xmath58 for other cases . in figure [ figsalphaa ] the joint probabilities for @xmath37 and @xmath39 are displayed , with a non - gaussian shape . comparing with ref .
@xcite , the peak values now happen for large values of @xmath37 and @xmath59 .
this figure , compared to figures [ figsalpha ] and [ figsa ] , is an illustration of the importance of the marginalization process because it changes the peak values and credible regions depending on whether two or one - dimensional parameter space is used . the unified scenario of quartessence , with no pressureless matter , is again favoured as , for example , the gcgm and cgm ( without fixed parameters ) predict @xmath62 and @xmath63 , respectively .
the same estimations from ref .
@xcite , @xmath64 , and ref .
@xcite , @xmath65 , show that the increased number of sne ia has substantially decreased the estimated error .
compared to ref .
@xcite , the dispersion is also highly decreased , which once more favours the quartessence scenario .
the case of flat universe has an even smaller dispersion for the quartessence model .
see figures [ figsgcgomegas ] and [ figscgomegas ] . like ref .
@xcite , the gcgm , the cgm and the @xmath0cdm remarkably lead to almost the same predictions concerning the dark energy component , @xmath61 , when all parameters are free : @xmath66 , @xmath67 and @xmath68 , respectively . by comparing with ref .
@xcite , now @xmath61 ( which behaves as @xmath69 for @xmath70cdm ) has lower values and narrower dispersions , like @xmath60 , see figures [ figsgcgomegas][figslcdmomegas ] .
the joint probability for @xmath60 and @xmath61 is now smoother for the gcgm and the cgm cases ( figure [ figcgccgomegam0c0 ] ) , with the @xmath71 , @xmath72 and @xmath73 contours of figures [ figcgccgomegam0c0 ] and [ figlcdmomegam0c0 ] showing significantly smaller regions .
so the analysis for @xmath0cdm is now quite in agreement with the results of ref .
@xcite . in comparison with ref .
@xcite , a closed universe is still clearly favoured , but with slightly smaller probability , i.e. , @xmath75 .
but more important , the dispersion for @xmath74 is now substantially narrowed , as shown by figures [ figsgcgomegas][figslcdmomegas ] . the probability to have a flat universe , @xmath76 , is now greater ( @xmath77 , @xmath78 and @xmath79 for gcgm , cgm and @xmath0cdm ) , and after setting the pressureless matter it increases , exception being the @xmath0cdm case . with respect to refs .
@xcite ( using the selected 26 sne ia data ) , the dispersion has also significantly decreased , for example : @xmath80 and @xmath81 versus @xmath82 @xcite and @xmath83 @xcite , for respectively gcgm and cgm . due to the larger number of sne ia used here with respect to refs .
@xcite , the dispersions have been quite decreased . for example , @xmath85 and @xmath86 estimated here for the gcgm and cgm versus @xmath87 @xcite and @xmath88 @xcite , respectively . the predicted age of the universe when no parameter is fixed has increased with respect to ref .
@xcite , fortunately not anymore dangerously near the recent estimations age of the globular clusters @xcite , @xmath89 .
the values for the deceleration parameter @xmath90 are increased ( less negative ) with respect to ref .
the estimated errors are significantly smaller than the ones of refs .
@xcite , for example : @xmath91 and @xmath92 versus @xmath93 @xcite and @xmath94 @xcite , for respectively gcgm and cgm . in all cases , @xmath95 ,
the probability to have an accelerating universe today , is equal to or very near @xmath96 .
another useful quantity is the scale factor at the moment the universe begins to accelerate , @xmath34 , keeping in mind that the scale factor is normalized with its present value @xmath35 equal to unity .
with respect to ref .
@xcite , @xmath34 decreases when no parameter is fixed . as expected , the larger number of supernovae in comparison with ref .
@xcite yields smaller credible intervals for @xmath34 , for example @xmath97 estimated here for gcgm versus @xmath98 of ref .
@xcite .
the probability the universe begins to accelerate before today , @xmath99 in tables [ tableparestgcg] [ tableparestlcdmp ] , is essentially @xmath58 , being approximately the same value of the probability to have an accelerating universe today , i.e. , @xmath95 .
theoretically they should be the same , so the fact that these independent probability calculations agree almost exactly shows the accuracy and reliability of the bayesian probability analyses of this work .
the present work has performed the most extensive analysis of the gcgm and cgm in what concerns the comparison of theoretical predictions with the type ia supernovae data , using the 157 data of the `` gold sample '' . by using the high productivity of * betocs * ( * * b**ay**e**sian * * t**ools for * * o**bservational * * c**osmology using * * s**ne ia @xcite ) , it was possible to make all the best - fittings , parameter estimations and figures shown here .
* the probability to have @xmath100 ( with more physical meaning ) is increased , as well as both @xmath101 and @xmath102 ; * value of @xmath8 now has slightly smaller dispersion ; * the dispersion of @xmath9 and @xmath10 are also highly decreased , favouring even more the quartessence @xcite scenario ; * the dispersion for @xmath11 is now substantially narrowed , and the probability to have a flat universe is now greater ; * predicted age of the universe ( when no parameter is fixed ) has increased , fortunately not anymore dangerously near the recent estimations age of the globular clusters .
one important result concerns the hubble parameter , @xmath7 .
it is not acceptable to fix its value because arbitrary parameter estimations and usually bad best - fittings are obtained .
as the hst ( hubble space telescope ) prior @xcite for @xmath7 implies minor effects on all results , we can choose a flat or hst prior .
the cgm ( traditional chaplygin gas model ) , where @xmath6 , remains competitive and preferred in many cases : when the five parameters are considered , the probability is @xmath41 , but increases as much as to @xmath103 for the quartessence scenario . for the parameter @xmath8 , both gcgm and cgm cases of fixed curvature and matter densities shows a value near but small than @xmath19 as the best value of @xmath104 , such that the @xmath12cdm case ( @xmath49 ) is almost ruled out .
the results indicate that , for the gcgm , cgm and @xmath12cdm , a closed universe is favoured .
the gcgm and cgm favour the unified scenario ( quartessence ) , where the pressureless matter density is essentially zero .
there are many current and future applications @xcite and developments of betocs : using different sne ia data sets ( snls @xcite , etc ) , different priors , different cosmological models @xcite , etc .
we plan to release variant versions for other observational cosmological data @xcite , like * betocx * ( * * b**ay**e**sian * * t**ools for * * o**bservational * * c**osmology using * x*-ray gas mass fraction of galaxy clusters ) , so it will be possible to cross the estimations from different observational data . | the type ia supernovae observational data is one of the most important in observational cosmology nowadays . here
we present the first public version of * betocs * ( * * b**ay**e**sian * * t**ools for * * o**bservational * * c**osmology using * * s**ne ia ) , which is a powerful and high productivity tool aimed to help the theoretical physicist community investigate cosmological models using type ia supernovae ( sne ia ) observational data .
betocs is applied to the generalized chaplygin gas model ( gcgm ) , traditional chaplygin gas model ( cgm ) and @xmath0cdm , ranging from 5 to 3 free parameters , respectively .
the `` gold sample '' of 157 supernovae data is used .
it is shown that the chaplygin gas scenario is viable ( in most cases the @xmath0cdm is disfavoured ) and the quartessence scenario ( that unifies the description for dark matter and dark energy ) is favoured .
the hubble parameter ( @xmath1 ) is important and should not be fixed and it can be estimated or marginalized with or without the hubble space telescope prior .
pacs number(s ) : 98.80.bp , 98.80.es , 04.60.gw |
methods of nonlinear time series analysis have been widely applied in physics , physiology , finance and biology .
complex network theory has become one of the most important developments in statistical physics @xcite .
recent studies have shown that complex network theory may be an effective method to extract the information embedded in time series @xcite .
many complicated dynamics systems in nature and society can be described by complex networks . in a complex system ,
its elements are represented by nodes , and their interactions are represented by directed or undirected edges .
based on the small - world and scale - free properties of complex networks @xcite , newman _
@xcite investigated extensively the structure and function of real - world complex networks in different areas .
the advancement of network theory provides us with a new perspective to perform time series analysis @xcite .
therefore , we can further understand the structural features and dynamics mechanism of complex systems by studying the basic topological properties of networks .
many algorithms have been proposed to construct different complex networks from time series @xcite , such as visibility graphs @xcite , space state networks @xcite , recurrence networks @xcite , and nearest - neighbor networks @xcite .
recurrence is a basic characteristic of many complex dynamical systems , which can be used to describe the dynamics behavior of systems .
@xcite developed the method of recurrence plots ( rps ) to visualize the recurrence property of complex dynamical systems .
a remarkable advantage of this method is that it is very suitable for short and even non - stationary time series .
applications of the method of rps can be found in various fields of research such as physiology , neuroscience , earth sciences , engineering , biology and finances . in order to quantify the structures of the recurrence plots , zbilut and webber
developed the recurrence quantification analysis ( rqa ) based on linear and diagonal line structures of recurrence plots @xcite .
most of these measures of complexity are available on tocsy software platform @xcite .
cross recurrence plot ( crp ) @xcite and joint recurrence plot ( jrp ) @xcite are bivariate and multivariate extensions of the rps , respectively .
recently , donner _
_ @xcite and marwan _ et al . _
@xcite proposed the recurrence network based on the method of rps .
this novel idea can be considered as a complementary view on the rqa , allowing us to gain additional information from corresponding network - theoretic measures which can not be obtained by rqa .
they considered the phase space state vectors @xmath7 ( defined in section ii ) as nodes of a complex network and identify the recurrence matrix of the original time series with the adjacency matrix of an associated complex network . in this way
, an undirected and unweighted network is represented by a binary adjacency matrix @xmath8 ( defined in section ii ) .
recurrence networks have been used to many model systems such as logistic map , hnon map , lorenz system , rssler system and bernoulli map @xcite , and real economic series @xcite .
it is well known that fractional brownian motion ( fbm ) described by the hurst index @xmath1 @xmath9 is a self - similar process with stationary increments @xcite .
the fbm with @xmath10 is in fact the classical brownian motion .
the correlation between increments of the fbm is negative if @xmath11 and positive if @xmath12 .
the sample paths of fbm series are relative more rough and variable for small hurst index @xmath1 , while the sample paths of fbm series are more smooth for large hurst index @xmath1 . in this work ,
we try to reveal the relationship between fbm and its related recurrence network from the perspective of network structure .
this prompted us to further study the fundamental topological and fractal properties of recurrence networks constructed from fbms .
we think that applying recurrence network analysis ( rna ) to fbm can yield results which are theoretically meaningful .
the reasons are : ( a ) visibility graphs and recurrence networks are two main classes of methods to map time series to networks .
recently some works have been done to study the relationship between the exponents of newly developed methods(including visibility graphs ) and the hurst index of the associated fbms @xcite .
( b ) donges _ et al .
_ @xcite have proposed an analytical framework for rna of time series of chaotic maps and stochastic processes ( such as uniformly distributed noise and gaussian noise ) .
fbms are generalization of gaussian noise .
( c ) donner _ et al . _
@xcite applied rna to study the real economic time series .
and it is well known that fbm has been used as a theoretical framework to study economic time series @xcite . in 1999 , riley _
@xcite applied the rqa technique for analyzing center of pressure ( cop ) signals and the nonstationarity of the cop is expected from a consideration of cop trajectories as fbm .
the remaining of this paper is organized as follows . in section 2
, we adopt the recurrence plot method to construct recurrence networks . in section 3
, we investigate the basic topological characteristics of the recurrence networks constructed from fbms .
we next introduce the random sequential box - covering method @xcite to calculate the fractal dimension of the recurrence networks in section 4 . in section 5
, we introduce an improved algorithm newly developed by our group @xcite , which is based on the modified fixed - size box - counting algorithm @xcite , to probe the multifractal behavior of recurrence networks .
we finally draw some conclusions in section 6 .
the method of recurrence plots is based on the theory of phase space reconstruction introduced by packard _
_ @xcite . in order to easily construct a more suitable space and
then to reveal more meaningful information from original time series , packard _
@xcite proposed the derivative reconstruction method and time delay method to reconstruct a finite - dimensional phase space .
however , in practice , we do not know any prior information of time series , therefore the latter method is widely used to reconstruct phase space of time series . for a given scalar time series @xmath13 , we construct the delay vectors @xmath14 for @xmath15 , where @xmath16 , @xmath17 is the embedding dimension and @xmath18 is the time delay .
takens embedding theorem ensures that we can recreate a topologically equivalent @xmath17-dimensional phase space from an infinite noise - free time series data by means of the time delay method @xcite .
the most significant advantage of phase space reconstruction is that it can preserve the geometrical invariants of the original system , such as the fractal dimension and the lyapunov exponents @xcite .
so we can investigate much about the dynamics characteristics of time series in phase space more easily .
the basic embedding theorem of phase space reconstruction assumes that we can choose the delay time @xmath18 without any limitation and the embedding dimension @xmath17 under the condition of @xmath19 , where @xmath20 is the fractal dimension of the underlying attractor .
however , since the observed time series data sets are finite and noisy in the most common case , the selection of the time delay @xmath18 and the embedding dimension @xmath17 is rather important for the quality of the phase space reconstruction .
the binary recurrence matrix @xmath21 is defined as @xmath22 where @xmath23 is a threshold distance , @xmath24 is a suitable norm ( e.g. , the @xmath25 - norm , the euclidean norm and the @xmath26 - norm ) in the considered phase space and @xmath27 is the heaviside step function ( i.e. @xmath28 if @xmath29 , and @xmath30 otherwise ) . the resulting matrix @xmath31 is a symmetric matrix where the elements @xmath32 or @xmath33 .
moreover , the matrix @xmath34 exhibits the line of identity ( the main diagonal ) @xmath35 . in a two - dimensional space ,
the recurrence plot is described by using different colors for different values of the recurrence matrix , e.g. , plotting a black dot at the coordinates @xmath36 if @xmath37 , and a white dot if @xmath38 .
thus the recurrence plot always has a black main diagonal line .
now we obtain the adjacency matrix @xmath39 of the recurrence network from the recurrence matrix @xmath34 as @xmath40 where @xmath41 is the kronecker delta function introduced here in order to avoid self - loops @xcite .
in complex network , each state vector @xmath7 of the reconstructed phase space represents a single node .
the topological structure of the recurrence network can be described with the adjacency matrix @xmath42 , and the elements @xmath43 and @xmath44 correspond to connection and disconnection , respectively . and
the size of the network is @xmath45 .
as mentioned above , the choice of the time delay @xmath18 and the embedding dimension @xmath17 plays an important role in the recurrence plot method
. if the time delay @xmath18 selected is too small , each coordinate of the vector @xmath7 will be so close to each other that the trajectories of the reconstructed phase space are compressed along the identity line , and this phenomenon is referred to as redundance .
if the time delay @xmath18 selected is too large , the coordinates of the vector @xmath7 are completely independent of each other and then the reconstructed attractor dynamics become causally disconnected , and this phenomenon is referred to as irrelevance @xcite . at present , auto - correlation function @xcite and mutual information method @xcite are the most common approaches for the estimation of the time delay @xmath18 .
@xcite proposed a recursive method of calculating mutual information and pointed out that the first local minimum of mutual information is the best criterion for choosing time delay @xmath18 in phase space reconstruction from time series data .
takens embedding theorem gives us a way to estimate the embedding dimension @xmath17 .
this is to say that all self - intersections of the orbit ( which is the reconstructed attractor ) in the reconstructed phase space will be eliminated for any embedding dimension @xmath46 .
the attractor will be completely unfolded in this embedding dimension @xmath17 .
but in practice , since the dimension @xmath20 is unknown in many cases , we can not directly calculate the embedding dimension @xmath17 from the takens embedding theorem .
many algorithms based on computing some invariants on the attractor have been proposed to calculate the minimum embedding dimension @xmath17 .
the most popular method for calculating the minimum embedding dimension is the false nearest - neighbors ( fnn ) algorithm @xcite . according to the takens embedding theorem
, we assume that @xmath17 is qualified as an embedding dimension . if any two points are close to each other in the @xmath17-dimensional phase space , then they are still close to each other in the @xmath47-dimensional phase space .
we call this two points true neighbors , otherwise , they are false neighboring points .
the main idea of the fnn algorithm is to measure the percentage of false nearest neighbors along a signal trajectory change with increasing embedding dimension .
the optimal embedding dimension means that there is no false neighbors and self - intersections in the phase space .
this indicates that the minimum embedding dimension is that for which the percentage of false nearest neighbors drops to zero or nearly zero for a given tolerance level for the first time @xcite .
the threshold distance @xmath23 is another key parameter of an rp .
the selection of the threshold @xmath23 depends strongly on the system considered .
if @xmath23 is too small , there may be almost no recurrence points and we can not learn anything about the recurrence structure of the underlying system . on the other hand , if @xmath23 is too large , almost every point is a neighbor of every other point , which leads to a lot of artifacts and redundant information @xcite .
therefore , special attention has been required on its choice . in practice ,
the most common method is to set @xmath23 for a certain proportion of the standard deviation @xmath48 of the original time series .
however , riley and van orden suggested choosing @xmath23 such that the recurrence point density remains low ( often smaller than 5% ) @xcite .
this method is often used in rqa .
the advantage of choosing an adaptive recurrence threshold by using a fixed recurrence rate in rqa is that it can preserve the recurrence point density and then allows us to compare the results of recurrence quantification analysis of different systems .
it is easy to imagine that there is no direct relationship between recurrence rate or link density and network connectivity . and
it is so difficult for us to find a fixed recurrence rate that all recurrence networks obtained at recurrence threshold @xmath23 calculated from the fixed recurrence rate are connected .
however , we restrict our attention to the consideration of connected network in this article .
therefor we do not adopt the fixed recurrence rate for obtaining such an appropriate recurrence threshold .
if we do so , then we will not be able to study the topological and fractal properties of the entire network .
in other word , we can only focus on the disconnected components or the largest connected component of each recurrence network .
the critical phenomenon is one of the most interesting findings of complex networks . in 1959 ,
erds _ et al . _
@xcite introduced the classical random graph model and studied the structural phase transition of the birth of the giant connected component in the network architectures .
every pair of nodes of random graph is connected with probability @xmath49 .
they found that there exists a critical probability @xmath50 where a giant connected component was formed and it can be spread across the entire network . if the probability @xmath49 is less than this critical probability @xmath50 , the network is composed of disconnected components .
similarly to the critical behavior of random - graph theory , we study the critical connectivity of the recurrence networks by examining the size of the largest connected component and then determine the critical recurrence threshold @xmath51 . here
we firstly employ the connection rate of network to quantify the connectivity of the recurrence network , which is defined as the number of nodes of the largest connected component divided by the size of the network .
secondly , we observe the change of connection rate of network with the increase of the parameter @xmath23 .
it can be easily imagined that the connection rate of network increases when the parameter @xmath23 becomes larger , as illustrated in fig .
1 . in fig . 1 , we set the recurrence threshold @xmath23 ranging from @xmath52 to @xmath53 with a step of @xmath54 , where @xmath48 is the standard deviation of the original fbm series .
finally , we choose an appropriate recurrence threshold @xmath51 where the connection rate of the network arrives at @xmath33 for the first time . as can be seen in fig . 1 , we set the critical recurrence threshold @xmath55 .
the recurrence network obtained at recurrence threshold @xmath51 will be connected .
the threshold @xmath23 beyond this critical value @xmath51 will result in redundant connections among nodes , and threshold @xmath23 below this critical value @xmath51 will lead to a number of disconnected components where each component only contains a connected subnetwork .
consequently , we select this suitable recurrence threshold @xmath51 to construct the recurrence network and then investigate its topological and fractal properties . in addition
, we also notice that the critical recurrence threshold @xmath51 decreases with the increase of the hurst index @xmath1 from 0.4 to 0.95 . we know that the fbm series is relatively more smooth and regular for large hurst index @xmath1 .
the consequence is that there are many data points in the original fbm series are mapped to state vectors which are close to each other in a dense region of the phase space so that the largest connected component of the recurrence network contains more nodes than the one of the network with the small hurst index @xmath1 .
the specific explanation for this phenomenon can be found in the second subsection of the section 3 . in this work ,
we consider fbm of length @xmath56 with different hurst indices @xmath1 ranging from 0.05 to 0.95 ( the step difference is 0.05 ) .
recurrence networks can be constructed by the following steps : 1 . for a given fbm time series
, we use the mutual information method @xcite and the fnn algorithm @xcite to calculate the time delay @xmath18 and the embedding dimension @xmath17 , respectively .
these two procedures are available on tocsy software platform @xcite .
the values of embedding dimensions @xmath17 we got vary from 5 to 7 , and those of the time delay @xmath18 vary from 10 to 20 .
we use the connection rate of the recurrence network to determine the suitable threshold distance @xmath51 @xcite .
we observe the connection rate of network reaching @xmath33 for the first time when the parameter @xmath23 ranging from @xmath52 to @xmath57 with a step of @xmath54 , where @xmath48 is the standard deviation of the original fbm series .
the connection rate of network can be calculated by matlab - bgl toolbox @xcite in matlab .
3 . based on the above steps
, we can calculate the recurrence matrix @xmath34 by eq .
then we obtain the binary adjacency matrix @xmath42 of an unweighted and undirected recurrence network by eq .
( 3 ) . as a result
, the recurrence network is also connected .
for clarity of visualization , we here only draw the recurrence network for the fbm series of length @xmath58 with hurst index @xmath59 . after obtaining the adjacency matrix @xmath42 of recurrence network according to the above three steps ,
we convert the adjacency matrix @xmath42 into pajek " @xcite format and then we adopt the pajek " which is a freely available software platform for visualizing and analyzing complex networks to visualize the recurrence network . as can be clearly seen in fig .
2 , the recurrence network constructed from the associated fbm series is connected .
in addition , we find that the connection rates of recurrence networks constructed from fbms with hurst indices @xmath60 , and @xmath61 are far less than 1 . in these cases ,
the recurrence networks are not connected .
so in the following sections , we only consider connected networks with hurst indices @xmath1 from 0.4 to 0.95 .
in this section , we numerically study the basic topological features of recurrence networks including the degree distribution , the clustering coefficient and the motif distribution .
recent works have shown that a large number of real - world networks are referred to as scale - free networks because the degree distributions @xmath62 follows a power - law @xmath63 where the degree exponent @xmath64 varying in the range @xmath65 @xcite .
the degree distributions of the visibility graphs constructed from fbms have power - law tails @xcite .
in addition , xie _ et al . _ @xcite found that horizontal visibility graphs constructed from fbms have exponential degree distributions @xmath66 and the degree exponent @xmath67 increases with hurst index @xmath1 .
it has been shown that many real - world complex networks share similar universal statistical features , such as the small - world character and the scale - free distribution , but may demonstrate different local structure characteristics @xcite .
an in - depth study of motif distribution can help us to understand the design principles of complex networks on the local structure level @xcite . in 2004 ,
@xcite developed an approach for comparing network local structures , basing on the significance profile ( sp ) .
they calculated the triad significance profile ( tsp ) for networks from different fields and found that networks with similar characteristic profiles are grouped into superfamilies .
degree distribution is one of the most fundamental and important topological properties of complex networks .
we simulated fbms with different hurst indices @xmath1 ranging from 0.40 to 0.95 with an increment of 0.05 .
for each value of @xmath1 , we generated 1000 realizations and then calculated the average degree distribution .
for example , we give the average degree distribution of recurrence networks constructed from fbms with hurst index @xmath59 in fig .
we find that all the recurrence networks constructed from fbm series exhibit exponential degree distributions .
we estimated the degree exponents @xmath67 and display the average degree exponents @xmath0 in fig .
4 . from fig .
4 , we can see that the parameter @xmath67 increases first and then decreases with the increase of @xmath1 .
the inflection point is at around @xmath68 .
we find that the relationship between @xmath1 and @xmath0 can be fitted by a cubic polynomial function which is also shown in fig .
4 . the clustering coefficient measures the density of triangles in a complex network . the definition of the clustering coefficient given by watts and strogatz in ref .
@xcite has been used quite widely in complex network .
the local clustering coefficient @xmath69 is defined as @xmath70 where @xmath71 is the degree of node @xmath72 , @xmath73 is the actual number of edges among @xmath71 nearest neighbor nodes of the node @xmath72 . the clustering coefficient of the whole network is then given by the average of @xmath69 over all the nodes in the entire network : @xmath74 from the geometrical point of view , the local clustering coefficient @xmath69 also can be defined as @xmath75 where a triple represents a vertex with edges connected any two neighbor vertices .
there are two different triples as shown in fig .
5 . the triple @xmath76 will occur if the three vertices are all close to each other .
conversely , the triple @xmath77 will occur if the vertex @xmath72 is connected to its two neighbor vertices @xmath78 and @xmath79 , but vertices @xmath78 and @xmath79 are not connected . in network topology ,
the triangle structure @xmath76 means transitivity and stability in the network . in this paper
, we study the dependence of the clustering coefficient @xmath80 of the recurrence networks against the hurst index @xmath1 of the original fbms . for each @xmath1 , we simulated the fbm @xmath81 times , hence we got 1000 fbm series .
for each fbm series , a recurrence network was constructed and its clustering coefficient @xmath80 was calculated .
we display the relationship between @xmath1 and the average clustering coefficient @xmath82 of recurrence networks in fig .
6 . from fig
. 6 , we can see that the average clustering coefficient @xmath82 increases with the hurst index @xmath1 .
the results show that there are more triangle structures in the recurrence network when the hurst index @xmath1 increases from 0.4 to 0.95 .
this means that the transitive and stable structure is more common for large hurst indices .
this phenomenon can be easily explained by following points . it is well known that the hurst index characterizes the raggedness and irregularity of fbm .
as already mentioned , the fbm is a self - similar process with stationary increments and possesses long - range dependence .
the hurst index @xmath1 @xmath83 is a measure of the intensity of long - range dependence in a fbm series .
more specifically , the fbm with @xmath4 indicates the absence of long - range dependence .
and the closer hurst index @xmath1 is to @xmath33 , the greater the extent of long - range dependence or persistence .
the hurst index @xmath1 less than @xmath84 corresponds to anti - persistence , which means that the process displays strong negatively correlated and fluctuates violently . for large hurst index @xmath1 , the fbm series is relatively more smooth and regular .
the result is that close data points in the original fbm series are mapped to state vectors @xmath7 which are close to each other in a dense region of the phase space so that these nodes @xmath7 are more likely to be connected to each other in recurrence network . as expected , there are many more transitive and stable structures @xmath76 are to appear in recurrence network constructed from the fbm series with large hurst index @xmath1 .
however , the fbm series with small hurst index @xmath1 is relatively more rough and irregular . as a consequence of this ,
these close points in the original fbm series are mapped to state vectors @xmath7 in a sparse region of the phase space that are less likely to be mutually connected , and therefore the triple @xmath77 will be more common . from this point of view , the recurrence network reflects the smoothness and regularity of the fbms to some extent . because it is very time - consuming to study next a few properties of large complex networks , we only generated 100 realizations of fbm for each value of hurst index @xmath1 and calculated the average in the following .
researches have shown that subgraphs or motifs are the building blocks of complex networks and the superfamilies of networks can be defined as the network motif patterns of occurrence @xcite .
distinct from the degree distribution and the clustering coefficient , the occurrence frequency of small subgraphs or motifs can characterize the local structural properties of real - world networks and then map the relation between the function and the local structure of real - world systems @xcite .
_ @xcite observed the superfamily phenomenon in the nearest - neighbor networks constructed from different time series .
they also found that the distribution of network motif ranks can be used to distinguish and to characterize different types of dynamics in periodic , chaotic and periodic with noise processes @xcite .
@xcite calculated the frequencies of occurrence of the six motifs of size @xmath85 within the horizontal visibility graphs constructed from fbms . in this work ,
the constructed recurrence networks are connected and undirected , so here we only consider motifs with 4 nodes as in ref . @xcite .
7 shows all six different network motifs of size @xmath85 in connected and undirected networks .
these motifs reflects the different local structure characteristics in network .
there are two extreme blocks : motif @xmath86 and motif @xmath87 .
the motif @xmath86 means the most irregular and nontransitive structure that will appear if the node is connected to its three neighbor nodes , but these neighbors are not connected to each other . on the contrary , the motif @xmath87 indicates the most transitive structure that will appear if the three neighbors are also connected to each other .
the relative occurrence frequencies of motifs @xcite are defined as @xmath88 where @xmath89 is the number of motif @xmath90 in the network . after constructing the recurrence networks based on the recurrence plot method described above , we calculated the occurrence frequencies of various motifs within the recurrence networks by the network motif detection tool provided by milo _
_ @xcite . for each hurst index @xmath1 , which runs from 0.40 to 0.95 with a step of 0.05 ,
100 realizations of fbms were generated .
then we constructed these time series into 100 recurrence networks and obtained the average of @xmath91 . in fig .
8 , we rearrange the relative occurrence frequencies @xmath92 in descending order for different hurst indices @xmath1 . as we can see from fig .
8 , these seemingly unrelated recurrence networks are divided into two superfamilies based on the very similar motif rank distributions .
8(a ) shows a superfamily determined by the motif rank @xmath93 , which includes the recurrence networks constructed from fbms with hurst indices @xmath94 , and @xmath95 . the rest of recurrence networks with hurst indices @xmath96 belong to another superfamily and the motif rank pattern is @xmath97 ( fig .
8(b ) ) . in addition , three particular motifs @xmath98 , and @xmath87 play an important role in the classification of recurrence networks we constructed because their order determines the motif rank distributions .
a similar phenomenon was also observed in refs .
they gave a detailed explanation for this common phenomenon .
9 further illustrates the dependence of @xmath92 with respect to the hurst index @xmath1 .
more specially , the occurrence frequency @xmath99 of motif @xmath86 strictly monotonically decreases with the increase of the hurst index @xmath1 ( see fig .
in contrast to the motif @xmath86 , the frequencies @xmath100 and @xmath101 increase with the increase of the hurst index @xmath1 ( see fig .
9(e ) and fig .
we also note that the fluctuation of the frequencies @xmath102 is comparatively small in fig .
9(d ) , which means that the motif @xmath103 almost does not appear in all of the recurrence networks .
these results indicate that there are more regular and transitive structures in the networks when the hurst index @xmath1 varies from @xmath104 to @xmath105 .
that is to say , the fully transitive motif @xmath87 will be common in these recurrence networks constructed from the fbm series with large hurst index @xmath1 than the ones with small hurst index @xmath1 .
on the contrary , the irregular and nontransitive structure such as @xmath86 will be more common for these network with small hurst index @xmath1 .
these trends are roughly coincide with the ones in the previous subsection .
it should be noted that two triples @xmath76 and @xmath77 in fig .
5 are just two motifs with @xmath106 nodes in undirected network . measuring and analyzing the density distribution of the motifs with @xmath85 nodes
may be seen as a generalization of the aforementioned transitivity concept .
therefore , as mentioned in the previous subsection , we can give a similar explanation for these trends in this subsection .
in this sense , recurrence networks capture the correlation of the associated fbms .
in 1967 , mandelbrot introduced the fractal idea in ref . @xcite . in fractal geometry , a fractal object is self - similar because it contains small parts similar to the whole @xcite .
recently , the fractal and self - similarity properties of complex networks have been studied extensively in various fields and systems @xcite .
_ @xcite found that many real - world networks such as the world - wide web ( www ) , social networks , protein - protein interaction ( ppi ) networks and cellular networks consist of self - repeating patterns .
they also believed that these complex networks are self - similar under a certain length - scale .
after the small - world character and scale - free property , self - similarity has become the third basic characteristic of complex networks . to gain further understanding of complex networks , numerous algorithms have been developed to calculate the fractal dimension of complex networks .
@xcite proposed to calculate the fractal dimension via a box - counting method .
@xcite investigated the skeleton and fractal scaling in complex networks using an improvement algorithm which is a modified version of the original method introduced by song _
then zhou _ et al . _
@xcite developed an alternative algorithm , based on the edge - covering box - counting , to detect self - similarity of cellular networks . in this section
, we adopt the random sequential box - covering method proposed by kim _
@xcite for calculating the fractal dimensions of recurrence networks . for a given network ,
let @xmath107 be the smallest number of boxes of lateral size @xmath108 which are needed to cover the entire network .
the fractal scaling implies the power - law relationship between @xmath107 and @xmath108 ; the fractal dimension @xmath109 is then given by @xmath110 usually , the fractal dimension @xmath109 can be obtained by fitting the linear relationship between @xmath107 and @xmath108 in a log - log plot . before introducing the random sequential box - covering algorithm @xcite
, we use floyd s algorithm @xcite of matlab - bgl toolbox @xcite to calculate the shortest - path distance matrix @xmath111 for each network according to the adjacency matrix @xmath8 of the recurrence network . the random sequential box - covering algorithm @xcite can be described as follows .
we start with all vertices labelled as not burned . then , 1 .
select a vertex randomly at each step ; this vertex serves as a seed and then consider the seed as the center of a box .
2 . for the center of the box , search all the neighbor vertices within distance @xmath108 and burn all vertices which are found but have not been burned yet . assign the newly burned vertices to the new box .
if no newly burned vertex is found , then this box is discarded .
3 . repeat steps ( i ) and ( ii ) until all vertices in the entire network are assigned to their respective boxes . in this work , we simulated fbm series with different hurst indices @xmath1 ranging from 0.4 to 0.95 in the step of 0.05 .
for each @xmath1 , we simulated 100 fbm time series and then constructed the obtained series into 100 recurrence networks .
apparent power - law behaviors of the two typical empirical recurrence networks constructed from fbm series with different hurst indices are shown in fig .
the fractal dimension @xmath109 is the absolute value of the slope of linear regression between @xmath112 and @xmath113 for each hurst index .
11 shows the relationship between hurst index @xmath1 and the average fractal dimension @xmath2 over 100 realizations . as we can see from fig .
11 , the average fractal dimension @xmath2 decreases with increasing @xmath1 .
furthermore , it is surprising that the curve shows a nice linear relationship : @xmath114 which approximates the theoretical relationship between hurst index @xmath1 and the fractal dimension @xmath20 of the graph of fbm @xmath115 .
our numerical results show that the fractal dimension of the recurrence networks constructed is very close to that of the graph of the original fbms .
in other words , the fractality of fbms is inherited in their recurrence networks .
we can explain here why the two fractal dimensions are so closely related that the hurst index @xmath1 of the fbm series can be approximately estimated from the perspective of the complex network . as it has already been mentioned , the pairs of nodes in the recurrence network can be connected only when the two corresponding state vectors @xmath7 in phase space are close enough so that the phase space distance between them is less than the critical recurrence threshold @xmath51 .
this is to say that recurrence networks we constructed are random geometric graphs faithfully representing the geometry of a set in phase space .
hence , neighborhood sizes around a node in the network as well as around the corresponding state vector in phase space will show the same scaling behavior .
this results in analogous behavior of the derived fractal dimensions .
so far , we have numerically studied the basic topological properties and the fractal dimensions of the recurrence network .
more recently , some chaotic model systems such as logistic map , hnon map , generalised baker s map and bernoulli map , periodic and two - dimensional quasiperiodic motions have been studied analytically by donner _
_ @xcite and donges _ et al . _
@xcite from the point of view of @xmath23-recurrence network .
to better understand the interrelationships between network properties and dynamical system , they defined some newly continuous measures based on well - known graph theoretical measures and gave two novel notions of dimension in phase space ( clustering dimension and transitivity dimension ) .
they have put forward a theoretical framework for these new measures .
their results indicated that @xmath23-recurrence networks show a strong relation between dynamical systems and graph theory @xcite .
their work provides us with a possible way to derive the analytical results of these topological properties and dimensions for recurrence network of fbm series in our future work .
in real world , two fractal objects may have the same fractal dimension but looks completely different .
for these real - world fractals , however , the tool of multifractal analysis shows better performance and seems more powerful than fractal analysis .
multifractal analysis has been successfully applied in a variety of fields such as financial modelling @xcite , biological systems ( e.g. @xcite ) and geophysical data analysis ( e.g. @xcite ) .
@xcite applied the modified fixed - size box - counting algorithm to explore the multifractal behavior of some theoretical networks , namely scale - free networks , small - world networks , random networks , and a kind of real networks , namely ppi networks of different species .
their numerical results indicate that multifractality exists in scale - free networks and ppi networks , while for small - world networks and random networks their multifractality is not clear - cut .
fixed - size box - covering algorithm @xcite is one of the most common and important methods of multifractal analysis . for a given measure @xmath116 with support set @xmath117 in a metric space , we consider the partition sum @xmath118^{q } , \eqno{(12)}\ ] ] where @xmath119 and the sum runs over all different nonempty boxes @xmath120 of a given size @xmath121 in a covering of the support set @xmath117 . from the definition above
, we can obtain @xmath122 and @xmath123 .
the exponent @xmath124 of the measure @xmath116 can be defined as @xmath125 and the generalized fractal dimensions of the measure @xmath116 are defined as @xmath126 for @xmath127 , and @xmath128 for @xmath129 , where @xmath130 .
the linear regression of @xmath131/(q-1)$ ] against @xmath132 for @xmath127 gives numerical estimates of the generalized fractal dimensions @xmath133 , and similarly a linear regression of @xmath134 against @xmath132 for @xmath129 .
in particular , @xmath135 is the box - counting dimension ( or fractal dimension ) , @xmath136 is the information dimension , and @xmath137 is the correlation dimension . if the @xmath124 or @xmath133 curve versus @xmath138 is a straight line , the object is monofractal .
however , if this curve is convex , the object is multifractal . in order to calculate the exponent @xmath124 and the generalized fractal dimensions @xmath133 and then study the multifractality of networks , the measure @xmath116 of each box
is usually defined as the ratio of the number of nodes covered by the box and the total number of nodes in the entire network .
@xcite proposed a modified fixed - size box - counting algorithm to calculate the @xmath124 and @xmath133 and then investigate the multifractal behavior of complex networks . recently
_ @xcite made some improvements based on the modified fixed - size box - counting algorithm proposed by wang _
the improved algorithm @xcite can be summarized as follows : 1 .
initially , make sure all nodes in the entire network are not covered and no node is selected as a center of a box .
2 . according to the size @xmath45 of networks constructed , set @xmath139 appropriately .
rearrange the nodes into @xmath81 different random orders . more specifically , in each random order , nodes which will be selected as a center of a box are randomly arrayed .
3 . set the radius @xmath140 of the box which will be used to cover the nodes in the range @xmath141 $ ] , where @xmath20 is the diameter of the network .
4 . treat the nodes in the @xmath142-th random order that we got in ( ii ) as the center of a box , search all the neighbor nodes by distance @xmath140 from the center and cover all nodes which are found but have not been covered yet .
if no newly covered nodes have been found , then this box is discard .
. repeat steps ( iv ) and ( v ) until all the nodes are covered by the corresponding boxes .
we denote the number of boxes in this box covering as @xmath143 .
repeat steps ( iv ) to ( vi ) for all 1000 random orders to find a box covering with minimal number of boxes @xmath143 .
8 . for the nonempty boxes @xmath120 in the first box covering with minimal number of boxes @xmath143 , define the measure of this box as @xmath144 , where @xmath145 is the number of nodes covered by the box @xmath120 , and @xmath45 is the size of the network .
then calculate the partition sum @xmath146^{q}$ ] .
9 . for different value of @xmath140 , repeat steps ( iii ) to ( viii )
to calculate the partition sum @xmath147 and then use the @xmath148 for linear regression .
a key step of linear regression is to obtain the appropriate range of @xmath149 $ ] .
then we calculate the exponents @xmath124 and generalized fractal dimensions @xmath133 in the scaling ranges . in our calculation
, we obtain the generalized fractal dimensions through a linear regression of @xmath150/(q-1)$ ] against @xmath151 for @xmath127 , and similarly a linear regression of @xmath152 against @xmath151 for @xmath129 , where @xmath153 .
12 shows the linear regression for the recurrence network constructed from fbm with hurst index @xmath154 .
the numerical results show that the best fit occurs in the range @xmath155 for this case ( @xmath59 ) .
we select the best linear fit scaling range to calculate the exponents @xmath124 and generalized fractal dimensions @xmath133 and then to determine the multifractality of recurrence networks from the shape of these curves . in this paper
, we detect the multifractal behavior of recurrence networks using our improved version of the modified fixed - size box - counting algorithm introduced by li _
_ @xcite . for each value of the hurst index @xmath1 , we averaged the results over @xmath156 realizations of the fbm series .
we summarize the corresponding numerical results in table 1 , which includes the hurst index @xmath1 , the average box - counting dimension @xmath157 , the average information dimension @xmath5 , the average correlation dimension @xmath6 , and @xmath158 , where the quantity @xmath159 was used to verify how @xmath133 changes along each curve . from table 1
, we can see that the average box - counting dimension @xmath157 , the average information dimension @xmath5 and the average correlation dimension @xmath6 are roughly decrease with the increase of the hurst index @xmath1 from @xmath104 to @xmath105 .
we show the average @xmath160 curves in fig . 13 and average @xmath161 curves in fig . 14 . from figs . 13 and 14 , we find that the @xmath160 and @xmath161 curves of recurrence networks are not straight lines .
so the multifractality exists in these recurrence networks constructed from fbm series .
meanwhile , we also find that the multifractality of these networks becomes stronger at first and then weaker which indicated by the value of @xmath162 when the hurst index of the associated time series becomes larger from 0.4 to 0.95 .
in particular , the recurrence network with the hurst index @xmath4 possess the strongest multifractality .
the dependence relationships of the average information dimension @xmath5 and the average correlation dimension @xmath6 on the hurst index @xmath1 are given in fig .
15 . as shown in fig .
15 , we find that these relationship can be well fitted by following linear formulas : @xmath163 and @xmath164
in this work , we constructed recurrence networks from fbm series based on the idea of recurrence plot .
we extracted the statistical properties of time series from the perspective of complex networks .
we studied the basic topological features of the recurrence networks constructed from fbm series with different hurst indices @xmath1 .
our numerical results indicate that the recurrence networks constructed exhibit exponential degree distributions . with the increase of @xmath1 , the average degree exponent @xmath0 increases first and then decreases .
the relationship between @xmath1 and @xmath0 can be fitted by a cubic polynomial function .
it was found that the average clustering coefficient @xmath82 increases with the hurst index @xmath1 , which means that there are more transitive and stable structures for networks constructed from fbms with large hurst index @xmath1 . at the microscopic level , we investigated the motif rank distribution of recurrence networks .
we found that the recurrence networks are grouped into two superfamilies based on the motif rank distribution .
the three key motifs @xmath98 , and @xmath87 determine the motif rank pattern and then classify networks . at the same time , we also paid attention to the dependence relationship of the average occurrence frequency @xmath92 with respect to the hurst index @xmath1 . from the aspect of fractality and self - similarity , we performed fractal and multifractal analyses using the random sequential box - covering method and the improved method based on modified fixed - size box - counting algorithm , respectively .
the numerical results show that the average fractal dimension @xmath2 decreases with hurst index @xmath1 and the linear relationship @xmath3 was obtained surprisingly .
moreover , our numerical results of multifractal analysis show that the multifractality exists in these recurrence networks , and the multifractality of these networks becomes stronger at first and then weaker when the hurst index of the associated time series becomes larger from 0.4 to 0.95 .
in particular , the recurrence network with the hurst index @xmath4 possess the strongest multifractality .
we also noted that the average information dimension @xmath5 and the average correlation dimension @xmath6 are roughly decrease with the increase of the hurst index @xmath1 from @xmath104 to @xmath105 .
the dependence relationships of the average information dimension @xmath5 and the average correlation dimension @xmath6 on the hurst index @xmath1 can be well fitted with linear functions . from the above results ,
we conclude that the inherent nature of time series affects the structure characteristics of the associated networks and the dependence relationship between them appears retained .
our works support that complex networks is a suitable and effective tool to perform time series analysis .
this project was supported by the natural science foundation of china ( grant nos . 11071282 and 11371016 ) , the chinese program for changjiang scholars and innovative research team in university ( pcsirt ) ( grant no .
irt1179 ) , the research foundation of education commission of hunan province of china ( grant no .
11a122 ) , the lotus scholars program of hunan province of china , the aid program for science and technology innovative research team in higher educational institutions of hunan province of china . r.v .
donner , j.f .
donges , y. zou , n. marwan , and j. kurths , in : _ proceedings of the 2010 international symposium on nonlinear theory and its applications nolta2010 ( krakow 2010 ) _ , tokyo : ieice 2010 , p87 - 90 . | many studies have shown that we can gain additional information on time series by investigating their accompanying complex networks . in this work ,
we investigate the fundamental topological and fractal properties of recurrence networks constructed from fractional brownian motions ( fbms ) .
first , our results indicate that the constructed recurrence networks have exponential degree distributions ; the average degree exponent @xmath0 increases first and then decreases with the increase of hurst index @xmath1 of the associated fbms ; the relationship between @xmath1 and @xmath0 can be represented by a cubic polynomial function .
we next focus on the motif rank distribution of recurrence networks , so that we can better understand networks at the local structure level .
we find the interesting superfamily phenomenon , i.e. the recurrence networks with the same motif rank pattern being grouped into two superfamilies .
last , we numerically analyze the fractal and multifractal properties of recurrence networks .
we find that the average fractal dimension @xmath2 of recurrence networks decreases with the hurst index @xmath1 of the associated fbms , and their dependence approximately satisfies the linear formula @xmath3 , which means that the fractal dimension of the associated recurrence network is close to that of the graph of the fbm .
moreover , our numerical results of multifractal analysis show that the multifractality exists in these recurrence networks , and the multifractality of these networks becomes stronger at first and then weaker when the hurst index of the associated time series becomes larger from 0.4 to 0.95 .
in particular , the recurrence network with the hurst index @xmath4 possess the strongest multifractality .
in addition , the dependence relationships of the average information dimension @xmath5 and the average correlation dimension @xmath6 on the hurst index @xmath1 can also be fitted well with linear functions .
our results strongly suggest that the recurrence network inherits the basic characteristic and the fractal nature of the associated fbm series .
= -18 mm = -0.1 cm = -0.1 cm * key words * : recurrence network ; fractional brownian motion ; fractal dimension ; multifractal analysis . *
pacs * : 89.75.hc , 05.45.df , 47.53.+n |
the discussion of the correct form of the energy - momentum tensor of electromagnetic waves in material media has been the subject of a hot debate for more than one century @xcite .
the two most famous proposals were made by minkowski and abraham on the beginning of last century @xcite and have distinct predictions for the momentum of light in a medium .
the abraham formulation predicts that an electromagnetic wave with total momentum @xmath0 in vacuum decreases its momentum to @xmath1 after entering a dielectric medium with refraction index @xmath2 , while the minkowski formulation predicts that the momentum of the wave is increased to @xmath3 after entering the medium .
several experiments on this subject were made in the past 60 years @xcite , many of them with the purpose of solving the debate , but the fact is that no decision can be made on experimental grounds . as it was pointed out by penfield and haus @xcite , the abraham , minkowski and other electromagnetic energy - momentum tensors are accompanied by distinct material energy - momentum tensors . and when the properties of the material energy - momentum tensors are carefully taken into account , both formulations predict the same experimental results .
a recent review of this debate can be seen in ref .
the eventual conclusion is that there are many compatible ways of dividing the total energy - momentum tensor of the system into an electromagnetic tensor and a material one .
nevertheless , we can ask whether there is , among all possible ways of dividing the momentum and the energy of electromagnetic waves in material media into electromagnetic and material parts , a division that could be considered as more natural . recently
, many attempts have been made to calculate the momentum transfer from electromagnetic waves to material media directly from the lorentz force law @xcite .
we believe that this is the most natural way of calculating the material momentum . in ref .
@xcite it is presented a review of some treatments based on lorentz force calculations for experiments on the topic , showing that consistent results are obtained .
we also believe that the most natural form for the material energy density is the sum of the potential and kinetic energy densities of the charges of the medium , which can be modified by the presence of an electromagnetic wave . the goal of this paper is to show , using reasonable models for material media , that if the material parts of the momentum density and of the energy density of electromagnetic waves in linear media are to be calculated in this way , the electromagnetic parts of the energy density @xmath4 , of the energy flux @xmath5 and of the momentum density @xmath6 must be @xmath7 where @xmath8 represents the macroscopic electric field , @xmath9 the macroscopic magnetic field , @xmath10 the permittivity of free space and @xmath11 the permeability of free space .
these forms are identical to the expressions in vacuum , but written in terms of the macroscopic fields .
the form for the electromagnetic momentum density above had appeared before , for instance in the book of livens @xcite .
however , the discussion of the energy of electromagnetic waves in linear media is practically absent from the literature . in sec .
[ discussion ] our proposal for the division is more carefully described and compared with previous treatments . in secs .
[ sec : clas - diel ] , [ sec : quant - diel ] and [ sec : quant - mag ] different models for dispersive non - absorptive dielectric and magnetic media are shown to be consistent with the proposed material energy density . in sec .
[ sec : flux ] we reproduce the model introduced by penfield and haus @xcite for the material energy flux and material momentum of an object with magnetic dipole moment in an applied electric field , showing the consistency of the division .
this momentum of a magnetic dipole in an electric field is frequently called `` hidden momentum '' @xcite and is a relativistic effect . in recent papers ,
barnett _ et al . _
@xcite associated the minkowski momentum with the electromagnetic canonical momentum and the abraham momentum with the electromagnetic kinetic momentum .
based on the model of penfield and haus , we suggest that the momentum density @xmath6 of eq .
( [ mom em ] ) should be associated with the electromagnetic relativistic momentum . finally , in sec .
[ sec : conclusions ] we present our concluding remarks .
the microscopic lorentz force density that acts on a density charge @xmath12 moving with velocity @xmath13 , forming a current density @xmath14 , is @xmath15 where @xmath16 and @xmath17 are the microscopic electric and magnetic fields . if the separation between the charges in the medium is much smaller than the wavelength of the incident electromagnetic wave , we can associate averaged electric and magnetic fields @xmath8 and @xmath9 to volumes containing a huge number of particles , but still with dimensions much smaller than the radiation wavelength .
the microscopic electric and magnetic fields can be written as @xmath18 , @xmath19 , where @xmath20 and @xmath21 depend on the microscopic distribution of charges and average to zero .
so the averaged force density in this volume can be written as @xmath22 where @xmath23 represents the averaged charge density and @xmath24 the averaged current density in the volume . in a linear medium without free charges , we can write @xmath25 and @xmath26 , where @xmath27 and @xmath28 are the polarization and magnetization of the medium @xcite . the electric and magnetic macroscopic fields are the fields that appear on the lorentz force equation above , so these are the fields that can transfer momentum and energy to material macroscopic bodies . from our point of view , these should be the fields that transport electromagnetic energy and momentum .
for this reason , these are the fields that appear in eqs .
( [ energy em ] ) and ( [ mom em ] ) for the electromagnetic energy density , energy flux and momentum density .
the abraham and minkowski momentum densities , on the other hand , are @xmath29 and @xmath30 , respectively , where @xmath31 , @xmath32 and @xmath33 is the speed of light in vacuum @xcite .
some of the previously cited papers that calculate the material momentum by the lorentz force consider that the electromagnetic momentum density should have the abraham value @xmath34 @xcite .
our proposal for the electromagnetic momentum density of eq .
( [ mom em ] ) agrees with the abraham form in a non - magnetic medium .
as it was discussed by mansuripur @xcite , to have momentum conservation considering the abraham form for the electromagnetic momentum density in magnetic media , a modification must be done in the lorentz force law .
mansuripur opts for this modification in the lorentz force to avoid the concept of `` hidden momentum '' , that we will discuss in sec .
[ sec : flux ] . in a recent paper @xcite
, we showed that the form of eq .
( [ mom em ] ) for the electromagnetic momentum density and the form of eq .
( [ lorentz ] ) for the lorentz force law are compatible with momentum conservation in a series of examples . ) .
] we also showed compatibility with the balazs _ gedanken _ experiment @xcite , showing compatibility with the theory of relativity when taking into account the `` hidden momentum '' . because in our formulation there is no need for a modification in the lorentz force law , we believe that our treatment is more natural .
now let us discuss the energy of electromagnetic waves in linear media .
in a closed system , the local conservation of energy can be written as @xmath35 where @xmath36 and @xmath37 represent the total energy density and total energy flux of the system .
for a system composed by an approximately monochromatic electromagnetic wave with frequency @xmath38 in a linear dispersive non - absorptive dielectric and magnetic medium , we can write @xcite @xmath39_{\omega=\omega_c } \frac{\varepsilon_0\langle|\mathbf{e}|^2\rangle}{2}+\\&&+\left[1+\frac{d(\omega\chi_\mathrm{m})}{d\omega}\right]_{\omega=\omega_c } \frac{\mu_0\langle|\mathbf{h}|^2\rangle}{2}\;,\\ \mathbf{s}_{\mathrm{tot } } = & & \mathbf{e}\times \mathbf{h}\;,\end{aligned}\ ] ] where @xmath40 represents the average of @xmath41 over one period of oscillation of the field , @xmath42 is the electric susceptibility and @xmath43 is the magnetic susceptibility of the medium .
so , if the electromagnetic parts of the energy density and of the energy flux of electromagnetic waves in linear media are given by eq .
( [ energy em ] ) , the material counterparts must be given by @xmath44_{\omega=\omega_c } \frac{\varepsilon_0\langle|\mathbf{e}|^2\rangle}{2}+\\&&+\left[-\alpha_\mathrm{m}+\omega\frac{d\alpha_\mathrm{m}}{d\omega}\right]_{\omega=\omega_c } \frac{\langle|\mathbf{b}|^2\rangle}{2\mu_0}\;,\\\label{flux mat } \mathbf{s}_{\mathrm{mat } } = & & -\mathbf{e}\times \mathbf{m}\;,\end{aligned}\ ] ] with @xmath45 , so that @xmath46 and @xmath47 .
in the next three sections we show that commonly used models for dielectric and magnetic linear media acquire the energy density of eq .
( [ dens en mat ] ) with the presence of an electromagnetic wave , and in sec . [ sec : flux ] we reproduce the model of penfield and haus @xcite that justifies the material energy flux of eq .
( [ flux mat ] ) .
the first model that we present is the drude - lorentz model for dielectric media .
a similar treatment was previously made by haus and kogelnik @xcite . in this model ,
the atomic nuclei are fixed and the electrons are bound by harmonic potentials @xmath48 , where @xmath49 represents the electron mass , @xmath50 the distance of each electron to its equilibrium position and @xmath51 the natural frequency of the harmonic oscillator . considering a non - absorptive medium , the equation of motion of the @xmath52 component of each electron can be written as @xmath53 where @xmath54 represents the electron charge and @xmath55 is the incident electric field .
the solutions of the above differential equation are @xmath56 because we are disregarding the absorption of radiation , we must consider that @xmath57 is far from @xmath51 .
if we call @xmath58 the density of electrons subjected to the potential @xmath59 , the electric susceptibility of the medium can be written as @xmath60 the material energy density corresponds tho the sum of the kinetic @xmath61 and potential @xmath62 energies of the electrons multiplied by their densities @xmath58 . using eq .
( [ x ] ) , we have @xmath63 using eq . ( [ chi_clas ] ) , it is straightforward to see that eq .
( [ dens en mat 1 ] ) can be written as the first term of eq .
( [ dens en mat ] ) : @xmath64 \frac{\varepsilon_0|\mathbf{e}|^2}{2}.\ ] ]
now let us consider a quantum model for a linear non - absorptive dielectric medium .
the medium is composed by molecules with fixed nuclei and @xmath65 electrons with charge @xmath54 and mass @xmath49 around .
the hamiltonian that couples these nuclei and electrons may be very complex , but our results do not depend on its specific form .
the eigenstates of the molecular hamiltonian will be denoted @xmath66 , and the respective eigenvalues @xmath67 .
an arbitrary pure state of the molecular system can be written as @xmath68 .
an electric field @xmath69 acts as a perturbation on the system .
our treatment is based on loudon s book @xcite .
the total hamiltonian can be written as @xmath70 where @xmath71 is the @xmath52 component of the center of mass position operator of the @xmath65 electrons and @xmath72 is the total charge of the electrons . from the schrdinger equation with the hamiltonian above applied on the state ( [ estado atomo ] ) , after taking the scalar product on both sides with @xmath73 , we obtain @xmath74 let us consider that the molecule is in its fundamental state @xmath75 before the interaction with the field starts and that the field produces a small perturbation , such that @xmath76 always .
so we will consider only the term @xmath77 in the summation above to the calculus of @xmath78 .
the electric susceptibility refers to the situation where the interaction between the field and the molecules has achieved an equilibrium condition , so we can take an indefinite integral as the solution of eq .
( [ eq dif cm ] ) : @xmath79,\ ] ] with @xmath80 , @xmath81 , @xmath76 .
we used the definition @xmath82 to define the origin of the coordinates system .
the oscillating dipole moment of the molecule can be written as @xmath83 where we disregarded terms @xmath84 with @xmath49 and @xmath2 both different from zero , @xmath85 was considered 1 and c.c .
stands for the complex conjugate . using eqs .
( [ dip ] ) and ( [ cm ] ) , we can see that a medium composed by a density @xmath86 of these molecules has an electric susceptibility @xmath87.\ ] ] because we are considering a non - absorptive medium , @xmath88 is real .
this approximation of non - absorptive media is valid only when @xmath57 is far from all @xmath89 for which @xmath90 .
the material energy density is the expectation value of the non - perturbed hamiltonian @xmath91 multiplied by the density of molecules , because the hamiltonian @xmath91 contains the kinetic and potential energy of the electrons : @xmath92 using eq .
[ cm ] , we have @xmath93.\ ] ] taking the temporal average over one period of oscillation of the field we have @xmath94 using eq .
( [ chi ] ) , it is straightforward to show that the energy density can be written as the first term of eq .
( [ dens en mat ] ) : @xmath95 \frac{\varepsilon_0\langle|\mathbf{e}|^2\rangle}{2}.\ ] ] if we have different kinds of molecules composing the medium , each one identified by a symbol @xmath96 , the electric susceptibility and the energy density of the medium will be @xmath97 and eq .
( [ dens en mat quant ] ) remains valid .
the quantum model that we use for a magnetic linear medium is based on van vleck s book @xcite and considers the response of electrons under the action of a central potential to an applied magnetic field .
we use cylindrical coordinates @xmath98 in this section . to simplify the treatment
, we will consider that the magnetic field is generated by a vector potential @xmath99 , such that @xmath100 , instead of considering the vector potential of an electromagnetic wave .
the lagrangean of an electron of mass @xmath49 and charge @xmath54 under the action of this magnetic field and an atomic central potential @xmath101 is @xmath102 the lagrangean does not depend on the variable @xmath103 , so the momentum conjugated to this coordinate is a constant of the movement of the system : @xmath104 so the @xmath105 component of the angular momentum of the electron , @xmath106 , is a function of the applied magnetic field : @xmath107 where @xmath108 represents the original angular momentum of the electron , when the magnetic field is zero .
the ratio between the dipole moment and the orbital angular momentum of an electron is @xmath109 @xcite , so the variation of the electron dipole moment is @xmath110 , with @xmath111 , @xmath112 being the unperturbed quantum state of the electron .
considering that the medium has atoms with random orientations , on taking the average of the orientations we can consider , for each quantum sate of the electrons , @xmath113 and @xmath114 , where @xmath115 is the distance from the origin of the central potential . considering a density @xmath58 for electrons in each quantum state @xmath116 , the magnetization of the medium can be written as @xmath117 so the coefficient @xmath118 , defined by the relation @xmath119 , so that @xmath120 , is @xmath121 the variation of the electrons kinetic energy density can be written as @xmath122 using eq . ( [ alpham ] )
, we can see that this energy density can be written as the second term of eq .
( [ dens en mat ] ) , @xmath123 since in this model @xmath124 does not depend on the frequency @xmath57 .
+ here we reproduce the model of penfield and haus @xcite that shows that a magnetic dipole under the action of an electric field can have associated energy flux and momentum , even if the dipole is not moving .
consider a rectangle of sides @xmath125 and @xmath126 that conducts an electric current @xmath127 in a region of space with an electric field @xmath8 .
the situation is depicted in fig .
[ fig - hidden ] .
the magnetic dipole moment of the object is @xmath128 .
the electric field does a positive work on the positive charges that go up in the right side of the figure and a negative work on the positive charges that go down in the left side ( if we consider that the moving charges are negative , the final conclusions will be the same ) .
so a charge @xmath54 moving to the left has an extra energy @xmath129 with respect to the situation when it moves to the right .
this fact generates an energy flux from right to left .
if we have a density @xmath86 of these dipoles , the energy flux can be found multiplying the energy per unity time @xmath130 that goes to the left in each dipole by the length @xmath125 along which the energy flows and by the dipole density @xmath86 .
so we have @xmath131 where @xmath28 is the magnetization of the medium .
this result agrees with eq .
( [ flux mat ] ) .
the simple model presented in this section is very distinct from an atomic system .
however , since the magnetic response of an atom due to a magnetic field oscillating at optical frequencies is due to the orbital motion of the electrons , a similar phenomenon should occur in that case .
now let us analyze the relativistic linear momentum of the object .
the momentum of the charges that move to the left in fig .
[ fig - hidden ] is higher because they have more energy .
the ratio between the momentum and the energy of a relativistic particle is @xmath132 , where @xmath13 is the particle velocity .
let us consider that the current @xmath127 is uniform along the circuit .
so if we have a linear charge density @xmath133 with velocity @xmath134 going to the left and a linear charge density @xmath135 with velocity @xmath136 going to the right , we have @xmath137 .
calling @xmath138 the energy of a particle that goes to the right in fig .
[ fig - hidden ] and @xmath54 the total moving charge of the circuit , the total linear momentum of the circuit is @xmath139}{qc^2}\right]\mathbf{\hat{z } } = -\frac{iabe}{c^2}\mathbf{\hat{z}}\nonumber \\ & = & \frac{\mathbf{m}\times \mathbf{e}}{c^2}.\end{aligned}\ ] ] so , if we have a magnetization @xmath28 in a medium with an electric field @xmath8 , there will be a `` hidden momentum '' density @xmath140 .
it is worth mentioning that the `` hidden momentum '' is a relativistic effect .
if we had considered that the momentum of a particle is @xmath141 , there would be no `` hidden momentum '' . in recent works ,
barnett _ et al . _
@xcite associated the abraham momentum with the electromagnetic kinetic momentum and the minkowski momentum with the electromagnetic canonical momentum . considering the `` hidden momentum ''
, we see that the relativistic momentum of an atom with magnetic moment @xmath142 can be written as @xmath143 , where @xmath144 , @xmath49 being the mass and @xmath13 the velocity of the atom .
if we have a bunch of these atoms forming a medium , the difference of the relativistic and the kinetic momenta of all atoms is equal to the difference of the total momentum of the field when we consider the abraham momentum @xmath145 , and the total momentum @xmath146 when we consider the momentum density of eq .
( [ mom em ] ) .
so we can write @xmath147 the last equality is derived in ref .
the total kinetic and canonical momenta are always conserved , so we conclude that , if we associate @xmath146 with the relativistic momentum of the field , the total relativistic momentum will be always conserved .
in this paper we defended a particular division of the energy and of the momentum of electromagnetic waves in linear media in electromagnetic and material parts as being the most natural .
the electromagnetic parts of the energy density , energy flux and momentum density are given by eqs .
( [ energy em ] ) and ( [ mom em ] ) , having the same form than in vacuum when written in terms of the macroscopic electric and magnetic fields .
the material momentum is calculated directly from the lorentz force law , the material energy corresponds to the kinetic and potential energies of the charges of the medium and the material energy flux is the result of the work done by the electric field on the charges that compose an object with magnetic dipole moment . in a previous work @xcite , we showed momentum conservation in various circumstances considering the electromagnetic momentum density of eq .
( [ mom em ] ) and calculating the material momentum with the lorentz force law . here we showed , using reasonable models for linear media , that we have energy conservation considering the energy density of eq .
( [ energy em ] ) and calculating the material energy density as the sum of the kinetic and potential energy densities of the charges of the medium . to justify the material energy flux , we used a model introduced by penfield and haus @xcite .
we also argued that the momentum density of eq .
( [ mom em ] ) can be associated with the relativistic electromagnetic momentum .
it is worth to mention that , although we used models for linear non - absorptive media , we believe that the proposed division can be extended to any kind of media . for example , on an absorbing medium the absorbed energy is usually converted into heat or chemical reactions , that can also be associated with the kinetic and potential energies of the particles of the medium . | we defend a natural division of the energy density , energy flux and momentum density of electromagnetic waves in linear media in electromagnetic and material parts . in this division ,
the electromagnetic part of these quantities have the same form as in vacuum when written in terms of the macroscopic electric and magnetic fields , the material momentum is calculated directly from the lorentz force that acts on the charges of the medium , the material energy is the sum of the kinetic and potential energies of the charges of the medium and the material energy flux results from the interaction of the electric field with the magnetized medium .
we present reasonable models for linear dispersive non - absorptive dielectric and magnetic media that agree with this division .
we also argue that the electromagnetic momentum of our division can be associated with the electromagnetic relativistic momentum , inspired on the recent work of barnett [ phys .
rev .
lett . * 104 * , 070401 ( 2010 ) ] that showed that the abraham momentum is associated with the kinetic momentum and the minkowski momentum is associated with the canonical momentum . |
dendrimers are an important class of artificial macromolecules with a treelike structure , which are synthesized by repeating units in a hierarchical self - similar fashion around a central core @xcite .
their unique structural features make them promising candidates for a broad range of potential applications such as light harvesting antennae @xcite and molecular amplifiers @xcite , the latter of which can be used as efficient platforms for drug delivery @xcite . in view of their practical significance ,
thus far , dendrimers have received extensive attention within the scientific community @xcite . in the context of light harvesting by dendrimers , it is the large number of absorbing elements at the periphery and an efficient transfer of the absorbed energy to the center that make dendrimers work as antennas @xcite .
generally , light harvesting can be described as a trapping process with a ( fluorescent ) trap located at the center . a primary quantity related to trapping
is average trapping time ( att ) , which is the average of mean first - passage time ( mfpt ) @xcite to the target over all starting nodes , where mfpt from a node to the trap is the expected time steps needed for a walker starting off from this node to visit the trap for the first time .
att is a quantitative indicator measuring the trapping efficiency , which has been much studied for diverse complex systems @xcite .
_ _ because of the theoretical and practical relevance , trapping in dendrimers has also been devoted to concerted efforts .
the problem was first addressed in refs .
@xcite , where the mfpt from a peripheral node to the central node was computed analytically . in refs .
@xcite , mfpt from any node to the central node , as well as the att to the center , were deduced .
these works unveiled the effect of structure on the mfpt and att to the central node . note that most previous works on trapping dendrimers based on discrete time random walks focused on the unbiased random walk .
however , sometimes it is more suitable to describe some particular problems by a biased random walk than the unbiased one @xcite , since the transition probability depends on not only network topologies but also other properties relevant to the diffusion dynamics @xcite . among numerous biased random walks ,
maximal entropy random walk ( merw ) @xcite , maximizes the entropy of paths , has been studied recently @xcite . in merw
, the transition probability incorporates the node centrality measured by the eigenvector associated with the largest eigenvalue of adjacent matrix of the graph where the dynamical process takes place .
this local transition rate leads to substantial effects on the diffusion behaviors such as stationary distribution @xcite and relaxation time @xcite .
the principle of entropy maximization has found some applications , including optimal sampling algorithm @xcite and demographic stability of population @xcite .
very recently , as a powerful tool , merw has been fruitfully applied in the analysis of complex networks @xcite . at present , it is still of theoretical and practical interest to explore possible applications of merw on other areas . in this paper , we study the trapping problem in cayley trees as models of dendrimers @xcite , based on merw that incorporates the centrality of nodes of the macromolecular graphs .
we concentrate on a particular case with a perfect trap placed at the central node , for which we derive an analytical exact expression for att .
we then provide an upper bound for att , whose leading scaling behaves with the system size @xmath1 as @xmath2 , a scaling much smaller than that corresponding to unbiased random walk in cayley trees , for which the att scales linearly with @xmath1 .
theses results show that for trapping process in dendrimers with a trap fixed at the center , merw is considerably more efficient in comparison to unbiased random walk .
in this section , we introduce the construction and some relevant features of dendrimers modeled by cayley trees , which are built in an iterative way .
the particular construction process allows for analytically determining the properties of cayley trees and deriving exact solutions for diverse dynamical processes on large but finite structures .
cayley trees after @xmath3 iterations ( generations ) , denoted by @xmath4 ( @xmath5 , @xmath6 ) are constructed as follows @xcite . at the initial generation ( @xmath7 )
, @xmath8 contains only a central node ; at @xmath9 , @xmath10 new nodes are generated and linked to the central node to form @xmath11 .
these @xmath10 new single - degree nodes constitute the peripheral nodes of @xmath11 . for any @xmath12
, @xmath4 is obtained from @xmath13 : for each peripheral node of @xmath13 , @xmath14 new nodes are created and are connected to the peripheral node .
figure [ cayley ] illustrates the structure for a specific dendrimer @xmath15 .
let @xmath16 denote the number of nodes in @xmath17 , which are created at @xmath18th generation .
then , we can verify that @xmath19 thus , the total number of nodes in @xmath4 is @xmath20 . ] the special construction of cayley trees also makes it possible to analytically determine the eigenvalues and their associated eigenvectors @xcite .
let @xmath21_{n \times n}$ ] denote the adjacency matrix of network @xmath4 , in which @xmath22 if nodes @xmath18 and @xmath23 are linked to each other , @xmath24 otherwise .
although for a general graph , it is often difficult to determine the eigenvalues and eigenvectors of its adjacency matrix , for @xmath4 the problem can be settled . note that @xmath4 has @xmath1 eigenvalues .
we represent these @xmath1 eigenvalues as @xmath25 , @xmath26 , @xmath27 , @xmath28 , and @xmath29 , respectively .
every eigenvalue @xmath30 takes the form @xmath31 @xmath32 where @xmath33 . in the sequel , what we are concerned with
is the largest eigenvalue and its corresponding eigenvector of @xmath34 , which are denoted by @xmath35 and @xmath36 , respectively .
the largest eigenvalue @xmath35 can be expressed as @xmath37 where @xmath38 with regard to the eigenvector @xmath36 , since all its entries @xmath39 for nodes in a given generation @xmath18 are identical , it can be written as : @xmath40 in order to express the entry @xmath41 , we introduce the following quantity : @xmath42 for @xmath43 .
it is easy to verify @xmath44}{\sin\theta}\,.\ ] ] from eq .
( [ b2 ] ) , we have @xmath45 in the following text , we choose @xmath46}$ ] .
thus , @xmath47}}\,.\ ] ] the selection of @xmath48 ensures the proper normalization of every entry of eigenvector @xmath36 : @xmath49=1\,.\ ] ]
after introducing the construction and related properties of cayley trees @xmath4 , in this section , we study merw in dendrimers with a perfect trap fixed at the central node .
our goal is to unveil the influence of maximal entropy random walk on the efficiency of trapping performed on this important family of polymer networks .
the merw considered here is a discrete time random walk @xcite . at each discrete time step ,
the walker jumps from its current position @xmath18 to any of its neighboring nodes @xmath23 with probability @xmath50 let @xmath51 denote the set of neighbors of a node @xmath18 .
then , for an arbitrary node @xmath18 , @xmath52 fulfills the condition @xmath53 , since @xmath54 the stationary distribution for merw on @xmath4 is @xcite @xmath55 the main subject we focus on here is trapping problem in @xmath4 described by merw with a deep trap located at the innermost node .
and the main quantity we are interested in is the att to the trap . for convenience of description , we label the central node of @xmath17 by @xmath56 , and consecutively label all other nodes as @xmath57 , and @xmath1 .
we use @xmath58 to represent the trapping time of node @xmath18 , which is defined as the expected time for a walker starting off from node @xmath18 to reach the trap in @xmath17 for the first time .
then , the att denoted by @xmath59 is the average of @xmath58 over all non - trap nodes distributed uniformly in @xmath17 , that is , s @xmath60 note that merw in any connected binary network can be represented as generic random walk in a corresponding weighted network @xcite . for @xmath17 , the generalized adjacency matrix ( weight matrix ) @xmath61_{n \times n}$ ] corresponding to the weighted networks , denoted by @xmath62 , associated with merw is defined as follows
. the entries @xmath63 if nodes @xmath18 and @xmath23 are adjacent in @xmath17 , and the element @xmath64 if nodes @xmath18 and @xmath23 are not directly connected by an edge in @xmath17 . in weighted networks
@xmath62 , the strength @xcite of a node @xmath18 is defined by @xmath65 , which is @xmath18th nonzero entry of the diagonal strength matrix @xmath66 , and the laplacian matrix of @xmath62 is defined to be @xmath67 . for generic random walk in @xmath62 ,
the transition probability for the walker from current state @xmath18 to one of its neighboring nodes @xmath23 is @xmath68 , which is the same as that of merw in @xmath17 .
therefore , the stationary distribution for generic random walk in @xmath62 is also the same as that corresponding to merw in @xmath4 . according to recently obtained result about trapping in weighted networks @xcite , the explicit expression for att @xmath59 in eq .
( [ d4 ] ) can be expressed as @xmath69 in eq .
( [ d5 ] ) , @xmath70 is the sum of strengths over all nodes in @xmath62 , namely @xmath71 ; @xmath72 are the @xmath1 eigenvalues of @xmath73 , rearranged as @xmath74 , and @xmath75 are the corresponding mutually orthogonal eigenvectors of unit length , where @xmath76 .
it should be mentioned that although the expression for att @xmath59 provided by eq .
( [ d5 ] ) seems compact , it requires computing the eigenvalues and eigenvectors of matrix @xmath73 .
since the network size @xmath1 grows exponentially with @xmath3 as shown in eq .
( [ c2 ] ) , for moderately large @xmath3 , the computation of spectra demands an impractically large computational effort .
moreover , by using eq .
( [ d5 ] ) it is very hard and even impossible to obtain useful information about the dependence of the leading behavior of @xmath59 on the network size @xmath1 .
it is thus of significant practical importance to seek other approaches for determining the att @xmath59 .
fortunately , the specific architecture of cayley trees allows for analytical calculation of @xmath59 and evaluation of its dominant scaling . according to the structure of dendrimers , all the nodes in @xmath17
can be classified into @xmath77 levels .
the central node is at level @xmath78 , and nodes born at the first generation @xmath18 are at level @xmath56 , and so forth . by symmetry ,
the trapping time for nodes at the same level is identical . in the case without confusion , we use @xmath58 to denote the trapping time for a node at @xmath18th level . then
, the following relations hold : @xmath79+p_i^{\rm{down}}[1+f_{i+1}(n)],&{0<i < n},\\ 1+f_{i-1}(n),&{i = n}. \end{cases}\ ] ] in eq .
( [ c10 ] ) , @xmath80 represents the transition probability for a walker hopping from a node at level @xmath18 to its ( unique ) father node at level @xmath81 , while @xmath82 denotes the transition probability from any node at level @xmath18 to its @xmath14 child nodes at level @xmath83 . for unbiased random walk in @xmath4 @xcite , @xmath84 and @xmath85 , obeying relation @xmath86
however , as will be shown below , for merw in @xmath4 , @xmath80 and @xmath82 are different from those corresponding to unbiased random walk , although @xmath86 also holds . and
the disparity for transition probability between these two kinds of random walks leads to quite different scalings for att . by definition of transition probability for merw in @xmath4 , see eq .
( [ d1 ] ) , @xmath80 and @xmath82 for merw in @xmath4 are @xmath87 and @xmath88 respectively . considering above - obtained results , @xmath80 and @xmath82 can be further recast as @xmath89 } { \sin\theta}}{(m-1)^{(n - i)/2}\frac{\sin[(n - i+1)\theta]}{\sin\theta } } \nonumber \\ & = & \frac{1}{2\cos\theta}\frac{\sin[(n - i+2)\theta]}{\sin[(n - i+1)\theta ] } \nonumber\\ & = & \frac{\sin[(n - i+2)\theta]}{\sin[(n - i+2)\theta]+\sin[(n - i)\theta]}\end{aligned}\ ] ] and @xmath90 } { \sin\theta}}{(m-1)^{(n - i)/2}\frac{\sin[(n - i+1)\theta]}{\sin\theta } } \nonumber\\ & = & \frac{1}{2\cos\theta}\frac{\sin[(n - i)\theta]}{\sin[(n - i+1)\theta ] } \nonumber \\ & = & \frac{\sin[(n - i)\theta]}{\sin[(n - i+2)\theta]+\sin[(n - i)\theta]}\,,\end{aligned}\ ] ] both of which evidently fulfil the relation @xmath86 , implying that our computation for @xmath80 and @xmath82 is correct . using eqs .
( [ c16 ] ) and ( [ c18 ] ) , eq .
( [ c10 ] ) becomes @xmath91[1+f_{i-1}(n)]}{\sin[(n - i+2)\theta]+\sin[(n - i)\theta ] } \\
+ \frac{\sin[(n - i)\theta][1+f_{i+1}(n)]}{\sin[(n - i+2)\theta]+\sin[(n - i)\theta ] } , & { 0<i < n } , \\ 1+f_{i-1}(n),&{i = n}.\\ \end{cases}\ ] ] therefore , for @xmath92 , we have @xmath93+\sin[(n - i)\theta]\}}f_i(n)\nonumber \\ & = & \sin[(n - i+2)\theta][1+f_{i-1}(n)]\nonumber \\ & & + \sin[(n - i)\theta][1+f_{i+1}(n)]\,,\end{aligned}\ ] ] which can be rewritten as @xmath94[f_i(n)-f_{i-1}(n)]\nonumber \\ & = & \sin[(n - i+2)\theta]+\sin[(n - i)\theta ] \nonumber \\ & & + \sin[(n - i)\theta][f_{i+1}(n)-f_i(n)]\,.\end{aligned}\ ] ] in order to determine the quantity @xmath95 , we define @xmath96 then , @xmath97}+\frac{\sin{i\theta}}{\sin[(i+2)\theta]}b_{i-1}(n)\ ] ] holds for all @xmath92 .
note that @xmath98-f_{n-1}(n ) \nonumber \\ & = & 1\,,\end{aligned}\ ] ] using which eq .
( [ c23 ] ) can be solved to yield @xmath99}{\sin[(i+2-k)\theta]}+1.\ ] ] making use of eqs .
( [ c22 ] ) and ( [ c25 ] ) , @xmath58 can be evaluated as @xmath100}{\sin[(n - j+2-l)\theta]}+1\bigg]\,,\end{aligned}\ ] ] where @xmath101 was used .
equation ( [ c26 ] ) provides a closed form expression for trapping time for an arbitrary node . inserting eq .
( [ c26 ] ) into eq .
( [ d4 ] ) , we can obtain the closed - form formula of att @xmath95 for merw in @xmath4 with a single trap positioned at the central node , which reads @xmath102}{n-1 } \nonumber \\ & = & \frac{\displaystyle{\sum_{i=1}^n}\left\{m(m-1)^{i-1}\displaystyle{\sum_{j=1}^i}\left[2\displaystyle{\sum_{k=0}^{n - j-1}}\displaystyle{\prod_{l=0}^k}\frac{\sin[(n - j - l)\theta]}{\sin[(n - j+2-l)\theta ] } + 1\right]\right\}}{\frac{m[(m-1)^n-1]}{m-2}}\nonumber \\ & = & \frac{m-2}{(m-1)^n-1}\sum_{i=1}^n \left\{(m-1)^{i-1}\sum_{j=1}^i \left[2\sum_{k=0}^{n - j-1}\prod_{l=0}^k\frac{\sin[(n - j - l)\theta]}{\sin[(n - j+2-l)\theta]}+1 \right]\right\}\,.\end{aligned}\ ] ] in fig . [ trapping ]
, we present results about the att @xmath95 to the central node for merw in @xmath4 .
the results are obtained by using eqs .
( [ d5 ] ) and ( [ c27 ] ) , respectively .
figure [ trapping ] shows that the results generated by eq .
( [ d5 ] ) and eq .
( [ c27 ] ) agree with each other , confirming our theoretical results for @xmath95 provided by eq .
( [ c27 ] ) .
( color online ) the att @xmath103 to the central node corresponding to merw in @xmath4 for various @xmath10 and @xmath3 .
the hollow symbols represent the results generated by eq .
( [ d5 ] ) , while the solid symbols stand for the analytical results given by eq .
( [ c27 ] ) . ]
although the expression for the att @xmath95 given in eq .
( [ c27 ] ) is exact , it is rather lengthy and awkward , from which we can not see obvious dependence of @xmath95 on the network size @xmath1 .
however , we will show below that , when the networks are large enough , from eq .
( [ c27 ] ) one can derive an upper bound for the leading scaling of @xmath95 in terms of the network size @xmath1 . first , let us examine the product term @xmath104}{\sin[(n - j+2-l)\theta]}\ ] ] in eq .
( [ c27 ] ) .
seemingly , there are @xmath105 multipliers in both the denominator and numerator in eq .
( [ c28 ] ) .
in fact , for any @xmath106 we can decrease the number of multipliers from @xmath105 to two by simplifying eq .
( [ c28 ] ) as follows : @xmath107}{\sin[(n - j+2-l)\theta ] } \nonumber \\ & = & \frac{\sin[(n - j)\theta]\sin[(n - j-1)\theta]\cdots\sin[(n - j - k)\theta]}{\sin[(n - j+2)\theta]\sin[(n - j+1)\theta]\cdots\sin[(n - j+2-k)\theta ] } \nonumber \\ & = & \frac{\sin[(n - j+1-k)\theta]\sin[(n - j - k)\theta]}{\sin[(n - j+2)\theta]\sin[(n - j+1)\theta]}\,.\end{aligned}\ ] ] since @xmath108 , for any @xmath109 ( @xmath110 ) and @xmath111 ( @xmath112 ) , every term in both the denominator and numerator of eq .
( [ c29 ] ) has the form @xmath113 , where @xmath114 is an integer between @xmath56 and @xmath77 .
thus , @xmath115 and @xmath116 .
according to eq .
( [ b0 ] ) , we have @xmath117 where @xmath118 is a certain positive constant .
note that if constant @xmath118 is chosen properly , the following relation also holds : @xmath119-[(n+1)(m-2)]}{(m-2)n+2m-2 } \nonumber \\ & > & \frac{1}{cn}\,.\end{aligned}\ ] ] on the other hand , when @xmath3 is large enough , @xmath120 .
then , each term @xmath121 $ ] in the numerator of eq .
( [ c28 ] ) satisfies @xmath122>\sin \frac{1}{cn } = \frac{1}{cn}$ ] .
thus , @xmath123}{\sin[(n - j+2-l)\theta]}\le \frac{1}{\frac{1}{cn } } \frac{1}{\frac{1}{cn } } \approx d\,n^2,\ ] ] where @xmath124 is another positive constant . substituting the result in eq .
( [ c32 ] ) into eq .
( [ c27 ] ) , we obtain @xmath125\bigg\ } \nonumber \\ & \le&\frac{m-2}{(m-1)^n-1}\sum_{i=1}^n(m-1)^{i-1}dn^4 \nonumber \\ & \le&\frac{m-2}{(m-1)^n-1}\frac{(m-1)^n-1}{m-2}dn^4 \nonumber \\ & \le&d\,n^4\,.\end{aligned}\ ] ] note that , eq . ( [ c2 ] ) enables us to represent @xmath3 in terms of the system size @xmath1 as @xmath126-\ln m}{\ln(m-1)}\,.\ ] ] hence , the upper bound for @xmath103 in eq .
( [ c33 ] ) can be expressed in the following form : @xmath127-\ln m}{\ln(m-1)}\right]^4 \nonumber \\ & \le & f(\ln n)^4\end{aligned}\ ] ] where @xmath128 is a certain positive constant .
equation ( [ c35 ] ) provides an upper bound of dominant behavior of @xmath103 for merw in massive networks @xmath4 with the immobile trap located at the central node , which shows that the efficiency of trapping in dendrimers described by merw is very high .
it is in sharp contrast with the att for unbiased random walk in the same networks , where the leading term of att increases lineally with the system size @xcite .
thus , merw is instrumental in improving the efficiency of trapping process taking place in dendrimers .
based on maximal entropy random walk ( merw ) we have performed an analytical research on trapping in dendrimers with a single trap placed at the central node .
we have derived an explicit expression for att as an indicator of the trapping efficiency , grounded on which we have deduced an upper bound for the leading scaling of att . the obtained upper bound scales with the network size @xmath1 as @xmath2 , and is considerably lower than the att for trapping in dendrimers based on unbiased random walk , which grows lineally with the network size @xmath1 .
this means that in dendrimers merw can dramatically lessen the att to the central node , making the diffusion process highly efficient .
this theoretical work might prove useful to address light harvesting by dendrimers , as well as various other dynamical processes occurring in nanoscale systems .
j. gmez - gardees and v. latora , phys .
e * 78 * , 065102(r ) ( 2008 ) .
w. parry , trans .
soc . * 112 * , 55 ( 1964 ) .
z. burda , j. duda , j. m. luck , and b. waclaw , phys .
lett . * * 1**02 , 160602 ( 2009 ) . | we use maximal entropy random walk ( merw ) to study the trapping problem in dendrimers modeled by cayley trees with a deep trap fixed at the central node .
we derive an explicit expression for the mean first passage time from any node to the trap , as well as an exact formula for the average trapping time ( att ) , which is the average of the source - to - trap mean first passage time over all non - trap starting nodes .
based on the obtained closed - form solution for att , we further deduce an upper bound for the leading behavior of att , which is the fourth power of @xmath0 , where @xmath1 is the system size .
this upper bound is much smaller than the att of trapping depicted by unbiased random walk in cayley trees , the leading scaling of which is a linear function of @xmath1 .
these results show that merw can substantially enhance the efficiency of trapping performed in dendrimers . |
@xcite were the first who formulated the idea about the analogy of flares on red dwarf stars and on the sun .
at present it seems obvious that the physical nature of the activity of the sun and red dwarf stars is similar ( @xcite ) .
@xcite put forward the pioneering idea about the important role of gas - dynamic effects in the formation of the emission of solar flares .
the authors showed that the radiative cooling of the shocked gas can be a source of the flare s emission in the optical range .
the shock wave is produced by impulsive heating of the chromosphere by suprathermal particles .
the kostyuk pikelner model predicts the formation of a relatively homogeneous gas layer , a few kilometers of thickness , which is responsible for the flare s optical emission in the continuum and spectral lines .
the layer is located between the front of the downward shock and the region of plasma heated from above .
this model was applied to the problem of flares on red dwarf stars in @xcite .
the authors fulfilled calculations of a one - dimensional radiative shock wave and demonstrated the formation of a region with a small gradient of the plasma parameters : temperature , pressure , and density ( see figures 3 and 4 in @xcite ) .
this result is confirmed by the calculations of the radiative cooling behind a shock front in the atmospheres of cool stars ( see @xcite and @xcite ) with a more detailed account for the elementary processes in the plasma : ionization , recombination , bremsstrahlung , excitation and de - excitation of discrete levels of atoms and ions as well as radiation scattering at the frequencies of spectral lines .
red dwarf stars have a complex structure of flare spectra : the intensities of the emission lines strongly increase , the continuum emission is also enhanced ; it is well approximated with planck curves obtained for different temperatures at the blue and red sides relative to the balmer jump ( see @xcite ) .
layers producing the blackbody emission should be in a state close to thermodynamic equilibrium , whereas the gas generating strong emission lines is essentially non - equilibrium .
thus , the flare plasma includes at the same time equilibrium and non - equilibrium regions .
therefore , we investigate the problem of the formation of an emission spectrum in a wide range of the cloud parameters . we consider a homogeneous gas layer of thickness @xmath1 with definite values of temperature and density and consisting of pure hydrogen .
we develop the model of @xcite in three points .
firstly , we consider the influence of the layer s radiation ( bremsstrahlung and recombination ) on the occupation of atomic levels .
this development is necessary , as the flare luminosity is stronger in the optical range than that of the quiescent atmosphere of the whole star .
secondly , we take into account that broadening of the high members of spectral series is caused exclusively by the linear stark effect .
thirdly , we consider the _ two - temperature _ approximation for a flare layer . in section
1 the problem is formulated . in section
2 we write the set of equations of ionization balance with regard to the continuum . in section 3 the calculation of the continuum emission is described . in section 4
we present the method of calculation of the quantum escape probability . in section 5 the intensities of the emission
spectral lines are calculated .
section 6 presents the results of the emission spectrum calculations of a layer with parameters possible in the flare regions of red dwarf stars .
we calculate the emission in spectral lines and continuum of an homogeneous pure hydrogen layer .
its temperature , density and dimensions are assumed to be given .
the value of the ion atom temperature , @xmath2 , is taken to be greater than the value of the electron temperature , @xmath3 , because both ions and atoms are heated more intensively than electrons at the shock front ( see , e.g. , @xcite ) .
we use the results of the radiative shock wave calculations by @xcite and @xcite to state the input parameter set .
these calculations were fulfilled under the conditions of the atmosphere of a red giant star with an effective temperature of @xmath4 .
the values of @xmath2 and @xmath3 at the shock front are respectively @xmath5 ev and @xmath6 ev .
the lyman-@xmath7 ( ly-@xmath7 ) optical depth , @xmath8 , in the cooling region , as it follows from @xcite , lies in the range @xmath9 the value of @xmath8 weakly depends on the density of the gas , as the cooling rate , @xmath10 , is proportional to the hydrogen atom concentration , @xmath11 .
thus , @xmath11 is cancelled : @xmath12 here , @xmath13 @xmath14 is the electron density , @xmath15 is the velocity of the gas flowing from the shock front , and @xmath16 is some time interval from the moment when a given gas layer intersects the shock front . the value of @xmath17 is equal to the summa of the proton and atom concentrations : @xmath18 we accept @xmath17 typical for the m - dwarf chromospheres : @xmath19 the layer thickness , @xmath1 , is defined to be such that the condition is satisfied . the electron density is equal to @xmath20 the values of @xmath2 and @xmath3 were obtained from a red giant star , which has a 100x smaller gravity compared to a red dwarf star .
the ion - atom temperature at the shock front is determined by the shock velocity , @xmath21 , and weakly depends on the density of the gas .
if the value of @xmath21 lies in the range @xmath22 km / sec , then it is true that @xmath23 the value of @xmath3 at the shock front is determined by the temperature @xmath24 of the unperturbed gas and heating due to the compression .
the evolution of @xmath3 is caused by the balance between the supply of energy in elastic collisions with ions and atoms and radiative losses in the lyman-@xmath7 line .
the radiative cooling in the hydrogen lines is the most intensive at @xmath25 ( see @xcite ) .
therefore , we assume that @xmath26
the following processes we take into account : the electron impact ionization , excitation , and de - excitation , the triple recombination , radiative recombination ( induced and spontaneous ) as well as the influence of the layer s bremsstrahlung and recombination radiation on its ionization and excitation states .
the impact coefficients are taken from @xcite .
we use the kramers approximation ( see , e.g. , @xcite ) for the photoionization cross sections from the excited levels .
the rates of the inverse processes are calculated following the detailed balance principle .
the spectral lines optical depth can be rather large in this problem , so the scattering of line radiation should be taken into account .
we introduce so called effective einstein transition coefficients for the transitions between hydrogen levels @xmath27 and @xmath28 by the following formulae : @xmath29 here , @xmath30 is a photon escape probability out of the process of scattering .
the values of @xmath30 are calculated in the section [ probab ] for the resonance as well as subordinate transitions . according to the standard notation , @xmath31 and @xmath32 are respectively the probabilities of spontaneous and induced transition from the upper level @xmath28 to the lower level @xmath27 , and @xmath33 is the absorption coefficient .
we write the set of the balance equations for the discrete levels of a hydrogen atom : [ b_01 ] @xmath34 where @xmath35 we assume that the maximum value of the principal quantum number , @xmath36 , is equal to an effective series limit .
it can be evaluated by the inglis teller formula ( @xcite ) , as @xmath37 the following designations are introduced : + @xmath38 the density number of the hydrogen atoms on the discrete level with the principal quantum number @xmath28 ; + @xmath39 the radiative and triple recombination coefficients on the @xmath28-th level , respectively ; + @xmath40 the transition rate from the level @xmath28 to the level @xmath27 ; + @xmath41 the escape rate from the level @xmath28 .
we write the explicit formulae for the matrix coefficients in the left side of the equation set .
+ excitation rate ( @xmath42 ) is : @xmath43,\ ] ] where + @xmath44 is the electron impact excitation ; + @xmath45 is the spectral intensity of the layer continuous radiation in the frequency of @xmath46 transition .
+ deactivation rate ( @xmath47 ) is : @xmath48,\ ] ] where @xmath44 is the impact deactivation coefficient .
the diagonal elements are @xmath49+\\+\sum_{j = k+1}^{n}\left[q_{kj}(t_e)n_e+4\pi{}i_{kj}b^*_{kj}\right]+\mathcal{p}_k+q_k\cdot{}n_e,\end{gathered}\ ] ] where @xmath50 is the electron impact ionization coefficient , and @xmath51 is the photoionization rate from the level @xmath28 : @xmath52 here , @xmath53 is the photoelectric threshold from level @xmath28 , @xmath54 is the plank constant , @xmath55 is the cross section in the kramers approximation : @xmath56 where @xmath7 is the fine structure constant , @xmath57 is the bohr radius . + the spontaneous recombination coefficient , @xmath58 , can be calculated from using the principle of detailed balance : [ b_10 ] @xmath59 where @xmath60 is the exponential integral of order 1 .
the coefficients of induced recombination , @xmath61 , is given in the next section .
it is true that @xmath62 summing the set over all discrete levels yields the identity @xmath63 as a result , we come to the set of homogeneous equations with singular matrix for the relative level occupations , @xmath64 : @xmath65 where @xmath66 to avoid the singularity , the last equation ( @xmath67 ) is replaced by the normalization condition @xmath68 we take into account the scattering of the ionizing radiation using the division of @xmath69 by @xmath70 , where @xmath71 is the optical depth on the photoionization threshold from the level @xmath28 .
continuum emission consists of two components : free - free and bound - free .
the following formula is valid for an homogeneous layer : @xmath72 where @xmath73 is the source function , which is equal to the ratio of the emission coefficient , @xmath74 , to the absorption coefficient , @xmath75 : @xmath76 the summation is over all the levels for which the frequency of the photoelectric threshold is less than @xmath77 : @xmath78 the menzel factor is designated by @xmath79 : @xmath80 it shows the deviation of the level population from its equilibrium value .
this factor is equal to 1 when excitation and ionization are described by the formulae of boltzmann and saha .
we use the kramers approximation for the absorption coefficients ( @xcite , 34.4 ) : [ s_05 ] @xmath81 \varkappa^{(i)}_{\nu{k}}&=\dfrac{2^{9}}{3\sqrt{3}}\alpha\,\pi^{5/2}a_0 ^ 5\dfrac{1}{k^3}b_k\left(\dfrac{\mathrm{ry}}{h\nu}\right)^3\left(\dfrac{\mathrm{ry}}{t_e}\right)^{3/2}\times\notag\\&\quad\times\left(1-\dfrac1{b_k}\,e^{-y}\right ) e^{\beta_k}n_en_p,\qquad{}y\geqslant\beta_k,\label{s_05b}\end{aligned}\ ] ] where @xmath82 .
we write the relation between the ionization coefficient from the level @xmath28 , @xmath83 , and the bremsstrahlung coefficient , @xmath84 : @xmath85 the value of @xmath86 is equal to the sum of bremsstrahlung optical depth , @xmath87 , and photoionization one , @xmath88 : @xmath89 the free - free absorption and emission coefficients are related by kirchhoff low : @xmath90 the menzel factor connects the radiative recombination coefficient , @xmath91 , and photoionization coefficients for the @xmath28-th level : @xmath92 using the last formula and taking into account , we can transform : @xmath93 @xmath94 where the following function @xmath95 is introduced : @xmath96 the induced recombination coefficient is equal to @xmath97 as is easily seen from , the continuum optical depth , @xmath86 , is the function of @xmath98 when the frequency is varied and the other parameters are unchanged .
the source of optical emission moves slowly and cohesively , i.e. there is no noticeable velocity gradient .
therefore , the biberman - holstein approximation ( @xcite ) is used to calculate the scattering of line radiation .
consider a stationary gas that occupies a volume @xmath99 bounded by a closed surface @xmath100 .
let @xmath101 is the radius - vector of a point within the volume . according to @xcite , the photon escape probability from this point outside the homogeneous plasma
can be written in the form : @xmath102 here , @xmath103 is the angular frequency , @xmath104 is the spectral line profile , and @xmath105 is the line absorption coefficient .
@xmath106 denotes the radius - vector of a point belonging to the surface @xmath100 .
let the gas occupies a layer of finite thickness @xmath1 . in this case
, the mean escape probability is equal to @xmath107 where @xmath108\,,\,\,\,\tau_\omega=\cfrac{k_\omega{}\mathcal{l}}{2}\,\,.\ ] ] in these expressions , @xmath109 is the optical depth at the center of the layer and @xmath110 is the third - order exponential integral ( see , e.g. , @xcite ) . as stated in @xcite , the very wide wings of the balmer lines in the optically _ thick _ flare plasma are due to the linear stark effect .
when the stark broadening by electrons and ions is treated in a semi - classical picture , the static limit is valid in the distant long - wavelength line wing ( @xcite , 38.1 ) : @xmath111 here , @xmath112 is the stark broadening parameter of the transition from level @xmath28 to level @xmath113 ( @xmath114 ) , @xmath115 is the normal holtsmark field , and @xmath116 , @xmath117 is the transition angular frequency .
the stark width is approximately equal to @xmath118 the spectral line profile is determined by radiation damping , doppler effect and pressure effects .
the extensive tables of the stark broadened profiles for hydrogen plasma at @xmath119 have been published by @xcite . in the case of @xmath120
we will use a simple model of the line profile with the doppler core and the holtsmark wings .
we consider , e.g. , the h-@xmath7 line profile in order to justify this model .
the following set of parameters is assumed : @xmath121 we introduce the value of @xmath122 such that @xmath123^{-1}\,,\,\,\mbox{where}\,\,v\,\,\mbox{is the hydrogen atom}\ ] ] velocity ( along a line of sight ) .
the value of @xmath124 can be evaluated as @xmath125 here , @xmath126 is the hydrogen atom mass .
the doppler width , @xmath127 , is equal to @xmath128 here , @xmath129 , @xmath130 @xmath131 @xmath132 and @xmath133 .
therefore , the _ static limit _
( [ st_lim ] ) can be used when @xmath134 , i.e. , in the line wing ( see @xcite , 38.1 , 36.6 ) . on the other hand ,
we introduce the parameter @xmath135 , which is equal to the sum of the radiative damping constant , @xmath136 , and the parameter @xmath137 : @xmath138 where @xmath139 is the concentration of perturbing particles .
@xmath140 sec@xmath141@xmath131 @xmath142 . if @xmath143 then @xmath144 .
therefore , the central part of the h-@xmath7 line profile is the _ doppler profile _ ( see @xcite , 36.6 ) . in general
, there is some analogy between the line profile with the doppler core and the holtsmark wings and the voigt profile when the voigt parameter @xmath145 is much less than 1 .
the normalized voigt function is @xmath146 where @xmath124 is the dimensionless frequency displacement from the line center in units of @xmath127 . for
@xmath147 it can be assumed that in the first approximation ( see , e.g. , @xcite ) @xmath148\ ] ] the value of @xmath149 is the solution of the equation @xmath150 the transition from the doppler core to the lorentz wings in the voigt profile is quite sharp ( @xcite ) . the following model _
symmetric _ profile is similar to the voigt profile ( [ voigt_f ] ) : @xmath151 where @xmath152 is a normalizing constant .
the function @xmath153 is defined by @xmath154\ ] ] in these expressions , the value of parameter @xmath155 is significantly less than unity : @xmath156 we can find the constant @xmath152 from the equations @xmath157
@xmath158 @xcite introduced the model profile phenomenologically . here , we give the theoretical foundation for its application . to make an estimate of the optical depth in the ly-@xmath7 line at @xmath159 , we use ( [ par3 ] ) .
according to ( [ stark_eff ] ) , the stark width of the ly-@xmath7 line is approximately equal to @xmath160 , the doppler width is @xmath161 ; whence it follows that @xmath162 thus , the approximation ( [ st_f ] ) is adequate in our problem . using ( [ eq4f ] ) and ( [ eq_4 ] ) , we find that @xmath163 and @xmath164 . if the optical depth at the line center is equal to @xmath165 , then it is true that @xmath166 therefore , the ly-@xmath7 photon lives the medium at the frequencies of the distant line wings . as a result , equation ( [ probability ] ) can be written in the form : @xmath167 it is obvious that @xmath168 here , @xmath169 is the absorption coefficient at the doppler profile center : @xmath170 @xmath171 is the oscillator strength for absorption ( @xmath172 , @xmath173 ) . taking into account that @xmath174 , we obtain the asymptotic formula for the mean escape probability : @xmath175 consider the integral @xmath176\,\cfrac{d\tau}{\tau^{1+\kappa}}\,,\,\,\,\kappa\in(0,1)\,.\ ] ] by using the formula of integration by parts , this equation can be represented in the form : @xmath177}{\tau^\kappa}\right|_0^{+\infty}\,.\ ] ] if the value of @xmath178 lies between 0 and 1 , then the r.h.s . of
the last equation tends to zero .
thus , we obtain that @xmath179 + where @xmath180 is the gamma function . in this problem , the value of @xmath181 is equal to @xmath182 . finally , we come to @xmath183 the exponent @xmath184 in the last equation is the feature of the linear stark effect . strictly speaking , the resulting expression for @xmath185 is valid only for a model two - level atom .
however , the escape probability approach can be applied also for a multilevel atom .
we will use the model profile ( [ st_f ] ) up to @xmath186 ( for more details , see @xcite ) .
then the principal quantum numbers , @xmath28 and @xmath113 , must satisfy the condition @xmath187 constant @xmath152 is approximately unity in the range @xmath188 ; therefore , it is true that @xmath189 where @xmath190 is the optical depth in the @xmath191 transition . since the stark width increases with increasing @xmath28 , the stark broadening dominates over the doppler broadening in the limit @xmath192 .
then we use the following model profile : @xmath193\ ] ] here , @xmath57 is some constant , @xmath194
. the normalization condition gives @xmath195 after some calculations we obtain that @xmath196 it is interesting to note that the profile with the doppler core and the stark - broadened wings is similar to the h-@xmath197 line profile in the flare on uv ceti ( dm5.6e ) published in @xcite .
the authors found a doppler profile to fit the line _ core _ : for h-@xmath197 during outburst , the doppler width was equal to 0.9 .
thus , it is true that @xmath198 ev and @xmath199 .
further , eason et al . found that @xmath200 @xmath131 @xmath201 and @xmath202 .
thus , we obtain that @xmath203 and @xmath204 .
the authors satisfactorily represented _
part _ of the long - wavelength wing with a stark profile from the paper by @xcite .
in this paper the stark profile was calculated taking into account only the ion broadening in context of the holtsmark distribution of the electric microfields , while the stark broadening by electrons was neglected . but as stated in @xcite , the electron broadening dominates in stellar flares .
we consider that the deviation of the observed h@xmath7 line profile from the stark profile in the red wing ( see fig . 9 in @xcite )
is almost entirely due to neglecting the contribution of electrons to the line broadening in the optically _ thick _ flare plasma .
to simplify the notation , we have in mind _
the balmer series_. the intensity of the emission in the spectral line frequencies , @xmath205 , is equal to @xmath206,\ ] ] where index @xmath28 identifies the upper level of the transition @xmath207 .
we have denoted by @xmath208 the optical depth in the frequencies of the @xmath209 transition ; the negative absorption being taken into account : @xmath210 here , @xmath211,\ ] ] the spectral line intensity , @xmath212 , is equal to the integral of over the line profile : @xmath213\,d\nu.\ ] ] in this section , we consider that the continuum absorption is negligible within the profile of a spectral line .
so , we write the expression for the source function taking into account only the line emission and absorption : @xmath214 the planck function varies very slowly within the line profile . therefore , the source function at the frequency of the line center , @xmath215 , can be taken outside the integral sign : @xmath216\,d\nu.\ ] ] for the same reason , we consider that the continuum optical depth within the line profile is equal to @xmath217 .
the limits of integration are the frequencies where the line emission merges into continuum .
they are defined by the following condition : @xmath218 the simplified formulae and do not introduce significant inaccuracies in @xmath219 , as the transition region is quite narrow . moreover ,
the emission intensity in this region is much lower than the intensity in the line core .
note that there are some factors simplifying the calculation of integral .
if the condition is satisfied , then the central optical depth , @xmath71 , is rather large . consequently , the intensity in the line core frequencies coincides with a source function . in those parts of the line profile where @xmath220 is not too large , we use for @xmath221 a simple approximation ( [ griem ] ) , assuming that for the frequency displacement up to @xmath222 , the layer remains opaque .
we introduce frequency displacement @xmath223 which satisfies the condition @xmath224 in the offset range @xmath225 it is assumed that the intensity is equal to the source function , and in the line wings we calculate the integral @xmath226 here , parameter @xmath71 is defined as @xmath227 and the upper limit is calculated according to .
thus , the total energy emitted in the line is @xmath228 where the stark width , @xmath229 , is determined by formula ( [ stark_eff ] ) .
we performed calculations of the spectral line and continuum intensity for different values of density and temperature of a homogeneous gas layer under the conditions of flares on red dwarf stars .
the layer thickness @xmath1 is fixed .
we calculated the optical depth at the center of the balmer series lines ( see table [ table 1 ] ) and the menzel factors ; in all cases the values of @xmath79 do not differ from unity ( see table [ table 2 ] ) .
it is important that there is no external source of radiation ionizing the hydrogen atom from the excited levels .
let us address the results of calculations of the emission spectrum .
figure 1 shows how the emission lines are `` immersed '' in a continuum with increasing density of the gas .
the horizontal axis represents the wavelength @xmath230 in ngstrms , vertical the logarithm of the ratio @xmath231 .
balmer lines are indicated by vertical line segments , its height is proportional to the line center intensity . in all the graphs , the following values of the parameters are assumed : @xmath232 ev , @xmath233 ev , and @xmath234 km .
the intensity of the emission at the centers of the balmer series lines is very close to the planck function at the electron temperature @xmath235 .
therefore , all the spectral lines have the same upper boundary corresponding to zero on the vertical axis .
as the rise of the continuum occurs , first h@xmath236 `` immerses '' , then h@xmath237 and finally high members of the balmer series .
this is due to the fact that continuum absorption coefficient increases toward longer wavelengths .
[ fig1 ] , @xmath238 , @xmath239 , @xmath240 . under each curve
its number is written .
the upper curve almost coincides with the planck function ; the continuum absorption is not negligible in this case.,title="fig:",width=302 ] we temporarily exclude spectral lines in order to more conveniently trace how the continuous spectrum depends on the layer parameters .
figure 2 represents the ratio @xmath241 for two values of the electron temperature and three values of density .
ion temperature is equal here to 10 ev , and the thickness of the layer to 10 km .
the values of the optical depth in a continuous spectrum at the wavelength 4170 are shown on each curve .
it is seen that increasing of the continuum opacity essentially depends on the electron temperature : at higher values of @xmath3 , the transition to the planck curve occurs at lower values of @xmath242 .
this fact is caused by increasing of the occupation of excited levels ( @xmath243 ) with increasing the electron temperature .
indeed , the value of @xmath244 , as it follows from equations ( [ s_03 ] ) and ( [ s_05b ] ) .
while the gas density is increased , the intensity of the continuous spectrum @xmath245 tends to the planck function , remaining less than it .
the curve portions that are closer to the planck curve are located in the red region of the spectrum and at the blue side of the balmer jump .
when the density increases further , the blue part of the spectrum ( redder at the balmer jump ) approaches the planck function .
the shape of the line core is largely determined by the ion - atom temperature , as the doppler width , @xmath246 , depends only on the square root of @xmath247 ( if the principal quantum numbers @xmath28 and @xmath113 are unchanged ) .
it is true that at higher values of @xmath2 , the approximation for the core by a gaussian curve is valid for a higher number of members of the balmer series .
ev , solid ones to @xmath248 ev , numbers i , ii , and iii correspond to the following values of the total number densities of the hydrogen particles : @xmath249 @xmath250 , @xmath251 @xmath250 , and @xmath252 @xmath250 , respectively .
next , the values of the optical depth at wavelength @xmath253 are shown for each curve .
, width=302 ] .decimal logarithms of the optical depth at the centers of the balmer lines ( @xmath232 ev , @xmath233 ev , and @xmath234 km ) [ cols="^,^,^,^,^ " , ]
the main result of our work is the calculated plasma emission spectrum in a broad interval of the continuum optical depths , from the transparent gas to the gas whose emission intensity is close to the planck function .
the proximity to the planck function in the case of strong self - absorption is due to the mechanism of impact ionization and excitation ( from electron collisions ) of the plasma , whereas radiative processes are secondary .
the approach used in this paper fits well for red dwarf stars with their effective temperature of about 3000 k. however , the case of g- and f - type dwarfs requires taking into account the photospheric emission , e.g. , photoionization of hydrogen atoms from exited levels . in the case of appreciable emission and absorption in the continuum its ( continuum )
influence must be taken into account in the calculation of spectral line intensities . here
our work develops the approach used in @xcite and @xcite , where the line intensities were calculated under the assumption of the transparent continuum bringing small contribution to the emission . in the framework of the considered problem
the intensity of the subordinate lines is close to planck for any layer parameters implemented during the flares on red dwarf stars .
therefore , the relative height of the lines depends on the optical depth in the continuum .
if it is small then strong lines are observed against the background of faint continuum , which is described by formulae valid for transparent gas . in the case of appreciable self - absorption in the continuum
its spectrum becomes close to planck ; thereby , the lines weaken .
however , in the homogeneous layer model no strong lines on the background of the planck continuum observed by @xcite are obtained .
probably , we see in the flares the total emission from several layers ; some of them are transparent in the continuum , while other ones are not .
we hope to get a better agreement with observations by calculations of the radiative cooling behind the shock front under the conditions of the red dwarf chromospheres , since the shocked gas can include at the same time equilibrium and non - equilibrium regions .
morchenko is grateful to prof .
zemtsov for valuable remarks .
this work was partly supported by the russian foundation for basic research ( project code 14 - 02 - 00922 ) , and the scientific school ( project code 1675.2014.2 nsh ) .
we thank the anonymous referee for very useful suggestions and comments .
gershberg , r.e . , pikelner s.b . :
comments astrophys .
space phys .
* 4 * , 113 ( 1972 ) .
gershberg , r.e . : solar - type activity in main - sequence stars , ( springer ) ( 2005 ) .
kostyuk , n.d . ,
pikelner , s.b .
: , 590 ( 1975 ) .
katsova , m.m . ,
kosovichev , a.g . , livshits , m.a . : astrophysics . *
17 * , 156 ( 1981 ) .
doi : 10.1007/bf01005196 fadeyev , yu.a . ,
gillet , d. : , 901 ( 2001 ) .
arxiv : astro - ph/0603510 .
doi:10.1086/506268 belova , o.m . ,
bychkov , k.v . ,
morchenko , e.s . ,
nizamov , b.a . : astron
. reports . *
58 * , 650 ( 2014 ) doi:10.1134/ s1063772914090029 kowalski , a.f . ,
hawley , s.l . ,
wisniewski , j.p . , osten , r.a . ,
hilton , e.j . , holtzman , j.a . ,
schmidt , s.j . ,
davenport , j.r.a .
: , 57 ( 2013 ) .
doi:10.1088/0067 - 0049/207/1/15 drake , s.a . ,
ulrich , r.k .
: , 351 , 1980 .
zeldovich , ya.b . ,
raizer , yu.p . : physics of shock waves and high - temperature hydrodynamic phenomena , ( academic press ) ( 1967 ) .
johnson , l. c. : , 227 ( 1972 ) .
sobelman , i.i . :
introduction to the theory of atomic spectra , ( pergamon press ) ( 1972 ) .
vidal , c.r .
: j. quant .
transfer * 6 * , 461 ( 1966 ) .
holstein , t. : phys . rev . *
72 * , 1212 ( 1947 ) .
biberman , l.m . ,
vorobev , v.s . ,
yakubov , i.t . :
kinetics of nonequilibrium low - temperature plasmas , ( plenum publishers ) ( 1987 ) .
smirnov , b.m . : physics of weakly ionized gas , ( nauka ) ( 1972 ) .
sobolev , v.v .
, grinin v.p . : in : pallavicini , r. , dupree , a.k .
cool stars , stellar systems , and the sun , 9th cambridge workshop asp conference series . * 109 * , 629 ( 1996 ) .
sobolev , v.v . ,
grinin , v.p . :
astrophysics . *
38 * , 15 ( 1995 ) .
doi:10.1007/bf02113956 stehle , c. : , 509 ( 1994 ) .
ivanov , v.v .
: transfer of radiation in spectral lines , ( u.s .
government printing office ) ( 1973 ) .
mihalas , d. : stellar atmospheres , ( w.h . freeman and co. )
eason , e.l.e . ,
giampapa , m.s . ,
radick , r.r . ,
worden , s.p . ,
hege , e.k . : , 1161 ( 1992 ) .
underhill , a.b . ,
waddell , j.h .
: national bureau standards circular no . 603 ( 1959 ) .
katsova , m.m . ,
livshits , m.a . , butler , c.j . ,
doyle , j.g .
: , 402 ( 1991 ) . | the emission spectrum has been calculated of a homogeneous pure hydrogen layer , which parameters are typical for a flare on a red dwarf .
the ionization and excitation states were determined by the solution of steady - state equations taking into account the continuum and all discrete hydrogen levels .
we consider the following elementary processes : electron - impact transitions , spontaneous and induced radiative transitions , and ionization by the bremsstrahlung and recombination radiation of the layer itself .
the biberman
holstein approximation was used to calculate the scattering of line radiation .
asymptotic formulae for the escape probability are obtained for a symmetric line profile taking into account the stark and doppler effects .
the approximation for the core of the h@xmath0 line by a gaussian curve has been substantiated .
the spectral intensity of the continuous spectrum , the intensity of the lines of the balmer series and the magnitude of the balmer jump have been calculated .
the conditions have been determined for which the balmer jump and the emission line intensities above the continuum decrease to such low values that the emission spectrum can be assumed to be continuum as well as the conditions at which the emission spectrum becomes close to the blackbody . |
let @xmath0 be an affine kac - moody algebra , @xmath1 a @xmath2-submodule of a finite direct sum @xmath3 of irreducible integrable highest weight @xmath4-modules , and @xmath5 the limit of the demazure operator for an element @xmath6 of the weyl group as @xmath7 .
the main theorem of this paper gives sufficient conditions on @xmath1 so that the formula @xmath8 holds , where @xmath9 is the character of @xmath1 .
when @xmath1 is the one - dimensional @xmath2-module generated by the dominant integral weight @xmath10 then is the weyl - kac character formula .
the above result is well - known when @xmath1 is a union of demazure modules for any kac - moody algebra @xmath0 .
let @xmath11 be the derived subalgebra of @xmath0 .
consider the @xmath11-module @xmath1 given by a tensor product of finite - dimensional @xmath12-modules that admit a crystal of level at most @xmath13 , with the one - dimensional subspace generated by a highest weight vector of an irreducible integrable highest weight @xmath12-module of level @xmath13 .
such modules @xmath1 can be given the structure of a @xmath2-module and as such , satisfy the above conditions .
then a special case of is a bosonic formula for the @xmath14-enumeration of level - restricted inhomogeneous paths by the energy function . in type @xmath15
this formula was conjectured in @xcite , stated there as a @xmath14-analogue of the goodman - wenzl straightening algorithm for outer tensor products of irreducible modules over the type @xmath16 hecke algebra at a root of unity @xcite . in the isotypic component of the vacuum ,
the bosonic formula coincides with half of the bose - fermi conjecture in ( * ? ? ? * ( 9.2 ) ) .
the authors would like to thank nantel bergeron , omar foda , masaki kashiwara , atsuo kuniba , masato okado , jean - yves thibon , and ole warnaar for helpful discussions .
most of the following notation is taken from ref .
let @xmath17 be a dynkin diagram of affine type with vertices indexed by the set @xmath18 as in @xcite , cartan matrix @xmath19 , @xmath20 the affine kac - moody algebra , and @xmath21 the cartan subalgebra .
let @xmath22 and @xmath23 be the simple coroots and roots , which are linearly independent subsets that satisfy @xmath24 for @xmath25 where @xmath26 is the natural pairing .
let @xmath27 be the root lattice .
let the null root @xmath28 be the unique element of the positive cone @xmath29 in @xmath30 , that generates the one - dimensional lattice @xmath31 .
let @xmath32 be the canonical central element , where the integers @xmath33 are the analogues of the integers @xmath34 for the dual algebra @xmath35 defined by the transpose @xmath36 of the cartan matrix @xmath16 .
let @xmath37 ( the degree derivation ) be defined by the conditions @xmath38 where @xmath39 is the kronecker delta ; @xmath40 is well - defined up to a summand proportional to @xmath41 .
then @xmath42 is a basis of @xmath21 .
let @xmath43 be the dual basis of @xmath44 ; the elements @xmath45 are called the fundamental weights . the weight lattice is defined by @xmath46 ; in the usual definition the scalar @xmath47 is absent .
the weight lattice contains the root lattice since @xmath48 for @xmath49 .
define @xmath50 .
say that a weight @xmath51 has level @xmath13 if @xmath52 .
consider the standard symmetric bilinear form @xmath53 on @xmath44 .
since @xmath54 is a basis of @xmath44 , this form is uniquely defined by @xmath55 for @xmath25 , @xmath56 for @xmath57 and @xmath58 .
this form induces an isomorphism @xmath59 defined by @xmath60 for @xmath57 and @xmath61 .
also @xmath62 .
the weyl group @xmath63 is the subgroup of @xmath64 generated by the simple reflections @xmath65 ( @xmath57 ) defined by @xmath66 the form @xmath53 is @xmath63-invariant .
suppose @xmath67 is a real root , that is , the @xmath68-weight space of @xmath0 is nonzero and there is a simple root @xmath69 and a weyl group element @xmath70 such that @xmath71 .
define @xmath72 by @xmath73 .
this is independent of the expression @xmath74 .
define @xmath75 by @xmath76 let @xmath11 be the derived algebra of @xmath0 , obtained by omitting " the degree derivation @xmath40 .
its weight lattice is @xmath77 .
denote the canonical projection @xmath78 by @xmath79 .
write @xmath80 and @xmath81 for @xmath57 .
the elements @xmath82 are linearly dependent .
write @xmath83 for the section of @xmath79 given by @xmath84 for all @xmath57 .
write @xmath85 .
define the level of @xmath86 to be @xmath87 .
consider the dynkin diagram @xmath88 obtained by removing the vertex @xmath89 from the diagram @xmath17 , with corresponding cartan matrix @xmath90 indexed by the set @xmath91 , and let @xmath92 be the simple lie algebra .
one has the inclusions @xmath93 .
let @xmath94 be the simple roots , @xmath95 the fundamental weights , and @xmath96 the root lattice for @xmath97 .
the weight lattice of @xmath97 is @xmath98 and @xmath99 . the image of @xmath100 into @xmath101 is denoted by @xmath102 .
we shall use the section of the natural projection @xmath103 given by the map @xmath104 that sends @xmath105 for @xmath106 . by abuse of notation , for @xmath100 , @xmath102 shall also denote the image of the element @xmath102 under the lifting map @xmath107 specified above .
. for @xmath109 , denote by @xmath110 the irreducible integrable highest weight @xmath111-module of highest weight @xmath112 .
let @xmath113 .
one has the formulas @xmath114 , @xmath115 , and @xmath116 .
observe that @xmath117 for @xmath51 let @xmath118 be the irreducible integral highest weight module of highest weight @xmath10 over the quantized universal enveloping algebra @xmath4 , @xmath119 the crystal base of @xmath118 , and @xmath120 the highest weight vector . by restriction from @xmath4 to @xmath12 , the module
@xmath118 is an irreducible integral highest weight module for @xmath12 of highest weight @xmath121 , with crystal @xmath119 that is @xmath122-weighted by composing the weight function @xmath123 with the projection @xmath79 .
conversely , any integrable irreducible highest weight @xmath12-module can be obtained this way .
a @xmath124-weighted @xmath125-crystal @xmath126 is a colored graph with vertices indexed by @xmath127 , directed edges colored by @xmath57 , and a weight function @xmath128 , satisfying the axioms below .
first some notation is required .
denote an edge from @xmath129 to @xmath130 colored @xmath131 , by @xmath132 or equivalently @xmath133 .
write @xmath134 ( resp .
@xmath135 ) for the maximum index @xmath136 such that @xmath137 ( resp .
@xmath138 ) is defined . 1 .
if @xmath139 then @xmath140 .
2 . @xmath141 .
an element @xmath142 is a highest weight vector if @xmath143 is undefined for all @xmath57 .
the @xmath131-string of @xmath127 consists of all elements @xmath138 ( @xmath144 ) and @xmath137 ( @xmath145 ) .
the nondominant part of the @xmath131-string is comprised of all elements which admit @xmath146 .
we also define the crystal reflection operator @xmath147 by @xmath148 it is obvious that @xmath149 is an involution .
observe that @xmath150 define the notation @xmath151 and @xmath152 .
if a @xmath4-module ( resp .
@xmath12-module , resp .
@xmath111-module ) has a crystal base then the latter is naturally a @xmath124-weighted ( resp .
@xmath122-weighted , resp .
@xmath101-weighted ) @xmath125-crystal ( resp .
@xmath125-crystal , resp .
@xmath153-crystal ) . given crystals @xmath154 and @xmath155 , contrary to the literature ( but consistent with the robinson - schensted - knuth correspondence in type a ) , define the following crystal structure on the tensor product @xmath156 .
the elements are denoted @xmath157 for @xmath158 ( @xmath159 ) and one defines @xmath160 when @xmath161 ( resp .
@xmath162 ) one defines @xmath163 an element of a tensor product of crystals is called a path .
the definitions here follow @xcite .
suppose that @xmath154 and @xmath155 are crystals of finite - dimensional @xmath12-modules such that @xmath156 is connected .
then there is an isomorphism of @xmath122-weighted @xmath125-crystals @xmath164 .
this is called the local isomorphism .
let the image of @xmath165 under this isomorphism be denoted @xmath166 .
then there is a unique ( up to a global additive constant ) map @xmath167 such that @xmath168 this map is called the local energy function .
consider @xmath169 with @xmath170 the crystal of a finite - dimensional @xmath12-module for @xmath171 .
assume that for all @xmath172 , @xmath173 is a connected @xmath122-weighted @xmath125-crystal .
given @xmath174 , denote by @xmath175 the @xmath176-th tensor factor in the image of @xmath129 under the composition of local isomorphisms that switch @xmath170 with @xmath177 as @xmath178 goes from @xmath179 down to @xmath180 .
then define the energy function @xmath181 where @xmath182 is the local energy function .
it satisfies the following property .
[ energy e ] ( * ? ? ?
1.1 ) suppose @xmath57 , @xmath127 and @xmath183 is defined . if @xmath184 then @xmath185 . if @xmath186 and @xmath129 has the property that for any of its images @xmath187 under a composition of local isomorphisms , @xmath188 with @xmath189 , then @xmath190 .
say that @xmath191 is classically restricted if @xmath129 is a @xmath97-highest weight vector , that is , @xmath183 is undefined for all @xmath106 .
for @xmath192 denote by @xmath193 the set of classically restricted @xmath127 of weight @xmath112 .
define the polynomial @xmath194 where @xmath195 is the energy function on @xmath126 . for @xmath0 of type @xmath15 @xmath196
is the generalized kostka polynomial @xcite .
let @xmath126 be the crystal of a finite - dimensional @xmath12-module .
say that @xmath126 is almost perfect of level @xmath13 @xcite if it satisfies the following weakening of the definition of a perfect crystal ( * ? ? ?
4.6.1 ) : 1 .
@xmath197 is connected .
2 . there is a @xmath198 such that there is a unique @xmath199 such that @xmath200 and for every @xmath127 , @xmath201 .
3 . for every @xmath127 , @xmath202 .
4 . for every @xmath203 of level @xmath13
, there is a @xmath204 such that @xmath205 .
@xmath126 is said to be perfect if the elements @xmath129 and @xmath130 in item 4 are unique .
from now on , fix a positive integer @xmath13 ( the level ) . for @xmath171 let @xmath170 be the crystal of a finite - dimensional @xmath12-module , that is almost perfect of level at most @xmath13 .
let @xmath206 , @xmath207 weights of level @xmath13 , and @xmath208 the set of paths @xmath209 such that @xmath210 is a highest weight vector of weight @xmath211 . in the special case that @xmath212 , the elements of @xmath208 are called the level-@xmath13 restricted paths of weight @xmath211 . [ level tensor ] @xcite ( * ? ? ?
* appendix a ) .
let @xmath0 be an affine kac - moody algebra in one of the infinite families .
let @xmath126 be the tensor product of crystals of finite - dimensional @xmath12-modules , that are almost perfect of level at most @xmath13 , and @xmath203 a weight of level @xmath13 .
then there is an isomorphism of @xmath122-weighted @xmath125-crystals @xmath213 where @xmath211 is of level @xmath13 .
this isomorphism of @xmath122-weighted crystals can be lifted to one of @xmath124-weighted crystals by specifying an integer multiple of @xmath214 for each highest weight vector in @xmath215 .
however for our purposes this should be done in a way that extends the definition of the energy function for @xmath126 . to this end , choose a perfect crystal @xmath216 of level @xmath13 , and assume that for all @xmath217 , @xmath218 is connected .
let @xmath219 be the unique element such that @xmath220 .
define the energy function @xmath221 by @xmath222 where @xmath223 is the energy function defined in . for @xmath224 , define an affine weight function @xmath225 .
this defines the @xmath124-weight of every highest weight vector in @xmath215 and hence a @xmath124-weight function for all of @xmath215 .
then one has the following @xmath124-weighted analogue of : @xmath226 where @xmath211 is of level @xmath13 .
this decomposition can be described by the polynomial @xmath227 our goal is to prove a formula for the polynomial @xmath228 .
let @xmath153 be the antisymmetrizer @xmath229 write @xmath230 where @xmath231 is the set of positive roots of @xmath0 and @xmath232 is the dimension of the @xmath68-weight space in @xmath0 .
let @xmath233 be the unique weight defined by @xmath234 for all @xmath57 and @xmath235 .
it satisfies @xmath236 where @xmath237 is the dual coxeter number .
define the operator @xmath238 where @xmath239 makes sense by expanding the reciprocals of the factors of @xmath240 in geometric series .
the computation is defined in a suitable completion of @xmath241 $ ] .
one has @xmath242 for all @xmath51 , which is the weyl - kac character formula ( * ? ? ?
* theorem 10.4 ) .
[ gen boson ] let @xmath0 be an affine kac - moody algebra in one of the infinite families , @xmath243 the crystal of a finite direct sum of irreducible integrable highest weight @xmath4-modules and @xmath244 a subset such that : 1 .
@xmath126 is closed under @xmath146 for all @xmath57 .
@xmath243 is generated by @xmath126 .
3 . for all @xmath127 and @xmath57 ,
if @xmath245 then the @xmath131-string of @xmath129 in @xmath243 is contained in @xmath126 .
then @xmath246 without loss of generality it may be assumed that @xmath247 for some @xmath51 . multiplying both sides of by @xmath248 ,
one obtains @xmath249 observe that both sides are @xmath63-alternating .
the @xmath63-alternants have a basis given by @xmath250 where @xmath51 . taking the coefficient of @xmath251 on both sides , @xmath252 where @xmath253 is the set of pairs @xmath254 such that @xmath255 observe that if @xmath256 is such that @xmath129 is a highest weight vector , then @xmath257 and @xmath258 , for both of the regular dominant weights @xmath259 and @xmath260 are in the same @xmath63-orbit and hence must be equal .
conditions 1 and 2 ensure that @xmath261 .
let @xmath262 .
it is enough to show that there is an involution @xmath263 with no fixed points , such that if @xmath264 then @xmath6 and @xmath265 have opposite sign . in this case
@xmath266 is said to be sign - reversing .
let @xmath267 be the set of pairs @xmath268 such that @xmath245 .
define the map @xmath269 by @xmath270 .
note that @xmath271 by condition 3 .
the condition for @xmath272 is @xmath273 since @xmath274 , @xmath275 , so that @xmath276 .
this shows that @xmath277 is well - defined .
it follows directly from the definitions that @xmath277 is a sign - reversing involution . since @xmath278 it suffices to define a global involutive choice of the canceling root direction for each pair @xmath279 , that is , a function @xmath280 such that if @xmath281 then 1 .
2 . @xmath283 .
let @xmath284 be an expression of @xmath10 as a sum of fundamental weights . by ( * ?
* lemma 8.3.1 ) , @xmath119 is isomorphic to the full subcrystal of @xmath285 generated by @xmath286 . given @xmath279 , let @xmath287 be the image of @xmath129 in the above tensor product of crystals of modules of fundamental highest weight .
let @xmath288 be minimal such that @xmath289 is not a highest weight vector .
then @xmath290 is a highest weight vector , say of weight @xmath211 .
let @xmath291 be a perfect crystal of the same level as @xmath292 . given any @xmath293 , the theory of perfect crystals ( * ? ? ?
* section 4.5 ) gives an isomorphism of @xmath124-weighted crystals @xmath294 where @xmath295 is determined by @xmath296 and @xmath297 and @xmath298 is @xmath124-weighted using the energy function .
let @xmath299 have image @xmath300 where @xmath301 .
assume that @xmath297 is large enough so that @xmath302 .
if one takes the image of @xmath303 in such a tensor product for @xmath304 the tensor factors @xmath305 through @xmath306 do not change .
let @xmath178 be minimal such that @xmath307 is not a highest weight vector .
observe that @xmath178 is independent of @xmath297 as long as @xmath297 is big enough .
then @xmath308 is a highest weight vector , say of weight @xmath309 .
so @xmath310 is such that @xmath311 for some @xmath57 ; let @xmath312 be the set of such @xmath57 .
fix an @xmath313 .
consider the same constructions for @xmath314 .
let @xmath315 be the image of @xmath130 in the above tensor product of irreducible crystals of fundamental highest weights .
then @xmath316 and @xmath317 is not a highest weight vector ; in particular it admits @xmath146 .
take l large enough so that the image of @xmath318 in @xmath319 also has the form @xmath320
. then @xmath321 and @xmath322 admits @xmath146 .
the level of the fundamental weight @xmath323 is @xmath33 . for the affine algebras @xmath324 and @xmath325 , @xmath326 for all @xmath57 .
for all others @xmath327 .
the theorem now follows from lemma [ lev one ] below , applied with the dominant integral weight @xmath309 .
[ lev one ] for the affine kac - moody algebra @xmath0 in one of the infinite families , there exist perfect crystals @xmath126 of level one and two ( the latter case only for @xmath328 ) having the following property .
let @xmath10 be a dominant integral weight of positive level , @xmath329 the set of elements @xmath330 such that @xmath331 is not a highest weight vector .
then there is a map @xmath332 such that if @xmath333 then 1 . @xmath334 .
2 . for any @xmath335 ,
writing @xmath336 , one has @xmath337 and @xmath338 . for the involutive property 2 , it is sufficient that @xmath339 is constant on the nondominant part of every string .
hence one only needs to consider @xmath340 perfect crystals of level one for @xmath324 ( @xmath341 ) , @xmath342 ( @xmath343 ) , @xmath344 ( @xmath345 ) , @xmath346 ( @xmath341 ) , @xmath347 ( @xmath343 ) and @xmath348 ( @xmath349 ) are listed in table [ table crystals ] ( see ( * ? ? ?
* section 6 ) ) .
note that there are no elements satisfying .
this guarantees the existence of the map @xmath339 with the desired properties .
.[table crystals]level one perfect crystals [ cols="^,<",options="header " , ] there are the following strings of length greater than one @xmath350{\footnotesize$\mbox{k}$ } \text(6,0)[]{\footnotesize$\mbox{k}$ } \longarrow(30,0)(60,0 ) \bboxc(96.5,0)(39,13 ) \line(90,-6.5)(90,6.5 ) \text(83,0)[]{\footnotesize$\mbox{k}$ } \text(103,0)[]{\footnotesize$\mbox{k+1}$ }
\longarrow(120,0)(150,0 ) \bboxc(180,0)(52,13 ) \line(180,-6.5)(180,6.5 ) \text(167,0)[]{\footnotesize$\mbox{k+1}$ } \text(193,0)[]{\footnotesize$\mbox{k+1}$ } \ptext(45,2)(0)[b]{k } \ptext(135,2)(0)[b]{k } \end{picture } } } & & \text{for $ 1\leq k <
n$ } \notag \\ & \raisebox{-0.15cm}{\scalebox{0.8 } { \begin{picture}(240,22)(-30,-10 ) \bboxc(6.5,0)(39,13 ) \line(0,-6.5)(0,6.5 ) \text(-7,0)[]{\footnotesize$\mbox{k}$ } \text(13,0)[]{\footnotesize$\overline{\mbox{k+1}}$ } \longarrow(30,0)(60,0 ) \bboxc(90,0)(26,13 ) \line(90,-6.5)(90,6.5 ) \text(83,0)[]{\footnotesize$\mbox{k}$ } \text(96,0)[]{\footnotesize$\overline{\mbox{k}}$ } \longarrow(120,0)(150,0 ) \bboxc(173.5,0)(39,13 ) \line(180,-6.5)(180,6.5 ) \text(167,0)[]{\footnotesize$\mbox{k+1}$ } \text(186,0)[]{\footnotesize$\overline{\mbox{k}}$ } \ptext(45,2)(0)[b]{k } \ptext(135,2)(0)[b]{k } \end{picture } } } & & \text{for $ 1\leq k < n$ } \label{string gen}\\ & \raisebox{-0.15cm}{\scalebox{0.8 } { \begin{picture}(240,22)(-30,-10 ) \bboxc(0,0)(52,13 ) \line(0,-6.5)(0,6.5 ) \text(-13,0)[]{\footnotesize$\overline{\mbox{k+1}}$ } \text(13,0)[]{\footnotesize$\overline{\mbox{k+1}}$ } \longarrow(30,0)(60,0 ) \bboxc(83.5,0)(39,13 ) \line(90,-6.5)(90,6.5 ) \text(77,0)[]{\footnotesize$\overline{\mbox{k+1}}$ } \text(96,0)[]{\footnotesize$\overline{\mbox{k}}$ } \longarrow(120,0)(150,0 ) \bboxc(180,0)(26,13 ) \line(180,-6.5)(180,6.5 ) \text(173,0)[]{\footnotesize$\overline{\mbox{k}}$ } \text(186,0)[]{\footnotesize$\overline{\mbox{k}}$ } \ptext(45,2)(0)[b]{k } \ptext(135,2)(0)[b]{k } \end{picture } } } & & \text{for $ 1\leq k <
n$ } \notag \\[2 mm ] & \raisebox{-0.15cm}{\scalebox{0.8 } { \begin{picture}(240,22)(-30,-10 ) \bboxc(0,0)(26,13 ) \line(0,-6.5)(0,6.5 ) \text(-7,0)[]{\footnotesize$\mbox{n}$ } \text(6,0)[]{\footnotesize$\mbox{n}$ } \longarrow(30,0)(60,0 ) \bboxc(90,0)(26,13 ) \line(90,-6.5)(90,6.5 ) \text(83,0)[]{\footnotesize$\mbox{n}$ } \text(96,0)[]{\footnotesize$\overline{\mbox{n}}$ } \longarrow(120,0)(150,0 ) \bboxc(180,0)(26,13 ) \line(180,-6.5)(180,6.5 ) \text(173,0)[]{\footnotesize$\overline{\mbox{n}}$ } \text(186,0)[]{\footnotesize$\overline{\mbox{n}}$ } \ptext(45,2)(0)[b]{n } \ptext(135,2)(0)[b]{n } \end{picture } } } & & \label{string n}\\[2 mm ] & \raisebox{-0.15cm}{\scalebox{0.8 } { \begin{picture}(240,22)(-30,-10 ) \bboxc(0,0)(26,13 ) \line(0,-6.5)(0,6.5 ) \text(-7,0)[]{\footnotesize$\overline{\mbox{1}}$ } \text(6,0)[]{\footnotesize$\overline{\mbox{1}}$ } \longarrow(30,0)(60,0 ) \text(90,0)[]{\footnotesize$\emptyset$ } \longarrow(120,0)(150,0 ) \bboxc(180,0)(26,13 ) \line(180,-6.5)(180,6.5 ) \text(173,0)[]{\footnotesize$\mbox{1}$ } \text(186,0)[]{\footnotesize$\mbox{1}$ } \ptext(45,2)(0)[b]{0 } \ptext(135,2)(0)[b]{0 } \end{picture}}}.\notag & & \end{aligned}\ ] ] note that none of the elements satisfies .
for type @xmath347 the crystal @xmath351 is perfect of level 2 ( * ? ? ?
1.6 and 6.7 ) .
the elements are given by @xmath352{\footnotesize$\mbox{x}$ } \text(19,0)[]{\footnotesize$\mbox{y}$ } \end{picture}}}$ ] with @xmath353 and @xmath354 .
the action of @xmath355 for @xmath356 is the same as for the above @xmath325 crystal of level one , and @xmath357 where @xmath358 is the involution that exchanges 1 and @xmath359 ( with appropriate reorderings ) .
the strings of length greater than one are the same as in and .
in addition there are the following @xmath89-strings of length 2 @xmath350{\footnotesize$\overline{\mbox{1}}$ } \text(6,0)[]{\footnotesize$\overline{\mbox{1}}$ } \longarrow(30,0)(60,0 ) \bboxc(90,0)(26,13 ) \line(90,-6.5)(90,6.5 ) \text(83,0)[]{\footnotesize$\mbox{2}$ } \text(96,0)[]{\footnotesize$\overline{\mbox{1}}$ } \longarrow(120,0)(150,0 ) \bboxc(180,0)(26,13 ) \line(180,-6.5)(180,6.5 ) \text(173,0)[]{\footnotesize$\mbox{2}$ } \text(186,0)[]{\footnotesize$\mbox{2}$ } \ptext(45,2)(0)[b]{0 } \ptext(135,2)(0)[b]{0 } \end{picture } } } & & \notag \\ &
\raisebox{-0.15cm}{\scalebox{0.8 } { \begin{picture}(240,22)(-30,-10 ) \bboxc(0,0)(26,13 ) \line(0,-6.5)(0,6.5 ) \text(-7,0)[]{\footnotesize$\overline{\mbox{2}}$ } \text(6,0)[]{\footnotesize$\overline{\mbox{1}}$ } \longarrow(30,0)(60,0 ) \bboxc(90,0)(26,13 ) \line(90,-6.5)(90,6.5 ) \text(83,0)[]{\footnotesize$\mbox{1}$ } \text(96,0)[]{\footnotesize$\overline{\mbox{1}}$ } \longarrow(120,0)(150,0 ) \bboxc(180,0)(26,13 ) \line(180,-6.5)(180,6.5 ) \text(173,0)[]{\footnotesize$\mbox{1}$ } \text(186,0)[]{\footnotesize$\mbox{2}$ } \ptext(45,2)(0)[b]{0 } \ptext(135,2)(0)[b]{0 } \end{picture } } } & & \label{string 0}\\ & \raisebox{-0.15cm}{\scalebox{0.8 } { \begin{picture}(240,22)(-30,-10 ) \bboxc(0,0)(26,13 ) \line(0,-6.5)(0,6.5 ) \text(-7,0)[]{\footnotesize$\overline{\mbox{2}}$ } \text(6,0)[]{\footnotesize$\overline{\mbox{2}}$ } \longarrow(30,0)(60,0 ) \bboxc(90,0)(26,13 ) \line(90,-6.5)(90,6.5 ) \text(83,0)[]{\footnotesize$\mbox{1}$ } \text(96,0)[]{\footnotesize$\overline{\mbox{2}}$ } \longarrow(120,0)(150,0 ) \bboxc(180,0)(26,13 ) \line(180,-6.5)(180,6.5 ) \text(173,0)[]{\footnotesize$\mbox{1}$ } \text(186,0)[]{\footnotesize$\mbox{1}$ } \ptext(45,2)(0)[b]{0 } \ptext(135,2)(0)[b]{0 } \end{picture}}}. & & \notag\end{aligned}\ ] ] the only elements fulfilling are @xmath352{\footnotesize$\mbox{1}$ } \text(19,0)[]{\footnotesize$\overline{\mbox{1}}$ } \end{picture}}}$ ] , @xmath352{\footnotesize$\mbox{1}$ } \text(19,0)[]{\footnotesize$\mbox{2}$ } \end{picture}}}$ ] , @xmath352{\footnotesize$\mbox{2}$ } \text(19,0)[]{\footnotesize$\overline{\mbox{1}}$ } \end{picture}}}$ ] , and @xmath352{\footnotesize$\mbox{2}$ } \text(19,0)[]{\footnotesize$\mbox{2}$ } \end{picture}}}$ ] which belong to a @xmath89-string and a @xmath360-string of length two .
it can be checked that setting @xmath361 for @xmath129 one of these four elements guarantees the involutive condition of @xmath339 . for type
@xmath342 the crystal @xmath351 is perfect of level 2 ( * ? ? ?
1.7 and 6.8 ) .
it consists of the elements @xmath352{\footnotesize$\mbox{x}$ } \text(19,0)[]{\footnotesize$\mbox{y}$ } \end{picture}}}$ ] with @xmath353 and @xmath362 ; @xmath363 is excluded .
the action of @xmath355 for @xmath356 is given by the tensor product rule using the action on the level 1 crystal of @xmath342 as given in table [ table crystals ] , and @xmath357 where @xmath358 is the involution that exchanges 1 and @xmath359 ( with appropriate reorderings ) .
the strings of length greater than one are those of equations and and in addition the following @xmath364-string of length four @xmath365{\footnotesize$\mbox{n}$ } \text(6,0)[]{\footnotesize$\mbox{n}$ } \longarrow(20,0)(50,0 ) \bboxc(70,0)(26,13 ) \line(70,-6.5)(70,6.5 ) \text(63,0)[]{\footnotesize$\mbox{n}$ } \text(76,0)[]{\footnotesize$\mbox{0}$ } \longarrow(90,0)(120,0 ) \bboxc(140,0)(26,13 ) \line(140,-6.5)(140,6.5 ) \text(133,0)[]{\footnotesize$\mbox{n}$ } \text(146,0)[]{\footnotesize$\overline{\mbox{n}}$ } \longarrow(160,0)(190,0 ) \bboxc(210,0)(26,13 ) \line(210,-6.5)(210,6.5 ) \text(203,0)[]{\footnotesize$\mbox{0}$ } \text(216,0)[]{\footnotesize$\overline{\mbox{n}}$ }
\longarrow(230,0)(260,0 ) \bboxc(280,0)(26,13 ) \line(280,-6.5)(280,6.5 ) \text(273,0)[]{\footnotesize$\overline{\mbox{n}}$ } \text(286,0)[]{\footnotesize$\overline{\mbox{n}}$ } \ptext(35,2)(0)[b]{n } \ptext(105,2)(0)[b]{n } \ptext(175,2)(0)[b]{n } \ptext(245,2)(0)[b]{n } \end{picture}}}.\ ] ] the same four elements as for @xmath347 satisfy and again setting @xmath361 for these ensures the involutive property of @xmath339 . for type
@xmath344 the crystal @xmath351 is perfect of level 2 ( * ? ? ?
1.8 and 6.9 ) .
it consists of the elements @xmath352{\footnotesize$\mbox{x}$ } \text(19,0)[]{\footnotesize$\mbox{y}$ } \end{picture}}}$ ] with @xmath353 and @xmath366 , the cases @xmath367 and @xmath368 being excluded .
the action of @xmath355 for @xmath356 is given by the tensor product rule using the action on the level 1 crystal of @xmath344 as given in table [ table crystals ] , and @xmath357 where @xmath358 is the involution that exchanges 1 and @xmath359 ( with appropriate reorderings ) .
again the strings of length greater than one are the same as in equations and plus the following @xmath364-strings @xmath350{\footnotesize$\mbox{n}$ } \text(6,0)[]{\footnotesize$\mbox{n}$ }
\longarrow(30,0)(60,0 ) \bboxc(96.5,0)(39,13 ) \line(90,-6.5)(90,6.5 ) \text(83,0)[]{\footnotesize$\mbox{n}$ } \text(103,0)[]{\footnotesize$\overline{\mbox{n-1}}$ }
\longarrow(120,0)(150,0 ) \bboxc(180,0)(52,13 ) \line(180,-6.5)(180,6.5 ) \text(167,0)[]{\footnotesize$\overline{\mbox{n-1}}$ } \text(193,0)[]{\footnotesize$\overline{\mbox{n-1}}$ } \ptext(45,2)(0)[b]{n } \ptext(135,2)(0)[b]{n } \end{picture}}}\\ & \raisebox{-0.15cm}{\scalebox{0.8 } { \begin{picture}(240,22)(-30,-10 ) \bboxc(0,0)(52,13 ) \line(0,-6.5)(0,6.5 ) \text(-13,0)[]{\footnotesize$\mbox{n-1}$ } \text(13,0)[]{\footnotesize$\mbox{n-1}$ } \longarrow(30,0)(60,0 ) \bboxc(83.5,0)(39,13 )
\line(90,-6.5)(90,6.5 ) \text(77,0)[]{\footnotesize$\mbox{n-1}$ } \text(96,0)[]{\footnotesize$\overline{\mbox{n}}$ } \longarrow(120,0)(150,0 ) \bboxc(180,0)(26,13 ) \line(180,-6.5)(180,6.5 ) \text(173,0)[]{\footnotesize$\overline{\mbox{n}}$ } \text(186,0)[]{\footnotesize$\overline{\mbox{n}}$ } \ptext(45,2)(0)[b]{n } \ptext(135,2)(0)[b]{n } \end{picture}}}\\ & \raisebox{-0.15cm}{\scalebox{0.8 } { \begin{picture}(240,22)(-30,-10 ) \bboxc(-6.5,0)(39,13 ) \line(0,-6.5)(0,6.5 ) \text(-13,0)[]{\footnotesize$\mbox{n-1}$ } \text(6.5,0)[]{\footnotesize$\mbox{n}$ } \longarrow(30,0)(60,0 ) \bboxc(90,0)(52,13 ) \line(90,-6.5)(90,6.5 ) \text(77,0)[]{\footnotesize$\mbox{n-1}$ } \text(103,0)[]{\footnotesize$\overline{\mbox{n-1}}$ } \longarrow(120,0)(150,0 ) \bboxc(186.5,0)(39,13 ) \line(180,-6.5)(180,6.5 ) \text(173,0)[]{\footnotesize$\overline{\mbox{n}}$ } \text(193,0)[]{\footnotesize$\overline{\mbox{n-1}}$ } \ptext(45,2)(0)[b]{n } \ptext(135,2)(0)[b]{n } \end{picture}}}\ ; .\end{aligned}\ ] ] in addition to the four elements @xmath352{\footnotesize$\mbox{1}$ } \text(19,0)[]{\footnotesize$\overline{\mbox{1}}$ } \end{picture}}}$ ] , @xmath352{\footnotesize$\mbox{1}$ } \text(19,0)[]{\footnotesize$\mbox{2}$ } \end{picture}}}$ ] , @xmath352{\footnotesize$\mbox{2}$ } \text(19,0)[]{\footnotesize$\overline{\mbox{1}}$ } \end{picture}}}$ ] , and @xmath352{\footnotesize$\mbox{2}$ } \text(19,0)[]{\footnotesize$\mbox{2}$ } \end{picture}}}$ ] also the elements @xmath369{\footnotesize$\overline{\mbox{n-1}}$ } \text(38,0)[]{\footnotesize$\overline{\mbox{n-1}}$ } \end{picture}}}$ ] , @xmath369{\footnotesize$\mbox{n-1}$ } \text(38,0)[]{\footnotesize$\overline{\mbox{n-1}}$ } \end{picture}}}$ ] , @xmath370{\footnotesize$\mbox{n}$ } \text(25,0)[]{\footnotesize$\overline{\mbox{n-1}}$ } \end{picture}}}$ ] , and @xmath370{\footnotesize$\overline{\mbox{n}}$ } \text(25,0)[]{\footnotesize$\overline{\mbox{n-1}}$ } \end{picture}}}$ ] satisfy .
the latter ones are contained in an @xmath371-string and an @xmath364-string . setting @xmath361 for the first four elements and @xmath372 for the last four elements ensures the involutive property of @xmath339 . the crystal @xmath373 is a level 2 perfect crystal for @xmath348 ( *
* sections 1.9 and 6.10 ) .
the elements of this crystal are @xmath374 , @xmath375{\footnotesize$\mbox{x}$ } \end{picture}}}}$ ] , and @xmath352{\footnotesize$\mbox{x}$ } \text(19,0)[]{\footnotesize$\mbox{y}$ } \end{picture}}}$ ] with @xmath376 and @xmath353 ; @xmath363 is excluded .
the action of @xmath355 for @xmath356 is given by the tensor product rule using the action on the level 1 crystal of @xmath348 as given in table [ table crystals ] , and the action of @xmath377 is given by @xmath378{\footnotesize$\emptyset$ } \longarrow(30,0)(60,0 ) \bboxc(70,0)(13,13 ) \text(70,0)[]{\footnotesize$\mbox{1}$ } \ptext(45,2)(0)[b]{0 } \end{picture } } } & \\ \raisebox{-0.15cm}{\scalebox{0.8 } { \begin{picture}(94,22)(-2,-10 ) \bboxc(19.5,0)(13,13 ) \text(19,0)[]{\footnotesize$\mbox{x}$ } \longarrow(30,0)(60,0 ) \bboxc(77,0)(26,13 ) \line(77,-6.5)(77,6.5 ) \text(70,0)[]{\footnotesize$\mbox{1}$ } \text(83,0)[]{\footnotesize$\mbox{x}$ } \ptext(45,2)(0)[b]{0 } \end{picture } } } & \qquad \text{for $ x\neq \overline{1}$}\\ \raisebox{-0.15cm}{\scalebox{0.8 } { \begin{picture}(94,22)(-2,-10 ) \bboxc(19.5,0)(13,13 ) \text(19,0)[]{\footnotesize$\overline{\mbox{1}}$ } \longarrow(30,0)(60,0 ) \text(70,0)[]{\footnotesize$\emptyset$ } \ptext(45,2)(0)[b]{0 } \end{picture } } } & \\ \raisebox{-0.15cm}{\scalebox{0.8 } { \begin{picture}(94,22)(-2,-10 ) \bboxc(13,0)(26,13 ) \line(13,-6.5)(13,6.5 ) \text(6,0)[]{\footnotesize$\mbox{x}$ } \text(19,0)[]{\footnotesize$\overline{\mbox{1}}$ } \longarrow(30,0)(60,0 ) \bboxc(70,0)(13,13 ) \text(70,0)[]{\footnotesize$\mbox{x}$ } \ptext(45,2)(0)[b]{0 } \end{picture } } } & \qquad \text{for $ x\neq 1 $ } \end{split}\ ] ] and undefined otherwise .
the strings of length greater than one are given by , and @xmath379{\footnotesize$\overline{\mbox{1}}$ } \text(6,0)[]{\footnotesize$\overline{\mbox{1}}$ } \longarrow(20,0)(50,0 ) \bboxc(65,0)(13,13 ) \text(66,)[]{\footnotesize$\overline{\mbox{1}}$ } \longarrow(80,0)(110,0 ) \text(125,0)[]{\footnotesize$\emptyset$ } \longarrow(140,0)(170,0 ) \bboxc(185,0)(13,13 ) \text(186,0)[]{\footnotesize$\mbox{1}$ } \longarrow(200,0)(230,0 )
\bboxc(250,0)(26,13 ) \line(250,-6.5)(250,6.5 ) \text(243,0)[]{\footnotesize$\mbox{1}$ } \text(256,0)[]{\footnotesize$\mbox{1}$ } \ptext(35,2)(0)[b]{0 } \ptext(95,2)(0)[b]{0 } \ptext(155,2)(0)[b]{0 } \ptext(215,2)(0)[b]{0 } \end{picture } } } \\[2 mm ] & \raisebox{-0.15cm}{\scalebox{0.8 } { \begin{picture}(280,22)(-30,-10 ) \bboxc(65,0)(13,13 ) \text(66,0)[]{\footnotesize$\mbox{n}$ }
\longarrow(80,0)(110,0 ) \bboxc(125,0)(13,13 ) \text(126,0)[]{\footnotesize$\mbox{0}$ } \longarrow(140,0)(170,0 ) \bboxc(185,0)(13,13 ) \text(186,0)[]{\footnotesize$\overline{\mbox{n}}$ } \ptext(95,2)(0)[b]{n } \ptext(155,2)(0)[b]{n } \end{picture}}}\end{aligned}\ ] ] there are no elements with property . the crystal @xmath373 is a level 2 perfect crystal for @xmath346 ( * ? ? ?
* sec . 1.10 and 6.11 ) .
the elements of this crystal are @xmath374 , @xmath375{\footnotesize$\mbox{x}$ } \end{picture}}}}$ ] , and @xmath352{\footnotesize$\mbox{x}$ } \text(19,0)[]{\footnotesize$\mbox{y}$ } \end{picture}}}$ ] with @xmath380 and @xmath353 .
the action of @xmath355 for @xmath356 is given by the tensor product rule using the action on the level 1 crystal of @xmath346 as given in table [ table crystals ] , and the action of @xmath377 is the same as in .
the strings of length greater than one are as in for @xmath349 , and .
again there are no elements with property .
[ northeast ] suppose @xmath0 is of type @xmath15 in lemma [ lev one ] .
the function @xmath339 amounts to a canonical choice of a simple root @xmath131 among those such that the given element admits @xmath146 .
consider @xmath381 such that @xmath382 .
in addition to the realization of the crystal @xmath383 by the space of homogeneous paths using the crystal given in the proof of lemma [ lev one ] , one may also consider the realization in @xcite by @xmath364-regular partitions .
suppose @xmath112 is the partition corresponding to @xmath129
. then up to the dynkin diagram automorphism that sends @xmath384 to @xmath385 modulo @xmath364 , the choice of violation @xmath339 corresponds to the corner cell of @xmath112 that is in the rightmost column of @xmath112 .
this choice of corner cell is used in @xcite to define the smallest demazure crystal of @xmath383 containing @xmath129 .
[ pi ] let @xmath0 be as in theorem [ gen boson ] , and @xmath126 , @xmath10 , and @xmath216 be as in .
suppose in addition that for all @xmath171 and @xmath386 , if @xmath387 under the local isomorphism @xmath388 and @xmath389 then @xmath390 .
then @xmath391 it is enough to verify the hypotheses of theorem [ gen boson ] , applied to @xmath392 .
@xmath393 is isomorphic to a direct sum of irreducible integrable highest weight modules by theorem [ level tensor ] .
@xmath394 is obviously closed under the @xmath146 .
it follows from ( * ? ? ?
* lemma 1 ) that @xmath395 generates @xmath393 . to check the third condition of theorem [ gen boson ] ,
let @xmath127 and @xmath57 be such that @xmath396 .
then @xmath397 .
this implies that the @xmath131-string of @xmath398 inside @xmath393 , consists of vectors of the form @xmath399 where @xmath400 .
finally , lemma [ energy e ] with @xmath126 replaced by @xmath401 guarantees that the affine weight function on @xmath393 determined by its value on highest weight vectors , agrees on the subset @xmath394 with the function @xmath402 where @xmath403 is the original weight function .
observe that even without the extra hypothesis on the action of @xmath404 in theorem [ pi ] , one obtains a bosonic formula .
the extra condition is only needed to show that the energy function @xmath405 gives rise to the correct affine weight for all elements of the form @xmath406 and not just on the highest weight vectors .
perhaps this extra condition is always a consequence of the other hypotheses .
now the formula is written more explicitly .
let @xmath407 and @xmath408 be of level @xmath13 .
a formula equivalent to is obtained by taking the coefficient of @xmath409 on both sides : @xmath410 { k}(b,{\lambda},{\lambda}',b_0)(q ) = \sum_{(w , b)\in { \mathcal{s } } } ( -1)^w\ ] ] where @xmath253 is the set of pairs @xmath254 such that @xmath411 let @xmath412 be the sublattice of @xmath101 given by the image under @xmath413 of the @xmath414-span of the orbit @xmath415 .
let @xmath416 be the group of translations by the elements of @xmath412 , where @xmath417 is translation by @xmath418
. then @xmath419 and @xmath420 . for @xmath418 and @xmath100 of level @xmath13
, one has ( * ? ? ?
* ( 6.5.2 ) ) @xmath421 the action of @xmath422 on the level @xmath13 weight @xmath10 is given by @xmath423 now @xmath424 where @xmath425 is the dual coxeter number and @xmath426 is the half - sum of the positive roots in @xmath97 .
recall that @xmath427 leaves @xmath428 invariant . in write @xmath429
where @xmath422 and @xmath418 , obtaining @xmath430 since both sides are weights of level @xmath13 , by equating coefficients of @xmath428 and projections into @xmath101 , one obtains the equivalent conditions @xmath431 and @xmath432 therefore one has the equality @xmath433 where @xmath127 satisfies @xmath434
let @xmath100 be a weight of level @xmath13 but not necessarily dominant .
consider the weight @xmath260 .
if it is regular ( not fixed by any @xmath70 ) then there is a unique @xmath70 such that @xmath437 .
it follows from the definition of @xmath5 that @xmath438 then for all @xmath57 , @xmath439 suppose @xmath184 .
then @xmath440 for @xmath186 , recall that @xmath441 then @xmath442 and @xmath443 these equations express the @xmath14-equivalence in @xcite .
let @xmath444 have standard basis @xmath453 and @xmath101 be the subspace of @xmath444 orthogonal to the vector @xmath454 .
then @xmath455 for @xmath456 , @xmath457 , @xmath53 is the ordinary dot product in @xmath444 , and @xmath427 is the symmetric group on @xmath364 letters acting on the coordinates of @xmath444 . since @xmath458 and @xmath446 is @xmath459-linear , one may replace every term @xmath460 by @xmath461 in .
define the map @xmath447^{{\overline{w}}}[q]\rightarrow{\mathbb{z}}[{\overline{p}}]^{{\overline{w}}}[q]$ ] given by @xmath462 .
define @xmath463 in @xmath447^{{\overline{w}}}[q]$ ] by the condition that the above linear map sends @xmath464 and @xmath465 to the same element . with this definition ,
we have @xmath466 it is not hard to see that this recovers the @xmath14-equivalence of schur functions given in @xcite . in this section
it is assumed that @xmath0 is of type @xmath15 , @xmath467 , and the tensor factors @xmath170 are perfect crystals of the form @xmath468 in the notation of @xcite with @xmath469 for all @xmath295 . by restriction to @xmath111 , @xmath170 is the crystal of the irreducible integrable @xmath111-module of highest weight @xmath470 . in this case
@xmath216 is not needed . to see this , recall that @xmath170 can be realized as the set of column - strict young tableaux of the rectangular shape having @xmath471 rows and @xmath472 columns with entries in the set @xmath473 . in @xcite the @xmath122-weighted @xmath125-crystal structure on the perfect crystals @xmath474
is computed explicitly .
in particular , if @xmath386 is a tableau then @xmath475 is at most the number of ones in the tableau @xmath129 , which is at most @xmath472 by column - strictness .
therefore @xmath476 never admits @xmath404 .
thus the energy function @xmath195 of has the property that for any @xmath477 such that @xmath478 , one has @xmath190 .
thus one obtains the bosonic formula in this case . since @xmath0 is of type @xmath15 , @xmath435 and @xmath479 .
take @xmath480 in .
the lattice @xmath412 is given by the root lattice @xmath481 of @xmath97 , which may be realized by @xmath482 .
let @xmath483 be the set of paths @xmath127 of weight @xmath484 .
then @xmath485 notice that @xmath486 is ( up to an overall factor ) the @xmath487 form of the supernomial @xmath329 of ref .
@xcite so that @xmath488 equals the left - hand side of ( * ? ? ?
* ( 9.2 ) ) up to an overall power of @xmath14 .
this shows that the left - hand side of ( * ? ? ?
* ( 9.2 ) ) is indeed the generating function of level-@xmath13 restricted paths . to establish the equality ( * ? ? ?
* ( 9.2 ) ) it remains to prove that also the right - hand side equals the generating function of level - restricted paths .
as in the previous section let @xmath0 be of type @xmath489 and assume that @xmath490 . fix @xmath491 and @xmath492 weights of level 1 .
it is easy to verify that @xmath208 consists of at most one element @xmath493 .
choose @xmath494 such that @xmath495 exists .
then by and we find that @xmath496 where @xmath497 is the set of paths @xmath127 of weight @xmath498 . a similar formula exists for @xmath499 : @xmath500 where @xmath483 is the set of paths @xmath127 of weight @xmath501 .
the right - hand side is the generating function of paths in @xmath126 of level zero since there are no level zero restricted paths unless @xmath126 is empty .
however , the arguments of sections [ sec general bose ] and [ sec inhomo paths ] do not imply that also the left - hand side is the generating function of level zero paths since it was assumed in the proof of theorem [ gen boson ] that the level of the crystals @xmath170 does not exceed @xmath13 .
we have assumed that @xmath502 which are crystals of level one .
however , it is possible to define a sign - reversing involution directly on @xmath490 without using the crystal isomorphisms that are used in the proof of theorem [ gen boson ] .
let @xmath127 .
there exists at least one @xmath503 such that @xmath504 is defined .
define @xmath505 which has the property that @xmath506 where as before @xmath507 .
hence define the involution @xmath508 .
it is again sign - reversing and has no fixed points when @xmath509 .
this proves that the left - hand side of is the generating function of level 0 restricted paths .
equation was conjectured in @xcite . for @xmath510 identity follows from the @xmath14-binomial theorem , for @xmath511 it was proven in ( * ? ? ?
* proposition 5.1 ) and for general @xmath364 it was conjectured in @xcite .
e. date , m. jimbo , a. kuniba , t. miwa , and m. okado , _ paths , maya diagrams and representations of @xmath513 _ , in `` integrable systems in quantum field theory and statistical mechanics '' , 149191 , adv . stud .
pure math .
, * 19 * , academic press , boston , ma , 1989 .
v. lakshmibai , _ standard monomial theory for @xmath515 _ , in `` operator algebras , unitary representations , enveloping algebras , and invariant theory '' ( paris , 1989 ) , 197217 , progr .
, 92 , birkhuser boston , boston , ma , 1990 . | we prove a bosonic formula for the generating function of level - restricted paths for the infinite families of affine kac - moody algebras . in affine type
a this yields an expression for the level - restricted generalized kostka polynomials . supported by the `` stichting fundamenteel onderzoek der materie '' . ] partially supported by nsf grant dms-9800941 . ] |
the goal of this note is to describe some basic operations that allow to construct new free divisors from given ones , and to classify toric free surfaces and binomial free divisors
. we mainly deal with weighted homogeneous polynomials over a field of characteristic @xmath0 , though several statements and constructions generalize to power series .
a ( formal ) _ free divisor _ is a reduced polynomial ( or power series ) @xmath1 in variables @xmath2 over a field @xmath3 such that its jacobian ideal @xmath4 is perfect of codimension @xmath5 in the polynomial or power series ring . for generalities about free divisors and their importance in singularity theory
we refer to , say , @xcite and the references therein .
a determinantal characterization of free divisors is due to k. saito @xcite : a reduced polynomial @xmath1 is a free divisor if and only if there exists a matrix @xmath6 of size @xmath7 with entries in the relevant polynomial or power series ring such that @xmath8 and @xmath9 , where @xmath10 is the usual gradient of @xmath1 . in that case @xmath6
is called a _ discriminant _ ( or _ saito _ ) _ matrix _ of the free divisor .
the normal crossing divisor @xmath11 , for some @xmath12 , provides a simple example of a free divisor .
indeed , it is an example of a _ free arrangement _ , that is , a hyperplane arrangement given by linear equations @xmath13 such that the product @xmath14 is a free divisor , see @xcite for more on free arrangements .
section 2 contains generalities and notation . in section 3
we study homogeneous polynomials that are annihilated by @xmath15 linearly independent _ euler vector fields _ , that is , polynomials @xmath1 such that the vector space generated by the linear derivatives @xmath16 is of dimension at most @xmath5 .
we show that such a polynomial is a free divisor provided the gradient @xmath17 vanishes as an element of the first homology module of the associated _ buchsbaum - rim complex_. as an application , we classify in theorem [ poly3 ] those _ free surfaces _ @xmath18 that are weighted homogeneous and annihilated by some euler vector field . in section 4
we present a _ composition formula _ or _ chain rule _ for free divisors .
such a formula implies , for instance , that if @xmath1 and @xmath19 are free divisors in distinct variables then @xmath20 is also a free divisor .
in section 5 we exhibit some _ triangular _ free divisors , that is , free divisors whose discriminant matrix has a triangular form .
it follows , for instance , that for natural numbers @xmath21 , the polynomial @xmath22 is a free divisor .
in section 6 we characterize _ binomial free divisors _ by showing that a binomial in @xmath23 variables @xmath24 is a free divisor if and only if it is , up to permutation and scaling of the variables , of the form @xmath25 with @xmath26 , @xmath27 , and @xmath28 .
in particular , any reduced binomial is a _ factor of a free divisor_. this observation leads us to ask whether any reduced polynomial is a factor of a free divisor .
we discuss this question in section 7 , where we show that the simplest approach will not work : if @xmath1 is a smooth form of degree greater than @xmath5 in more than @xmath5 variables then @xmath29 is not a free divisor . in the final section 8 ,
we point out that homogeneous free divisors _ extend into the tangent bundle _ : along with @xmath1 , the polynomial @xmath30 in twice as many variables @xmath31 is again a free divisor .
moreover , it will again be _ linear _ , if this holds for @xmath1 .
we want to point out that similar `` extension problems '' for free divisors have been considered by others as well , especially in @xcite .
the authors began discussing the results presented here when they met at the cimpa school on commutative algebra , 26 december 2005 to 6 january 2006 , in hanoi , vietnam .
we want to thank the colleagues who organized that school for the stimulating atmosphere and generous hospitality .
special thanks are due to eleonore faber who not only produced the pictures included here in sections 3 and 5 , but also provided the ( counter-)example in remark [ faber ] .
let @xmath32 be the polynomial ring @xmath33 $ ] or formal power series ring @xmath34\!]$ ] over a field @xmath3 of characteristic @xmath0 .
let @xmath35 denote the module of vector fields ( or @xmath3-linear derivations ) of @xmath32 , with @xmath36 being shorthand for the corresponding partial derivative , @xmath37 . for @xmath38 , we further abbreviate @xmath39 , so that the _ gradient _ of @xmath1 with respect to the chosen variables is given by the vector @xmath40 . for @xmath41 , we call the linear vector field @xmath42 the _ euler vector field _ associated to @xmath43
. it is an _ euler vector field for @xmath1 _ , if @xmath44 , for some @xmath45 .
a vector @xmath46 induces naturally a @xmath47-grading on @xmath33 $ ] by setting @xmath48 .
accordingly , one can assign to any non - zero polynomial @xmath1 a degree @xmath49 , and that polynomial is _
@xmath50homogeneous _ , that is , homogeneous with respect to this grading , if all the nonzero monomials in @xmath1 are of degree @xmath49 .
if @xmath38 is @xmath50-homogeneous , then @xmath51 .
the jacobian . ] ideal @xmath52 of @xmath1 is , by definition , @xmath53 .
note that @xmath54 precisely when there exists a derivation @xmath55 such that @xmath56 .
this happens , for example , if @xmath1 is homogeneous of non - zero degree with respect to some weight @xmath46 .
it is well known that , in general , @xmath52 defines the _ singular locus _ of the hypersurface ring @xmath57 , equivalently , the hypersurface @xmath58 in affine @xmath59space @xmath60 .
a ( formal ) _ free divisor _ is a polynomial ( or power series ) @xmath1 , whose jacobian ideal @xmath52 is _
perfect_. ] of codimension @xmath5 in @xmath32 .
in particular , @xmath1 is then _ squarefree _ , equivalently , the hypersurface ring @xmath57 is _ reduced _ , and we then simply also call @xmath1 _ reduced _ and the singular locus of that hypersurface is a cohen - macaulay subscheme of codimension two in @xmath61 . as simplest examples ,
any separable polynomial in @xmath62 $ ] defines a free divisor , and so does any reduced @xmath63 $ ] .
k. saito , who introduced the notion , gave the following important criterion for @xmath1 to be a free divisor : ( saito @xcite ) [ saito ] let @xmath38 be reduced .
then @xmath1 is a free divisor if and only if there exists a @xmath7 matrix @xmath6 with entries in @xmath32 such that @xmath64 and @xmath65
. the matrix @xmath6 appearing in this criterion is called a _ discriminant _ ( or _ saito _ ) _ matrix _ of @xmath1 .
if the entries of @xmath6 can be chosen to be linear polynomials , then @xmath1 is called a _ linear _ free divisor .
note that @xmath1 is then necessarily a homogeneous polynomial of degree @xmath59.the _ normal crossing divisor _
@xmath66 is a simple example of a linear free divisor .
it follows immediately from this criterion that a free divisor @xmath38 remains a free divisor in any polynomial or power series ring over @xmath32 . when viewed as an element of such larger ring , @xmath1 is called the _ suspension _ of the original free divisor from @xmath32 .
a different way to state the criterion , and to link it with the definition we chose , denote @xmath67 those vector fields @xmath68 such that @xmath69 , equivalently , @xmath70 is a well defined element of @xmath32 . with this notation , one has a short exact sequence of @xmath32modules @xmath71&\operatorname{der}(-\log f)\ar[r]&\theta\ar[r]^-{df}&j(f)/(f)\ar[r]&0 } \end{aligned}\ ] ] and a reduced @xmath1 is a free divisor if , and only if , @xmath72 is a free @xmath32module , necessarily of rank @xmath59 .
a discriminant matrix is then simply the matrix of the inclusion @xmath67 , when bases of these free modules are chosen .
now we turn to our results .
in this section we assume that 1 . @xmath38 is a nonzero squarefree polynomial that belongs to the ideal of its derivatives , @xmath73 .
2 . the @xmath3-vector space of euler vector fields annihilating @xmath1 has dimension at least @xmath15 . in other words , there exist @xmath15 linearly independent euler vector fields @xmath74 , for @xmath75 , such that @xmath76
. denote by @xmath6 the @xmath77 scalar matrix @xmath78 and by @xmath79 the matrix @xmath80 of the same size . under these assumptions
the jacobian ideal of @xmath1 is equal to the ideal of its partial derivatives and has codimension at least two . to show that it defines a cohen - macaulay subscheme of codimension two
, it suffices thus to find a _
burch matrix _
, necessarily of size @xmath81 , for the partial derivatives . by assumption
, we have a matrix equation in @xmath32 of the form @xmath82 we need one more syzygy ! more precisely ; see , for example , @xcite ; to get a hilbert burch matrix for @xmath83 , we want a column vector @xmath84 with entries from @xmath32 , such that we have an equality of sequences of elements from @xmath32 of the form @xmath85 where @xmath86 is obtained from @xmath79 by appending the column vector @xmath50 , and @xmath87 denotes the sequence of appropriately signed maximal minors of the indicated @xmath81 matrix .
define a @xmath32linear map from @xmath88 to @xmath88 through @xmath89 where @xmath90 denotes the @xmath81matrix obtained from @xmath79 by adding the column @xmath50 .
clearly , @xmath91 , and the sequence of free ( graded ) @xmath32modules @xmath92 is the beginning of the _ buchsbaum rim complex _ for the matrix @xmath79 ; see , for example , ( * ? ? ?
* appendix a.2 ) . by the given setup ,
the vector @xmath93 is a cycle in this complex , and the required vector @xmath50 exists if , and only if , the class of @xmath17 is zero in the first homology group @xmath94 of this buchsbaum rim complex .
now , if the ideal of the maximal minors of @xmath79 has the maximal possible codimension , equal to @xmath95 , then the entire buchsbaum
rim complex is exact and so , in particular , @xmath96 .
the minor of @xmath79 obtained by deleting rows @xmath97 and @xmath98 is the monomial @xmath99 , where @xmath100 is the minor of @xmath6 obtained by deleting the rows corresponding to @xmath97 and @xmath98 .
the ideal generated by these minors will have maximal codimension if , and only if , all the maximal minors of @xmath6 are non - zero . summing up , we have the following result . under the assumption _
( a ) _ and _ ( b ) _ , and with the notation as above , [ burim ] 1 .
the polynomial @xmath1 is a free divisor if , and only if , the class of @xmath17 in the first homology @xmath94 of the buchsbaum rim complex associated to @xmath79 vanishes .
2 . if all the maximal minors of @xmath6 are non - zero , then @xmath1 is a free divisor .
consider @xmath101 with @xmath102 nonzero and @xmath103 different exponents with @xmath104 , for each @xmath97 , to ensure that @xmath1 is reduced .
the euler vector field @xmath105 then annihilates @xmath1 if , and only if , @xmath106 and @xmath107 .
assuming @xmath108 for some pair of indices @xmath109 , the space of euler vector fields annihilating @xmath1 has dimension @xmath15 .
the corresponding @xmath110 coefficient matrix @xmath6 then satisfies @xmath111 , where @xmath112 is the obvious @xmath113 matrix of scalars .
linear algebra tells us that the maximal minors of @xmath6 are then , up to sign and a common non - zero constant , equal to the maximal minors of @xmath112 . by virtue of proposition [ burim](2 )
we can conclude that if @xmath108 for all pairs @xmath109 , then the binomial @xmath1 is a free divisor . in section [ binocla ] below we will give a complete characterization of homogeneous binomial free divisors . in three variables the considerations above
lead to a complete characterization of free divisors that are weighted homogeneous and annihilated by an euler vector field . to write down the corresponding hilbert
burch matrices in a compact form , the following tool will be useful .
let @xmath114 be a natural number , @xmath115 $ ] a polynomial ring over a field @xmath3 of characteristic zero , and @xmath116 a subset of the variables @xmath117 .
define a @xmath3linear endomorphism @xmath118 on @xmath32 through the following action on monomials : @xmath119 where @xmath120 denotes the usual total degree of @xmath121 with respect to the variables @xmath122 . in words
, @xmath123 has the polynomials that are homogeneous of total degree @xmath43 in the variables @xmath122 as eigenvectors of eigenvalue @xmath124 .
if @xmath125 is the set of all variables then the corresponding @xmath3linear endomorphism will be simply denoted by @xmath126 .
as is well known , the endomorphism just defined can be used to split in characteristic zero the tautological koszul complex on the variables . here
we will use the following form .
let @xmath127 be the indicated vector space over @xmath3 and @xmath128 an isomorphic copy of it .
let @xmath129 be the exterior algebra over @xmath32 on variables @xmath130 , the graded @xmath32module underlying the usual koszul complex . the @xmath32linear derivation @xmath131 defines a differential on @xmath132 for any choice of @xmath133 .
let @xmath134 denote the subspace generated by those variables @xmath122 among the @xmath135 , for which @xmath136 , and denote by @xmath137 the corresponding variables among the @xmath130 in the isomorphic copy of @xmath138 .
if @xmath139 is a _ cycle _ for @xmath140 , then the class of @xmath141 in @xmath142 is zero if , and only if , @xmath143 in @xmath144 . in that case
, @xmath145 provides a boundary , @xmath146 .
[ poly3 ] let @xmath3 be a field of characteristic zero and @xmath147 $ ] a reduced polynomial in three variables such that @xmath1 is contained in the ideal of its partial derivatives , @xmath148
. assume further that there is a triple @xmath149 of elements of @xmath3 that are not all zero such that the euler vector field @xmath150 satisfies @xmath151 .
we then have the following possibilities , up to renaming the variables : 1 . if @xmath152 , then @xmath1 is a free divisor with hilbert burch matrix @xmath153 where @xmath154 denotes the corresponding second order derivative of @xmath1 .
if @xmath155 , but @xmath156 , then @xmath1 is a free divisor if , and only if , @xmath157 .
if that condition is verified and @xmath158 , then @xmath159 is an element of @xmath32 and a hilbert burch matrix is given by @xmath160 3 . if @xmath161 , then @xmath1 is independent of @xmath162 and so , as the suspension of a reduced plane curve , is a free divisor .
we simply need to verify that the hilbert
burch matrix is correct .
one may either use now the preceding lemma , or calculate directly , as we will do .
we just verify that , in case ( 1 ) , the minor obtained when deleting the first row is correct , leaving the remaining calculations to the interested reader .
it suffices to check the case when @xmath163 is a monomial with @xmath164 and @xmath165 .
then , @xmath166 as required . to apply this result
, we need to detect euler vector fields annihilating given polynomials , and the following remark is useful for this purpose .
assume @xmath1 is a polynomial that is homogeneous with respect to two weights @xmath167 .
for every @xmath168 , the polynomial @xmath1 is then homogeneous with respect to @xmath169 , of degree @xmath170 .
taking @xmath171 and @xmath172 , we conclude that @xmath1 is homogeneous of degree @xmath0 with respect to @xmath173 , and so the corresponding euler vector field annihilates @xmath1 .
if further some degree @xmath170 is not zero , then @xmath1 satisfies the assumption @xmath174 from the beginning .
this remark can be applied as follows .
set @xmath175 with @xmath176 , @xmath177 and @xmath178 .
assume that the @xmath179 are non - zero and distinct so that @xmath1 is reduced .
then @xmath1 is a free divisor if , and only if , not both @xmath180 and @xmath181 equal @xmath0 , equivalently , @xmath182 . to prove the statement ,
take @xmath183 and @xmath184 , so that @xmath1 becomes homogeneous with respect to both @xmath185 and @xmath50 ,
satisfying @xmath186 hence , by the remark above , @xmath187 , and the euler vector field associated to @xmath188 annihilates @xmath1 .
clearly , the second and the third coordinates of this vector are non - zero , while the first one equals @xmath189 .
now , if @xmath180 or @xmath181 is non - zero , then @xmath190 and we conclude by theorem [ poly3 ] , either part ( 1 ) or ( 2 ) , that @xmath1 is a free divisor . on the other hand ,
if @xmath191 then @xmath1 contains a pure power of @xmath117 and so @xmath192 . we may then conclude by theorem [ poly3](2 ) that @xmath1 is not a free divisor .
some isolated members of this family of examples have been identified as free divisors before : @xmath193 the quadratic cone with , respectively , one or two planes , of which one is tangent , or @xmath194 the whitney umbrella with an adjoint plane ; see @xcite .
a remarkable feature of this example is that it exhibits free surfaces with arbitrarily many irreducible components that are not suspended , in that we can , for example , extend the family of examples involving quadratic cones to @xmath195 for @xmath196 with @xmath197 and scalars @xmath198 satisfying @xmath199 .
such @xmath1 will clearly have @xmath200 many irreducible components , @xmath201 among them planes . [
cols="^,^ " , ] [ sumpow ] given positive integers @xmath202 , for @xmath203 , set @xmath204 .
take @xmath205 as a seed and set @xmath206 to obtain @xmath207 for @xmath208 .
the resulting product @xmath209 of brieskorn pham polynomials is a free divisor by proposition [ trianfree ] .
one can easily calculate the entries of the discriminant matrix .
to illustrate , we treat the case where each exponent is @xmath210 , so that @xmath211 .
the first column can be taken as representing the usual euler vector field that is the unique extension of the euler vector field for @xmath212 .
the second column can be taken to correspond to the vector field @xmath213 that in turn corresponds to the automorphism interchanging @xmath214 and @xmath215 . as for this @xmath68
one has @xmath216 , lemma [ trilemma ] shows that the corresponding matrix entries below the second row will be zero as well .
now we indicate how to obtain the entries of columns @xmath217 through @xmath97 . counting from the top , start with @xmath218 , thus , putting @xmath219 as the entry in the @xmath220 row as first nonzero entry in column @xmath221 , and note that @xmath222 , so that @xmath223 . by lemma [ trilemma ] ,
the entry below it will be @xmath224
now @xmath225 again , and induction shows that a relevant discriminant matrix can be taken in the form @xmath226
the goal of this section is to investigate binomials @xmath227 , with @xmath228 , and exponent vectors @xmath229 with @xmath230 , @xmath231 , that are free divisors .
this forces each entry of @xmath232 to be in @xmath233 and we can absorb the constants @xmath234 into the variables to reduce to the form @xmath235 , where @xmath236 is a product of distinct variables and @xmath237 are coprime monomials .
we further assume @xmath238 $ ] , with @xmath3 as usual a field of characteristic @xmath0 , and we may suppose that @xmath239 involves all the variables , as otherwise it is just a suspension of a divisor that satisfies this requirement . with these preparations
we show the following result .
[ binofree ] the binomial @xmath235 as above is a free divisor if 1 . at most one of the variables appearing in @xmath240
does not appear in @xmath236 , and 2 . at most one of the variables appearing in @xmath241
does not appear in @xmath236 .
note that if @xmath239 is required to involve all variables , then these conditions imply @xmath242 . if @xmath239 is a _ homogeneous _ binomial , that is , @xmath243
, then the preceding sufficient conditions are also necessary .
for the first claim , we can write , up to a permutation of the variables and setting @xmath244 and @xmath245 , @xmath246 where @xmath247 and @xmath248 with @xmath26 , @xmath27 and @xmath249 .
let @xmath250 be the @xmath3 vector space generated by the monomials @xmath251 and @xmath252 involved in @xmath239 .
obviously , @xmath250 is @xmath5-dimensional , the elements @xmath253 form a basis , and @xmath250 contains @xmath254 for each @xmath255 .
so we get relations @xmath256 whence we also get relations @xmath257 collecting this information in the @xmath258 matrix @xmath259 it follows from ( 1 ) and ( 4 ) that the first @xmath260 entries of @xmath261 are congruent to @xmath0 modulo @xmath239 , while ( 5 ) implies that the last entry of @xmath261 equals @xmath0 already in @xmath32 .
finally , it is straightforward that @xmath262 whence we conclude from saito s criterion in theorem [ saito ] that @xmath239 is a free divisor .
next we show that if @xmath239 is a _ homogeneous _ free divisor then conditions ( a ) , ( b ) are satisfied .
we argue by contradiction .
suppose that @xmath239 is a free divisor that involves all variables , but fails one of the conditions ( a ) or ( b ) . by symmetry , and after permutating the variables , we may assume that @xmath239 is of the form : @xmath263 where we set @xmath264 as before , and @xmath265 . with @xmath266
again the jacobian ideal of @xmath239 , note that @xmath267 . since @xmath266 is perfect of codimension @xmath5 , either @xmath268 is a minimal prime of @xmath266 or @xmath269 . in the former case
, @xmath270 implies @xmath271 , and that is impossible . in the latter case , @xmath272 and
so @xmath273 must be divisible by @xmath274 for some @xmath97 .
this contradicts the homogeneity of @xmath239 .
a particular case of theorem [ binofree ] has recently been presented independently by simis and tohaneanu ( * ? ? ?
2.11 ) : in our notation from the proof above , they take a homogeneous binomial of the form @xmath275 , with @xmath276 , and @xmath136 for @xmath277 in @xmath278 , so that @xmath279 is homogeneous of degree @xmath280 and the only potentially missing variable in the first summand is @xmath214 .
the authors then affirm that @xmath281 are homogeneous free divisors .
theorem [ binofree ] shows that in each case , @xmath282 is a homogeneous free divisor as well .
the results of the previous sections show that : 1 .
any reduced homogeneous binomial has a multiple that is a free divisor by theorem [ binofree ] .
2 . if @xmath3 is algebraically closed , then any quadric @xmath283 can be put in standard form @xmath284 .
hence it has a multiple that is a free divisor by example [ sumpow ] .
3 . if @xmath285 are free divisors in distinct sets of variables , then @xmath286 divides the free divisor @xmath20 by corollary [ corofg ] .
so we are led to ask : let @xmath1 be a ( homogeneous ) reduced polynomial .
does there exist a free divisor @xmath19 such that @xmath1 divides @xmath19 ?
this question is also raised and adressed in @xcite . in light of the discussion above , the first case to look at
is that of cubics in @xmath217 variables .
again , by example [ sumpow ] , we know that the fermat cubic @xmath287 divides the free divisor @xmath288 .
so , what about other smooth cubics or smooth hypersurfaces in general ?
what we can prove is a negative result : it asserts that a smooth form , in @xmath289 variables of degree larger than 2 , times a product of @xmath59 linearly independent linear forms is never a free divisor .
[ divdiv1 ] let @xmath1 be a smooth form of degree @xmath290 in @xmath289 variables and @xmath291 linearly independent linear forms . set @xmath292 and denote @xmath293 $ ] the jacobian ideal of @xmath19 .
then one has : 1 .
@xmath19 is not a free divisor , instead 2 .
@xmath294 . in particular ,
if @xmath295 then @xmath296 . since @xmath297 and @xmath289 implies @xmath298 , assertion ( 1 ) follows indeed from ( 2 ) as claimed . to prove ( 2 ) in theorem [ divdiv1 ]
, we need to set up some notation . to avoid confusion
, @xmath299 will denote the vector with coordinates @xmath300 , while @xmath301 denotes the ideal or module generated by the @xmath300 . for a form @xmath1
, we set @xmath302 , with @xmath303 as before .
[ divdiv2 ] let @xmath1 be a form in @xmath33 $ ] .
if @xmath304 is reduced , then the ideals @xmath305 and @xmath306 of @xmath32 have the same projective dimension . in particular , @xmath19 is a free divisor if , and only if , @xmath307 is perfect of codimension @xmath5 .
set @xmath308 and note that @xmath309 . if @xmath310 is a syzygy of @xmath311 , then @xmath312 is thus a syzygy of @xmath313 . by the hilbert
burch theorem , the syzygy module of @xmath313 is generated by @xmath314 with @xmath315 , whence there exist polynomials @xmath316 such that @xmath317 since @xmath19 is squarefree , @xmath318 does not divide @xmath1 , whence that variable must divide @xmath179 for each @xmath97 . in other words , @xmath319 for suitable @xmath320 , and then @xmath321 is a syzygy of @xmath322 .
therefore , the @xmath32-linear map @xmath323 sending @xmath324 to @xmath325 induces an isomorphism between the syzygy module of @xmath322 and the syzygy module of @xmath326 . because @xmath1 is homogeneous , one has the euler relation @xmath327 , whence @xmath328 using the euler relation once more , one obtains as well @xmath329 , thus , @xmath330 , and then also @xmath331 accordingly , these ideals agree .
it follows that the first syzygy module of the ideal @xmath305 and that of the ideal @xmath332 differ only by a free summand whose rank is in fact the @xmath3-dimension of the vector space of euler vector fields annihilating @xmath1 .
so the statement follows .
[ exdiv ] let us illustrate the preceding result . 1 .
consider @xmath333 with @xmath334 , with @xmath335 pairwise coprime monomials of same degree , and set @xmath336 .
then @xmath337 , because here the ideal @xmath338 is the complete intersection ideal @xmath339 .
2 . let @xmath1 be the _ cayley form _ in @xmath59 variables , the elementary symmetric polynomial of degree @xmath340 , that can be written @xmath341 and consider @xmath304 .
+ denoting @xmath342 the ideal generated by all square - free monomials of degree @xmath343 , it is well known that @xmath342 is perfect of codimension @xmath344 .
the radical of the jacobian ideal of @xmath1 is easily seen to be @xmath345 .
so @xmath1 is irreducible and , for @xmath346 , singular with singular locus of codimension @xmath217 .
+ on the other hand , one checks that @xmath347 and lemma [ divdiv2 ] therefore verifies that @xmath19 is a free divisor , as was also observed in @xcite , where further a discriminant matrix is given .
3 . for a given form @xmath1 , smooth and in generic coordinates ,
the elements @xmath348 tend to form a regular sequence . in that case
, the resolution of the first syzygy module of @xmath305 is thus given by the corresponding tail of the koszul complex on @xmath348 , shifted in degree , and therefore @xmath349 embeds as the nonzero artinian submodule @xmath350 into @xmath351 , forcing @xmath296 . as a concrete example , take a fermat hypersurface @xmath352 , with @xmath353 .
[ divdiv4 ] for a subset @xmath6 of @xmath354 , set @xmath355 . with notation as in lemma [ divdiv2 ] , one obviously has @xmath356 accordingly , either @xmath357 or the projective dimension of @xmath349 is at least the codimension of @xmath358 .
in particular , if @xmath359 , then no such monomial is in @xmath348 , and we see again that @xmath296 .
the last example leads to the following result .
[ divdiv7 ] assume @xmath360 $ ] with @xmath289 is smooth of degree @xmath297 , and let @xmath361 be linearly independent linear forms . with @xmath362 one then has @xmath363 changing coordinates we may assume @xmath364 .
set @xmath365 . in view of example [ exdiv]([divdiv4 ] ) to lemma [ divdiv2 ] , it is enough to show that @xmath366 . if @xmath367 this is obvious . if @xmath368 , then @xmath369 , and we argue as follows .
suppose by contradiction that @xmath370 with @xmath371 .
let @xmath372 be a monomial in the support of @xmath1 that is different from @xmath373 . from ( [ dis * ] ) it follows that @xmath374 .
if we show that the support of @xmath1 contains at least @xmath59 monomials different from @xmath373 whose exponents are linearly independent , we can conclude that @xmath375 for all @xmath97 , thus , contradicting ( [ dis * ] ) . since @xmath1 is smooth , for each @xmath97 there
exists some @xmath376 , such that the monomial @xmath377 is in the support of @xmath1 .
we claim that the exponents of @xmath378 , for @xmath379 , are indeed linearly independent . to prove this , consider the linear map @xmath380 defined as @xmath381 .
any such map is easily seen to satisfy @xmath382 , whence the eigenvalues of @xmath383 are either @xmath0 or roots of unity . in particular , no integer @xmath384 with @xmath385 is a root of the characteristic polynomial @xmath386 of @xmath383 .
therefore we have that @xmath387 at @xmath388 , and this proves the claim . as for a last ingredient ,
note the following .
[ divdiv8 ] if @xmath360 $ ] is smooth , then the codimension of @xmath389 is at least @xmath390 .
let @xmath391 be a minimal prime of @xmath392 in @xmath32 .
if @xmath232 is the number of variables @xmath318 contained in @xmath391 , then that prime ideal contains at least @xmath393 of the @xmath394 .
hence @xmath391 contains two regular sequences : one of length @xmath232 and the other of length @xmath393 .
so the codimension of @xmath395 is at least @xmath390 .
the _ proof of theorem _ [ divdiv1 ] is now obtained by combining lemma [ divdiv2 ] , proposition [ divdiv7 ] , and lemma [ divdiv8 ] .
as far as we know , in example [ exdiv]([divdiv4 ] ) , it might be even true that for _ any _ smooth @xmath1 in _ any _ system of coordinates , @xmath396 , so that then , in particular , always @xmath296 .
however , for a smooth @xmath1 , the ideal @xmath338 can be of codimension @xmath390 , but , of course , only for @xmath59 even .
for example , @xmath397 is smooth and the codimension of @xmath398 is @xmath5 .
nevertheless , in this case @xmath399 still has depth @xmath0 since @xmath400 .
let @xmath401 $ ] as before , and set @xmath402 $ ] . define a map @xmath403 by @xmath404 for every @xmath38 .
clearly @xmath405 is a @xmath3-linear derivation . for a matrix @xmath406 with entries in @xmath32
we set @xmath407 .
[ jets ] let @xmath38 be a homogeneous free divisor of degree @xmath408 .
then @xmath409 is a free divisor in @xmath410 , in @xmath411 variables and of total degree @xmath412 , that is linear if @xmath1 is so .
first note that @xmath409 is reduced because @xmath413 is irreducible . by contradiction , if @xmath413 were reducible then , since @xmath413 is homogeneous of degree @xmath414 in the @xmath125 s , the partial derivatives of @xmath1 had a non - trivial common factor contradicting the fact that @xmath1 is reduced .
secondly we identify a discriminant matrix for @xmath409 .
since @xmath1 is homogeneous , a discriminant matrix for @xmath1 can be constructed as follows .
because @xmath52 is a perfect ideal of codimension @xmath5 , we can find a hilbert - burch matrix @xmath415 for @xmath52 , of size @xmath416 , such that the @xmath417-minor of @xmath79 obtained by removing the @xmath97-th row is @xmath418 .
adjoining @xmath419 as a column to the matrix @xmath79 , we obtain the matrix @xmath420 that is by construction a discriminant matrix for @xmath1 .
we now claim that the following @xmath421 block matrix @xmath422 is a discriminant matrix for @xmath409 .
its determinant is clearly @xmath409 by definition of @xmath423 and @xmath424 .
the product rule yields @xmath425 now @xmath426 , and so it remains to show that @xmath427 expanding returns the vector @xmath428 concerning its first part , note that @xmath429 , whence @xmath430 therefore , ( [ disff * ] ) holds and @xmath431 is confirmed as a free divisor .
the assertions on degree and number of variables are obvious from the construction .
viewing @xmath38 as the function @xmath433 $ ] , its differential fits into the exact zariski
jacobi sequence of khler differential forms @xmath434 \omega^{1}_{k[t]/r}&\ar[l]\omega^{1}_{r / k}\cong \oplus_{i}rdx_{i } & & \ar[ll]_-{df\partial/\partial t}\omega^{1}_{k[t]/k}\otimes_{k[t]}r\cong rdt}\ ] ] and one may interpret @xmath435 as the ring of regular functions on the tangent bundle @xmath436 over @xmath437 .
this identifies @xmath438 with the regular functions on the total space of the affine relative tangent `` subbundle '' @xmath439 , the kernel of the jacobian map @xmath440 that consists of the vector fields vertical with respect to ( the fibres of ) @xmath1 over the affine line @xmath441 $ ] .
accordingly , the hypersurface @xmath442 defined by @xmath431 is the union of that affine `` bundle '' with @xmath443 , the restriction of the total tangent bundle @xmath444 to @xmath445 , in turn the fibre over @xmath0 of the function @xmath1 .
equivalently , @xmath443 is the suspended free divisor obtained as the inverse image of @xmath445 along the structure morphism @xmath446 .
thus , @xmath447 .
@xmath448t_{x / s}\ar@{_{(}->}[rd]\ar[rr]&&\{0\}\ar@{_{(}->}[rd]\\ \ \hphantom{h}h\ar@{^{(}->}[rr]&&t_{x}\ar[rr]^-{df}\ar[rd]&&t_{s}\ar[rd]\\ & \ar@{_{(}->}[ul](fp)^{-1}(0)\ar[rd ] \ar@{^{(}->}[ru]&&x\ar[rr]^-{f}&&s\\&&f^{-1}(0)\ar@{^{(}->}[ru]\ar[rr]&&\{0\}\ar@{^{(}->}[ru]}\end{aligned}\ ] ] interesting examples are hard to visualize as they will live in four or more dimensions . however , the intersection of the two ( unions of ) components , @xmath449 is easy to understand : geometrically , over @xmath450
it fibres into the union of the hyperplanes perpendicular to @xmath451 for some @xmath452 on @xmath58 , that is , @xmath453 1 .
[ remff*1 ] given a homogeneous free divisor @xmath1 in a polynomial ring of dimension @xmath59 , one can iterate the use of theorem [ jets ] to get an infinite family @xmath456 of homogeneous free divisors , defined by @xmath457 and @xmath458 , where @xmath459 is , of course , to be understood relative to the polynomial ring containing @xmath460 . by construction , @xmath461 belongs to a polynomial ring of dimension @xmath462 , its degree equals @xmath463 , and it is a linear free divisor if , and only if , @xmath1 is linear . + taking @xmath464 as a seed , we obtain the sequence of linear free divisors @xmath465 in @xmath466 $ ] .
2 . [ remff*2 ] theorem [ jets ] holds also for free divisors that are weighted homogeneous of degree @xmath467 with respect to some weight vector @xmath468 . in the proof one simply replaces the column vector @xmath469 in the discriminant matrix with @xmath470 .
again , linearity is preserved . with @xmath471 ,
let @xmath472 $ ] , assign weights @xmath473 , and set @xmath474 .
then @xmath475 is a free divisor in @xmath476 variables of weighted homogeneous degree @xmath477 that will be linear along with @xmath1 .
_ nonlinear sections of nonisolated complete intersections_. new developments in singularity theory ( cambridge , 2000 ) , 405445 , nato sci .
ii math .
, * 21 * , kluwer acad .
publ . , dordrecht , 2001
. _ commutative ring theory_. translated from the japanese by m. reid .
second edition .
cambridge studies in advanced mathematics , * 8*. cambridge university press , cambridge , 1989 .
. + isbn : 0 - 521 - 36764 - 6 | we present several methods to construct or identify families of free divisors such as those annihilated by many euler vector fields , including binomial free divisors , or divisors with triangular discriminant matrix .
we show how to create families of quasihomogeneous free divisors through the chain rule or by extending them into the tangent bundle .
we also discuss whether general divisors can be extended to free ones by adding components and show that adding a normal crossing divisor to a smooth one will not succeed . |
the mixing of neutral @xmath0 mesons provides a fertile area for precision tests of the standard model . the fact that the mixing is a second order weak process and is also suppressed by small ckm angles in the standard model makes it sensitive to new physics . in order to take full advantage of experimental measurements of the mixing
, one needs to determine the hadronic matrix elements of the effective weak operators . for @xmath0 mixing the relevant operators are local four - quark operators with @xmath1 , where @xmath2 is @xmath2-quark number , and the relevant states are @xmath3 and @xmath4 mesons or @xmath5 and @xmath6 mesons .
a first - principle evaluation of such operator matrix elements is possible with lattice qcd .
lattice computations usually involve an extrapolation in light quark masses to the physical up and down masses , and always require an extrapolation in lattice spacing @xmath7 to @xmath8 , the continuum .
these extrapolations can be controlled by using a version of chiral perturbation theory that includes the effects of the discretization errors associated with the choice of lattice action .
in two recent lattice calculations of @xmath0 mixing @xcite , staggered light quarks are combined with non - staggered heavy quarks using nrqcd @xcite or the fermilab action @xcite , respectively . in such cases ,
the appropriate chiral theory is `` rooted , heavy - meson staggered chiral perturbation theory '' ( ) @xcite . in this paper ,
i calculate @xmath0 mixing to one - loop order in . roughly speaking
, i work to leading order in the heavy - quark expansion , although i do include the large @xmath9 effects : the hyperfine splitting of @xmath0 and @xmath10 and the flavor splitting of @xmath11 and @xmath12 .
this is a systematic approximation in the power counting introduced by boyd and grinstein @xcite and discussed recently in ref .
@xcite for the lattice calculation of heavy - light meson decay constants .
if instead one prefers a power counting strictly in @xmath9 , which sets the splittings to zero , it is easy to take that limit of the results given in this paper . in the calculation , it is important to take into account the exact form of the lattice operator used to approximate the continuum one .
references @xcite construct the four - quark operators as the local product of two local bilinears , each formed from a heavy antiquark field and a light quark field with the `` naive '' lattice discretization . as proposed in ref .
@xcite and discussed below , the naive field is constructed in turn from the simulated staggered fields ( or more precisely , the naive propagator is constructed from the staggered propagator ) . both the use of naive fields and the local nature of the four - quark operator influence the form of the corrections at one loop .
it is not hard to understand the qualitative effect of the lattice locality of the four - quark operator . because of lattice doubling symmetry
, a ( single - component ) staggered quark field actually corresponds to the 16 continuum degrees of freedom of four `` tastes '' of four - component dirac particles . on the lattice ,
the spin and taste degrees of freedom can be made explicit in position space by combining the 16 staggered components associated with an elementary hypercube @xcite . for our four - quark operators ,
the two light staggered quarks are tied to the same space - time point , so that their spin and tastes are coupled .
the coupling produces undesired contributions to the operator , with `` wrong spin '' and `` wrong taste . ''
these undesired contributions appear at @xmath13 in the lattice spacing .
fortunately , in the matrix elements considered here , continuum @xmath14 taste symmetry suppresses wrong - taste contributions and therefore wrong - spin contributions . on the lattice ,
@xmath14 taste symmetry is violated at @xmath15 , so the undesired contributions come in at that order . since @xmath15 corrections appear at one loop in , that is the order at which we find wrong - spin , wrong - taste contributions to @xmath0 mixing .
similarly , it is clear that the effect of using naive quarks in the operators and interpolating fields must be to sum over tastes , since the naive quarks have no explicit taste index .
however , the details are non - trivial .
it turns out that the heavy - light meson propagator is simply an average over the initial and final tastes , which are equal to each other .
the three - point function involves a complicated sum over tastes of the staggered quarks in the interpolating fields and four - quark operator , and there is coupling between the spin matrices in the operator and the taste sum .
these details play a key part in the discussion below .
a calculation in can be thought of as `` staggering '' the corresponding continuum calculation , which here would be in heavy - meson chiral perturbation theory .
in fact , when the continuum calculation includes partial quenching effects , it is often possible to deduce the proper staggered version without having to recalculate explicitly any of the diagrams ( see , for example , ref .
@xcite ) . in the current case
, a partially quenched continuum calculation does exist @xcite .
however , the complications due to the naive - to - staggered translation and the wrong spin - taste contributions make it necessary to perform the staggered calculation from scratch . nevertheless , ref .
@xcite is extremely useful here , and provides a check of the current results in the @xmath16 limit .
the @xmath0 mixing matrix elements for any four - quark operator that can appear in the standard model and in possible extensions such as supersymmetry can be written in terms of the matrix elements of the following five operators @xcite @xmath17\nonumber \\ \co_2 & = &
( \bar b l q ) \;[\bar b l q]\nonumber \\ \co_3 & = & ( \bar b l q]\ ; [ \bar b l q ) \nonumber\\ \co_4 & = & ( \bar b l q)\ ; [ \bar b r q ] \nonumber\\ \co_5 & = & ( \bar b l q]\ ; [ \bar b r q ) \ , \eqn{susy - basis}\end{aligned}\ ] ] where pairs of round or square parenthesis indicate how the color indices are to be contracted , and @xmath18 and @xmath19 are the right and left projectors : @xmath20 and @xmath21 .
operators , @xmath22 , @xmath23 and @xmath24 appear in the standard model , with @xmath22 ( which mixes with @xmath23 under renormalization ) governing the mass differences of the neutral @xmath0 eigenstates , @xmath25 and @xmath26 .
operators @xmath27 and @xmath28 appear in extensions of the standard model .
additional operators with @xmath29 can also contribute beyond the standard model , but parity implies that their mixing matrix elements in qcd are equal to those of the above operators .
in addition to parity , fierz transformations are needed in order to write the mixing matrix elements of any four - quark operator with these quantum numbers in terms of those in ; for a detailed explanation see ref .
@xcite .
i note that the corresponding projectors @xmath30 in ref .
@xcite do not have the factor of @xmath31 , so the operators there are differently normalized .
since in any case unknown low energy constants will enter in the chiral theory , this normalization difference is unimportant here .
the fact that is a complete set of operators for @xmath0 mixing implies that wrong - spin contributions to the operators do not in fact lead to any new low energy constants in the chiral theory .
wrong - spin contributions to operator @xmath32 merely lead to the appearance of the low energy constants associated with operators @xmath33 in the one - loop expression for the @xmath32 matrix element .
thus , a staggered lattice calculation that computes the matrix elements of all the operators in will not suffer from increased systematic or statistical errors due to the wrong - spin issue .
existing calculations @xcite study the matrix element @xmath22 exclusively . in the case of ref .
@xcite , the one - loop contributions of wrong - spin operators were not known at the time , so one presumably should include some additional systematic error in their result . in the case of ref .
@xcite , it was not possible to make a complete study of this effect because the matrix elements of the other operators were not computed .
however , an associated systematic error was estimated . in the relevant staggered simulations , the fourth - root of the fermion determinant
is taken in order to eliminate the four - fold multiplicity of tastes in the sea . the rooted theory
then suffers from nonlocal violations of unitarity at non - zero lattice spacing @xcite . however , there are strong theoretical arguments @xcite , as well as other analytical and numerical evidence @xcite , that the local , unitary theory of qcd is recovered in the continuum limit . furthermore , it is straightforward to take rooting into account in the chiral theory .
one simply needs to multiply each sea quark loop by a factor of @xmath34 @xcite .
this can be done either by following the quark flow @xcite to locate the loops , or more systematically by replicating the sea quarks @xmath35 times and taking @xmath36 in the result of the chiral calculation @xcite .
since i will need to work out the quark flows in any case , i use the former method below .
the remainder of this paper is organized as follows .
briefly reviews the basics of , focusing in particular on those aspects that will be important here .
in , i discuss the connection between naive and staggered quarks , and how it influences the structure of the four - quark operators and the interpolating heavy - light meson fields .
the calculation of the one - loop diagrams is detailed in .
i also briefly explain why taste - violations coming from mixing under renormalization do not need to be considered at this order .
compiles the final formulae for the chiral and continuum extrapolation of the matrix elements of the operators defined in ; corresponding results for the b ( `` bag '' ) parameters are collected in appendix [ sec : bparams ] .
i conclude in and make some additional comments about existing and future lattice computations of @xmath0 mixing . a preliminary account of the current calculation appears in ref .
@xcite .
though i denote heavy quarks as @xmath2 quarks and heavy mesons as @xmath0 mesons throughout , the current calculation in also applies to the local matrix elements in neutral @xmath37 mixing , with the usual caveat that the omitted @xmath38 terms ( @xmath39 is a generic heavy quark ) are larger in that case
. however , long distance contributions are presumably much more important in the @xmath37 case @xcite than in the @xmath0 case .
such contributions are beyond the scope of this work , and are likely to be difficult to compute on the lattice .
see however ref .
@xcite for a lattice approach to long - distance effects in the kaon system .
here , i give a brief summary of some of the basic features and definitions from staggered chiral perturbation theory , both of heavy - light mesons and of light mesons ( `` pions '' ) . in this summary ,
i follow ref .
@xcite fairly closely , but adapt the notation slightly , to make it more similar to that of ref .
the reader is referred to the literature @xcite for more details . for convenience in making connection to ref .
@xcite , i write the lagrangian and do the perturbative calculations in minkowski space .
let @xmath40 be the field that annihilates the pseudoscalar meson containing a heavy quark @xmath2 and a light quark @xmath41 ( the @xmath42 for @xmath43 ) , while @xmath44 does the same for the vector meson ( @xmath45 for @xmath43 ) . to take advantage of heavy - quark spin symmetry , pseudoscalar and vector fields are combined in the field @xmath46\ , \eqn{hdef}\ ] ] which destroys a meson , while @xmath47\frac{1 + \vslash}{2 } \ , \eqn{hbardef}\ ] ] creates a meson .
here @xmath48 is the meson velocity , and the @xmath49 on @xmath50 and @xmath51 indicates that they are minkowski - space matrices : @xmath52 , @xmath53 , and @xmath54 , with @xmath55 and @xmath56 the euclidean ( hermitian ) dirac matrices .
the label @xmath57 indicates the `` flavor - taste '' index of the light quark in the meson .
for @xmath58 flavors of light quarks , @xmath57 can take on @xmath59 values .
later , i will write @xmath57 as separate flavor ( @xmath60 ) and taste ( @xmath7 ) indices , @xmath61 . under @xmath62 heavy - quark spin symmetry , the heavy - light field transforms as @xmath63 with @xmath64 , while under the @xmath65 chiral symmetry , @xmath66 with @xmath67 defined below .
we keep the light flavor and taste indices implicit here .
the light mesons are combined in a hermitian field @xmath68 . for @xmath58 staggered flavors , @xmath69 is a @xmath70 matrix given by : @xmath71 i show the @xmath72 portion of @xmath69 explicitly , and in fact detailed final results below will assume @xmath72 .
each entry in is a @xmath73 matrix , written in terms of the 16 hermitian basis elements of the clifford taste algebra .
it is convenient to take the generators of this algebra to be @xmath74 @xcite , where @xmath75 denotes complex conjugation .
thus we write , for example , @xmath76 with @xmath77 $ ] .
it is useful to divide the indices @xmath78 into pairs of indices : @xmath79 , where @xmath80 labels the so(4 ) representation ( p , a , t , v , i ) and @xmath81 labels the element within each representation .
thus @xmath81 runs from 1 to @xmath82 , where @xmath82 is the dimension of each representation ( 1,4,6,4,1 , respectively ) .
we then can write @xmath83 the component fields of the flavor - neutral elements of @xmath69 ( namely @xmath84 , @xmath85 , ) are real ; the other ( flavor - charged ) fields ( @xmath86 , @xmath87 , ) are complex .
the mass matrix is the @xmath88 matrix @xmath89 where the portion shown is again for the @xmath72 case . from @xmath69 one
constructs the unitary chiral field @xmath90 $ ] , with @xmath91 the tree - level pion decay constant . in our normalization , @xmath92 .
terms involving the heavy - lights are conveniently written using @xmath93 $ ] .
these fields transform trivially under the @xmath62 spin symmetry , while under @xmath65 we have @xmath94 with global transformations @xmath95 and @xmath96 . the transformation @xmath97 , defined by , is a function of @xmath69 and therefore of the coordinates .
it is convenient to define objects involving the @xmath98 field that transform only with @xmath97 and @xmath99 .
the two possibilities with a single derivative are @xmath100 \ , \\
\mathbb{a}_{\mu } & = & \frac{i}{2 } \left [ \sigma^{\dagger } \partial_\mu \sigma - \sigma \partial_\mu \sigma^{\dagger } \right ]
\ .\end{aligned}\ ] ] @xmath101 transforms like a vector field under the @xmath65 chiral symmetry and , when combined with the derivative , can form a covariant derivative acting on the heavy - light field or its conjugate : @xmath102 with implicit sums over repeated indices . the covariant derivatives and @xmath103 transform under the chiral symmetry as @xmath104 we can write the leading order ( lo ) chiral lagrangian as @xmath105 where @xmath106 is the standard staggered chiral perturbation theory ( ) lagrangian for the light - light mesons , and @xmath107 is the contribution of the heavy - lights . in minkowski space , we have @xmath108 \nonumber \\ * & & { } + c_4\frac{1}{2 } \sum_{\nu } [ { \ensuremath{\operatorname{tr}}}(i\xi_\nu\xi_5\sigma i\xi_\nu\xi_5\sigma ) + h.c . ]
+ c_6\ \sum_{\mu<\nu } { \ensuremath{\operatorname{tr}}}(\xi_{\mu\nu}\sigma \xi_{\mu\nu}\sigma^{\dagger } ) \nonumber \\ * & & { } + c_{2v}\frac{1}{4 } \sum_{\nu } [ { \ensuremath{\operatorname{tr}}}(\xi_{\nu}\sigma ) { \ensuremath{\operatorname{tr}}}(\xi_{\nu}\sigma ) + h.c . ] + c_{2a}\frac{1}{4 } \sum_{\nu } [ { \ensuremath{\operatorname{tr}}}(i\xi_\nu\xi_5 \sigma ) { \ensuremath{\operatorname{tr}}}(i\xi_\nu\xi_5\sigma ) + h.c . ] \nonumber \\ * & & { } + c_{5v}\frac{1}{2 } \sum_{\nu } { \ensuremath{\operatorname{tr}}}(\xi_{\nu}\sigma ) { \ensuremath{\operatorname{tr}}}(\xi_{\nu}\sigma^{\dagger } ) + c_{5a}\frac{1}{2}\sum_{\nu}{\ensuremath{\operatorname{tr}}}(i\xi_\nu\xi_5\sigma ) { \ensuremath{\operatorname{tr}}}(i\xi_\nu\xi_5\sigma^{\dagger } ) \label{eq : vsigma } \ , \\
\cl_{\rm hl } & = & -i { \ensuremath{\operatorname{tr}}}(\overline{h}^{(b ) } h^{(b)}\ ; v{\negmedspace\cdot\negmedspace}\leftvec d ) + g_\pi { \ensuremath{\operatorname{tr}}}(\overline{h}^{(b)}h^{(b)}\gamma^{\mu}_{(m)}\gamma_5^{(m ) } \mathbb{a}_{\mu } ) \ .
\label{eq : l - hl}\end{aligned}\ ] ] here @xmath109 denotes a trace over flavor - taste indices and , where relevant , dirac indices .
the product @xmath110 is treated as a matrix in flavor - taste space : @xmath111 .
the covariant derivative @xmath112 acts only on the field immediately preceding it . for convenience ,
i work with diagonal fields ( @xmath113 , @xmath37 , ) and leave the anomaly ( @xmath114 ) term explicit in .
we can take @xmath115 and go to the physical basis ( @xmath116 , @xmath117 , ) at the end of the calculation @xcite .
at tree level , the light - light meson composed of quarks of flavor @xmath60 and @xmath118 , and with @xmath119 taste representation @xmath80 , is @xmath120 here @xmath121 is the taste splitting , which can be expressed in terms of @xmath122 , @xmath123 , @xmath124 and @xmath125 in @xcite .
the residual @xmath119 taste symmetry @xcite at this order implies that the mesons within a given taste representation are degenerate in mass .
i now list some key expressions from the feynman rules given in ref .
@xcite , but adapted to the current notation . using separate indices for flavor ( @xmath126 ) and taste ( @xmath127 ) , the ( quark - line ) connected pion propagator in minkowski space is @xmath128 \ .\ ] ] here , i have used the fact that @xmath129 is hermitian is to replace the complex conjugation in by interchange of indices on the right - hand side , which will be convenient later .
similarly , the disconnected ( hairpin ) propagator is @xmath130 \ , \ ] ] where @xmath131 with the hairpin strength @xmath132 given by @xmath133 @xmath134 and @xmath135 denote valence mesons made from @xmath136 or @xmath137 quarks , respectively , with @xmath138 and @xmath139 their masses . for the sea mesons ,
the masses @xmath140 and @xmath141 do not include the mixing effects of the hairpins . the re - diagonalized states after including the hairpins are @xmath117 and @xmath142 .
for concreteness i have assumed the 2 + 1 case : @xmath143 .
the propagators for the heavy - light mesons are @xmath144 where @xmath145 is the @xmath10-@xmath0 mass splitting .
the @xmath146 vertex ( including the @xmath147 from @xmath148 ) is : @xmath149 where the repeated indices @xmath150 and @xmath151 are summed . for @xmath0 mixing
, we also need corresponding fields that destroy and create mesons with @xmath152 quarks , @xmath153-like and @xmath154-like mesons .
these fields and their interactions can be obtained from the previous ones using charge conjugation @xcite .
the individual meson fields are indicated by , for example , @xmath155 and @xmath156 .
( note that the light quark label @xmath57 does not distinguish between quarks and antiquarks . )
the combined fields are @xmath157\frac{1 - \vslash}{2}\ , \eqn{hdef - bbar } \\
\overline{h}^{(\bar b)}_q & = & \frac{1 - \vslash}{2 } \left [ \gamma^\mu_{(m ) } p^{\ast ( \bar b)\dagger}_{\mu , q } +
i \gamma_5^{(m ) } p^{(\bar b)\dagger}_q\right ] \ .\eqn{hbardef - bbar}\end{aligned}\ ] ]
the propagators for the @xmath155 and @xmath156 fields are the same as those for the @xmath158 and @xmath40 fields , .
the @xmath159 vertex is @xmath160
the naive light quark action may be rewritten as four copies of the staggered action : @xmath161 where @xmath162 is the naive quark field and @xmath163 is a `` copied '' staggered field , with each dirac component @xmath164 separately having the staggered action .
i call `` copy symmetry '' the @xmath14 that acts on the copy index @xmath147 .
unlike the @xmath14 vector taste symmetry , which acts on individual staggered fields ( written in the spin - taste basis ) and is exact for an interacting theory only in the continuum limit , copy symmetry is an exact lattice symmetry .
thus the propagator of a copied staggered field is @xmath165 where @xmath166 is the normal ( uncopied ) staggered field .
this implies @xmath167 in the simulations using staggered quarks , the naive field is never constructed _ per se _ ; instead is used to translate staggered propagators into naive propagators @xcite .
an interpolating field @xmath168 for a @xmath169 meson is @xmath170 i assume that in practical applications @xmath168 will always be summed over a time - slice , either explicitly , or implicitly by using translation invariance . to leading order in @xmath7
, @xmath171 varies smoothly ( up to gauge transformation ) between neighboring spatial sites , but @xmath172 does not , due to taste doubling . on the other hand , in the spin - taste basis , which we arrive at by summing the staggered fields over hypercubes , the staggered fields are smooth on the doubled lattice .
we are thus led to focus on the average of @xmath168 over a spatial cube .
let @xmath173 with @xmath174 even , and let @xmath175 be a 4-vector with all components 0 or 1 . for @xmath176 even ( @xmath177 ) the averaged field is @xmath178\ .
\eqn{h - av0 } \end{aligned}\ ] ] inserted gauge links for gauge invariance of point - split quantities are implicit . for @xmath176 odd ( @xmath179 ) ,
the result is the same except the term on the last line of changes sign .
using the fact that @xmath180 , it is not hard to see that this second term just gives the usual staggered oscillating ( in time ) state with opposite parity .
i have dropped higher order terms in @xmath7 coming from the variation of the heavy - quark field over the cube . for simplicity , we simply assume from now on that all components of @xmath60 are even ( @xmath181 ) , and that the oscillating state has been removed by the fitting procedure .
we then have @xmath182 we now convert to a spin - taste basis for the staggered fields .
the standard construction for a single staggered field is @xcite @xmath183 where @xmath184 is a spin index and @xmath7 is a taste index . as is well known @xcite , this decomposition is correct only to lowest order in @xmath7 and generates a spurious @xmath185 term in the spin - taste action , but it is good enough for our purposes .
here we need a copied version : @xmath186 with spin indices implicit , then becomes @xmath187 where repeated indices are summed . with
, this implies that the contraction of @xmath188 with @xmath189 ( _ i.e. _ , the heavy - light propagator ) is automatically averaged over tastes : @xmath190 where a sum over @xmath7 is implicit .
analysis of the four - quark operator is more complicated because the two bilinears from which it is constructed are not separately summed over space ; only the four - quark operator is summed .
however , we can write it in terms of separately summed bilinears by using the identities @xmath191 here @xmath192 is any of the 16 independent hermitian matrices @xmath193 in , which obey @xmath194 .
we get , for operator @xmath195 : @xmath196 here we have dropped the `` wrong parity '' part , which does not contribute if oscillating terms are removed by the fitting procedure .
note that contributions with @xmath197 have incorrect spin ( @xmath198 instead of @xmath199 ) , and coupling of taste ( @xmath200 ) and copy ( @xmath201 ) indices .
there are @xmath202 and higher corrections to , coming from the variations of the heavy quark field over the spatial cube and from corrections to the spin - taste construction in position space .
as discussed in , however , such terms do not contribute to non - analytic terms in chiral perturbation theory until nnlo .
using , copy symmetry ( ) implies @xmath203 where @xmath204 is the quark propagator ( in a given background ) for taste @xmath7 into taste @xmath205 .
if taste symmetry is exact , @xmath206 , and only the correct spin ( @xmath207 ) contributes .
thus the desired matrix element will be obtained in the continuum limit . at one loop , however , taste violations allow @xmath208 not to be proportional to @xmath209 , and terms with incorrect spins can contribute .
for example , the taste - violating hairpin with vector taste can give a term proportional to @xmath210 .
then , since @xmath211 , the spin of the operator is @xmath212 instead of @xmath213 may be simplified by taking advantage of the exact @xmath119 taste symmetry of the staggered chiral theory at one loop . within any @xmath119 multiplet @xmath214 with dimension @xmath215 ( _ e.g. _ , the vector - taste multiplet v with @xmath216 ) , the value of all one - loop diagrams would be unchanged if we replaced any multiplet element in the taste factors @xmath217 with another element from the same multiplet . and
@xmath218 are copy , not taste , indices at this point , they will become taste indices shortly .
] we therefore write @xmath219 , where @xmath220 labels the element within multiplet @xmath221 . replacing the sum over @xmath192 in with a double sum over @xmath222 , and using the @xmath119 symmetry
, we then have @xmath223 within a given multiplet , this decouples the sum over spins from the sum over tastes .
thus , for a given continuum operator @xmath224 , the plan is to calculate the one - loop diagrams for each of the operators @xmath225 between external ( interpolating ) fields @xmath226 and @xmath227 , where , from , @xmath228 each diagram for operator @xmath229 then gets an additional factor @xmath230 coming from , where @xmath231 whether explicitly indicated or not , all repeated indices will then be summed ; this includes taste and copy indices ( @xmath232 and @xmath233 ) as well as the indices with dual , spin - taste meaning ( @xmath234 ) .
we now set up one - loop for the operators described above .
we follow ref .
@xcite as much as possible , but must take into account the complications of copy and taste indices .
it is convenient first to express the operators @xmath229 , given in in terms of the basis of . from the relations among operators listed for example in we find @xmath236 @xmath237 @xmath238 @xmath239 @xmath240 the chiral representatives of the standard operators on the right hand side of
are given in .
there , the only relevant quantum number of the light quarks is their flavor , and both bilinears have the same flavor , which is labeled @xmath57 .
here we also need to label the taste and , for the moment , the copy index of the light quarks , and these are not in general the same for both light quarks in the operator .
so we adopt the notation @xmath241 , where @xmath60 labels the quark flavor only , @xmath205 ( or other letters near the beginning of the alphabet ) labels the quark taste , and @xmath242 ( or other letters near the middle of the alphabet ) labels the quark copy . from @xcite
, we then have @xmath243 \ , , \nonumber\\ o^{xck;xd\ell}_{2(3 ) } & = & \beta_{2(3 ) } \big ( \sigma p^{(b)\dagger } \big)_{x , c , k } \big ( \sigma p^{(\bar{b } ) } \big)_{x , d,\ell } + \beta_{2(3)}^\prime \big ( \sigma p^{\ast ( b)\dagger}_{\mu } \big)_{x , c , k } \big ( \sigma p^{\ast ( \bar{b}),\mu } \big)_{x , d,\ell}\ , , \eqn{chiral - ops1 } \\ o^{xck;xd\ell}_{4(5 ) } & = & \frac{\beta_{4(5)}}{2 } \left [ \big ( \sigma p^{(b)\dagger } \big)_{x , c , k } \big ( \sigma^{\dagger } p^{(\bar{b } ) } \big)_{x , d,\ell } + \big ( \sigma^{\dagger } p^{(b)\dagger } \big)_{x , c , k } \big ( \sigma p^{(\bar{b } ) } \big)_{x , d,\ell } \right ] \nonumber\\ & & + \frac{\beta_{4(5)}^\prime}{2 } \left [ \big ( \sigma p^{\ast ( b)\dagger}_{\mu } \big)_{x , c , k } \big ( \sigma^{\dagger } p^{\ast ( \bar{b}),\mu } \big)_{x , d,\ell } + \big ( \sigma^{\dagger } p^{\ast ( b)\dagger}_{\mu } \big)_{x , c , k } \big ( \sigma p^{\ast ( \bar{b}),\mu } \big)_{x , d,\ell}\right ] \ , . \nonumber\end{aligned}\ ] ] the method used to obtain these operators in ref .
@xcite is a standard spurion analysis .
the factors of @xmath98 and @xmath244 are present in order make the light - quark spurions , which transform by left or right chiral rotations in , into objects that transform with @xmath97 ( defined in ) and can combine with the heavy - meson fields @xmath245 and @xmath246 to make invariants . although many insertions of dirac matrices are possible in forming the invariants , they all reduce down to the simple forms in when expressed in terms of @xmath247 , @xmath248 , @xmath249 , and @xmath250 . as pointed out in ref .
@xcite , this follows from heavy - quark spin symmetry , which relates the amplitude for @xmath0@xmath251 mixing to that of @xmath252@xmath253 . in ,
i have used the fact that we are only interested in the parity - even part of these operators to set the two coefficients detmold and lin call @xmath254 and @xmath255 simply to @xmath256 .
so where they have @xmath257 we will have simply @xmath254 , and similarly for @xmath258 and @xmath259 .
we use the latin @xmath260 for the chiral operators to distinguish them from quark - level operators @xmath261 . the external interpolating @xmath251 and @xmath0 fields
are taken to be , respectively , @xmath262 strictly speaking , there should be a second set of terms on the right hand sides in in which the pairs of indices @xmath263 and @xmath264 are interchanged .
these come about because the light quark field in either of the bilinears of the operator can be the one that creates the light antiquark in the @xmath251 or annihilates the light quark in the @xmath0 .
however , since is symmetric under this interchange , the extra terms give identical results to the ones we have already , and therefore can be dropped at this point .
note that the @xmath265 are low energy constants with arbitrary normalization .
we can now use copy symmetry to simplify the equations , and ultimately eliminate the copy indices entirely .
however , this can not be done without taking into account the quark flow through the diagrams : copy symmetry works at the level of the light quark propagators , and not on meson propagators _ per se_. the point is that a given light quark in an external meson field can end up in either a @xmath266- or @xmath267-labeled meson field in , and the copy symmetry has a different effect on the diagram in the two cases . if the light quark in the external @xmath268 field contracts with the quark in a @xmath266-labeled meson field ( which has taste and copy indices @xmath263 ) , then the combination of copy symmetry ( which gives @xmath269 ) and the @xmath270 in forces @xmath271 .
on the other hand , if the same light quark contracts with the quark in a @xmath267-labeled meson field ( indices @xmath264 ) , then we have @xmath272 .
so @xmath230 will end up with a different taste structure in the two cases . for convenience ,
i prefer to rename the taste indices ( @xmath273 ) in the second case so that @xmath230 remains the same . in doing so
i adopt the convention that the taste-@xmath7 quark in the external @xmath266-labeled field always contracts with the taste-@xmath205 quark in the operator , but that taste-@xmath205 quark may be in either the @xmath266- or the @xmath267-labeled meson field of the operator .
similarly , the taste-@xmath274 quark in the external @xmath267 field always contracts with the taste-@xmath275 quark in the operator . keeping the above points in mind
, we now eliminate the copy indices completely from the calculation . with the exception of the wave - function renormalization diagrams , which of course do not involve the four - quark operators at all
, the final procedure is as follows : * for a particular operator @xmath224 of interest , we first write the related operators @xmath229 as linear combinations of the standard operators , following in
. it will be convenient then to define @xmath276 and @xmath277 as the @xmath278 and @xmath279 corresponding to operator @xmath229 . for @xmath276 follows immediately from .
+ & + ' '' '' + @xmath280 & @xmath281 & @xmath282 & @xmath283 & @xmath284 + 1 & @xmath285 & @xmath286 & @xmath287 & @xmath288 & @xmath285 + 2 & @xmath289 & @xmath290 & @xmath291 & @xmath285 & @xmath289 + 3 & @xmath292 & @xmath290 & @xmath293 & @xmath285 & @xmath292 + 4 & @xmath294 & @xmath295 & @xmath296 & @xmath295 & @xmath297 + 5 & @xmath298 & @xmath299 & @xmath296 & @xmath299 & @xmath300 + * we then calculate the chiral diagrams , using copy - free versions of for the chiral operators , namely @xmath301 \hspace{3mm}[{\bf or\ } c\leftrightarrow d ] , \nonumber\\ o^{xc;xd}_{2(3 ) } & = & \beta_{2(3 ) } \big ( \sigma p^{(b)\dagger } \big)_{x , c } \big ( \sigma p^{(\bar{b } ) } \big)_{x , d } + \beta_{2(3)}^\prime \big ( \sigma p^{\ast ( b)\dagger}_{\mu } \big)_{x , c } \big ( \sigma p^{\ast ( \bar{b}),\mu } \big)_{x , d } \hspace{3mm}[{\bf or\ } c\leftrightarrow d ] , \nonumber \\
o^{xc;xd}_{4(5 ) } & = & \frac{\beta_{4(5)}}{2 } \left [ \big ( \sigma p^{(b)\dagger } \big)_{x , c } \big ( \sigma^{\dagger } p^{(\bar{b } ) } \big)_{x , d } + \big ( \sigma^{\dagger } p^{(b)\dagger } \big)_{x , c } \big ( \sigma p^{(\bar{b } ) } \big)_{x , d } \right ] \eqn{chiral - ops2}\\ & & \hspace{-7 mm } + \frac{\beta_{4(5)}^\prime}{2 } \left [ \big ( \sigma p^{\ast ( b)\dagger}_{\mu } \big)_{x , c } \big ( \sigma^{\dagger } p^{\ast ( \bar{b}),\mu } \big)_{x , d } + \big ( \sigma^{\dagger } p^{\ast ( b)\dagger}_{\mu } \big)_{x , c } \big ( \sigma p^{\ast ( \bar{b}),\mu } \big)_{x , d}\right ] \hspace{3mm}[{\bf or\ } c\leftrightarrow d ] . \nonumber\end{aligned}\ ] ] the external interpolating fields are @xmath302 for a diagram with a given quark flow , one should use either the explicit version of the operators in or the alternative @xmath273 forms to ensure that the external taste-@xmath7 light quark contracts with the taste-@xmath205 light quark in the operators ( which also guarantees that tastes @xmath274 and @xmath275 contract ) . *
each diagram is then multiplied by the overall factor @xmath303 and the repeated taste ( @xmath232 ) and spin ( @xmath221 ) indices are summed .
we are now ready to compute the one - loop diagrams in using the feynman rules given above in and . because of the complications due to taste and copy indices , it is not possible in general simply to modify the continuum results of ref .
@xcite to insert staggered corrections , as in ref .
we must calculate most chiral diagrams from scratch .
the exception is the wave - function renormalization ( parameterized below by the function @xmath304 ) , which is simple enough that the modification process ( `` staggering '' ) works . in the wave - function case ,
the naive - to - staggered translation gives no complications because , from , it only requires setting initial and final tastes equal and averaging over them .
the taste - averaging has no effect because discrete taste symmetry ( shift symmetry ) implies that a two - point function is in any case proportional to the identity in taste space .
, as can be seen most easily by using the formulation of shift symmetry in ref .
the construction is built into staggered chiral theory , so the statement is also true to all orders in .
the fact that we have used the ( approximate ) position space taste construction @xcite in the translation from naive to staggered operators is irrelevant , since for our purposes only the lowest order translation is needed . ] in addition to wave - function renormalization , there are two types of one - loop diagrams : tadpole graphs , , and sunset diagrams , .
these are parameterized by functions @xmath305 and @xmath306 respectively .
contributions from incorrect spins can enter in the tadpoles and sunset diagrams ; we call such contributions @xmath307 , and @xmath308 .
the complete matrix elements are given by @xmath309 and @xmath310 for @xmath311 .
nonrelativistic normalization , which is standard in heavy - light chiral perturbation theory , is assumed for the states @xmath312 and @xmath313 in these expressions . with relativistic normalization , an extra factor of @xmath314 , the mass of the @xmath315 meson ,
would appear on the right - hand sides .
[ cols="^,^ " , ] in terms of @xmath316 , the correct - spin and wrong - spin contributions of are then @xmath317 where the upper sign is for @xmath318 and the lower sign is for @xmath319 .
the difference in chiral structure of between @xmath320 and @xmath321 in matters for ( b ) , because the two pion fields come from different @xmath98 or @xmath244 factors .
the correct - spin contribution @xmath322 comes from the @xmath323 term in the sum , while the @xmath324 terms give the wrong - spin piece @xmath325 .
the quantities @xmath276 are listed in .
the final tadpole diagram is the disconnected contribution to ( b ) , shown in . in this case , the taste structure of the disconnected propagator is @xmath326 , which , when combined with the taste matrices in the overall factor , , gives @xmath327 .
the hairpin propagator is non - zero only in the @xmath284 , @xmath283 , and @xmath281 channels .
so we get a singlet contribution to the correct - spin operator ( @xmath328 ) and taste - violating contributions for two of the wrong - spin operators ( @xmath329 ) .
the result is @xmath330}_{x_i}\big(\{m^{(2)}_{x_i}\ } ; \{\mu_i\}\big)\ ; \frac{\partial{\cal i}_{x , i}}{\partial m^2_{x_i } } -\hspace{-3mm}\sum_{j \in \{m^{(2)}_i\}}d^{[2,2]}_{j , x_i}\big(\{m^{(2)}_{x_i}\};\{\mu_i\}\big){\cal i}_{j , i } \bigg\}\eqn{t - discb}\ , \\
\tilde\ct_{x,{\rm \figrefeq{tad - quark - discb}}}^{(n ) } & = & \hspace{-1mm}\mp\frac{i}{4\beta_nf^2}\bigg\ { \beta_n^{(v)}\ , a^2\delta'_{v}\bigg[r^{[3,2]}_{x_v}\big(\{m^{(3)}_{x_v}\ } ; \{\mu_v\}\big)\ ; \frac{\partial{\cal i}_{x , v}}{\partial m^2_{x_v } } \nonumber \\ & & \hspace{40mm}-\hspace{-3mm}\sum_{j \in \{m^{(3)}_v\}}d^{[3,2]}_{j , x_v}\big(\{m^{(3)}_{x_v}\};\{\mu_v\}\big){\cal i}_{j , v } \bigg ] \eqn{ttilde - discb}\\ & & \hspace{3 mm } + \beta_n^{(a)}\ , a^2\delta'_{a}\bigg[r^{[3,2]}_{x_a}\big(\{m^{(3)}_{x_a}\ } ; \{\mu_a\}\big)\ ; \frac{\partial{\cal i}_{x , a}}{\partial m^2_{x_a } } -\hspace{-3mm}\sum_{j \in \{m^{(3)}_a\}}d^{[3,2]}_{j , x_a}\big(\{m^{(3)}_{x_a}\};\{\mu_a\}\big){\cal i}_{j , a } \bigg ] \bigg\}\ .
\nonumber\end{aligned}\ ] ] again the upper sign is for @xmath318 and the lower sign is for @xmath319 ; the reasoning is the same as in . and
[ fig : tad - quark - discb],scaledwidth=65.0% ] the sunset diagram , , is very similar to the tadpole contribution that connects the incoming @xmath0 and outgoing @xmath251 , ( b ) .
again , there are two quark flows : , a connected graph similar to , and , a disconnected graph similar to
. the taste structures of the sunset graphs are identical to the corresponding tadpole graphs . as in
, the @xmath273 version of is used in .
the main difference between the sunset and tadpole graphs is the actual integral , which here involves two heavy - light propagators and factors of @xmath332 , and so gives @xmath333 instead of @xmath334 .
in addition , the @xmath98 and @xmath244 matrices in are all set to 1 in the sunset case , so there is no difference in overall sign between the results for operators @xmath335 and those for operators @xmath336 . otherwise , everything is the same as for the tadpole case .
i find , for , @xmath337 the disconnected sunset graph , , gives @xmath338}_{x_i}\big(\{m^{(2)}_{x_i}\ } ; \{\mu_i\}\big)\ ; \frac{\partial\ch^{\delta^*}_{x , i}}{\partial m^2_{x_i } } -\hspace{-3mm}\sum_{j \in \{m^{(2)}_i\}}d^{[2,2]}_{j , x_i}\big(\{m^{(2)}_{x_i}\};\{\mu_i\}\big)\ch^{\delta^*}_{j , i } \bigg\}\eqn{q - disc } \ , \\
\tilde\cq_{x,{\rm \figrefeq{sunset - quark - disc}}}^{(n ) } & = & \hspace{-1mm}-\frac{ig^2_{b^*b\pi}}{4\beta'_n f^2}\bigg\ { \beta^{\prime(v)}_n\ , a^2\delta'_{v}\bigg[r^{[3,2]}_{x_v}\big(\{m^{(3)}_{x_v}\ } ; \{\mu_v\}\big)\ ; \frac{\partial\ch^{\delta^*}_{x , v}}{\partial m^2_{x_v } } \nonumber \\ & & \hspace{40mm}-\hspace{-3mm}\sum_{j \in \{m^{(3)}_v\}}d^{[3,2]}_{j , x_v}\big(\{m^{(3)}_{x_v}\};\{\mu_v\}\big)\ch^{\delta^*}_{j , v } \bigg ] \eqn{qtilde - disc}\\ & & \hspace{3 mm } + \beta^{\prime(a)}_n\ , a^2\delta'_{a}\bigg[r^{[3,2]}_{x_a}\big(\{m^{(3)}_{x_a}\ } ; \{\mu_a\}\big)\ ; \frac{\partial\ch^{\delta^*}_{x , a}}{\partial m^2_{x_a}}\hspace{2 mm } -\hspace{-3mm}\sum_{j \in \{m^{(3)}_a\}}\hspace{-2 mm } d^{[3,2]}_{j , x_a}\big(\{m^{(3)}_{x_a}\};\{\mu_a\}\big)\ch^{\delta^*}_{j , a } \bigg ] \bigg\}\ . \nonumber\end{aligned}\ ] ] and @xmath331 .
[ fig : sunset - quark - conn],scaledwidth=65.0% ] and @xmath331 .
[ fig : sunset - quark - disc],scaledwidth=65.0% ] under renormalization , there are continuum - like mixings of the desired 4-quark lattice operators @xcite , but there may in addition be perturbative mixing with operators of incorrect spin and taste .
further , there are discretization corrections to our identification of the spin and taste of the operators , which can arise from variations in the heavy - quark field over a spatial cube , as well as higher order terms in the position - space spin - taste formalism @xcite
. however , incorrect spin - taste operators generated by any of these causes can not contribute to nonanalytic terms in the matrix element until nnlo . in order to see this , consider the standard power - counting in ( rooted ) staggered chiral perturbation theory ( @xmath339 ) : @xmath340 the @xmath341 mixing four - fermion operators , when translated into the chiral effective theory , are of @xmath13 in the aforementioned power - counting scheme .
thus their lo ( tree - level ) contributions to the relevant matrix elements are of @xmath13 , and their nlo ( one - loop ) contributions are of @xmath342 . perturbative mixing with wrong - taste operators can occur at one - loop order in @xmath343 . in the chiral effective theory
, these wrong - taste operators would thus enter with coefficients of @xmath344 . as was shown in ref .
@xcite , which considered the contribution of wrong - taste operators to neutral kaon mixing in , @xmath345 is numerically of the same size as the taste - breaking factor @xmath346 on the @xmath347 fm milc asqtad ensembles @xcite .
thus the appropriate way to include the strong coupling constant in the power - counting is : @xmath348 one - loop chiral diagrams involving the wrong - taste operators from perturbative mixing would therefore contribute to matrix elements only at nnlo , @xmath349 , higher order than i am considering here .
wrong - taste operators may also occur because of the @xmath202 corrections to the taste identification of the operators and the interpolating fields , , coming either from the variation of the heavy quark field over the hypercube ( see ) or from @xmath202 corrections to spin - taste identification of .
since the matrix elements are taste conserving at tree level in , any such taste - violating effects must appear twice , inducing @xmath15 corrections .
these effects can then be absorbed into @xmath350-dependent analytic terms ( see below ) .
nontrivial terms could appear at one loop in chiral perturbation theory , but the extra factor of @xmath350 implies that such terms are again effectively nnlo .
consequently , only the lo taste breaking in the 4-quark operators , together with taste - breaking terms in the lo pion chiral lagrangian , will modify the one - loop continuum chiral logarithms , and these modification are what have been calculated above .
note however , that in more highly improved versions of staggered quarks , taste violations ( the @xmath350 terms in ) will be reduced relative to the asqtad case , but perturbative mixing may not be similarly reduced . in such cases , it may be more reasonable to consider the @xmath351 ) perturbative corrections to be of lo in the chiral expansion , and their one - loop chiral corrections to be nlo , the same order as the corrections computed in .
i make some more comments on this point in .
we can now combine results from different graphs to get the complete tadpole and sunset contributions . for the correct - spin tadpole contribution , adding gives @xmath352}_{x_i}\big(\{m^{(2)}_{x_i}\ } ; \{\mu_i\}\big)\ ; \frac{\partial{\cal i}_{x , i}}{\partial m^2_{x_i } } \nonumber \\ & & -\hspace{-3mm}\sum_{j \in \{m^{(2)}_i\}}d^{[2,2]}_{j , x_i}\big(\{m^{(2)}_{x_i}\};\{\mu_i\}\big){\cal i}_{j , i } \bigg ] + a^2\delta'_{v}\bigg[r^{[3,2]}_{x_v}\big(\{m^{(3)}_{x_v}\ } ; \{\mu_v\}\big)\ ; \frac{\partial{\cal i}_{x , v}}{\partial m^2_{x_v } } \nonumber \\ & & -\hspace{-3mm}\sum_{j \in \{m^{(3)}_v\}}d^{[3,2]}_{j , x_v}\big(\{m^{(3)}_{x_v}\};\{\mu_v\}\big){\cal i}_{j , v } \bigg ] + \big(v\rightarrow a\big)\bigg\ } , \eqn{t - correct - tot123}\end{aligned}\ ] ] @xmath353}_{x_v}\big(\{m^{(3)}_{x_v}\ } ; \{\mu_v\}\big)\ ; \frac{\partial{\cal i}_{x , v}}{\partial m^2_{x_v } } \nonumber \\ & & -\hspace{-3mm}\sum_{j \in \{m^{(3)}_v\}}d^{[3,2]}_{j , x_v}\big(\{m^{(3)}_{x_v}\};\{\mu_v\}\big){\cal i}_{j , v } \bigg ] + \big(v\rightarrow a\big)\bigg\}. \eqn{t - correct - tot45}\end{aligned}\ ] ] the incorrect - spin tadpole contributions come from . adding them , and using the values of @xmath276 in and of @xmath316 in
, we have @xmath354}_{x_v}\big(\{m^{(3)}_{x_v}\ } ; \{\mu_v\}\big)\ ; \frac{\partial{\cal i}_{x , v}}{\partial m^2_{x_v}}\hspace{3 mm } -\hspace{-3mm}\sum_{j \in \{m^{(3)}_v\}}\hspace{-2 mm } d^{[3,2]}_{j , x_v}\big(\{m^{(3)}_{x_v}\};\{\mu_v\}\big){\cal i}_{j , v } \big ] \nonumber \\ & & \hspace{5mm}-a^2\delta'_{a}\big[r^{[3,2]}_{x_a}\big(\{m^{(3)}_{x_a}\ } ; \{\mu_a\}\big)\ ; \frac{\partial{\cal i}_{x , a}}{\partial m^2_{x_a } } \hspace{3 mm } -\hspace{-3mm}\sum_{j \in \{m^{(3)}_a\}}\hspace{-2mm}d^{[3,2]}_{j , x_a}\big(\{m^{(3)}_{x_a}\};\{\mu_a\}\big){\cal i}_{j , a } \big ] \bigg)\bigg\}\ , \eqn{t1-wrong - tot}\end{aligned}\ ] ] @xmath355}_{x_v}\big(\{m^{(3)}_{x_v}\ } ; \{\mu_v\}\big)\;\frac{\partial{\cal i}_{x , v}}{\partial m^2_{x_v } } \nonumber \\ & & \hspace{15mm}-\hspace{-3mm}\sum_{j \in \{m^{(3)}_v\}}\hspace{-2mm}d^{[3,2]}_{j , x_v}\big(\{m^{(3)}_{x_v}\};\{\mu_v\}\big){\cal i}_{j , v } \big ] -a^2\delta'_{a}\big[r^{[3,2]}_{x_a}\big(\{m^{(3)}_{x_a}\ } ; \{\mu_a\}\big)\ ; \frac{\partial{\cal i}_{x , a}}{\partial m^2_{x_a } } \nonumber \\ & & \hspace{15mm}-\hspace{-3mm}\sum_{j \in \{m^{(3)}_a\}}\hspace{-2 mm } d^{[3,2]}_{j , x_a}\big(\{m^{(3)}_{x_a}\};\{\mu_a\}\big){\cal i}_{j , a } \big ] \bigg)\bigg\}\ , \eqn{t2-wrong - tot}\end{aligned}\ ] ] @xmath356}_{x_v}\big(\{m^{(3)}_{x_v}\ } ; \{\mu_v\}\big)\;\frac{\partial{\cal i}_{x , v}}{\partial m^2_{x_v } } \nonumber \\ & & \hspace{15mm}-\hspace{-3mm}\sum_{j \in \{m^{(3)}_v\}}\hspace{-2mm}d^{[3,2]}_{j , x_v}\big(\{m^{(3)}_{x_v}\};\{\mu_v\}\big){\cal i}_{j , v } \big ] -a^2\delta'_{a}\big[r^{[3,2]}_{x_a}\big(\{m^{(3)}_{x_a}\ } ; \{\mu_a\}\big)\ ; \frac{\partial{\cal i}_{x , a}}{\partial m^2_{x_a } } \nonumber \\ & & \hspace{15mm}-\hspace{-3mm}\sum_{j \in \{m^{(3)}_a\}}\hspace{-2 mm } d^{[3,2]}_{j , x_a}\big(\{m^{(3)}_{x_a}\};\{\mu_a\}\big){\cal i}_{j , a } \big ] \bigg)\bigg\}\ , \eqn{t3-wrong - tot}\end{aligned}\ ] ] @xmath357}_{x_v}\big(\{m^{(3)}_{x_v}\ } ; \{\mu_v\}\big)\;\frac{\partial{\cal i}_{x , v}}{\partial m^2_{x_v } } \nonumber \\ & & \hspace{15mm}-\hspace{-3mm}\sum_{j \in \{m^{(3)}_v\}}\hspace{-2mm}d^{[3,2]}_{j , x_v}\big(\{m^{(3)}_{x_v}\};\{\mu_v\}\big){\cal i}_{j , v } \big ] -2a^2\delta'_{a}\big[r^{[3,2]}_{x_a}\big(\{m^{(3)}_{x_a}\ } ; \{\mu_a\}\big)\ ; \frac{\partial{\cal i}_{x , a}}{\partial m^2_{x_a } } \nonumber \\ & & \hspace{15mm}-\hspace{-3mm}\sum_{j \in \{m^{(3)}_a\}}\hspace{-2 mm } d^{[3,2]}_{j , x_a}\big(\{m^{(3)}_{x_a}\};\{\mu_a\}\big){\cal i}_{j , a } \big ] \bigg)\bigg\}\ , \eqn{t4-wrong - tot}\end{aligned}\ ] ] @xmath358}_{x_v}\big(\{m^{(3)}_{x_v}\ } ; \{\mu_v\}\big)\;\frac{\partial{\cal i}_{x , v}}{\partial m^2_{x_v } } \nonumber \\ & & \hspace{15mm}-\hspace{-3mm}\sum_{j \in \{m^{(3)}_v\}}\hspace{-2mm}d^{[3,2]}_{j , x_v}\big(\{m^{(3)}_{x_v}\};\{\mu_v\}\big){\cal i}_{j , v } \big ] -2a^2\delta'_{a}\big[r^{[3,2]}_{x_a}\big(\{m^{(3)}_{x_a}\ } ; \{\mu_a\}\big)\ ; \frac{\partial{\cal i}_{x , a}}{\partial m^2_{x_a } } \nonumber \\ & & \hspace{15mm}-\hspace{-3mm}\sum_{j \in \{m^{(3)}_a\}}\hspace{-2 mm } d^{[3,2]}_{j , x_a}\big(\{m^{(3)}_{x_a}\};\{\mu_a\}\big){\cal i}_{j , a } \big ] \bigg)\bigg\}\ .
\eqn{t5-wrong - tot}\end{aligned}\ ] ] in the continuum limit , when all taste propagators become degenerate and @xmath359 , the wrong - spin terms clearly vanish . for the correct - spin sunset diagrams , adding gives @xmath360}_{x_i}\big(\{m^{(2)}_{x_i}\ } ; \{\mu_i\}\big)\ ; \frac{\partial{{\cal h}^{\delta^*}_{x , i}}}{\partial m^2_{x_i } } \nonumber \\ & & -\hspace{-3mm}\sum_{j \in \{m^{(2)}_i\}}d^{[2,2]}_{j , x_i}\big(\{m^{(2)}_{x_i}\};\{\mu_i\}\big){\cal h}^{\delta^*}_{j , i } \bigg ] \bigg\}\ .
\eqn{q - correct - tot}\end{aligned}\ ] ] the incorrect - spin sunset contributions come from . again using the values of @xmath361 in and of @xmath276 in
, we have @xmath362}_{x_v}\big(\{m^{(3)}_{x_v}\ } ; \{\mu_v\}\big)\ ; \frac{\partial{\cal h}^{\delta^*}_{x , v}}{\partial m^2_{x_v } } -\hspace{-3mm}\sum_{j \in \{m^{(3)}_v\}}\hspace{-3mm}d^{[3,2]}_{j , x_v}\big(\{m^{(3)}_{x_v}\};\{\mu_v\}\big){\cal h}^{\delta^*}_{j , v } \big ] \nonumber
\\ & & \hspace{-7mm}-a^2\delta'_{a}\big[r^{[3,2]}_{x_a}\big(\{m^{(3)}_{x_a}\ } ; \{\mu_a\}\big)\ ; \frac{\partial{\cal h}^{\delta^*}_{x , a}}{\partial m^2_{x_a } } -\hspace{-3mm}\sum_{j \in \{m^{(3)}_a\}}\hspace{-4mm}d^{[3,2]}_{j , x_a}\big(\{m^{(3)}_{x_a}\};\{\mu_a\}\big){\cal h}^{\delta^ * } _ { j , a } \big ] \bigg)\bigg\}\ , \eqn{q1-wrong - tot}\end{aligned}\ ] ] @xmath363}_{x_v}\big(\{m^{(3)}_{x_v}\ } ; \{\mu_v\}\big)\;\frac{\partial{\cal h}^{\delta^*}_{x , v}}{\partial m^2_{x_v } } -\hspace{-3mm}\sum_{j \in \{m^{(3)}_v\}}\hspace{-2mm}d^{[3,2]}_{j , x_v}\big(\{m^{(3)}_{x_v}\};\{\mu_v\}\big){\cal h}^{\delta^*}_{j , v } \big]\nonumber \\ & & \hspace{-7mm}-a^2\delta'_{a}\big[r^{[3,2]}_{x_a}\big(\{m^{(3)}_{x_a}\ } ; \{\mu_a\}\big)\ ; \frac{\partial{\cal h}^{\delta^*}_{x , a}}{\partial m^2_{x_a } } -\hspace{-3mm}\sum_{j \in \{m^{(3)}_a\}}\hspace{-2 mm } d^{[3,2]}_{j , x_a}\big(\{m^{(3)}_{x_a}\};\{\mu_a\}\big){\cal h}^{\delta^*}_{j , a } \big ] \bigg)\bigg\}\ , \eqn{q2-wrong - tot}\end{aligned}\ ] ] @xmath364}_{x_v}\big(\{m^{(3)}_{x_v}\ } ; \{\mu_v\}\big)\;\frac{\partial{\cal h}^{\delta^*}_{x , v}}{\partial m^2_{x_v } } -\hspace{-3mm}\sum_{j \in \{m^{(3)}_v\}}\hspace{-2mm}d^{[3,2]}_{j , x_v}\big(\{m^{(3)}_{x_v}\};\{\mu_v\}\big){\cal h}^{\delta^*}_{j , v } \big ] \nonumber \\ & & \hspace{-7 mm } -a^2\delta'_{a}\big[r^{[3,2]}_{x_a}\big(\{m^{(3)}_{x_a}\ } ; \{\mu_a\}\big)\ ; \frac{\partial{\cal h}^{\delta^*}_{x , a}}{\partial m^2_{x_a } } -\hspace{-3mm}\sum_{j \in \{m^{(3)}_a\}}\hspace{-2 mm } d^{[3,2]}_{j , x_a}\big(\{m^{(3)}_{x_a}\};\{\mu_a\}\big){\cal h}^{\delta^*}_{j , a } \big ] \bigg)\bigg\}\ , \eqn{q3-wrong - tot}\end{aligned}\ ] ] @xmath365}_{x_v}\big(\{m^{(3)}_{x_v}\ } ; \{\mu_v\}\big)\;\frac{\partial{\cal h}^{\delta^*}_{x , v}}{\partial m^2_{x_v } } \nonumber \\ & & \hspace{5mm}-\hspace{-3mm}\sum_{j \in \{m^{(3)}_v\}}\hspace{-2mm}d^{[3,2]}_{j , x_v}\big(\{m^{(3)}_{x_v}\};\{\mu_v\}\big){\cal h}^{\delta^*}_{j , v } \big ] -2a^2\delta'_{a}\big[r^{[3,2]}_{x_a}\big(\{m^{(3)}_{x_a}\ } ; \{\mu_a\}\big)\ ; \frac{\partial{\cal h}^{\delta^*}_{x , a}}{\partial m^2_{x_a } } \nonumber \\ & & \hspace{10mm}-\hspace{-3mm}\sum_{j \in \{m^{(3)}_a\}}\hspace{-2 mm } d^{[3,2]}_{j , x_a}\big(\{m^{(3)}_{x_a}\};\{\mu_a\}\big){\cal h}^{\delta^*}_{j , a } \big ] \bigg)\bigg\}\ , \eqn{q4-wrong - tot}\end{aligned}\ ] ] @xmath366}_{x_v}\big(\{m^{(3)}_{x_v}\ } ; \{\mu_v\}\big)\;\frac{\partial{\cal h}^{\delta^*}_{x , v}}{\partial m^2_{x_v } } \nonumber \\ & & \hspace{5mm}-\hspace{-3mm}\sum_{j \in \{m^{(3)}_v\}}\hspace{-2mm}d^{[3,2]}_{j , x_v}\big(\{m^{(3)}_{x_v}\};\{\mu_v\}\big){\cal h}^{\delta^*}_{j , v } \big ] -2a^2\delta'_{a}\big[r^{[3,2]}_{x_a}\big(\{m^{(3)}_{x_a}\ } ; \{\mu_a\}\big)\ ; \frac{\partial{\cal h}^{\delta^*}_{x , a}}{\partial m^2_{x_a } } \nonumber \\ & & \hspace{10mm}-\hspace{-3mm}\sum_{j \in \{m^{(3)}_a\}}\hspace{-2 mm } d^{[3,2]}_{j , x_a}\big(\{m^{(3)}_{x_a}\};\{\mu_a\}\big){\cal h}^{\delta^*}_{j , a } \big ] \bigg)\bigg\}\ .
\eqn{q5-wrong - tot}\end{aligned}\ ] ] in comparing to the continuum results of ref .
@xcite , one needs to be aware of the many differences in notation .
in particular , the particles called @xmath134 and @xmath367 in ref
. @xcite are called @xmath117 and @xmath134 , respectively , here . taking the notational differences into account , it is straightforward to check that the continuum limits of the above equations reproduce the results in .
the analytic terms in are of the form @xmath368 where @xmath369 and @xmath370 are continuum low - energy constants , and @xmath371 are lattice low - energy constants , summarizing the effects of taste - violating analytic chiral operators at nlo .
as indicated , these constants depend on @xmath58 , on the operator whose matrix element is being calculated .
it is straightforward to see that these are in fact the analytic terms that may appear by considering the effects of adding mass or taste - violating spurions to the chiral operators in .
the parameters @xmath369 , @xmath370 , and @xmath371 , together with the parameters @xmath372 and @xmath373 in are to be determined from the fits to lattice data .
i have calculated neutral @xmath0 mixing to one - loop in staggered chiral perturbation theory for the complete set of standard model and beyond - the - standard - model operators , .
my results are given by , with expressions for the various terms listed in and .
the construction of the operators on the lattice as local products of local heavy - light bilinears , coupled with the use of staggered light quarks , results in the appearance of `` wrong spin / taste operators '' that are @xmath13 in the lattice spacing .
their contributions to the matrix elements considered here are suppressed to nlo because violation of taste symmetry is required . at that order
, they induce mixing of the operators , summarized in the quantities @xmath374 , , and @xmath375 , . however , as long as all five @xmath224 are analyzed simultaneously , there are no new low - energy constants induced by these effects : the constants @xmath376 and @xmath377 are all already present in the continuum .
effective operator mixings may come from three additional sources , weak - coupling perturbative corrections , corrections to the position - space spin - taste construction , and corrections to the taste identification of the operators and interpolating fields .
i argue that any nonanalytic terms that arise from such mixings are effectively nnlo , higher order than what has been considered here .
because the light staggered quark is converted to a naive quark in the lattice representatives of the operators , the relationship between the staggered and naive quark fields plays an important role in my analysis .
in particular , a naive quark is equivalent to four copies of staggered quarks , and the resulting `` copy symmetry '' can be used to simplify the calculations .
an interesting resulting feature is that the taste structure of the operators is not the same for all diagrams , but depends on the quark flow .
lattice computations in refs .
@xcite focused on the calculation of the quantity @xmath378 , which comes from the matrix element of operator @xmath22 .
the chiral effects of the wrong spins were not known at the time of the hpqcd calculation @xcite and were therefore omitted from the analysis and error estimate . in the fermilab / milc calculation @xcite ,
the complete expressions were available , but the matrix elements of operators other than @xmath22 were not calculated , preventing a direct inclusion of the wrong - spin effects .
however , it was possible to estimate the error of omitting these effects by using a small subset of new data to investigate the other matrix elements .
the result , @xmath379 included a 3.2% error from this effect , which was the second largest source of error . in the ongoing second - generation fermilab / milc project @xcite ,
matrix elements of all five operators @xmath224 are being computed , which means that the complete expressions can be used in the analysis , and there will be no `` wrong - spin error . '' of course , a chiral / continuum extrapolation error will remain . for future lattice computations of mixing with `` highly improved staggered quark '' ( hisq ) ensembles @xcite , taste violations are sufficiently reduced that the power counting used here , , may no longer be appropriate . depending on the range of lattice spacings
studied and the statistical errors in the data , taste violations may in fact be so small that continuum heavy meson might prove adequate for describing the data .
more likely , one will want to use the forms calculated here , but it may be necessary in addition to include the nlo chiral effects of the operators that enter through weak - coupling perturbative mixing , since such effects may no longer be much smaller than taste - violating nlo effects .
including such effects in nlo will however be straightforward , since the chiral logarithms for the complete set of operators , , have already been calculated above .
i am indebted to j. laiho , r.s . van de water , and c. bouchard for help with various aspects of the calculation , and thank a. kronfeld and all my other colleagues in the milc and fermilab / lattice collaborations for discussions .
this work has been partially supported by the department of energy , under grant number de - fg02 - 91er40628 .
i also thank the galileo galilei institute for theoretical physics for the hospitality and the infn for partial support while this work was in progress .
it is sometimes convenient to express the mixing matrix elements in terms of b ( or `` bag '' ) parameters .
a fairly common set of definitions is given for example in ref .
@xcite : where @xmath381 is the decay constant of the @xmath315 meson , @xmath314 is its mass , @xmath382 , and @xmath383 , @xmath384 , @xmath385 , and @xmath386 .
relativistic normalization of the states is assumed in these expressions .
the expression for the decay constant in , including heavy - meson hyperfine and flavor splittings , is given in eq .
( 6.20 ) of ref .
@xcite . to convert it to the current notation
, we just must replace the chiral logarithm functions @xmath387 and @xmath218 with @xmath188 and @xmath334 using . using quantities defined above in ,
we may write the result as @xmath388\;.\ ] ] it is not surprising that @xmath389 , @xmath390 , and @xmath391 appear , because the wave function and the tadpole contributions of are factorizable : they affect only one meson and one bilinear of the 4-quark operators , so they are exactly the contributions that appear in the decay constant . as always , heavy - light chiral perturbation theory is expressed in terms of the nonrelativistically normalized states of heavy quark effective theory .
thus the low energy constant , @xmath392 , that describes the decay constant at tree level in chiral perturbation theory includes a factor of @xmath393 . similarly , with the relativistic normalization used , one factor of @xmath314 needs to be included in our expressions for these matrix elements in terms of the parameters @xmath376 and @xmath377 . taking these normalization factors into account , and using , we have @xmath394 where @xmath395 in the second equation , and the upper ( plus ) sign is for @xmath396 , while the lower ( minus ) sign is for @xmath319 . here
the wave function and tadpole contributions of have canceled , and @xmath397 is a new correct - spin tadpole term that comes only from the nonfactorizable tadpole diagrams , .
combining gives @xmath398}_{x_i}\big(\{m^{(2)}_{x_i}\ } ; \{\mu_i\}\big)\ ; \frac{\partial{\cal i}_{x , i}}{\partial m^2_{x_i}}\hspace{1mm}-\hspace{-3 mm } \sum_{j
\in \{m^{(2)}_i\}}\hspace{-3mm}d^{[2,2]}_{j , x_i}\big(\{m^{(2)}_{x_i}\};\{\mu_i\}\big){\cal i}_{j , i } \bigg ] \bigg\}\;. \eqn{s}\end{aligned}\ ] ] the other chiral logarithm functions in are given in .
the analytic terms in have the same form as in ( with , of course , redefined low energy constants ) .
g. p. lepage and b. a. thacker , nucl .
suppl . * 4 * , 199 ( 1988 ) ; b. a. thacker and g. p. lepage , phys .
d * 43 * , 196 ( 1991 ) .
a. x. el - khadra , a. s. kronfeld and p. b. mackenzie , phys .
d * 55 * , 3933 ( 1997 ) [ arxiv : hep - lat/9604004 ] .
a. bazavov , c. bernard , c. m. bouchard , c. detar , m. di pierro , a. x. el - khadra , r. t. evans and e. d. freeland _ et al .
_ [ fermilab lattice and milc collaborations ] , phys .
d * 85 * , 114506 ( 2012 ) [ arxiv:1112.3051 ] .
m. wingate , j. shigemitsu , c. t. h. davies , g. p. lepage and h. d. trottier , phys .
d * 67 * , 054505 ( 2003 ) [ hep - lat/0211014 ] .
f. gliozzi , nucl .
b * 204 * , 419 ( 1982 ) ; a. duncan , r. roskies and h. vaidya , phys .
b * 114 * , 439 ( 1982 ) ; h. kluberg - stern , a. morel , o. napoly and b. petersson , nucl .
b * 220 * , 447 ( 1983 ) . c. aubin and c. bernard , phys .
d * 76 * , 014002 ( 2007 ) [ arxiv:0704.0795 [ hep - lat ] ] . w. detmold and c. j. d. lin , phys .
d * 76 * , 014501 ( 2007 ) [ hep - lat/0612028 ] .
f. gabbiani , e. gabrielli , a. masiero and l. silvestrini , nucl .
b * 477 * , 321 ( 1996 ) [ hep - ph/9604387 ] .
s. prelovsek , phys . rev .
d * 73 * , 014506 ( 2006 ) [ hep - lat/0510080 ] . c. bernard , m. golterman and y. shamir , phys .
d * 73 * , 114511 ( 2006 ) [ hep - lat/0604017 ] .
y. shamir , phys .
d * 71 * , 034509 ( 2005 ) [ hep - lat/0412014 ] . c. bernard , phys .
d * 73 * , 114503 ( 2006 ) [ hep - lat/0603011 ] .
y. shamir , phys .
d * 75 * , 054503 ( 2007 ) [ hep - lat/0607007 ] . c. bernard , m. golterman and y. shamir , phys .
d * 77 * , 074505 ( 2008 ) [ arxiv:0712.2560 [ hep - lat ] ] .
s. r. sharpe , pos * lat2006 * , 022 ( 2006 ) [ hep - lat/0610094 ] .
a. s. kronfeld , pos * lat2007 * , 016 ( 2007 ) [ arxiv:0711.0699 [ hep - lat ] ] .
m. golterman , pos * confinement8 * , 014 ( 2008 ) [ arxiv:0812.3110 [ hep - ph ] ] .
a. bazavov , d. toussaint , c. bernard , j. laiho , c. detar , l. levkova , m. b. oktay and s. gottlieb _ et al .
_ , rev .
phys . * 82 * , 1349 ( 2010 ) [ arxiv:0903.3598 [ hep - lat ] ] .
l. wolfenstein , phys .
b * 164 * , 170 ( 1985 ) .
n. h. christ [ rbc and ukqcd collaboration ] , arxiv:1012.6034 [ hep - lat ] .
w. -j .
lee and s. r. sharpe , phys .
d * 60 * , 114503 ( 1999 ) [ hep - lat/9905023 ]
. s. r. sharpe and r. s. van de water , phys .
d * 71 * , 114505 ( 2005 ) [ hep - lat/0409018 ] .
s. sharpe and n. shoresh , phys .
d * 64 * , 114510 ( 2001 ) [ hep - lat/0108003 ] .
e. follana _
et al . _ [ hpqcd / ukqcd collaboration ] , phys .
rev . d*75 * , 054502 ( 2007 ) [ hep - lat/0610092 ] ; a. bazavov _ et al . _
[ milc collaboration ] , phys .
d * 82 * , 074501 ( 2010 ) [ arxiv:1004.0342 ] and arxiv:1212.4768 . | i calculate , at one loop in staggered chiral perturbation theory , the matrix elements of the complete set of five local operators that may contribute to @xmath0 mixing both in the standard model and in beyond - the - standard - model theories .
lattice computations of these matrix elements by the fermilab lattice / milc collaborations ( and earlier by the hpqcd collaboration ) convert a light staggered quark into a naive quark , and construct the relevant 4-quark operators as local products of two local bilinears , each involving the naive light quark and the heavy quark .
this particular representation of the operators turns out to be important in the chiral calculation , and it results in the presence of `` wrong - spin '' operators , whose contributions however vanish in the continuum limit . if the matrix elements of all five operators are computed on the lattice , then no additional low energy constants are required to describe wrong - spin chiral effects . |
gillespie s stochastic simulation algorithm has recently gained popularly as a method for simulating the evolution of biochemical systems .
its advantage lies in its ability to capture the inherent stochasticity present in a chemical reaction system , and provide a full statistical description of the evolution of the system a point not addressed by traditional mass action theory , which , through a mathematical model of ordinary differential equations ( odes ) , is able to capture only ensemble averaged behaviour and assumes a continuum of reactant concentrations .
despite the fact that the deterministic mass action theory and the stochastic model of gillespie s algorithm are equivalent in the limit of large system sizes ( both assuming the well - mixed assumption ) , it is widely accepted that the stochastic model is more appropriate for biochemical applications , for the reason that biochemical systems commonly involve very low numbers of reactant molecules .
however , recent studies in the phenomenon of dynamic disorder of biomolecules reveal further stochasticity in certain biochemical systems that has yet to be accounted for by either of the two models : biochemical systems with randomly fluctuating rate constants .
dynamic disorder refers to the fluctuation of the conformational state of a biomolecule , which may be attributed to the minimization of energy landscapes @xcite .
these fluctuations are also associated with causing the fluctuation of the reaction rate constant of the biomolecule due to its changed conformational states @xcite .
the effects of fluctuations due to dynamic disorder has been actively investigated by experimentalists @xcite , whilst models to describe this phenomenon , both on the molecular level and on the macroscopic kinetics level , have also been studied @xcite .
it is interesting to note that a subdiffusion of the conformational fluctuations based on fractional brownian motion model was proposed by @xcite , in which a key assertion was the molecule s long - range memory of its conformational fluctuations resulting in fluctuations on a broad range of time scales . in all the findings
, it is widely realized that the effects of dynamic disorder on the reaction kinetics of biochemical reaction systems differ from the behaviour of systems where dynamic disorder is absent , and are not detectable by ensemble experiments or models .
one case in point is shown in @xcite , where experiments on the @xmath0-galactosidase enzyme reaction have uncovered interesting statistical properties heavy - tailed behaviour and correlation in the product formation rates which are not predicted by models with non - random rate constants .
in particular , careful statistical analysis in @xcite lead the authors to propose that the rate constants in this enzyme reaction system take on a continuum of values and has a stationary gamma distribution . in light of the findings in the single enzyme reaction ,
the purpose of this paper is to propose a modification of gillespie s algorithm to allow for random rate constants to be incorporated into the simulation .
whilst we retain the idea of the original algorithm , which is to simulate a sample trajectory for the time evolution of the system by simulating the successive occurrences of every reaction in the system , we extend the underlying stochastic model to account for both the simulation of the waiting time between reactions , as well as the simulation of the time evolution of the fluctuating rate constant . in essence , the stochastic model for the evolution of the chemical system is a joint distribution @xmath1 that models the interdependency between the waiting time @xmath2 and the changing rate constant @xmath3 .
the joint distribution is a matter of modelling , thus would of course depend on the physical properties of the reactant molecules .
however , what we propose here is a general framework that forms a starting point for how one might go about the modification of the gillespie s algorithm .
subsequently , we specialize the general framework to the single enzyme example investigated in @xcite , and illustrate a successful way to incorporate physical constraints on the slowly interconverting conformers into model for the joint distribution .
it is here that the basic idea of markov chain monte carlo is applied as a model to enforce the physical constraints , and numerical simulations from the algorithm thus derived are in good concert with the experimental results . before proceeding ,
we note how this paper differs from other works relating to the gillespie s algorithm
. one major limitation of the algorithm is its slowness , thus one broad area of research focuses on speeding up the algorithm .
examples of these methods include the @xmath2-leaping method @xcite , used to simulate more than one reaction at once , and the quasi - ssa and multiscale techniques , used for simulating stiff systems that possess reactions that occur on multiple time scales @xcite .
simulations can also fully or partially employ the use of the equivalent chemical master equation @xcite .
such speed - ups are not intended to change the fundamental stochastic properties of the chemical system , but typically to merely make approximations to the stochastic model or its chemical master equation to achieve better computational efficiency .
other works have studied more fundamental modifications of the system model , such as fluctuating rate constants of white noise type @xcite , or biochemical systems possessing rate constants that vary with time as a result of external factors such as cell growth or temperature changes @xcite .
however , to the best of our knowledge , no works have addressed the general issue of randomly fluctuating rate constants , particularly the kind exhibited in the presence of dynamic disorder where the rate constants vary over a wide range of time scales comparable to or slower than the time scale of the reactions .
finally , we remark that there seems to be no easy way to reduce the problem of dynamic disorder into the original framework of gillespie s algorithm ; even if the conformational changes were to be represented as a finite number of basic reactions in the chemical system , such a reduction is inadequate for reproducing the full dynamics of a system that actually has a continuous range of conformational states and rate constants ( see @xcite and its accompanying supplementary material ) .
before launching into the modifications of the gillespie algorithm , we briefly review the ideas in the original algorithm .
the gillespie algorithm is derived from a probabilistic model of chemical reactions at the level of molecular interactions . given a system of chemical reactions @xmath4 , indexed by @xmath5 , of the form
@xmath6 we associate with each reaction @xmath4 a constant @xmath7 with the property that + @xmath8 the constant @xmath7 takes the interpretation in the ssa as the rate constant of the chemical reaction . denote by @xmath9 the stoichiometric coefficient representing the total number of possible combinations of @xmath4 reactant molecules available .
the _ propensity _ @xmath10 describes the rate for reaction @xmath4 to occur , in the sense that @xmath11 is the probability that reaction @xmath4 will occur in the next infinitesimal time interval @xmath12 .
it is then shown that the waiting time @xmath2 for the next reaction in the system to occur is exponentially distributed with rate @xmath13 and density @xmath14 additionally , @xmath15 is the probability that the reaction that occurred is @xmath4 .
the ssa produces a sample trajectory starting from an initialized the reactant state space at time @xmath16 , followed by iterative simulation of subsequent reaction steps .
each reaction step is simulated by generating a random number @xmath2 from the density and a random index @xmath5 for the reaction @xmath4 , then updating the reactant state space and time step accordingly .
having been motivated to accommodate random rate constants into the ssa , we present a framework that allows us to extend the ssa to situations where the rate constant is a random variable .
for ease of presentation , we first consider a single reaction scheme involving only one reactant .
subsequently , extending the derivation to reactions with more than one reactant , or to a system of reactions is straightforward .
thus , consider @xmath17 the rate constant @xmath18 is a continuous - time markov process , which represents the time evolution of the fluctuating rate constant .
further properties of the process @xmath19 can be any physically relevant assumptions . in this paper
, we are primarily interested to have @xmath19 possess a stationary distribution with density given by @xmath20 .
the stationarity of @xmath19 is a natural assumption in view of analogous steady state assumptions on the conformational state of a biomolecule @xcite .
the choice for the stationary density is a modelling issue that depends upon the reaction in question , or may be suggested from experimental data .
for example , the gamma distribution was proposed in @xcite , @xmath21 with parameters @xmath22 and where @xmath23 is the gamma function .
although the quantity @xmath3 is a randomly fluctuating process and is no longer constant , we will continue to use the term ` rate constant ' to refer to values that @xmath19 takes at a given time @xmath24 .
we will also refer to the distribution of @xmath19 as the ` rate distribution ' .
the starting point for the modification of the ssa is the idea that , given the rate constant at the time of the last reaction to be @xmath25 , we want to derive a model for the joint density @xmath26 of the waiting time @xmath2 for the next reaction to occur and the rate constant @xmath27 at the time when the next reaction occurs . upon elucidating a model for @xmath26 ,
the algorithm proceeds similarly to the original ssa in a reaction - stepwise fashion , except that now a pair of random numbers @xmath28 is drawn according to @xmath26 , and both the time step and rate constant must be updated . in this way
, the algorithm sees only the rate constants @xmath29 at reaction times , and ignores any underlying properties of the underlying process @xmath19 .
, we are implicitly assuming the markovian property for both the reaction system and the process @xmath19 . the former case can be justified if we assume a well - mixed reaction system , in the sense that each reacting particle s location is uniformly distributed within the reaction volume .
the latter case is a modelling assumption that is made in this paper for simplicity and illustrative purposes . under other compelling physical motivation ,
one may be compelled to develop a non - markovian model for the @xmath29s . ] determining the joint density @xmath26 is an issue of modelling , and should based on properties of the reactants and chemical system . in general settings , using a joint distribution is a central aspect of the modified algorithm because it captures a wide range of cases where @xmath2 and @xmath27 may or may not be correlated .
one case is a scenario where the process @xmath19 is modelled to evolve as an independent variable on which the distribution of @xmath2 depends as a dependent variable ; the other end of the spectrum are situations where @xmath28 are independent , given @xmath25 .
however , a more delicate and interesting structure of interdependence between @xmath2 and @xmath27 could arise in scenarios where , on the one hand the fluctuations of @xmath27 depend on the length of waiting time allotted , while on the other hand the waiting time depends on the dynamics of the rate constant . to make the problem of modelling the joint density more tractable ,
it is convenient to factorize the joint density @xmath26 into conditional densities , @xmath30 here , we use @xmath31 as a generic notation for a density . either the first or second equality may be used for constructing the model .
suppose we are given a model for @xmath32 and @xmath33 .
the algorithm then proceeds reaction - stepwise as shown in the following table .
the algorithm works analogously if we know @xmath34 and @xmath35 instead . in order to use the modification of the gillespie s algorithm
, one has to first construct a model based on one of the two conditional density factorizations in . in this paper
, we will focus on the latter factorization .
thus , we consider the conditional density @xmath36 , a.k.a . the transition kernel of @xmath19 , to be a model for how @xmath27 fluctuates over the waiting time interval of length @xmath2 ,
given that it starts at @xmath25 .
in general , the transition kernel @xmath36 should represent any physically relevant description of the @xmath3-dynamics .
such a description should be constructed to incorporate information about the stationary distribution , as well as the physical constraints on the dynamics of @xmath3 , that for instance could arise systems where reactants exhibit slow interconversion between conformational states ( e.g. @xcite ) , or in systems with rapid fluctuation of rate constants ( e.g. @xcite ) . in section [ section : single enzyme reaction ] , we will show a case study of slowly interconverting conformers in which we construct the transition kernel with help from the metropolis - hastings algorithm . however , for this section , we will assume that @xmath36 is given to us .
suppose @xmath36 has been determined .
we use the second equality in equation to obtain @xmath37 where @xmath38 is the conditional expected value of @xmath27 given @xmath39 , and @xmath40 is the stoichiometric number associated with the number of reactant molecules . in particular , we have derived in appendix [ appdx : tau , c0 ] that @xmath41 this formula indicates that the effective propensity of the reaction is @xmath42 , in the sense that @xmath43 is the probability that the reaction occurs in the infinitesimal time interval @xmath44 .
we observe that the distribution of @xmath2 depends only on the transition dynamics of the process @xmath19 specifically , it depends only on the _ conditional mean _ of @xmath19 in the waiting time interval .
this is an important observation , because it simplifies the modelling demands to rest only on providing a model of @xmath33 .
once @xmath33 is determined , the model for @xmath32 is automatically available , thereby reducing the complexity of the model .
since the cumulative distribution function of @xmath32 is @xmath45 sampling from the density @xmath46 can be achieved by inverting the expression @xmath47 for @xmath2 , where @xmath48 is a uniform random variable on @xmath49 $ ] . in other words , defining @xmath50 , we find @xmath2 by the transformation @xmath51 however , a closed form formula for @xmath52 may not always be readily available , except for special forms of the function @xmath53 ( see e.g. , @xcite ) , and hence numerical approximation procedures will have to be applied to compute @xmath2
. this may become computationally intensive , but we will not discuss these computational issues here .
[ [ semi - markov ] ] semi - markovian approximations + + + + + + + + + + + + + + + + + + + + + + + + + + + + + kou et al .
@xcite described a so - called semi - markovian approximation in which the fluctuation of the rate constant exhibits large variability from the time of one reaction to the next .
this is characterized by rapidly fluctuating dynamics of the rate constant in a shorter time scale than the reaction waiting times .
thus , the approximation aspect of this model assumes the process @xmath19 to be characterized by infinitesimally small correlation lengths , so that the rate constant at the next reaction is independent of its value at the last reaction .
this approximation leads to setting @xmath54 .
it is no surprise that the semi - markovian approximation yields no correlation between successive reaction waiting times @xcite . [
[ time - independent - transition - kernels . ] ] time independent transition kernels . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + if @xmath33 does not depend on @xmath2 , then we have @xmath55 and @xmath56 .
then it is equivalent to use either the first or second equalities in .
however , in general cases , it is not trivial to derive a formula for @xmath57 .
[ [ reduction - to - the - original - ssa . ] ] reduction to the original ssa .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + equation reduces to gillespie s original ssa with a non - random rate constant @xmath58 by the special choice of @xmath59 , where @xmath60 represents the dirac mass at 0 .
this leads to setting @xmath61 in equation .
it is clear that the algorithm will always pick @xmath62 almost surely .
[ [ non - random - but - time - varying - rate - constants . ] ] non - random but time varying rate constants .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + if the rate constant is a non - random function of time , @xmath19 , we can still apply the framework without the stationarity assumption .
the conditioning will be on @xmath25 at the current time @xmath24 , and thus we set @xmath63 .
this case had been studied in @xcite .
we illustrate an application of the modified gillespie algorithm to the simulation of single enzyme reactions , en route proposing a method to construct a model for the transition kernel @xmath36 using the metropolis - hastings algorithm . here
, we consider a simple version of the enzyme reaction , where an enzyme @xmath64 binds reversibly with the substrate @xmath65 to form a complex @xmath66 , which then dissociates to release the product @xmath67 .
the reaction is given by the typified enzyme kinetics scheme @xmath68 where @xmath69 and @xmath70 are non - random rate constants associated with the original gillespie algorithm , for the complex formation and dissociation reactions , and @xmath3 is the random rate constant for the product formation reaction with density @xmath20 .
when @xmath20 is a non - random constant @xmath58 , classical michaelis - menten kinetics provides a deterministic relation between the rate of product formation and the substrate concentration , @xmath71}{{\rm d}t } = \frac{v_{{\rm max}}[s]}{[s]+k_m}\ ] ] where @xmath72+[es])\bar{c}$ ] is the maximum enzyme velocity and @xmath73 is the michaelis constant . here and in future , the notation @xmath74,[s]$ ] , etc .
, denotes the concentration of the reactant species @xmath75 , etc . in the case of a single enzyme system , it is more appropriate when investigating the rate of formation of the product to consider the waiting time between the formation of successive products @xmath67 , which we refer to as the turnover time @xmath2 .
the rate of product formation can be reformulated as @xmath76}{[s]+k_m}\ ] ] where @xmath77 is the ensemble average of @xmath2 ( see @xcite ) . when dynamic disorder is present , the rate constant @xmath3 fluctuates according to a distribution @xmath20 . in this case , the rate of product formation in single enzyme reactions under a quasi - static condition of dynamic disorder has been shown to take an analogous form @xcite @xmath78}{[s]+c_m}\ ] ] where @xmath79 are the harmonic mean of @xmath3 and the effective michaelis constant , respectively . in the next section , we derive an algorithm to simulate the single enzyme reaction system using the method of modifying the ssa as described in the previous section .
in addition to making a simple extension to multiple reactions , we propose a method to model the joint distribution @xmath80 that can be applied quite generally .
a major consideration in the modelling comes from the paper by english et al @xcite , in which it was proposed that the single enzyme reaction exhibits dynamic disorder with the rate constant @xmath3 distributed according to the gamme distribution .
more importantly , the dynamic disorder was suggested to be a result of conformational fluctuations of the enzyme that occur at time scales much larger than product formation time lengths .
we focus our attention for the time being on the product formation reaction step , @xmath81 .
equation ( [ eqn : dynamic disorder mm ] ) holds with the assumption that interconversion of conformers occur much slower than the reaction .
this assumption is incorporated into the modified ssa by restricting the rate constant at the next time step , @xmath27 , to change by only a small amount between each reaction .
specifically , we use the conditional density as described in [ c dynamics ] , and model @xmath36 to be a transition kernel whose support lies in an @xmath82-small interval around @xmath25 . as a first pass , we also make a simplifying assumption that @xmath36 does not depend much on @xmath2 , and consider in its place a conditional density @xmath83 whose support lies in a fixed @xmath82-small interval around @xmath25 . alternatively , removing the simplifying assumption can be done by modelling the @xmath82-small interval to grow with @xmath2 . upon deciding on the support interval for @xmath36
, we then make use of the metropolis - hastings algorithm to construct a markov chain for the successive values of @xmath27 , as follows .
define the density @xmath84 and define the _ acceptance probability _
@xmath85 for @xmath86 .
then the transition probability for @xmath27 does not depend on @xmath2 and is given by @xmath87 it is clear that the @xmath27 generated in this way will lie in the interval @xmath88 , and it is easy to check that @xmath83 is the transition kernel of a markov chain with @xmath20 as its unique stationary distribution .
that is , if we pick a random number @xmath89 from the forcing distribution @xmath90 , and accept it with probability @xmath91 , we get that @xmath89 satisfies the transition kernel @xmath36 , and moreover @xmath89 is @xmath20-distribution provided @xmath25 is @xmath20-distributed . from
, we automatically have a formula for the waiting time distribution @xmath92 where @xmath93 , the stoichiometric number of the complex @xmath66 , is 1 if the enzyme is in complex form @xmath66 and 0 if it is in free form @xmath64 , and @xmath94 thus , the effective propensity of the product formation step is @xmath95 . before moving on
, it should be remarked that the use of the metropolis - hastings algorithm is solely as a way to construct a markov chain possessing a stationary distribution .
although its original development was to simulate samples from a stationary distribution that is otherwise difficult to simulate from , this purpose plays no role in its application here .
quite the contrary , the stationary distribution may be an easily simulated distribution , as is the case in our example with the gamma distribution .
rather , any markov chain can be used as a model , so long as it satisfies the physical constraints a maximum range of fluctuations and a stationary distribution .
the use of the metropolis - hasting algorithm so happens to be a convenient choice because it provides an easy way to specify the maximum range of fluctuations via the density @xmath96 whilst maintaining the stationarity due to the acceptance probability @xmath97 .
and we will see in the simulation results that this modelling choice corroborates the experimental results .
finally , the gillespie algorithm is adapted to integrate the procedure from the previous subsection into the algorithm for the entire enzyme reaction system which involves more than one reaction . in this reaction system , let @xmath98 be the number of substrate molecules , and @xmath99 .
the propensities of the reversible enzyme - substrate complex formation reaction are @xmath100 , @xmath101 and , given the current rate constant @xmath25 , the propensity of the product - forming reaction is @xmath102 .
the update quantities to be determined for each reaction step are the waiting time @xmath2 , the new rate constant @xmath27 for the product formation reaction , as well as the reaction that occurs .
the waiting time @xmath2 to the next reaction is chosen to satisfy @xmath103 for a uniform random number @xmath48 .
note that here , due to our assumption that @xmath90 is time - independent , @xmath53 is constant is @xmath104 and thus the evaluation of @xmath2 is a direct algebraic evaluation .
next , the new rate constant @xmath27 is chosen by picking @xmath3 from the density @xmath105 , and accepting @xmath3 with probability @xmath106 .
if it is accepted , set @xmath107 , otherwise it is rejected and set @xmath108 .
finally , the reaction that occurs is chosen in a standard way of comparing the relative propensities of the reactions _ _ at the time of the next reaction__@xcite : for the three reactions , complex formation , complex dissociation , and product formation reactions , the probabilities of their occurrence are proportional to @xmath109 respectively .
this leads to the modified ssa for simulating slowly interconverting conformers in single enzyme reactions . for an arbitrary chemical system of reactions @xmath110 with random rate constants drawn from the distribution @xmath111 and effective propensities
@xmath112 , the generalization of the modified gillespie algorithm is obvious . fig .
[ algo : slow interconversion ] summarizes the algorithm for each iteration of the reaction step . to simulate a single enzyme system
, we run the simulation starting with one enzyme molecule and @xmath113 number of free substrate molecules .
we assume a buffered substrate solution , that is , @xmath98 remains constant even if a complex formation reaction occurs .
the rate constants were @xmath114 , @xmath115 , and @xmath3 follows a gamma distribution with parameters @xmath116 , @xmath117 in equation .
the fluctuation of @xmath3 is at most @xmath118 in , which we fixed for all values of @xmath98 . with these parameters ,
each product formation occurs on average once in 30 complex dissociation reactions .
while the simplifying assumption is limited , the simulation results nonetheless show several features similar to those obtained experimentally by english et al .
one key feature is the heavy - tailed property of the distribution of the turnover time @xmath2 , a phenomenon that is not observed if the rate constant is non - random . in the latter case
, the distribution of the turnover time is known to exhibit exponential decay @xcite .
figure [ fig : histogram turnover fixepsi ] shows the heavy - tailed property in a histogram of turnover times when the modified algorithm was used ( dots ) , compared against the exponential decay obtained from the model without dynamic disorder ( crosses ) with a non - random rate constant @xmath119 $ ] . for each value of @xmath98 ,
the heavy - tailed property is most apparent the rare regime of large turnover times .
for the fixed @xmath82 model in section [ section : slow interconversion ] ; the middle panel shows the autocorrelation for a model without dynamic disorder ; the bottom panel shows the autocorrelation for the model with time - dependent @xmath3-dynamics in section [ section : time dependent c - dynamics ] . ] the crux of the work by english et al .
is the discovery of correlations between successive turnover times of a single enzyme molecule , where they found that short turnover times are more likely to be followed by short turnover times , and vice versa .
the top panel of figure [ fig : autocorrelation ] shows the autocorrelation function of the turnover time series @xmath120 , computed as @xmath121 and converting @xmath122 to @xmath123 using @xmath124 . by construct
, our model ensures that such positive correlations are produced , but that the systems containing different numbers of substrate molecules exhibit the same autocorrelation behavior shifted horizontally .
this is attributable to the fact that the variance of the forcing distribution @xmath105 governs the correlation between successive @xmath3 , so that the same value of @xmath82 applied for each value of @xmath98 necessarily results in the same autocorrelation behavior .
this result differs from english s experimental result , which show an increasing degree of autocorrelation for increasing substrate concentrations .
heuristically , the experimental results can be rationalized by noting that , when @xmath98 is low , the complex - forming reaction becomes rate limiting , wider fluctuations of @xmath3 are attainable during the longer waiting time for the reaction , resulting in smaller correlation between turnover times .
this heuristic indicates that the fixed @xmath82 model is inadequate .
it is fair to say that the simplified model in section [ section : slow interconversion ] with fixed @xmath82 is a poor model for the true dynamics of @xmath3 . to improve the model
, it becomes necessary to incorporate the time dependence in the @xmath3-dynamics . as a next step toward this goal
, we consider a model in which the amount of fluctuation of the rate constant @xmath3 depends on the length of the reaction waiting time . in place of , we use an interval that increases linearly with time , @xmath125 as usual , the acceptance probability is given by @xmath126 for @xmath127 , and the transition probability is @xmath128 .
the waiting time distribution , given in expression , is @xmath129 where the effective propensity @xmath130 now does depend on @xmath2 , @xmath131 the algorithm in diagram [ algo : slow interconversion ] is then applied to the model .
one immediate difficulty with incorporating time - dependence into the gillespie algorithm is that sampling the waiting time distribution involves inverting the equation . in general
, @xmath132 may be a complicated function and an explicit expression for the waiting time @xmath2 is difficult to obtain . consequently it becomes necessary to solve for @xmath2 numerically , and this procedure becomes computationally intensive . in our simulations , assuming that @xmath2 is sufficiently small , we solve for @xmath2 by linearizing the lhs of around 0 . in this way
, we speed up the algorithm albeit at the expense of admitting some error . -dynamics model ( large blue markers ) and the model without dynamic disorder ( small pink markers ) are shown . ]
figure [ fig : histogram turnover epsitau ] shows the histogram of the turnover times for increasing numbers of substrate molecules , @xmath133 . here , we took @xmath134 .
similar to figure [ fig : histogram turnover fixepsi ] , the heavy tail behaviour is also observed .
the key difference of the time - dependent model is seen in the autocorrelation graph in figure [ fig : autocorrelation ] , where the autocorrelation increases with increasing @xmath98 .
compared to the fixed @xmath82 model , the autocorrelation behaviour is in closer agreement qualitatively with the experimental results of english et al . as suggested by equations ( [ eqn : single enzyme mm ] ) and ( [ eqn : dynamic disorder mm ] ) , the lineweaver - burke plots ( fig .
[ fig : lineweaver - burke plots ] ) obtained for the models with and without dynamic disorder reveal the linear relationship between the mean turnover time @xmath135 and the inverse substrate concentrations .
the slope and @xmath136-intercept for the linear best fits for this data are shown in the following table . also shown are the estimated values of @xmath137 and @xmath138 for the models with dynamic disorder , and true and estimated values of @xmath58 and @xmath139 for the model without dynamic disorder .
the time - dependent @xmath3-dynamics model gives a more accurate estimate of @xmath138 , though the estimate of @xmath137 is still not accurate . [ cols="^,^,^,^ " , ] the modified algorithm requires four random number draws per reaction step , as compared to two for the original gillespie algorithm .
given the added complexity of the model , this is not an excessively large computational burden .
also , the modified algorithm suffers from the same issue of computational efficiency that the original gillespie algorithm faces , due to have to evolve the system reaction by reaction .
techniques to speed up the computing time , such as the _ next reaction method _ , can be applied to improve the computational efficiency in this modification of the algorithm , and will be the subject of future work .
however , the largest computational cost arises from having to invert equation to find @xmath2 .
approximations of @xmath2 , like the linearization done in section , may be employed if the error can be properly quantified .
we have provided a framework for the modified gillespie algorithm to address the problem of stochastic simulation of biochemical systems possessing dynamic disorder .
although the modelling and implementation shown in the single enzyme reaction examples leave much room for refinement , it nonetheless is able to pinpoint some concrete modelling ingredients that are corroborated by experimental data , and is a versatile method that can be adapted to many model dynamics .
this gives us a good indication of the direction that further modelling efforts can take . with this framework ,
model calibration against real data will be possible .
in this appendix , we derive the formula for the conditional probability @xmath140 , assuming that we have available a model for the transition kernel @xmath36 for the process @xmath19 .
suppose that at time @xmath24 the rate constant is @xmath25 .
let @xmath141 be the conditional probability that _ no _ reaction occurs within the next @xmath2 time .
then the conditional probability that the next reaction occurs in the infinitesimal time interval @xmath44 is approximately @xmath142 , and @xmath143 for an arbitrary @xmath2 , recalling the transition kernel @xmath36 for the underlying process
@xmath19 , we condition on the value of @xmath144 at time @xmath145 , @xmath146 where @xmath147 . in the event that the reaction occurs in @xmath148
, we will have that @xmath149 .
y. cao , d. t. gillespie , l. petzold , _ multiscale stochastic simulation algorithm with stochastic partial equilibrium assumption for chemically reacting systems _
, j. comp .
* 206(2 ) * ( 2005 ) , 395 - 411 .
w. min , x. s. xie , b. bagchi , _ two - dimensional reaction free energy surfaces of catalytic reaction : effects of protein conformational dynamics on enzyme catalysis _ , j. phys .
b , * 112(2 ) * ( 2008 ) , 454 - 466 . l. paola .
_ a time - dependent extension of gillespie algorithm for biochemical stochastic @xmath155-calculus _ , in _ proceedings of the 2006 acm symposium on applied computing _ , new york , usa , sac ( 2006 ) , 137144 .
s. macnamara , a. m. bersani , k. burrage , r. b. sidje , _ stochastic chemical kinetics and the total quasi - steady - state assumption : application to the stochastic simulation algorithm and chemical master equation _ , j. chem .
* 129 * ( 2008 ) , 095105 . | in an experimental study of single enzyme reactions , it has been proposed that the rate constants of the enzymatic reactions fluctuate randomly , according to a given distribution . to quantify the uncertainty arising from random rate constants ,
it is necessary to investigate how one can simulate such a biochemical system . to do this
, we will take the gillespie s stochastic simulation algorithm for simulating the evolution of the state of a chemical system , and study a modification of the algorithm that incorporates the random rate constants .
in addition to simulating the waiting time of each reaction step , the modified algorithm also involves simulating the random fluctuation of the rate constant at each reaction time .
we consider the modified algorithm in a general framework , then specialize it to two contrasting physical models , one in which the fluctuations occur on a much faster time scale than the reaction step , and the other in which the fluctuations occur much more slowly .
the latter case was applied to the single enzyme reaction system , using in part the metropolis - hastings algorithm to enact the given distribution on the random rate constants .
the modified algorithm is shown to produce simulation outputs that are corroborated by the experimental results .
it is hoped that this modified algorithm can subsequently be used as a tool for the estimation or calibration of parameters in the system using experimental data . |
many innovative industries produce goods or services that have an interpreted , or `` cultural '' component to the value placed on them by consumers , in that tastes differ in different markets : just because a product is `` big in japan '' does not mean it is `` big '' elsewhere .
the same could be said at the level of national industries : as figure 1 illustrates , for example , although there may be a market for my home country s music in japan , there may not be one in italy .
we are told that the division of labor is limited by the extent of the market , but for information goods the more fundamental issue may be what the extent of the market is limited by : why do only certain other countries have substantial demand for my industry s innovative products ?
the question has substantial practical consequences .
for individual producers and distributors of information goods , selling to foreign markets can be highly profitable due to negligible marginal costs for information goods ( wildman and siwek 1988 , parker and van alstyne , 2005 ) ; for whole economies , the presence of high - value export industries can drive overall growth and development ( amsden , 2001 ; hausmann , hwang , and rodrik , 2007 ; hidalgo et al , 2007 ) . despite the importance of the question
, however , received economic theory does not provide a clear answer .
[ fig : ivj ] the economic viewpoint is that specialization promotes exchange , and that specialization must be maintained to remain competitive ( ricardo , 1891 , krugman , 1980 , 1991 ; storper and christopherson , 1987 ; storper , 1992 ; malmberg and maskell , 1997 ) . by this logic ,
anything that would reduce the distinctiveness of a country s products would diminish its ability to export them .
accordingly , there is some fear in certain quarters that homogenizing effects of consuming imports especially of american goods would lessen the distinctiveness and global profitability of many countries information industries ( henry , 2001 ; world trade organization council for trade in services , 1998 ; unesco , 2005 ; african union , 2005 ) .
is this a well - founded fear ?
if distinctiveness is an asset for national information industries , is it better for them to be relatively insulated from imports ? or is it actually better to be interconnected and influenced by the outside ?
the opportunities and threats are greatest for smaller industries , in which potential outcomes for individual producers and distributors range from entering more lucrative foreign markets to being crowded out by global superstars . but understanding the dynamic effects of cross - market trade would also pose substantial strategic questions for multi - national enterprises as well . if warner music wants to expand its sales of its belgian intellectual property in the united states , for example , how much of their american or british repertoire should they license to distribute on online music streaming services in belgium ? would too much outside influence dilute or diminish the distinctive belgianness and lessen the export value of belgian repertoire ? or , on the contrary , would such influence make the music more familiar and valuable to new markets abroad ? more generally ,
where do new markets for interpreted information goods come from ?
although mainstream economic thought does not present a clear answer , the very term cultural good suggests an alternative framing .
cultural goods are cultural in that they derive their value from subjective , but socially contingent , meanings .
the export of cultural goods from one country to another can be seen as a form of international communication , in which the goods are the messages .
the meanings and value of cultural goods ( as with any messages ) are not universal or intrinsic , but depend instead on the language or expectations of the message s recipients for interpretation .
therefore , i view this not as a problem of the functioning of a market , but rather as a macroscopic network constituted by semiotically valent exchanges . in the model
i propose , national markets are nodes and flows of information goods between them are links .
i conduct a longitudinal network analysis of the global pattern of trade in a specific information industry recorded music in the period 1976 to 2010 . using this framing ,
the following sections develop and test the hypothesis that new export markets are developed through a process of transitive closure in the network of international trade .
when two countries markets experience the same social influences , it brings them close enough together for new homophilous ties to be formed .
in other words , consumption of foreign products at home helps , not hurts , home - market producers to enter new foreign markets , with the caveat that it only helps them enter markets that have a history of consuming the same foreign products that were consumed in the home market . selling in a market changes what is valued in that market , and new market development can be seen as a consequence of having social influences in common .
in general , it is believed that people are more `` homogenous within than between groups '' ( burt 2004 ) , and that explanations for this pattern can be grouped into two broad types : homophily and social influence ( c.f .
aral , muchnik , and sundararajan , 2009 ) .
these two explanations are not mutually exclusive , but explain complementary phenomena .
homophily is the tendency for individuals to form ties with others like themselves . social influence
, on the other hand , suggests that connected individuals will become more alike .
both have salient interpretations in the present context .
a tie suggests that the products of the exporting country are intelligible enough to consumers in the importing country to support a relatively high volume of trade .
indeed , by the very act of modeling international trade as a network of meaningful exchange ties , homophily is assumed by construction . in seeking to understand how the pattern of trade _ changes over time _
, however , we are interested in the role that social influence plays in concert with homophily . importantly , in the context of music trade , the primary vehicle for influence is the music recording , rather than interpersonal contact with individuals in other countries .
music is a form of symbolic communication , so the export of music from one country to another amounts to a sort of macroscopic communication tie .
although music listeners in the importing country typically do not directly interact with music listeners in the exporting country , they are on the receiving end of communicative exchange , and may be influenced by listening to a subset of the exporting country s music .
messages transmitted from one person to another in the process of communication , whether linguistic or non - linguistic , have no intrinsic meaning but must be interpreted as corresponding to meanings by their recipients ( e.g. maines 1977 , lakoff 1987 , carlile , 2002 ) .
the sounds that make up a word in one language may be nonsense in another language , for example .
there is a requirement , therefore , for a learned interpretive layer a sort of semantic knowledge that is an aspect of language that maps messages to meanings : a tacit knowledge ( polanyi 1966 , polanyi and prosch 1977 ) or absorptive capacity ( cohen and levinthal , 1990 ) on the demand side . to illustrate , consider a hypothetical repertoire of music .
if musical phrases in this body of work conventionally end in a certain way , a songwriter could play with the listener s learned expectation that a specific phrase will end the same way .
for example , by withholding the expected harmony and prolonging the penultimate chord , one might increase the dramatic tension . a piece of music that includes this technique depends for effect in part on the listener expecting the final harmony , only to have that expectation manipulated by the withholding of that harmony . if such music were brought to listeners from a different country who did not `` speak the same musical language , ''
a long penultimate chord would not have any particular meaning because there would be no expectation in mind of what was supposed to happen .
therefore a musical innovation - in this case the displacement of the expected harmony in time - does not create an increase in quality by creating objectively better combinations of musical elements , but rather depends on the listener s expectations to achieve an effect . over time ,
new and innovative music , in turn , becomes a reference point for composition and consumption of future music . to continue with the example of the withheld harmony ,
if it were used and heard enough , it could eventually become a conventional element in a phrase ending , and its effect on listeners would not in that case include surprise or increased tension .
this is not to say that every individual would have the same reaction ; people respond to music differently . in the case of the withheld harmony , though everyone may have learned to expect the same conventional phrase endings , some subset of the population may prefer those expectations to be met .
these individuals would favor a more conventional style , while others may enjoy the innovation .
although their tastes differ , the fact that they share that reference point means that each other s preferred music is nonetheless mutually intelligible . to people in other markets who are not familiar with the conventional phrase endings
, there would be no difference in effect if the final harmony was withheld or not .
looking at music through this lens , what exposure to the same music would provide , therefore , is not a predictable influence on tastes that pertains to all listeners , but rather a common reference point , to which people may relate differently . to the degree that words are used in shared experiential contexts , they should contribute to establishing shared meanings among the people that use them to communicate .
however , to the extent that contexts and associations differ , the meanings associated with words will tend to diverge .
this suggests that shared meaning and perhaps shared value is possible only to the extent that the communicators have shared experience . in summary ,
i have argued thus far that being exposed to music in common is a precondition for sharing aspects of a `` musical language , '' which is to say the set of expectations through which new music is understood and evaluated .
the economic literature has not focused on trade in cultural goods , but does nonetheless include several important , if peripheral , treatments of the topic .
these contributions do not constitute a coherent perspective , but can be integrated by interpretation through the above framework and deserve consideration here .
linder ( 1961 ) suggested that trade propensities among countries should be understood as forming a matrix , representing the degree to which they produced products appropriate to each other s aggregate tastes . in more recent studies of trade in cultural goods
, a minority but significant perspective on the difficulty in developing new export markets revolves around empirical evidence for the idea of a `` cultural discount , '' goods from abroad are not valued as highly as domestic goods ( hoskins , mirus , and rozeboom 1989 ; huang , 2007 ; lee , 2006 ; park , 2006 ; rauch and trindade , 2009 ) .
these findings are entirely consistent with linder s hypothesis : tastes differ , and they are generally thought to be related to a variety of factors more or less exogenous to the phenomenon of trade itself .
others , especially schulze ( 1999 ) , go on to consider how tastes change , and how trade might itself affect tastes .
one factor in changing tastes is a variety of network externality , in that consumers are thought to derive more value out of listening to music , the more people they have to share the experience with ( c.f .
salganik , dodds , and watts 2006 , salganik and watts 2009 , who describe a short - run version of a similar phenomenon in terms of social influence ) .
schultz ( 1999 ) uses the term `` consumption capital '' to describe tastes formed through experience , explaining that `` eventually , the reinforcing effect of consumption capital accumulation makes it a part of the national culture .
this implies that trade in art should be a positive function of cultural proximity and that current trade is a positive function of past trade . ''
witt ( 2001 ) expresses a similar argument in evolutionary terms : what has been consumed in the past conditions what is valued and consumed in the future .
the conditioning is by means of the associations that the mind draws consciously or subconsciously between the old and the new . for higher - level and specific manifestations of wants
( that is , those that are not innate such as food and shelter ) , it is by these `` learned associations '' that tastes become specific to certain products and types of products . as tastes
are conditioned by consumption they in turn make up the selection mechanism by which products gain in prominence and ability to further influence tastes ( figure [ fig : cycle ] ) .
these minority perspectives in economic theory are clearly congruent with the reading of the sociologies of communication i advance above .
the task of the following section is to convert this perspective into a testable hypothesis .
if importing music influences the local musical language , how should we expect this to alter the observed pattern of trade ?
the most obvious implication would be toward a pattern of reciprocal trade .
if country a influences country b s musical language by exporting to it , the most obvious new market for country b s new output would be country a , because the two countries would have reference points in common . empirically , such a tendency to reciprocating trade ties would not be possible to attribute to the social influence - based mechanism i propose : reciprocal trade could be explained by homophily alone ( which , as noted above , we assume must be present ) or other dyadic latent variable without reference to social influence .
latent variables are always problematic in empirical work , but they are a particular concern here .
if we start with a pair of countries that trade with each other , we know that they are already similar enough to support one homophilous tie , and there would be no need to invoke social influence in explaining why a second , reciprocated tie would be formed .
i suggest that looking beyond bilateral relationships would lessen these empirical difficulties , and that rather than look for exports back to the original exporter , we should look for exports to other countries that were also influenced by the original exporter ( figure [ fig : tmts ] ) . to illustrate , consider a simple invented example .
imagine some country like the united states exports music to two other countries , say , thailand and nigeria .
some thai producers and some nigerian producers will be influenced by imports from the us , and some will not .
however , assuming that a large proportion of consumers in those countries have consumed american music , we can expect the local musical language to be influenced by american music .
if this process is happening in both nigeria and thailand simultaneously , we would expect their domestic musical output to be more intelligible to each other in the aggregate than before the influence of the united states .
given such an influence in common , we would therefore expect a subset of producers in nigeria to have an easier time selling their products to thai consumers ( and vice versa ) because of these shared reference points .
if consumer tastes are also influenced by imports , in other words , then there should be an increased likelihood not just of exporting after importing , but of exporting to _ certain countries _ that have common international influences from trade . in the three country case
, there are initially trade flows a @xmath0 b and a @xmath0 c , but no trade flows between b and c. there is therefore no prior evidence from the presence of trade that b and c have a high propensity to trade with each other , so a new trade flow b @xmath0 c or c @xmath0 b could be considered a more meaningful finding .
there is admittedly still a selection bias , because b and c both import from a , but since it is more indirect , it can be assumed to be lesser than the selection bias inherent in looking for reciprocal trade .
the tri - lateral approach is also advantageous in regard to effective sample size .
if , say , the united states exported to n=30 countries , there are 30 possible reciprocal trade flows one could look for , but n!/(2!(n- 2 ) ! ) = 435 cases of triadic closure we could look for , raising our power in controlling for exogenous variables . in sum , i argue that for information goods with an interpreted component , the more two countries have imported from the same international sources , the more likely one is to begin to export to the other .
i therefore test the hypothesis that the pattern of trade in music evolves through a process of transitive closure , or more precisely , hypothesis _ new markets for a country s music are more likely , the more transitive mediated triads the new trade relationship would complete . _
the hypothesis of this paper must be tested with longitudinal data .
records of exports and imports are extracted from the united nations commodity trade database ( comtrade 2012a ) .
i use data classified under standard international trade code revision 2 , commodity number 89832 ( sound recordings ) , which includes pre - recorded physical sound carrying media of any form .
cds , cassettes , phonograph records , and mini - discs are all included in this category , and blank media and computer software are not .
for these data , the value of an export flow is reported by each exporting country in united states dollars in `` free on board '' terms , which includes costs of the exported goods and the costs incurred in getting them to the country s border , but excludes shipping and insurance costs incurred in transporting them to the destination country .
data availability begins in 1976 , with 18 countries reporting exports to specific destinations .
there is a gradually increasing trend in the number of countries reporting the destination of their exports until it reaches a maximum in 2005 and 2006 at 150 countries reporting .
some pre - processing of data is required .
most notably , export data from taiwan is not available for political reasons .
however , it is possible to approximate these data by examining other countries reports of imports from `` other asia , nes , '' which is a term reserved in practice for taiwan .
some trade flows are reported with other indefinite partner designations ( for example , `` world , '' or `` areas , nes '' ) but unlike `` other asia , nes , '' these do not indicate a definite destination ( comtrade 2012b , 2012c ) .
i therefore do not include these indefinite data in the present study .
because my hypothesis is intended to test the idea that imports of music affect the music that is created within a country , we must exclude goods that merely pass through a country on their way somewhere else .
it is therefore necessary to subtract re - exports from total exports to arrive at a figure for net exports of music recordings produced domestically ( note that being produced domestically does not necessarily imply that the music was originally composed domestically ) .
other authors have tended to prefer to use imports , rather than exports as the source of data on trade flows for reasons of their greater accuracy , but this approach would not have allowed the subtraction of re - exports , which is the more significant distortion of the data for my purposes . the hypothesis is concerned with new trade flows , but we only wish to focus on economically meaningful ones .
it is therefore necessary to dichotomize the data to represent when a trade flow crosses over some significant threshold from being absent to being present. the method used to dichotomize the data could substantially affect the results of statistical models run on this data .
i therefore repeat the analyses on multiple versions of the binarized data .
the primary approach to dichotomization attempts to code a trade flow as present ( a `` 1 '' ) , if it is large relative to the market of the importing country . simply because of the difference in sizes , one million dollars worth of exports to the united states would certainly be much less influential on the local language than the same amount of music exported to say , belgium , or malawi . to establish a suitable means of comparison ,
i first took available data on music market sizes ( dollars spent by consumers ) from the international federation of the phonograph industry ( ifpi 1992 , 1998 , 1999 , 2005 ) and regressed them on population and gdp ( with all variables log - transformed ) ( table [ tab : sizeofdomesticmarket ] ) .
these variables had good predictive power ( @xmath1 of .89 ) and allow for reasonable approximation of music market sizes for the many countries that do not report data through the ifpi but do report population and gdp statistics .
i then chose a threshold , @xmath2 , such that if the value of the trade flow was greater than @xmath2% of the predicted size of the importing country s music market , the flow would be coded as `` 1 , '' and `` 0 '' otherwise . to confirm that results were not dependent on the value of @xmath3 i repeated the analyses , varying @xmath2 from 1% to 35% of the estimated size of importing music market .
these results are presented in table [ tab : effectoftieinclusionthresholdontransitivityparameterestimate ] , in the following section .
lrrrr & & + & estimate & p - value & estimate & p - value + intercept & -15.99 & @xmath4 & -15.464 & @xmath4 + ln gdp / capita & 1.23 & @xmath4 & 1.200 & @xmath4 + ln population & 0.988 & @xmath4 & 0.959 & @xmath4 + +
n & 47 & & 42 & + adj .
0.935 & & 0.893 & + [ tab : sizeofdomesticmarket ] in an additional series of analyses , rather than consider the relative size of trade flows , i used a dichotomization scheme based on absolute sizes .
for these runs , i used the following algorithm .
starting with the largest reported flow by dollar value and proceeding from largest to smallest , trade flows are coded as present ( as a 1 ) , until the sum of their dollar value equals 95% of the total global dollar value of reported trade .
the thousands of the smallest trade flows that together sum to less than five percent of total reported world trade in recorded music , are thus coded as absent ( as 0 ) . for each data year , i represent my trade data in a square matrix , representing all possible origin - destination pairs of countries as row - column entries .
each country is represented as a row ( as the exporting country ) and as a column ( the importing country ) .
these matrices taken together can be viewed as a representation of the phenomenon to be explained : the evolving network of trade in music .
starting with all available comtrade data , any country that was coded as importing or exporting in any of the data years is considered part of the study population of countries for all of years in a given set of analyses .
that is , the set of countries under consideration is constant across time , and omits all countries that do not report trade in music at a high level ( as defined by the inclusion algorithms described above ) during the study period .
one exception to this generalization is that a number of countries were divided into several smaller countries during the study period , such as yugoslavia . in this case , i treat the largest ( by population ) `` child '' country ( i.e. serbia ) as if it were the same country as the `` parent '' ( yugoslavia ) .
the smaller `` children '' ( e.g. croatia ) are left out of statistical estimations until they become independent states by coding their entire row and column of the trade matrices with `` structural zeros '' .
these encode an absent trade flow that is absent for exogenous reasons ( in this case , because the country did not exist ) , and are not considered by the statistical model ( see ripley , snijders , and preciado 2011 for more detail ) .
the other exception is germany , which unified the former east and west germanies . in this case , i treat west germany and unified germany as one country and east germany as if it could no longer trade for exogenous reasons ( with structural zeros ) . if we wish to study when one country is likely to begin to export to the other , then this is to say that we are particularly interested in the formation of trade flows the change from 0s to 1s in the set of matrices representing the changing pattern of trade . to evaluate
what conditions are associated with the changes from 0s to 1s , we must also take into account those 0s that remain 0s .
the absence of a trade flow between a given pair of countries in the study population provides information for empirical analysis , just as the presence of a trade flow does . in this study , therefore , each entry in the matrices representing trade in music corresponds to one observation in the dependent variable .
digital delivery of music recordings has , of course , been a major disruptive force in the music industry .
the ifpi reports that in 2006 , digital music consumption amounted to only 10% of total sales in the united states .
although growing rapidly , it had not yet become the overwhelming influence it is today during the study period . on the other hand ,
the effect of illegally copied and shared music on the volume of legitimate international trade up through 2006 is not known .
after dichotomizing the data , it does not appear to have had a dramatic effect , in that the number of trade ties continues to rise over time , even through the 2001 and 2006 data years ( see table [ tab : networkdescriptives ] ) . to confirm that there was no substantial effect on
the model proposed and tested here , i run an additional model excluding those final two observation years , and compare to the full data set .
in addition to the multivariate analyses , discussed below , i present results for non - parametric comparisons of the association between amount of common influence and changes to the trade network .
for each observation except the first ( before which there is no observed history ) , each potential ordered pair of countries is placed into one of four categories .
one category of ordered pairs has ties that have been maintained from the previous period ( `` maintained ties '' ) .
the second is ties that were present in the previous period but absent in the current period ( `` lost ties '' ) .
the third is `` new ties , '' and the fourth is ties that are not present in the current or previous time periods ( `` no ties '' ) . for each matrix entry in each category , i step backwards from the observation and count the number of previous observation years in which common influences were experienced by the two countries .
in addition to descriptive statistics for each category , i employ a discrete two - sample kolmogorov - smirnov ( ks ) tests to examine whether the distribution of common influences varies by category .
the two - sample ks test is a non - parametric test for differences between the cumulative distribution functions of empirical data , especially useful when the underlying distributions are not known a priori ( massey 1951 ) .
the empirical benchmark in describing international trade is the so - called `` gravity model '' ( e.g. anderson 1979 ) , and i look to this tradition as a starting point for inclusion of exogenous explanatory variables .
the gravity model shows that the level of trade between two countries is proportional to the product of their gdps ( analogous to the interaction of their economic masses ) and the inverse square of the geographic distance between them . in cultural industries
specifically , there are economic arguments that the size of the exporter s market is most important in determining the pattern of trade ( wildman and siwek 1988 , dupagne and waterman 1998 , lee and waterman 2007 ) .
i therefore represent the market sizes separately for the exporter ( aka `` ego '' ) and the importer ( `` alter '' ) . in the multivariate analyses below
, the effect of the covariates enters multiplicatively , so there is no substantial difference between this and the original gravity model .
the gravity model literature also typically includes variables for the geographic contiguity of pairs of countries , whether they have a history of a colonial relationship together , and whether they share a spoken language .
i follow this precedent and include variables for contiguity , direct colonial tie ( country a colonized country b ) , and indirect colonial tie ( countries a and b were both colonized by country c ) , and whether greater than 9% of the population of each pair of countries shares a common language .
data for commonality of language , geodesic distance , contiguity , and colonial relationship are taken from the cepii database of international trade statistics , available online at www.cepii.fr .
descriptive statistics for covariates and dichotomous trade networks are in tables [ tab : descriptivestatisticsforcovariates ] and [ tab : networkdescriptives ] , respectively . -1.5 cm lrrrrrrrr & min .
& 1st qrt . &
median & mean & 3rd qrt . &
& variance & unit + + ln(distance ) & 4.09 & 8.43 & 8.96 & 8.77 & 9.33 & 9.90 & 0.65 & kilometers + colony & 0 & 0 & 0 & 0.01 & 0 & 1 & 0.01 & indicator + contiguous & 0 & 0 & 0 & 0.02 & 0 & 1 & 0.02 & indicator + common language & 0 & 0 & 0 & 0.15 & 0 & 1 & 0.13 & indicator + common colonizer & 0 & 0 & 0 & 0.11 & 0 & 1 & 0.10 & indicator + music market 1976 & 1.25 & 6.39 & 8.06 & 8.36 & 10.37 & 15.35 & 7.33 & us$ ( 2005 ) + music market 1981 & 1.52 & 6.66 & 8.40 & 8.57 & 10.63 & 15.52 & 7.26 & us$ ( 2005 ) + music market 1986 & 1.81 & 6.70 & 8.49 & 8.71 & 10.76 & 15.72 & 7.18 & us$ ( 2005 ) + music market 1991 & 2.21 & 7.09 & 8.68 & 8.91 & 10.82 & 15.86 & 6.60 & us$ ( 2005 ) + music market 1996 & 2.67 & 7.13 & 8.83 & 9.04 & 11.05 & 16.04 & 6.71 & us$ ( 2005 ) + music market 2001 & 2.97 & 7.40 & 9.04 & 9.25 & 11.18 & 16.26 & 6.60 & us$ ( 2005 ) + music market 2006 & 3.57 & 7.67 & 9.43 & 9.52 & 11.46 & 16.41 & 6.58 & us$ ( 2005 ) + [ tab : descriptivestatisticsforcovariates ] llrrrrrrr & & 1976 & 1981 & 1986 & 1991 & 1996 & 2001 & 2006 + + & number of ties & 12 & 259 & 335 & 631 & 988 & 1125 & 1351 + & max outdegree & 3 & 63 & 71 & 105 & 128 & 142 & 136 + & # with od=1 & 5 & 11 & 11 & 8 & 18 & 24 & 12 + & # with od=0 & 182 & 158 & 150 & 136 & 111 & 97 & 100 + + & number of ties & 2 & 93 & 115 & 261 & 413 & 472 & 611 + & max outdegree & 1 & 25 & 32 & 60 & 80 & 79 & 81 + & # with od=1 & 2 & 10 & 13 & 16 & 15 & 22 & 20 + & # with od=0 & 188 & 170 & 165 & 151 & 137 & 123 & 119 + + [ tab : networkdescriptives ] i am hypothesizing an effect of the existence of certain trade flows on other trade flows .
of course , we must also explain the presence or absence of the first trade flows as well , which will be explained ( in part ) by the presence or absence of still other trade flows , and so on .
this sort of endogenous causation ( trade flows depend on trade flows ) , implies that each observed trade flow should play dual roles in the analysis . as an import ,
it is a causal factor relevant to the importing country s future exports . as an export from the originating country
, it is itself an outcome to be explained .
the observations in the dependent variable are , by hypothesis , interdependent .
interdependence of this sort is a clear violation of the independence assumptions of standard multivariate statistical models , but is the bread and butter of network models , in which it is known as a `` structural effect , '' in that the structure of the network endogenously influences its own subsequent structure .
conceptually intertwined with the main hypothesis are at least three other phenomena of interdependency .
each of these is predicted to be correlated with the transitive mediated triads of the main hypothesis , above .
these therefore act as competing hypotheses , in that they could theoretically explain observed transitive closure .
the first is the tendency toward reciprocity of already existing trade ties , as discussed above .
a correlation between reciprocity and transitivity could arise in a dense area of the network .
thus reciprocity must be controlled for .
the second phenomenon is having a common colonizer ( mentioned in `` exogenous variables , '' above ) .
if two countries were colonized by the same third country , it would not be surprising if they traded at an above average level with their colonizer and each other , thus creating triads
. these triads might be due to a similar process of influence to that in the main hypothesis ( if the colonizer s music influenced both importing countries ) , but it could also be due to completely unrelated factors .
for example , such countries could have shared legal systems that support trade , a history of bilateral shipping instituted by the colonizer , or other factors in common .
the third phenomenon is one that has been observed in international trade in differentiated industries in general : the tendency for major exporters to trade disproportionately with each other . in the `` new '' trade theory ( krugman , 1980 ) , countries that have successfully specialized in a variety or varieties of goods within an industry are likely export to many partner countries , selling their domestic products , as well as import from many partner countries , who make complementary products .
the result is the tendency for countries with many export markets to be each other s trading partners to a greater degree than would be predicted by the numbers of their trading partners alone . in the language of network analysis ,
the number of outgoing links is known as `` out - degree '' and this sorting of countries within a network by the number of outgoing links is known as out - degree assortativity ( hereafter `` assortativity '' for brevity ) . where a few countries have disproportionately many outgoing trade ties , and they disproportionately trade with each other
, this could appear statistically like the effect of a transitive process , even if transitivity played no role in the actual formation or dissolution of ties .
literature on network dynamics suggests that assortativity and transitivity are not just difficult to distinguish , but may in fact limit the values the other can take ( serrano and bogu , 2005 ; holme and zhao , 2007 ) .
the presence of a bounded range of possible transitivity confirms that even without a true tendency to transitivity , an assortativity effect can itself produce a considerable amount of incidental transitivity . over time
, both degree assortativity and transitivity can lead to a similar clustered pattern of ties in the network .
however , the two effects are constituted by different micro - processes . to rule out this alternative explanation of observed transitive closure , rather than include both transitivity and assortativity effects ( in which case , analogous to other multicollinearity problems
, it would be difficult to determine which structural effect was responsible for the observed pattern of ties ) , i run analyses with finer - grained ( yearly ) data that include an assortativity parameter but no transitivity parameter , testing if such a specification could account for the data .
the presence of endogenous structural effects in a model of network evolution mean that the status of each trade tie depends ( in multiple ways ) on the status of each other trade tie .
finding parameter estimates and evaluating their statistical significance analytically ( analogous to fitting a least - squares line algebraically , for example ) is simply not feasible for this type of data .
snijders and colleagues ( 2001 , 2005 , snijders et al .
2007 , ripley , snijders , and preciado 2011 ) therefore elaborated a markov chain monte carlo type computer simulation method of arriving at maximum likelihood parameter estimates and standard errors .
snijders longitudinal model of network evolution is known as the `` siena '' model , which stands for `` simulation investigation for empirical network analysis , '' now available as the package `` rsiena '' for the r statistical computing environment . to convey the nature of the analytical method used in this research , i present a simplified description of the siena model here ; for a more complete and detailed description , the reader is referred to the original sources ( snijders , 2001 , 2005 ; snijders et al .
2007 ; ripley , snijders , and preciado 2011 ) . in siena ,
the actual set of changes to the structure of the observed network in this case , changes in who exports to whom is considered the outcome of stochastic process , which is to be modeled . other changes to the network
could have resulted from the same underlying forces making up the stochastic process , but the modeling effort seeks parameter values that make the observed data more likely than any other combination of parameter values ( thus `` maximum likelihood '' ) . let us denote the current state of the network as @xmath5 , drawn from a distribution of such networks , @xmath6 , where @xmath7 when country @xmath8 exports to country @xmath9 at time @xmath10 and @xmath11 otherwise .
the probability that the network changes from one state to another , @xmath12 is a function of the sum of the influences from the effects in the model .
the observed data are characterized in terms of `` change statistics , '' @xmath13 , where @xmath10 indicates the period and @xmath14 denotes which statistic is being counted , for example the number of newly reciprocated trade ties or the number of newly completed transitive mediated triads .
the formation or dissolution of each trade tie can influence multiple change statistics simultaneously for example a new tie could be both a newly reciprocal tie and complete a transitive mediated triad . in the model , these statistics are multiplied by parameters @xmath15 that represent the contribution of the structure in question to the likelihood of forming a tie . at a high level ,
the model states that the tendency for the network @xmath6 to take on a certain form , @xmath5 can be described in terms of an objective function evaluated from the point of view of nodes , @xmath8 , and depends to various extents , @xmath15 , on the change in the number of certain structures , @xmath16 in @xmath5 : @xmath17 note that time is modeled if there is a `` true '' tendency toward the completion of transitive mediated triads ( tmts ) , above and beyond the amount that they appear as a by - product of other effects in the model , then the parameter for this effect , @xmath18 should be significantly greater than zero , and the network is more likely to evolve by adding these triads than otherwise . to estimate model parameters siena uses an iterative process of simulating networks based on the current parameter values , comparing the simulated networks to the observed data , and then updating the parameter values until model convergence is achieved . starting from initial parameter values , @xmath19
, the algorithm simulates changes to the network over time and calculates the change statistics in the simulated networks ( i.e. how many reciprocated ties or transitive mediated triads were formed in the simulations over the @xmath10 periods ) call these @xmath20 .
these are then compared with the actual change statistics observed from the empirical data , @xmath21 .
the values of @xmath19 are then incremented or decremented and more simulations are conducted .
this process of simulation , comparison , and parameter update is repeated until the simulated evolution of the network looks like the observed evolution of the network , or @xmath22 = s_{kt}^{obs}.\ ] ] the basic definitions for the change statistics for the structural variables are as follows .
the transitive mediated triads change statistic is equal to @xmath23 for @xmath24 for reciprocity , the change statistic is @xmath25 and for out - degree assortativity , it is @xmath26 for @xmath27 .
in addition to these basic definitions , there are also versions that separately measure the formation of ties and the maintenance of existing ties ( see ripley , snijders , and preciado 2011 for definitions ) .
non - parametric tests for a tendency toward triadic closure are positive , and in multivariate analyses the parameter for transitive mediated triads is positive and significant in all of its forms . the main hypothesis of this paper is therefore supported .
the pattern of trade evolves through a process of transitive closure : new markets for a country s music are more likely , the more transitive mediated triads the new trade relationship would complete .
more detailed results are presented , below .
two sets of unconditional comparisons were performed ( table [ tab : unconditionalcomparisonsofcommonhistorybytiecategory ] ) . the first compares all ordered pairs of countries at all time points that are not connected by a trade tie . within these ,
i compare the set that go on to form a trade tie to the set that remains without a tie in the next observation .
the other set considers ordered country pairs that are tied and compares those that lose that tie to those that maintain it in the next observation .
kolmogorov - smirnov tests show that the sets that had ties at the second observation ( those that created a new tie or maintained a tie ) had experienced significantly more social influence in common than those that did not have ties at the second observation ( those that lost ties or which maintained a lack of tie ) .
the only comparison that was not statistically significant was the length of individual influences held in common ( as opposed to the total of all influences or the number of influences ) at the 5% threshold for tie inclusion .
all other comparisons at both the 1% and 5% thresholds were significant at the @xmath28 level .
pairs of countries that formed ties or maintained ties had more social influence in common than pairs of countries that lost or did not form ties .
-1 cm llrrrr + & & maintained ties & lost ties & new ties & no ties + + & mean total common influence ( years ) & 20 & 15.9 & 10.6 & 4.1 + & mean duration of a common influence & 8.8 & 8.8 & 8.2 & 7.8 + & mean number of common influences & 2.3 & 1.8 & 1.3 & 0.5 + & n observed & 4484 & 2216 & 4884 & 365756 + + & total influence & & + & length of single influences in common & & + & number of influences in common & & + + + & & maintained ties & lost ties & new ties & no ties + + & mean total common influence ( years ) & 6.3 & 4.6 & 3.4 & 1 + & mean duration of a common influence & 8.1 & 8 & 7.6 & 7.2 + & mean number of common influences & 0.8 & 0.6 & 0.5 & 0.1 + & n observed & 1742 & 970 & 2184 & 372444 + + & total influence & & + & length of single influences in common & & + & number of influences in common & & + + + [ tab : unconditionalcomparisonsofcommonhistorybytiecategory ] the descriptive statistics in table [ tab : unconditionalcomparisonsofcommonhistorybytiecategory ] also give a sense of the time scale at work . broadly speaking , social influence can be seen to operate after one or two observations ( between 5 and 10 years ) on average .
those that formed ties had on average 10 years of common influence , which would be equivalent to two sources of common influence for one observation period , or one source for two .
reported parameter estimates from the multivariate siena models can be understood as log odds ratios that weight the contribution of the covariate that they modify to the probability that a trade tie exists .
the magnitude of an effect on the probability of a tie attributable to a difference in covariate values is @xmath29 , where @xmath30 denotes the difference in covariates , and @xmath31 the estimated parameter .
for example , if we wanted to compare the difference in tie probability between contiguous and non - contiguous countries ( a dummy variable ) , @xmath30 would be 1 , and the effect magnitude would be @xmath32 .
if we wanted to compare how much more likely a tie was to originate between countries one standard deviation closer than the mean distance between all countries , we would plug in -0.804 ( minus one standard deviation ( see table [ tab : descriptivestatisticsforcovariates ] for descriptive statistics ) ) for @xmath30 . as such , the size of the effect depends on the range and variability of each individual variable , as well as the estimated parameter .
parameters can not be directly compared to those estimated for other variables without reference to covariate values .
the main results for this study are presented in table [ tab : modelsestimatedfrom19762006dataobservationstakenatfiveyearintervals ] .
this series of models is estimated from observations taken at 5 year intervals from 1976 to 2006 .
the first column represents the basic model , with all variables parametrized in the `` evaluation '' form .
the evaluation form means that parameters weight the importance of the covariates in contributing to the presence of a trade tie , without distinguishing between new ties and existing ties .
the second model divides each variable into `` creation '' ( tendency to form new ties ) and `` endowment '' ( tendency to maintain existing ties ) parametrization .
when the absolute value of the parameter estimate is more than twice the standard error , the parameter can be considered significant at the @xmath33 level .
llrrrr & & est . & s.e . & est . & s.e .
+ & -5.01 & 0.08 & -5.50 & 0.09 + & 0.18 & 0.06 + & endowment & & & 2.09 & 0.25 + & creation & & & -0.29 & 0.13 + & 0.26 & 0.01 + & endowment & & & 0.11 & 0.04 + & creation & & & 0.33 & 0.03 + & -0.47 & 0.02 + & endowment & & & -0.73 & 0.07 + & creation & & & -0.31 & 0.04 + & 0.58 & 0.06 + & endowment & & & 0.35 & 0.21 + & creation & & & 0.67 & 0.14 + & 0.28 & 0.06 + & endowment & & & 0.55 & 0.23 + & creation & & & 0.25 & 0.12 + & 0.76 & 0.04 + & endowment & & & 2.01 & 0.14 + & creation & & & 0.00 & 0.08 + & 0.26 & 0.07 + & endowment & & & 2.57 & 0.28 + & creation & & & -0.83 & 0.16 + & -0.02 & 0.01 + & endowment & & & 0.05 & 0.03 + & creation & & & -0.06 & 0.02 + & 0.62 & 0.01 + & endowment & & & -1.29 & 0.05 + & creation & & & 2.71 & 0.07 + & 30.92 & 4.43 & 50.10 & 7.13 + & 10.73 & 1.18 & 17.61 & 2.02 + & 43.91 & 4.97 & 68.54 & 7.68 + & 39.66 & 3.15 & 56.90 & 4.54 + & 27.10 & 1.60 & 37.42 & 2.44 + & 23.99 & 1.33 & 32.43 & 1.98 + [ tab : modelsestimatedfrom19762006dataobservationstakenatfiveyearintervals ] to give a sense of the importance of the main result , table [ tab : magnitudeofcommoninfluenceeffect ] shows calculated effect magnitudes for different combinations of geographic distance and number of common influences .
this comparison shows that a single common influence is estimated to have a greater positive correlation with the probability of a new exporting relationship than would a hypothetical change in geographic distance of one standard deviation on the log scale .
unlike the other variables in the model , the parameter for common influences ( transitive mediated triads ) may enter multiplicatively if there is more than one common influence .
each common influence multiplies the odds of tie formation by about 1.39 .
so , two influences in common would mean the odds of new ties is @xmath34 , and three influences means that new ties are @xmath35 times more likely than without any influences in common .
considering the observed numbers of influences in common , this could represent a very large effect .
for example , in these data more than 800 ordered country pairs had 8 or more common influences , representing an increase in new tie probability of at least 13.9 times the set of countries with no influences in common ( see table [ tab : magnitudeofcommoninfluenceeffect ] ) .
lrrrrr distance ( miles ) & + & 0 & 1 & 2 & 3 & 4 + 12456 ( max ) & 0.70 & 0.98 & 1.36 & 1.89 & 2.63 + 8866 ( + 1 sd ) & 0.78 & 1.08 & 1.51 & 2.10 & 2.92 + 3983 ( mean ) & 1 & 1.39 & 1.93 & 2.69 & 3.74 + 1790 ( -1 sd ) & 1.28 & 1.78 & 2.48 & 3.45 & 4.80 + 804 ( -2 sd ) & 1.64 & 2.28 & 3.18 & 4.42 & 6.15 + 37 ( min ) & 4.27 & 5.93 & 8.25 & 11.48 & 15.97 + + [ tab : magnitudeofcommoninfluenceeffect ] plugging parameter estimates into @xmath29 to get effect magnitudes ( see beginning of section [ effectsizes ] , above ) , other factors in the formation of new ties are as follows .
new ties that reciprocate existing ties are only 75% as likely as non - reciprocal ties , ruling out reciprocity as a competing explanation for the finding of a tendency to transitivity . geographic distance is associated with fewer new ties ; countries that are a single standard less distant on the log scale are 1.28 times more likely to form a trade tie .
contiguous countries are an additional 1.28 times more likely to form a tie than non - contiguous countries .
unsurprisingly , a past direct colonial relationship makes a new tie almost twice as likely as otherwise , but interestingly , countries that were colonized by the same colonizer were less than half as likely ( 44% ) to form new ties than those that did not share a colonizer .
this rules out the competing hypothesis of common colonizers for explanation of new ties that complete triadic structures in the pattern of trade .
the effect with the greatest magnitude was the estimated size of the music market : doubling the size of a music market multiplies the probability of creating a new outgoing tie by more than 6.5 .
there was a very slight decrease in the probability of exporting to countries with larger music markets .
doubling market size of the destination country is associated with a reduction of new tie probability by about 4% .
the correlates of the maintenance of existing ties are somewhat different .
reciprocal ties are about 8 times more likely to be maintained than non - reciprocal ties . each common influence multiplies the odds of tie maintenance by about 1.11 ( a much smaller effect than in tie formation ) . for geographic distance
, a single standard deviation on the log scale has a greater effect for maintaining ties about 1.44 times more likely for more proximate countries , with an additional 1.73 times the probability of maintaining ties for contiguous countries .
a direct colonial tie makes maintaining ties about 1.42 times as likely , but countries that were colonized by the same colonizer are more than 13 times more likely to maintain their existing music trade .
having a language in common had no significant correlation with the creation of ties , but it is associated with maintaining existing ties around 7.4 times more of the time .
in contrast to tie formation , countries with larger music markets were much less likely to maintain their ties .
doubling music market size was associated with ties being maintained 58% less of the time .
in addition to the results already presented , i conducted many robustness checks .
i varied the threshold used to code a trade flow as present or absent from 1% to 33% of the importing country s estimated music market size ( table [ tab : effectoftieinclusionthresholdontransitivityparameterestimate ] ) .
the estimated parameter for transitive mediated triads increases steadily from .26 to 1.66 as the threshold increases .
this is not just reassuring but indeed strengthens the argument that social influence is occurring , because larger trade flows could be interpreted to exert more influence than small trade flows .
lrrrrrrrrr & + & 1% & 5% & 9% & 13% & 17% & 21% & 35% & 29% & 33% + estimate&0.26 & 0.50 & 0.68 & 0.73 & 0.92 & 1.08 & 1.43 & 1.43 & 1.66 + standard error & 0.01 & 0.04 & 0.07 & 0.13 & 0.16 & 0.19 & 0.25 & 0.29 & 0.39 + + [ tab : effectoftieinclusionthresholdontransitivityparameterestimate ] both economic theory and network science provide reasons to suspect that assortativity may be responsible for observed transitive closure . on these data
at least , that suspicion is not justified .
i re - estimated models on 15 one year periods and specified with an out - degree assortativity parameter , but with the transitivity parameter fixed at zero .
both linear and square root assortativity specifications were considered .
rao score tests of the null hypothesis that the parameter for transitive mediated triads is equal to zero ( because transitivity is accounted for by assortative processes ) were conducted , and the null hypothesis was rejected ( see table [ tab : scoretestofmodelrestrictedsttransitivemediatedtriadparameterisfixedatzero ] ) .
that is , assortativity can not explain the level of transitivity that we observe in the data .
-.2 cm .score test of model restricted s.t .
transitive mediated triad parameter is fixed at zero [ cols= " < , > , > , > " , ] + [ tab : scoretestofmodelrestrictedsttransitivemediatedtriadparameterisfixedatzero ] the following paragraphs report a series of additional robustness checks for which i report the estimate for the transitive mediated triads parameter .
all of these parameter estimates should be compared to 0.26 , which was the main effect reported in table 5 . to confirm that the choice of 5-year intervals did not unduly influence the results
, i repeated the analysis with lags of three , four , six and seven years .
the parameters from these analyses were all at least as high as the one estimated on 5-year intervals . , with parameter estimates for lags of 3 , 4 , 6 , and 7 years equal to 0.30 , 0.28 , 0.27 , and 0.30 , respectively . additionally ,
as mentioned above , it is likely that some countries in the sample simply do not have the industrial infrastructure to export .
so although social influence through importing may be occurring , we would not see evidence of it in their exports . to test the robustness of this specification
, i also ran a series of models including only the subset of countries which export to at least one other country during the study period .
i also varied the time lag between observation years from one to seven years .
these models show substantially the same results as the models with the full sample : including only those with at least one export market produced a parameter estimate of 0.36 .
the set of models run on ties coded according to their absolute , rather than relative , magnitude also agreed with the main results . after discarding the smallest trade flows in absolute terms such that their combined total is only 5% of aggregate music trade ,
the parameter estimate for transitive mediated triad creation was 0.28 .
the model which omitted the years which may have been subject to the distorting influence of digital distribution of music recordings produced similar findings to the full model , with the effect of common influences , distance , and colonial relationship estimated to have slightly stronger effects than the full model ( transitive mediated triad parameter estimated at 0.35 ) .
digital distribution may indeed have begun to have a real effect on the estimated models by moving some of the trade outside of the data set , but this distortion does not substantially affect the conclusions of the analysis . indeed , the relative size of the effect is stronger when we omit the years when the data may be distorted in this way .
chu - shore ( 2010 ) presents a cross - sectional analysis of the network structure of multiple industries . perhaps due to the limits of cross - sectional data , chu - shore ( 2010 ) drew different conclusions from those in this paper , but they can be understood complement the present study . as a sort of falsification exercise
, we could imagine what a similar analysis would find if it were conducted on commodities or purely utilitarian industries in which social influence were implausible , such as exports of coal , or gear cutters . in those industries ,
chu - shore found little to no evidence of a transitivity effect .
in other goods in which subjective valuation plays a strong role , such as books and furniture , there was a strong transitivity effect , lending further support to the argument here . as one example of such a falsification exercise conducted on longitudinal data , i repeated the analysis of this paper on trade in coal ( specifically anthracite , commodity number 3221 in the sitc revision 2 system ) .
as expected , there was no tendency toward transitive mediated triads : the estimated parameter was -0.50 and not significant . rather than triadic closure , trade in coal is characterized by hubs . to converge successfully ,
this model required the addition of the `` out - degree activity '' parameter ( tendency toward degree dispersion , which we would expect in an industry driven by comparative advantage due to natural endowments ) .
the out - degree activity parameter for coal was positive and significant at 0.51 .
in contrast , for music data , an added out - degree activity parameter was estimated at only 0.01 and the estimate for transitive mediated triads was unchanged at 0.26 .
many other notable network - based studies of international trade have been carried out in recent years ( mcnerney , fath , & silverberg , 2013 ; riccaboni & schiavo , 2014 ; ward , ahlquist & rozenas , 2013 ) , and while most are not directly relevant to the present results , bosker and westbrock ( 2014 ) and chaney ( 2014 ) both provide supply - side reasons to predict triadic closure in international trade flows .
both papers study aggregate trade ( bosker and westbrock ( 2014 ) does distinguish between final and intermediate goods ) , rather than trade in a single industry .
setting aside the above evidence ( chu - shore , 2010 ) that network structure depends on the information - intensiveness of the industry , it would be reassuring to rule out these explanations for transitivity .
conveniently , the two papers both predict different forms of triadic closure than the transitive mediated triad i observe in music trade .
bosker and westbrock s ( 2014 ) model implies closure as follows .
consider country @xmath9 , which imports from country @xmath8 , and assume the existence of country @xmath14 .
further assume that country @xmath14 is geographically close to @xmath8 and @xmath9 such that some level of trade is likely . in order to produce goods for export to @xmath9 , country
@xmath36 s demand for intermediate goods from country @xmath14 is expected to increase . in other words , given tie @xmath37 , they predict the formation or strengthening of tie @xmath38 . this would amount to tendencies toward transitive closure , of the non-mediated type ( given the existence @xmath39 ) and/or three - cycles ( given @xmath40 ) .
if we further assume the existence of @xmath41 ( the reciprocal tie to @xmath37 ) , they would also predict transitive mediated triads . in sum bosker and westbrock ( 2014 )
predict transitive ( non - mediated ) closure , three cycles and possible a combination of reciprocity and transitive mediated triads . chaney ( 2014 ) posits a process of trade partner discovery very similar to the models of vazquez ( 2003 ) and saramki & kaski ( 2004 ) , in which trading partners can be found either at random , or by transitive closure .
the actual direction of the trade flow is not determined by the transitive closure process , and multiple types of closed triads are consistent with the model .
chaney s model would predict the existence of transitive mediated triads , transitive ( non - mediated ) triads , and three - cycles .
the model i present here only predicts a significant tendency to transitive mediated triads , and not the other results .
additional robustness checks indicate no significant tendency to either three - cycles ( parameter estimate @xmath42 ) or transitive non - mediated triads ( parameter estimate @xmath43 ) .
the results above already indicate a tendency against reciprocity , providing further evidence against bosker and westbrock ( 2014 ) .
these negative results argue strongly against the supply - side models of bosker and westbrock ( 2014 ) and chaney ( 2014 ) in explaining the current results .
this is not surprising , given that those papers study aggregate trade , and these results in no way call into question the validity of their findings .
consuming foreign products is correlated with the future ability to sell in new markets but only in those markets that were also influenced by those same foreign products .
markets come to have compatible demand for information goods through exposure to the same social influences : for interpreted information goods , between - market trade evolves through a process of transitive closure . by focusing on triadic structures in the pattern of trade
, we can see how social influence complements homophily in the formation of new markets for a country s ( or other `` home industry s '' ) products .
given the highly subjective and varied nature of music recordings , we can safely assume that all ties in the network are homophilous .
taste in the importing country has to be compatible with the music being offered by exporters , or there would be no trade .
however , what homophily can not explain is how the pattern of trade will change and which new ties will come into being .
a process of social influence predicts triadic closure by providing common reference points for markets that previously lacked them .
quite analogously to the everyday experience of individual people , for markets , common experience increases the probability of new relationships .
the present study has considered the question of how a local industry for interpreted information goods begins to sell their products in a new market , focusing on whether social influence from outside markets would help or hurt this effort .
the results show strong positive effects of social influence from imports in developing new export markets , but only in those countries that have experienced the same social influences : the pattern of trade in music evolves through a process of transitive closure .
more fundamentally , the results support the conclusion that the more social influence two markets have in common , the more mutual demand those markets will have for each other s interpreted goods .
adam smith s ( 1776 ) oft - quoted maxim is `` the division of labor is limited by the extent of the market . '' in a sense , this research attempted to consider an even more fundamental question : what is the extent of the market limited by ?
the answer from these data is that for interpreted information goods , the extent of the market is limited by the amount of social influence in common .
african union . 2005 .
`` nairobi plan of action for cultural industries in africa '' first ordinary session of au conference of ministers of culture nairobi , kenya cohen , w. m. and levinthal , d.a .
`` absorptive capacity : a new perspective on learning and innovation . ''
_ administrative science quarterly _
35(1 , special issue : technology , organizations and innovation ) : 128 - 152 .
henry , r. 2001 .
`` caribbean music in the international context : dancing to the beat of a small place . ''
music industry workshop , third united nations conference on the least developed countries , proceedings of the youth forum ( pp .
59 - 64 ) .
brussels : united nations .
lee , s - w . , and waterman , d. 2007 .
`` theatrical feature film trade in the united states , europe , and japan since the 1950s : an empirical study of the home market effect . '' _ journal of media economics _ , 20(3),167 - 188 .
snijders , t.a.b .
`` models for longitudinal network data '' chapter 11 in p. carrington , j. scott , & s. wasserman ( eds . ) , _ models and methods in social network analysis_. new york : cambridge university press snijders , t.a.b . , steglich , c.e.g .
, schweinberger , m. , and huisman , m. 2007 . _ manual for siena version 3_. university of groningen , ics , groningen .
university of oxford , department of statistics , oxford .
storper , m. , and christopherson , s. 1987 .
`` flexible specialization and regional industrial agglomerations : the case of the u.s . motion picture industry . '' _ annals of the association of american geographers _ , 77(1 ) , 104 - 117 . | where do new markets come from ?
i construct a network model in which national markets are nodes and flows of recorded music between them are links and conduct a longitudinal analysis of the global pattern of trade in the period 1976 to 2010 .
i hypothesize that new export markets are developed through a process of transitive closure in the network of international trade .
when two countries markets experience the same social influences , it brings them close enough together for new homophilous ties to be formed .
the implication is that consumption of foreign products helps , not hurts , home - market producers develop overseas markets , but only in those countries that have a history of consuming the same foreign products that were consumed in the home market .
selling in a market changes what is valued in that market , and new market formation is a consequence of having social influences in common . |
black holes ( bhs ) are the end products of a complete gravitational collapse of the massive star surrounded by a boundary from which nothing can escape even light . despite their dark nature , bhs exist in the environment of the brightest objects of the universe .
black hole does not shine but the accreting gas that spirals around , emits radiations and then disappears when reaches the event horizon .
generally , a bh casts a shadow in an emitting medium which arises due to the bending of light that originates near the horizon and travels through the accretion flow .
the shape of a shadow depends on the photon capture sphere resulting from strong bending of light perceived by a distant observer .
the shadow can actually depict the compact object ( bh or naked singularity ) as its appropriate observations can provide information about the parameters as well as spacetime geometry around the compact object .
black holes are usually of two types : stellar mass bhs that are observed when they accrete matter from x - ray binaries and the supermassive bhs in the center of galaxies .
the direct observation of bh is not possible , therefore , event horizon is the defining characteristic which casts a relatively large shadow with respect to a distant observer .
the interest in this topic raised with the possibility of viewing the shadow of sgr a * , a central supermassive bh candidate of the milky way .
it is shown that the shadow of sgr a * is observable with very long baseline interferometry at submillimeter wavelengths with the assumption of thin accretion flow in related region @xcite .
further developments in this framework can be viewed in the literature @xcite .
the concept of a bh shadow initiated with bardeen @xcite as well as cunningham and bardeen @xcite who studied shadow of the kerr bh .
this technique can also be found in @xcite , where coordinates of the observer s sky have been calculated .
different aspects related to shadows for the kerr - newman bh were discussed in @xcite .
hioki and maeda @xcite discussed that bh can be distinguished from naked singularity by examining the shadows of kerr bh as well as naked singularity .
they provided a technique to measure the spin and angle of inclination by defining two observables .
the properties of different bhs shadows have been extensively studied in the literature @xcite .
the shadow of binary bhs is another interesting subject .
bohn et al .
@xcite showed that the shadow of binary bhs is different from the superposition of two singleton bh shadows .
shipley and dolan @xcite explored the shadows of two extremely charged binary bhs and stated that they are interesting examples of chaotic scattering in nature .
another fascinating feature of the bh shadows is that it can serve to test the no hair theorem @xcite .
the occurrence of curvature singularities in gravitational collapse has led to important developments in general relativity ( gr ) . to resolve the problem of singularities , quantum gravity proved to be very successful . since gr breaks down at short distances , the non - commutative ( nc ) theory provides a useful approach to discuss the spacetime dynamics at planck scale .
gross and mende @xcite explored the behavior of string interactions at short distances .
maggiore @xcite found generalized uncertainty principle which is consistent with the results obtained in string theory .
he suggested that the minimum length ( of order planck length ) emerges naturally from any theory of quantum gravity .
the structure of nc geometry is based on universal minimal length scale which is equivalent to planck s length . in this framework
, the spacetime quantization requires that its coordinates become hermitian operators which do not commute , i.e. , @xcite @xmath0=\iota \theta^{ab},\ ] ] where @xmath1 is a real - valued antisymmetric matrix with @xmath2 , @xmath3 is a constant of dimension @xmath4 . in the limit @xmath5
, the ordinary spacetime is recovered .
physically , @xmath1 represents a small patch in @xmath6-plane of the observable area as the planck s constant ( @xmath7 ) illustrates the smallest fundamental cell of the observable phase space in quantum mechanics .
the quantum gravity bhs incorporates the effects of quantum gravity in the vicinity of origin , where the classical curvature singularity appears .
many singularity free bhs have been obtained in the framework of quantum gravity such as bhs inspired by nc geometry . in the early 90 s , doplicher et al .
@xcite provided a detailed study about the spacetime non - commutativity which shows that gr together with the uncertainty principle of quantum mechanics leads to a class of models of nc spacetime .
there are some comprehensive reviews about nc field theory and the spacetime with nc coordinates @xcite .
the nc form of different bhs such as charged @xcite , higher dimensional @xcite , rotating @xcite have also been obtained .
the effects of non - commutativity on the properties of bhs have extensively been studied .
nicolini et al .
@xcite investigated radiating nc schwarzschild bh and found that non - commutativity can cure singularity problems .
nozari and mehdipour @xcite explored quantum tunneling from nc schwarzschild bh and showed that the information would be preserved by a stable bh remnant due to nc effects .
the lensing properties of nc bhs have also been explored .
ding et al .
@xcite studied the influence of nc parameter on the strong field lensing of nc schwarzschild bh and found that the nc parameter has a similar behavior as that of charge of reissner - nordstr@xmath8 m bh .
wen - wei @xcite found that the nc parameter affects the shape of bh shadow and increases its deformation as compared to kerr bh .
many other important aspects of spacetime non - commutativity have been discussed in literature @xcite . generally , bhs are characterized by non - zero angular momentum which breaks the spherical symmetry resulting in rotational symmetry around axis . the gravitational field around bh
is highly affected by its rotation , therefore , it is natural to expect some important effects on the shadow cast by the bh .
it is observed that the shadow of non - rotating bh has circular shape while the shadow of rotating bh is deformed due to its spin @xcite .
another motivation comes from the fact that sgr a * - the most remarkable candidate for bh which likely have spin whose axis is perpendicular to the galactic plane .
the predicted size of the shadow of sgr a * is @xmath9 30 @xmath10 arcseconds @xcite . according to no hair theorem , any stationary ,
asymptotically flat bh vacuum solution of the field equations can be characterized by three variables : mass , charge and angular momentum per unit mass ( @xmath11 ) .
this leads to the formation of kerr - newman ( kn ) bh as an outcome of non - spherical gravitational collapse .
it is suggested that an isolated rotating bh can not have an electromagnetic field except it endowed with a net electric charge @xcite .
it is possible that kerr newman bh could be the end point of magnetized , massive rotating stars as a result of catastrophic gravitational collapse @xcite .
it is also proposed that the charged rapidly rotating galactic bhs ( kerr newman bhs ) emit gamma rays ( in a bipolar out flow ) @xcite .
the effect of plasma on the properties of bh has extensively been studied .
muhleman and johnston @xcite discussed the deflection of light rays by gravity as well plasma by considering radio waves traveling near the sun .
he used a product of the plasma and gravitational refractive indices in weak field approximation .
breuer and ehlers @xcite explored the hamiltonian for light rays in a magnetized pressureless plasma .
a similar derivation was found by perlick @xcite for the pressureless , non - magnetized plasma .
perlick et al .
@xcite discussed the effect of plasma on the shadow of ellis wormholes as well as spherically symmetric bh and found that the decrease in the shadow by the presence of plasma .
rogers @xcite explored gravitational lensing and circular orbits of the schwarzschild bh for plasma density distributions @xmath12 with @xmath13 .
there is a large body of literature available @xcite on the effect of plasma on the shadows cast by the bhs . in this paper
, we study the shadow of charged rotating nc bh in the equatorial as well as non - equatorial plane .
the paper is organized as follows . in the next section ,
we discuss the null geodesics for the nc bh .
section * 3 * is devoted to study the apparent shape of the nc geometry inspired shadows and the behavior of observables . in section * 4 * , we explore the influence of plasma on the shape of shadows .
the results are concluded in the last section .
non - commutativity is an intrinsic property of the spacetime which implies the existence of a natural ultra - violet cutoff or equivalently a minimal ( measurable ) length in quantum field theory .
the nc spaces have a long history .
the first model of nc spacetime ( admitting a minimal length scale and invariant under the lorentz transformation ) was proposed by snyder @xcite which was an attempt to introduce a short distance cutoff in quantum field theory .
the interest in nc theory raised with the seminal paper of seiberg and witten @xcite in which they extended nc theory to the string theory with a constant @xmath14-field . some important examples of non - commutativity includes lei algebra type deformation ( the most studied example is @xmath15-minkowski spacetime @xcite ) and canonical type deformation ( a simple and widely discussed example of nc spacetime ) @xcite . in nc geometry ,
the pointlike smeared gravitating source ( @xmath16 is the dimension of spacetime ) is assumed to remove divergencies in gr .
modesto and nicolini @xcite obtained regular charged rotating bh solution in the background of nc geometry ( based on canonical deformation ) by employing newman - janis mechanism .
they found that the resulting bh solution is unique and exhibits local regular behavior at the origin as well as standard rotating geometries at large distances .
the ricci scalar as well as kretschmann invariant remain finite at the origin .
the charged nc bh in boyer - lindquist coordinates is given as @xmath17dt d\phi \\\label{1 } & -&\sin^2\theta[\sigma+a^2\sin^2\theta(2-\frac{\delta -a^2\sin^2\theta}{\sigma})]d\phi^2,\end{aligned}\ ] ] where @xmath18.\end{aligned}\ ] ] the event horizon of ( [ 1 ] ) is the largest root of @xmath19 given by @xmath20 a detailed discussion about the nature of horizons and thermodynamical properties are available in @xcite .
here we would like to discuss the horizons briefly .
the function @xmath19 produces horizons for @xmath21 , i.e. , for extremal and non - extremal cases , ( [ 1 ] ) represents a bh but for @xmath22 , there is no horizon and the metric ( [ 1 ] ) describes a regular charged spinning object such as charged spinning mini gravastar @xcite . in this paper
, we do not consider horizonless case .
for this purpose , we find maximum values for the spin and charge parameter corresponding to different values of non - commutative parameter by solving @xmath23 and @xmath24 numerically ( figure * 1 * ) .
the particle motion can be described by the lagrangian + @xmath25 where @xmath26 , @xmath27 is the particle s @xmath28-velocity and @xmath29 is the affine parameter .
the energy @xmath30 and angular momentum @xmath31 are given by @xmath32 the hamilton - jacobi equation yields @xmath33 we find that the lagrangian is independent of @xmath34 and @xmath35 , therefore @xmath36 and @xmath37 are conserved and hence describe stationary and axisymmetric characteristics of charged rotating nc bh .
equation ( [ 9 ] ) takes the form @xmath38 for the metric ( [ 1 ] ) , it leads to @xmath39 , \\\label{6n } \sigma\frac{\partial \phi}{\partial\lambda}&=&(l\csc^2\theta - ae)+\frac{a}{\delta}[e(r^2+a^2)-al ] , \\\label{4n1 } \sigma\frac{\partial r}{\partial\lambda}&=&\pm\sqrt{r } , \\\label{4n } \sigma\frac{\partial \theta}{\partial\lambda}&=&\pm\sqrt{\theta},\end{aligned}\ ] ] where @xmath40 ^ 2-\delta[k+(l - ae)^2],\\\label{15 } \theta&=&k+\cos^2\theta(a^2e^2-l^2\csc^2\theta),\end{aligned}\ ] ] and @xmath41 is the separation constant introduced by carter @xcite .
equation ( [ 4n1 ] ) can be represented as @xmath42 where @xmath43 ^ 2-\delta[k+(l - ae)^2].\ ] ] the above equation can describe radial motion of the particles .
figure * 2 * shows that orbits of the photons are initially unstable for small values of @xmath44 , it approaches to maximum values ( which shows the stable orbits ) and finally becomes unstable with increasing @xmath44 .
we also note that stable points for the photons in the charged nc geometry corresponding to large values of charge are greater than that of small values of charge . the conditions for unstable circular orbit @xmath45 and @xmath46 @xcite lead to @xmath47 where @xmath48 , @xmath49 are the impact parameters and @xmath50.\end{aligned}\ ] ] solving eqs.([14.1 ] ) and ( [ 14.2 ] ) , we have @xmath51 } [ m(r)(a^2 - 3r^2)+r(r^2+a^2 ) \\\label{17 } & \times&(m'(r)+1)+2q^2(r)-q(r)q'(r)(r^2+a^2 ) ] , \\\nonumber \eta&=&\frac{r^2}{a^2[m(r)+r(m'(r)-1)]^2}[m(r)(4a^2 - 9m(r)r+6r^2)r-2m'(r ) \\\nonumber & \times&(2a^2+r^2 - 3m(r)r)r^2-r^4(m'^2(r)+1)-4q^2(r ) \\\nonumber & \times&(a^2+q^2(r)-3m(r)r+m'(r)r^2+r^2)-q'(r)(4a^2
+ 4q^3(r ) \\\label{17aa } & -&6m(r)q'(r)r-2m'(r)q(r)r^2-q(r)r+2q(r)r^2)].\end{aligned}\ ] ] the above equations are important as they are related to the celestial coordinates of the image seen by an observer at infinity and determine the contour of the bh shadow ( will be discussed in the next section ) . + +
here , we study some features of bhs related to its shadows .
photons that emit from a background source of a bh suffer deflection due strong gravitational field .
the photons having small impact parameter fall into the bh and form a dark region in sky .
an observer viewing that part of the sky would observe a black spot having larger radius than the event horizon around the bh .
this dark region is called the shadow of the bh .
the shape of the bh shadow for a far away observer can be determined by the following celestial coordinates @xcite @xmath52 where @xmath53 is the angle of inclination between the rotation axis of bh and observer s line of sight .
the celestial coordinates @xmath54 and @xmath55 represent the apparent perpendicular distances of the image around the bh , seen from the axis of symmetry and from its projection on equatorial plane , respectively .
these coordinates provide the apparent position of the image in the plane which passes through the center of bh and is orthogonal to the line joining the bh and observer . using eqs.([6n])-([4n ] ) in ( [ 18])-([19 ] ) , we have @xmath56 here @xmath57 and @xmath58 are given by eqs.([17])-([17aa ] ) .
the shadow of a charged rotating nc bh is shown in figure * 3*. the silhouette of the shadow is influenced by nc mass , charge , spin and the angle of inclination .
the shape of the shadow changes with respect to the nc charge .
the high value of nc charge causes a decrease in the shadow . for the low values of charge ,
the shape deviates from the standard circle .
moreover , for the highly spinning bh , the shadow is more deformed than the slowly spinning bh .
there is more deformation for @xmath59 as compared to @xmath60 . here
, we study observables for the charged rotating nc bh . the observables can be used to determine important astronomical information about the collapsed objects ( bhs and naked singularities ) such as to characterize the apparent shape of the shadow .
the silhouette of bh shadow determines apparent image of the photon capture sphere seen by the observer at large distance .
hioki and meada @xcite proposed that the shadows s silhouette passes through four points , the top @xmath61 , bottom @xmath62 , right @xmath63 and left point @xmath64 .
they presented two parameters as observable quantities .
firstly , the radius @xmath65 determines the size ( approximate ) of the shadow defined as the radius passing through three different points , i.e. , top , bottom and the most right points of a reference circle .
the second is @xmath66 which measures the deformation and is defined as @xmath67 , where @xmath68 is the difference between the most left points , i.e. , @xmath64 and @xmath69 .
the point @xmath69 cuts the horizontal axis opposite to @xmath63 .
this discussion can be explained through the figure 1 given in @xcite .
the plots of radius of shadow and deformation parameter are shown in figure * 4*. we observe that the deformation parameter has increasing behavior along spin for the varying values of nc charge . for large value of nc charge , the deformation gradually decreases which means that the shape of shadow will approximately be a circle .
the deformation is high for the high values of spin .
the radius of the shadow has decreasing behavior with respect to spin .
in this section , we explore the effects of plasma on the shadow of rotating charged bh in the background of nc geometry . for the axially symmetric bh ,
the refraction index of the plasma is @xmath70 , where the photon frequency @xmath71 is measured by the observer with velocity @xmath27 .
the effective energy of the photon is @xmath72 , ( @xmath73 is the 4-momentum of the photon ) .
the relation between the refraction index and 4-momentum of the photon is given as @xcite @xmath74 where @xmath75 in the absence of plasma . in order to find analytical results , it is useful to write refraction index with specific plasma frequency @xmath76 @xcite @xmath77 the hamilton - jacobi equation for the photon around the bh surrounded by a plasma has the following form @xcite @xmath78.\ ] ] this leads to the following equations of motion @xmath79 , \\\label{6 m } \sigma\frac{\partial \phi}{\partial\lambda}&=&(l\csc^2\theta - ae)+\frac{a}{\delta}[e(r^2+a^2)-al ] , \\\label{4m1 } \sigma\frac{\partial r}{\partial\lambda}&=&\pm\sqrt{r_{p } } , \\\label{4 m } \sigma\frac{\partial \theta}{\partial\lambda}&=&\pm\sqrt{\theta_{p}},\end{aligned}\ ] ] where @xmath80 ^ 2+(n^2 - 1)(r^2+a^2)^2e^2 \\\nonumber & -&\delta[k+(l - ae)^2],\\\label{15 } \theta_{p}&=&k+\cos^2\theta(a^2e^2-l^2\csc^2\theta)-a^2(n^2 - 1)e^2\sin^2\theta.\end{aligned}\ ] ] for the sake of simplicity , we chose a specific form of @xmath76 with the assumption that the spacetime is filled with cold plasma ( non - magnetized ) whose electron plasma frequency is a function of the radial coordinate @xmath81 where @xmath82 and @xmath83 represent the electron charge and mass respectively , and @xmath84 denotes the number density of the electrons in plasma .
following @xcite , we consider @xmath84 as the power law density such that @xmath85 which yields @xmath86 , where @xmath87 is an arbitrary constant . here
we take @xmath88 . to study the effect of plasma on the shadow of charged rotating nc bh
, we follow the same procedure as in section * 2*. we find the parameters @xmath57 and @xmath58 of the form @xmath89 where @xmath90 , \\\nonumber z&=&\frac{n^2(r^2+a^2)^2}{\delta}+\frac{2rn^2(r)(r^2+a^2)- n(r)n'(r)(r^2+a^2)^2}{(m(r)-r - m'(r)r - q(r)q'(r))}.\end{aligned}\ ] ] the celestial coordinates under the influence of plasma can be written as @xmath91 figure * 5 * indicates the effect of plasma for @xmath92 , @xmath93 , @xmath94 and varying charge as well as @xmath95 .
we observe that the shadow with plasma is greater than the case of without plasma .
the shadow increases with the increase of plasma while increase in the value of charge tends to decrease the deformation .
the study of the apparent shapes of a bh shadow has been the subject of interest for the last few years as it may be possible to observe a bh in the center of galaxy in near future @xcite .
the shadow appears due to the effects of strong gravity near the bh , so it may be helpful to explore its nature . for classical kerr and kerr - newman bhs ,
the shadow is affected by the spin , charge and the inclination angle .
since the nc geometry has many applications such as resolving the singularity problems in gr , it would be interesting to study the effect of nc charge parameter on the shadow as well the particle orbits . in this paper
, we have examined the shadow cast by a charged rotating nc bh . in order to study the effect of nc charge on the silhouette of shadow
, we have first obtained null geodesics and analyzed the behavior of the effective potential .
it is found that the photons are stable for small values of radius but they approach to maxima for the large radius which correspond to unstable orbits .
after reaching the maxima , @xmath96 starts diverging .
we have used null geodesics to find the celestial coordinates @xmath54 and @xmath55 which describe the nature of the shadow .
it is found that the charged nc bh shadow is not only affected by nc charge but also due to spin as well as the angle of inclination .
we observe that the silhouette of shadow changes its shape with respect to varying charge parameter .
the shape deviates from the circle for the small charge while for the large nc charge , it approximately maintains the shape of a circle .
this behavior is almost similar to the rotating uncharged nc bh @xcite ( where the shape of the circle is maintained with increase in the nc parameter ) as well as charged bhs @xcite .
the high values of spin as well as angle of inclination deform the shape of the shadow .
we have also discussed the radius of shadow and deformation parameter for this bh .
the large value of nc charge corresponds to less deformation while for rotating uncharged nc bh @xcite the distortion parameter increases with nc parameter .
this leads to an idea that nc charge helps to maintain the shape of circle .
the radius of the shadow decreases as spin increases .
finally , we have examined the shadow of charged rotating nc bh in the presence of plasma .
the existence of plasma increases the size of the shadow and deformation in the shape reduces due to the presence of nc charge which is analogous to @xcite .
we conclude that the presence of nc charge not only affects the particle orbits but also affects shadow of the bh .
cash , w. , shipley , a. , osterman , s. and joy , m. : nature * 407*(2000)160 ; hirabayashi , h. et al . : arxiv : astro - ph/0501020 ; doeleman , s.s .
nature * 455*(2008)78 ; doeleman , s.s .
: astrophys .
j. * 695*(2009)59 .
amarilla , l. et al .
d * 81*(2010)124045 ; amarilla , l. and eiroa , e.f .
d * 87*(2013)044057 ; abdujabbarov , a. et al .
: astrophys .
space sci .
* 334*(2012)429 ; grenzebach , a. , perlick , v. and lammerzahl , c. : phys . rev .
d * 89*(2014)124004 ; papnoi , u. : phys . rev .
d * 90*(2014)024073 ; yang , l. and li , z. : int . j. mod . phys .
d * 25*(2016)1650026 .
garcia - compean , h. and soto - campos , c. : phys .
d * 74*(2006)104028 ; nozari , k. and fazlpour , b. : mod
lett . a * 22*(2007)2917 ; banerjee , r. majhi , b.r . and samnta , s. : phys . rev .
d * 77*(2008)124035 ; zhang , b. et al .
: europhys .
lett * 94*(2011)2002 ; miao , y.g . , xue , z. and zhang , s.j .
d * 21*(2012)1250018 .
bisnovatyi - kogan , g.s . and
tsupko , o.y .
cosmol . * 15*(2009)20 ; mon . not .
* 404*(2010)1790 ; morozova , v. , ahmedov , b. and tursunov , a. : astrophys .
space sci .
* 346*(2013)513 ; er , x. and mao , s. : mon . not
* 437*(2013)2180 ; abdujabbarov , a. et al .
: arxiv:1512.05206v2 ; atamurotov , f. , ahmedov , b. and abdujabbarov , a. : phys .
d * 92*(2015)084005 . | this paper investigates the shadow of a charged rotating non - commutative black hole . for this purpose
, we first formulate the null geodesics and study the effects of non - commutative charge on the photon orbit .
we then explore the effect of spin , angle of inclination as well as non - commutative charge on the silhouette of the shadow .
it is found that shape of the shadow deviates from the circle with the decrease in the non - commutative charge .
we also discuss observable quantities to study the deformation and distortion in the shadow cast by the black hole which decreases for small values of non - commutative charge .
finally , we study the shadows in the presence of plasma .
we conclude that the non - commutativity has a great impact on the black hole shadow .
* keywords : * non - commutative geometry ; black hole ; shadow . + * pacs : * 95.30.sf ; 04.70.-s . ; 04.62.+v . |
anomalous x - ray pulsars ( axps ) and soft gamma - ray repeaters ( sgrs ) are a small class of slowly rotating ( 5 - 12 s ) neutron stars with emission properties much at variance with those of ordinary x - ray pulsars , both the young radio pulsars and the x - ray binary pulsars .
they are called `` anomalous '' because their high x - ray luminosity ( @xmath0 erg / s ) can not be easily explained in terms of the conventional processes which apply to other classes of pulsars , i.e. accretion from a binary companion or injection of rotational energy in the pulsar wind / magnetosphere . on the other hand , measurements of spin periods and period derivatives , when the latter are interpreted as due to electromagnetic dipolar losses ,
suggest that these objects may host `` magnetars '' , i.e. neutron stars endowed with an ultra - strong magnetic field ( @xmath1 g , see duncan & thompson 1992 ; thompson & duncan 1993 , 1996 ) .
the magnetar scenario appears so far very promising in explaining both the main energy source of these objects ( the decay of the super - strong field ) and the emission of the short , energetic bursts .
moreover , the magnetar scenario can account for the properties of giant flares , extremely energetic transient events ( @xmath2 erg / s ) detected from sgrs . however ,
alternative scenarios to explain the enigmatic properties of these sources have been invoked . among these , models involving accretion from a fossil disk , left by the supernova event which gave birth to the neutron star , are still largely plausible ( van paradijs et al . 1995 ; chatterjee , hernquist & narayan 2000 ) . very recently
such a ( maybe passive ) disk has been indeed observed in the ir around axp 4u 0142 + 61 ( wang , chakrabarty & kaplan 2006 ) .
magnetars candidates are strong x - ray sources , and their x - ray spectra in the 0.5 - 10 kev band are usually fit with a two component model consisting of ( besides interstellar absorption ) a thermal blackbody with a typical temperature @xmath30.30.4 kev and a power - law with a relatively steep photon index , @xmath424 ( see kaspi 2006 in this volume , and woods & thompson 2006 for recent reviews ) . in some cases , sgr spectra have been fit with a single power - law component , but recent longer observations showed that , also for these sources , a blackbody component is often required ( mereghetti et al .
. despite the fact that the blackbody plus power - law spectral model has been largely applied to magnetar spectra for many years , a reliable physical interpretation of these two components is still missing .
quite recently , an attempt to interpret the observed spectrum and long term variations of 1e 1048.15937 in terms of a ( non resonant ) comptonization model has been presented by tiengo et al .
furthermore , the recent discovery of magnetar counterparts in the radio to the @xmath5-ray bands ( camilo et al .
2006 ; hulleman et al .
2000 ; kuiper et al .
2004 ) enforced the idea that their multiwavelength spectral energy distribution is by far more complicated than a simple blackbody plus power - law distribution ( see also den hartog et al .
contribution in this volume ) . and @xmath6=0.3 ( from lyutikov & gavriil 2006).,scaledwidth=45.0%,scaledwidth=35.0% ]
following the original suggestion by thompson , lyutikov & kulkarni ( 2002 ) , it has been proposed by lyutikov & gavriil ( 2006 ) that the presence of a warm non - relativistic plasma in the magnetosphere of magnetars , might be responsible , through efficient resonant electron cyclotron scattering , of distorting the thermal x - ray emission from the star surface . in particular , if a large volume of the neutron star magnetosphere of the is partially filled by @xmath7-currents , the thermal ( or quasi - thermal ) cooling radiation emerging from the star surface will experience repeated scatterings at the cyclotron resonance . the efficiency of the process is quantified by the scattering optical depth , @xmath8 .
following again lyutikov & gavriil ( 2006 ) , and assuming a dipolar configuration for the magnetic field , this can be written as @xmath9 where @xmath10 is the cross section for electron scattering in the magnetized regime , @xmath11 is the electrons number density , @xmath12 is the angle between the photon propagation direction and the local magnetic field , and @xmath13 here @xmath14 is the radial coordinate , @xmath15 is the electron cyclotron frequency , and @xmath16 is the local value of the magnetic field . at variance with the case of unmagnetized thomson scattering , in presence of a magnetic field @xmath10 is resonant and is given by @xmath17 where @xmath18 is the natural width of the first cyclotron harmonic and @xmath19 is the cross - section for unmagnetized thomson scattering .
the crucial point is that , in this model , the plasma distribution is spatially extended .
since the value of the resonant energy depends on the local value of the magnetic field , repeated scatterings of photons could lead to the formation of a high energy tail instead of a narrow line . at relatively large distances from the surface ,
the magnetic field drops to typical values such that scattering is resonant for soft x - ray photons , below @xmath20 kev .
on the other hand , at these energies the resonant scattering optical depth greatly exceeds that for thomson scattering , @xmath21 , @xmath22 where @xmath23 is the classical electron radius .
this implies that even a relatively small amount of plasma suspended in the magnetosphere of the neutron star may considerably modify the emergent spectra .
there is now increasing support to the idea that the magnetar magnetospheres might be twisted ( thompson , lyutikov & kulkarni 2002 ) , in which case they may be partially filled by @xmath24 currents with densities well in excess of the goldreich - julian density ( which is expected in the case of a simple dipolar configuration , see goldreich & julian 1968 ) .
a further effect is related to the fact that the spatial region , over which such currents are distributed , is permeated by a highly inhomogeneous magnetic field .
although for soft x - ray photons the scattering conserves energy in the electron rest frame , thermal / bulk electron motion produces an energy transfer in the observer rest frame .
multiple resonant cyclotron scatterings of thermal photons will , on average , up - scatter the transmitted radiation , resulting in the formation of a non - thermal spectrum .
the emerging spectrum will then appear distorted and , as it is the case for thermal non - magnetic comptonization , its shape may be well represented by the superposition of thermal and non - thermal components ( see fig.[fig : models ] ) .
here we present an application of these models to magnetar spectra .
theoretical models are computed as in lyutikov & gavriil ( 2006 ) , under the assumption that scattering occurs in a non - relativistic , static , warm medium and neglecting electron recoil , i.e. in the limit @xmath25 ( where @xmath6 , a model parameter , is the thermal velocity of electrons in units of the speed of light ) and for low - energy photons , @xmath26 . within the range of parameter values allowed by the numerical code ,
the non - thermal effects are most prominent for models in which the plasma is mildly relativistic , @xmath27 , and the optical depth is reaching its boudary value @xmath28 .
[ tab:1 ] lcc & bb + pl & rcs model + n@xmath29 ( @xmath30@xmath31 ) & 1.12@xmath32 & 0.49@xmath33 + kt ( kev ) & 0.64@xmath34 & 0.53@xmath34 + bb norm & 1.03@xmath35 &
+ @xmath36 & 3.3@xmath37 & + pl norm & 1.10@xmath38 &
+ @xmath6 & & 0.41@xmath33 + @xmath39 & & 1.9@xmath40 + rcs norm & & 0.22@xmath41 + @xmath42 & 0.99 & 1.08 + flux ( ergs@xmath43@xmath31 ) & 7.9@xmath44 & 7.9@xmath45 +
in order to perform a quantitative comparison between fits to magnetar spectra performed with the model described above and with the canonical blackbody plus power - law model , we developed a code to upload the rcs model into xspec .
we created a grid of intensity tables ( through a monte - carlo simulation ) for a set of values of the three model parameters , i.e. @xmath6 , @xmath39 and @xmath46 , where @xmath46 is the temperature of the seed thermal surface emission ( assumed to be a blackbody ) .
the parameters ranges used were @xmath47 ( step 0.1 ) , @xmath48 ( step 1 ) and @xmath49 kev@xmath50 kev ( step 0.05kev ) .
for each model , the spectrum has been computed in the energy range 0.112kev ( step 0.05kev ) .
the final xspec _ atable _ spectral model has therefore four parameters , the three listed above plus the last one being the normalization constant , which are simultaneously varied during the spectral fitting following the standard @xmath51 minimization technique . note that a previous attempt to apply the lyutikov & gavriil ( 2006 ) model to spectral data was presented by the authors themselves .
however , in that case @xmath6 and @xmath39 were fixed a priori during each fit and eventually varied manually .
the fit itself was done only by varying @xmath46 and the normalization factor , while in this case we are able to fit simultaneously all the four parameters , resulting in a more precise determination of their best fitting values and in a more reliable @xmath51 determination .
here we present the preliminary results of the spectral modelling of magnetars with the resonant cyclotron scattering model ( rcs ) .
we concentrate on the soft x - ray spectra ( 110 kev ) of three sources , 1e 104856 , 1rxs j17084009 and sgr 180620 , and we use data obtained with the xmm - newton satellite .
detailed information about the observations , data reduction and analysis can be found in mereghetti et al .
( 2004 ) , rea et al .
( 2005a ) and rea et al .
( 2005b ) , for the three sources , respectively .
the fit of the x - ray spectrum of 1e104856 is presented in figure[fig:1e1048 ] and table1 . in this case
we find that the rcs model is successful in reproducing the data , as well as the canonical blackbody plus power - law model ( bb+pl ) .
the two fits have the same number of degrees of freedom .
the value of the column density found with the rcs model is lower ( although both consistent with durant & vankerkwijk 2006 ) , while the temperature of the thermal component @xmath46 and the total flux are consistent with that found with the bb+pl modelling .
these results are in agreement with the previous findings by lyutikov & gavriil ( 2006 ) , who presented a similar attempt by using chandra data and varying manually some of the model parameters to reproduce the spectrum .
as mentioned before , the technique used here is more reliable in providing a determination of the best fitting parameters . in this case
, the source spectrum can be reproduced by invoking scattering by @xmath24 with thermal velocity @xmath52 and optical depth @xmath53 . on the other hand , the cases of 1rxsj17084009 and sgr180620 are more problematic .
these sources have much harder x - ray spectra ( in the 110 kev band ) , that can not be accounted for by the rcs model alone ( see fig.[fig : rcs_try ] ) at least when the spectral deformation resulting from the resonant scattering process is computed in the simple way proposed by lyutikov & gavriil ( 2006 ) .
however , it is worth noting that , at variance with 1e104856 , these two sources exhibit strong high - energy power - law tails which extend up to the @xmath5-rays ( kuiper et al .
2006 ; molkov et al .
2005 ; mereghetti et al .
motivated by that , we fitted the whole 150kev spectra with the rcs model plus a power - law . for both sources , we obtained a good fit with a reduced @xmath51 of 1.08 and 1.1 , respectively .
although extremely promising , these analysis are very preliminary and they still require a careful assessment of intercalibration issues which arise when different detectors are used . therefore , we do not discuss the results in more detail nor we present here best fitting parameter values . these results will be presented in more details in a forthcoming paper ( rea et al .
2006 , in prep . )
the preliminary results reported here show that the x - ray spectra of the axp 1e104856 can be successfully reproduced by the rcs model .
the quality of the fit is comparable to that obtained by using the canonical blackbody plus power - law model , with the advantage that the rcs model spectra provide a better physical understanding of the presence of the thermal/ non - thermal components in the observed spectrum . on the other hand , this is not totally unexpected since this source has a non - thermal tail which is relatively soft with respect to other axps , and a similar result has already been found by lyutikov & gavriil , although their fitting technique was less accurate . for the other two sources considered here , 1rxsj17084009 and sgr180620 , we find that the rcs model alone can not account for the relatively hard non - thermal component observed below @xmath5410 kev .
on the other hand , in both cases a composite model consisting of a rcs model plus a power - law can fit the x - ray spectra in the 1100 kev band .
this may suggest that , in the 610kev energy range , the x - ray emission of these sources is contaminated by the hard non - thermal component which has beed recently observed with integral at much higher energies ( kuiper et al .
a similar hard tail has not been observed , at least so far , in the case of 1e104856 . a detailed paper with a systematic application of the rcs model to the whole class of magnetars and a study of their spectral and flux variability in this scenario is in preparation , accompanied by a more complete discussion ( rea et al . 2006 , in prep . ) .
we thank valentina bianchin and gavin ramsay for their kind help in building the xspec rcs model .
we also thank fotis gavriil for having kindly allowed us to look into his preliminary model , andrea tiengo , gianluca israel and the anonymous referee for useful comments .
nr thanks the mullard space science laboratory , where this work was partially done , for the warm hospitality .
goldreich , p. & julian , w.h .
apj * 157 * , 869 ( 1969 ) hulleman , f. , van kerkwijk , m. h. & kulkarni , s. r. nature * 408 * , 689 ( 2000 ) kuiper , l. , hermsen , w. & mendez , m. apj * 613 * , 1173 ( 2004 ) kuiper , l. , hermsen , w. , den hartog , p. r. , collmar , w. apj * 645 * , 556 ( 2006 ) lyutikov , m. & gavriil , f. mnras * 368 * , 690 ( 2006 ) mereghetti , s. , tiengo , a. , stella , l. , et al .
apj * 608 * , 427 ( 2004 ) mereghetti , s. , tiengo , a. , esposito , p. , et al .
apj * 628 * , 938 ( 2005a ) mereghetti , s. , gotz , d. , mirabel , i. f. , hurley , k. a&a * 433 * , l9 ( 2005b ) molkov , s. , hurley , k. , sunyaev , r. , et al . a&a * 433 * , l13 ( 2005 ) rea , n. , oosterbroek , t. , zane , s. , et al .
mnras * 361 * , 710 ( 2005a ) | the x - ray spectra of the magnetar candidates are customarily fitted with an empirical , two component model : an absorbed blackbody and a power - law . however
, the physical interpretation of these two spectral components is rarely discussed .
it has been recently proposed that the presence of a hot plasma in the magnetosphere of highly magnetized neutron stars might distort , through efficient resonant cyclotron scattering , the thermal emission from the neutron star surface , resulting in production of non - thermal spectra . here
we discuss the resonant cyclotron scattering ( rcs ) model , and present its xspec implementation , as well as preliminary results of its application to anomalous x - ray pulsars and soft gamma - ray repeaters .
example.eps gsave newpath 20 20 moveto 20 220 lineto 220 220 lineto 220 20 lineto closepath 2 setlinewidth gsave .4 setgray fill grestore stroke grestore |
the following result is known in the literature as ostrowski s inequality @xcite .
[ t1]let @xmath0 \rightarrow \mathbb{r}$ ] be a differentiable mapping on @xmath1 with the property that @xmath2 for all @xmath3 then @xmath4 \left ( b - a\right ) m , \label{1.1}\]]for all @xmath5 .$ ] the constant @xmath6 is best possible in the sense that it can not be replaced by a smaller constant . in @xcite
, the author has proved the following ostrowski type inequality .
[ t2]let @xmath0 \rightarrow \mathbb{r}$ ] be continuous on @xmath7 $ ] with @xmath8 and differentiable on @xmath9 let @xmath10 and assume that @xmath11 then we have the inequality @xmath12 for any @xmath13 where for @xmath14@xmath15 @xmath16 ^{\frac{1}{p}},\;\text{is the } p-\text{logarithmic mean}\;p\in \mathbb{r}\backslash \left\ { -1,0\right\ } , \ ] ] and @xmath17 another result of this type obtained in the same paper is : [ t3]let @xmath0 \rightarrow \mathbb{r}$ ] be continuous on @xmath7 $ ] ( with @xmath8 ) and differentiable on @xmath9 if @xmath18 then we have the inequality @xmath19 ^{b - x}}{\left [ i\left ( a , x\right ) \right ] ^{x - a}}\right ] + 2\left ( x - a\right ) \ln x\right ] \label{1.3}\ ] ] for any @xmath13 where for @xmath20@xmath21 if some local information around the point @xmath22 is available , then we may state the following result as well @xcite .
[ t4]let @xmath0 \rightarrow \mathbb{r}$ ] be continuous on @xmath7 $ ] and differentiable on @xmath9 let @xmath23 and assume , for a given @xmath13 we have that @xmath24then we have the inequality @xmath25 m_{p}\left ( x\right ) .\end{gathered}\ ] ] for recent results in connection to ostrowski s inequality see the papers @xcite,@xcite and the monograph @xcite .
the main aim of this paper is to point out some generalizations of the results incorporated in theorems [ t2]-[t4 ] by the use of cauchy mean value theorem .
applications for other particular instances of functions are given as well .
we may state the following theorem .
[ t2.1]let @xmath26 \rightarrow \mathbb{r}$ ] be continuous on @xmath7 $ ] and differentiable on @xmath9 if @xmath27 for each @xmath28 and @xmath29then for any @xmath5 $ ] one has the inequality @xmath30 let @xmath31 $ ] with @xmath32 applying cauchy s mean value theorem , there exists a @xmath33 between @xmath34 and @xmath35 such that @xmath36 from where we get @xmath37 for any @xmath38 .$ ] using the properties of the integral , we deduce by ( [ 2.3 ] ) , that @xmath39 since @xmath27 on @xmath40 it follows that either @xmath41 or @xmath42 for any @xmath43 if @xmath41 for all @xmath44 then @xmath45 is strictly monotonic increasing on @xmath1 and @xmath46 if @xmath47 for all @xmath44 then @xmath48\ ] ] and the inequality ( [ 2.2 ] ) is proved .
the following midpoint inequality is a natural consequence of the above result .
[ c2.2]with the above assumptions for @xmath49 and @xmath50 one has the inequality @xmath51 1 .
if in the above theorem , we choose @xmath52 then from ( [ 2.2 ] ) we recapture ostrowski s inequality ( [ 1.1 ] ) .
if in theorem [ t2.1 ] we choose @xmath53 , @xmath54 or @xmath55 with @xmath56 then we obtain theorem [ t2 ] and theorem [ t3 ] respectively .
one may obtain many inequalities from theorem [ t2.1 ] on choosing different instances of functions @xmath57 [ p2.3]let @xmath0 \subset \mathbb{r}\rightarrow \mathbb{r } $ ] be continuous on @xmath7 $ ] and differentiable on @xmath58 if there exists a constant @xmath59 such that @xmath60then one has the inequality : @xmath61\end{gathered}\]]for any @xmath13 where @xmath62 and @xmath63 is the exponential men , i.e. , @xmath64 in particular , we have @xmath65 \gamma . \label{2.8}\ ] ] the proof is obvious by theorem [ t2.1 ] on choosing @xmath66 and we omit the details .
another example is considered in the following proposition .
[ p2.4]let @xmath0 \subset \left ( 0,\frac{\pi } { 2}\right ) \rightarrow \mathbb{r}$ ] be continuous on @xmath7 $ ] and differentiable on @xmath9 1 . if there exists a constant @xmath67 such that @xmath68 then one has the inequality @xmath69\end{gathered}\ ] ] for any @xmath13 where @xmath70 is the cos - mean value , i.e. , @xmath71 in particular we have @xmath72 \gamma _ { 1}. \label{2.11}\ ] ] 2 . if there exists a constant @xmath73 such that @xmath74 then one has the inequality @xmath75 , \end{gathered}\ ] ] for any @xmath13 where @xmath76 is the @xmath77mean value , i.e. , @xmath78 in particular , we have @xmath79 \gamma _ { 2}. \label{2.14}\ ] ] the following result also holds .
[ t2.5]let @xmath26 \rightarrow \mathbb{r}$ ] be continuous on @xmath7 $ ] and differentiable on @xmath80 @xmath81 if @xmath82 for @xmath83 then we have the inequality @xmath84 we obviously have : @xmath85 .
\notag\end{aligned}\]]applying cauchy s mean value theorem on the interval @xmath86 we deduce ( see the proof of theorem [ t2.1 ] ) that @xmath87for any @xmath88 and , similarly @xmath89for any @xmath90 consequently @xmath91 and @xmath92 since @xmath93 has a constant sign in either @xmath94 or @xmath95 it follows that @xmath96 is strictly increasing or strictly decreasing in @xmath94 and @xmath97 thus @xmath98 \\ & \\ \int_{a}^{x}g\left ( t\right ) dt - g\left ( x\right ) \left ( x - a\right ) & \text{if } g\text{\hspace{0.05in}is decreasing}\end{array } \right . \\ & = & \left| g\left ( x\right ) \left ( x - a\right ) -\int_{a}^{x}g\left ( t\right ) dt\right|\end{aligned}\ ] ] and , in a similar way @xmath99 consequently , by the use of ( [ 2.16 ] ) , we deduce the desired inequality ( [ 2.15 ] ) .
the following particular case may be of interest .
[ c2.6]let @xmath26 \rightarrow \mathbb{r}$ ] be continuous on @xmath7 $ ] and differentiable on @xmath100 if @xmath27 on @xmath101 then we have the inequality @xmath102 the following result also holds .
[ p2.7]let @xmath0 \rightarrow \mathbb{r}$ ] be continuous on @xmath7 $ ] and differentiable on @xmath80 @xmath81 assume that , for @xmath103 we have @xmath104 then we have the inequality @xmath105 .\end{gathered}\ ] ] the proof follows by theorem [ t2.5 ] applied for @xmath106 @xmath107 we omit the details .
[ r2.8]if @xmath49 is as in proposition [ p2.7 ] and @xmath108 then , by ( [ 2.21 ] ) , we get @xmath109 .\end{gathered}\ ] ] [ r2.9]if @xmath49 is as in proposition [ p2.7 ] and @xmath110 then , by ( [ 2.21 ] ) , we get @xmath111 m_{p}\left ( x\right ) , \end{gathered}\ ] ] which is the result obtained in ( [ 1.4 ] ) .
1 . let @xmath112 consider the function @xmath113 \rightarrow \mathbb{r}$ ] , @xmath114 @xmath115 then @xmath116 @xmath117@xmath118 and by corollary [ c2.6 ] , we may state the following proposition . + [ p3.1]let @xmath0 \subset \left ( 0,\infty \right ) \rightarrow \mathbb{r}$ ] be continuous on @xmath7 $ ] and differentiable on @xmath119 if @xmath120 then we have the inequality @xmath121 + the particular case @xmath122 is of interest and so we may state the following corollary . +
[ c3.2]let @xmath0 \subset \left ( 0,\infty \right ) \rightarrow \mathbb{r}$ ] be continuous on @xmath7 $ ] and differentiable on @xmath119 if @xmath123 then we have the inequality : @xmath124 \left ( b - a\right ) .\end{gathered}\ ] ] 2 .
let @xmath112 consider the function @xmath113 \rightarrow \mathbb{r}$ ] , @xmath125 then @xmath126 @xmath127@xmath128 and by corollary [ c2.6 ] we may state the following proposition . + [ p3.3]let @xmath0 \subset \left ( 0,\infty \right ) \rightarrow \mathbb{r}$ ] be continuous on @xmath7 $ ] and differentiable on @xmath119 if @xmath129 then we have the inequality : @xmath130 } { l\left ( a , a\left ( a , b\right ) \right ) a\left ( a , b\right ) } \right .
\\ + \left .
m_{2}\left ( \frac{a+b}{2}\right ) \cdot \frac{\left [ l\left ( a\left ( a , b\right ) , b\right ) -a\left ( a , b\right ) \right ] } { l\left ( a\left ( a , b\right ) , b\right ) a\left ( a , b\right ) } \right ] .\end{gathered}\ ] ] 3 .
let @xmath112 consider the function @xmath113 \rightarrow \mathbb{r}$ ] , @xmath131 then @xmath132 , @xmath133@xmath134 and by corollary [ c2.6 ] we may state the following proposition .
+ [ p3.4]let @xmath0 \subset \left ( 0,\infty \right ) \rightarrow \mathbb{r}$ ] be continuous on @xmath7 $ ] and differentiable on @xmath119 if @xmath135 then we have the inequality : @xmath136 ^{m_{1}\left ( \frac{a+b}{2}\right ) } , \left [ \frac{i\left ( a\left ( a , b\right ) , b\right ) } { a\left ( a , b\right ) } \right ] ^{m_{2}\left ( \frac{a+b}{2}\right ) } \right ) \right\ } .\end{gathered}\ ] ]
we may state the following theorem .
[ t4.1]let @xmath26 \rightarrow \mathbb{r}$ ] be continuous on @xmath7 $ ] and differentiable on @xmath1 and @xmath137 \rightarrow \lbrack 0,\infty ) $ ] an integrable function such that @xmath138 if @xmath82 for each @xmath28 and @xmath139then for any @xmath22 one has the inequality @xmath140 if @xmath41 for all @xmath44 then @xmath45 is strictly monotonic increasing on @xmath1 and @xmath144 + \int_{x}^{b}w\left ( t\right ) g\left ( t\right ) dt-\int_{a}^{x}w\left ( t\right ) g\left ( t\right ) dt.\end{aligned}\ ] ] if @xmath47 for all @xmath44 then @xmath145 + \int_{x}^{b}w\left ( t\right ) g\left ( t\right ) dt-\int_{a}^{x}w\left ( t\right ) g\left ( t\right ) dt\right ] , \end{gathered}\]]and the inequality ( [ 2.2 ] ) is proved .
[ t4.3]let @xmath26 \rightarrow \mathbb{r}$ ] be continuous on @xmath7 $ ] and differentiable on @xmath80 @xmath81 if @xmath137 \rightarrow \lbrack 0,\infty ) $ ] is integrable and @xmath150 and @xmath27 for @xmath151 then we have the inequality @xmath152 s.s .
dragomir , some new inequalities of ostrowski type , _ rgmia res . rep .
_ , * 5*(2002 ) , supplement , article 11 , ` [ on line : http://rgmia.vu.edu.au / v5(e).html]`to apper in _ gazette , austral . | some ostrowski type inequalities via cauchy s mean value theorem and applications for certain particular instances of functions are given . |
calculations of the reflected spectrum expected from x ray illumination of an accretion disk around a black hole are extremely important .
the reflected continuum and especially its associated iron k@xmath0 fluorescence line ( see e.g. basko , sunyaev & titarchuk 1974 ; lightman & white 1988 ; george & fabian 1991 ; matt , perola & piro 1991 ) are perhaps the only spectral features expected from optically thick material in the vicinity of a black hole , so can be used to test theoretical models of accretion in strong gravity ( fabian et al .
1989 , dabrowski et al .
1997 ) .
this is especially relevant since there is much current debate on the structure of the accretion flow .
the `` standard '' accretion disk models of shakura & sunyaev ( 1973 : hereafter ss ) are clearly incomplete .
they can produce uv and soft x ray emission , but not the hard x ray power law out to 200 kev .
this is observed to carry a significant fraction of the luminosity in active galactic nuclei ( agn ) and galactic black hole candidates ( gbhc ) in their low / hard state .
something other than the ss disk is required .
the two current candidates are either magnetic reconnection regions above the accretion disk , or that some part of the accretion flow is in an alternative ( non disk ) configuration .
the idea of magnetically powered active regions above a disk ( e.g. galeev , rosner & vaiana 1979 ) gained considerable plausibility from the recent discovery that the accretion disk viscosity is driven by a magnetic dynamo ( balbus & hawley 1991 ) .
buoyancy could then make the magnetic loops rise to the top of the disk , where the optical depth is low so that the energy released in reconnection can not thermalize ( e.g. di matteo 1998 ) .
such active regions could give the hard x ray spectra which are seen from agn and galactic black hole candidates ( haardt , maraschi & ghisellini 1994 ; stern et al .
1995 ; beloborodov 1999a , b ) .
spectral transitions between the low / hard state and the high / soft state in gbhc might then be associated with the disk becoming radiation pressure dominated .
this occurs at a few percent of the eddington accretion rate , roughly as observed in cyg x1 and other gbhcs .
perhaps magnetic buoyancy transfers more energy out of the disk and into a corona in gas dominated disks than in radiation pressure dominated ones ( nayakshin 1999a ) .
current ( though still incomplete ) mhd simulations do show strong magnetic fields in the low density regions above and below the disk ( hawley 2000 ) , although they do not yet carry enough power to reproduce the observed low / hard state ( miller & stone 2000 ) .
the idea of an alternative accretion flow became a serious contender with the rediscovery of another stable solution of the accretion flow equations for mass accretion rates below @xmath1 per cent of eddington .
the standard ss disk model derives the accretion flow structure in the limit where the gravitational potential energy is radiated locally .
this assumption is relaxed in the advection dominated accretion flow ( adaf ) models , so that the energy can be transported radially ( advected ) , as well as radiated locally .
there are two key ( and currently uncheckable ) assumptions underlying this : firstly that the viscosity mechanism gives the accretion energy to the protons ( as opposed to both protons and electrons ) and secondly that the protons transfer this energy to the electrons only via ( rather slow ) electron ion ( coulomb ) collisions .
this gives an accretion flow which is optically thin , geometrically thick and _ hot _ , with typical electron temperatures of order 200 kev as observed ( e.g. narayan & yi 1995 ) .
this model provides a rather natural explanation for the low / hard to high / soft spectral state transitions in gbhcs , as adafs can not exist at high mass accretion rates .
the electron
ion collision rate increases with the density of the flow , and so above a mass accretion rate of @xmath1 per cent of eddington ( as long as the disk viscosity parameter @xmath2 ; esin , mcclintock & narayan 1997 , quataert & narayan 1999 ) the electrons can drain most of the energy from the protons .
advection is then unimportant as the electrons radiate the gravitational energy released locally , and the flow collapses into an ss disk . in the one model
, there is an optically thick disk extending down to the last stable orbit at 3 schwarzschild radii . in the alternative ,
the inner disk is truncated , with the optically thick , cool flow replaced by an optically thin , x ray hot plasma .
the reflected spectrum gives an observational test between these two scenarios .
firstly the amount of reflection should be different .
a disk with active regions above it should subtend a solid angle of roughly @xmath3 , whereas a truncated disk with the x ray hot plasma filling the `` hole '' in the disk will subtend a rather smaller solid angle . secondly
, for the ss disk there is material down to the last stable orbit .
reflection features from the inner disk should then be strongly smeared and skewed by the combination of special and general relativistic effects expected in the vicinity of a black hole ( fabian et al .
this material is not present in the truncated disk models , so they should have narrower lines .
the seminal asca observation of the agn mcg63015 showed that the line is so broad that accretion disk models require that the material extends down to _ at least _ the last stable orbit in a schwarzschild metric ( tanaka et al .
1995 ) , and that it subtends a solid angle of @xmath4 with respect to the x ray source ( lee et al 1999 ) .
this very clear cut result gives strong support for the existence of an ss accretion disk , although other interpretations in terms of optically thick , highly ionized cloudlets close to the black hole rather than a smooth disk flow may be possible ( karas et al 2000 ) .
however , the line and reflected continuum seen from the gbhcs in their low / hard state ( where the x ray continuum bears a remarkable similarity to that from seyfert 1 type agn ) are _ not _ the same as that seen in mcg63015 .
the gbhc have reflection amplitudes which are rather less than expected for a hard x ray source above a disk ( e.g. gierlinski et al . 1997
; done & 1999 ) .
the iron line does not generally show the extreme smearing associated with material within 5 schwarzschild radii , although there is detectable broadening of the line showing that there is cool , optically thick material within @xmath5 schwarzschild radii ( , done & smith 1997 ; 1998 ; 1999 ; done & 1999 ) .
it is also emerging that while some seyfert 1 agn do have extremely broadened iron lines like mcg63015 ( e.g. ngc 3516 ; nandra et al .
1999 ) , others look more like the gbhc in not having the extreme skewed lines expected from the very inner disk ( e.g. ic4329a ; done , madejski & 2000 ) , and that many do not show reflection subtending the expected @xmath3 solid angle ( smith & done 1996 ; zdziarski , lubiski & smith 1999 ) .
can this be interpreted as a strong evidence that the accretion flow is not always given by an optically thick disk ?
such a conclusion is plainly consistent with the data .
perhaps the objects with extreme line profiles have high mass accretion rates , where the adaf s can not exist .
perhaps these are the analogue of the gbhc in their high / soft state ( done & 1999 ; zdziarski et al . 1999 ) .
alternatively , ross , fabian & young ( 1999 ) argue that the lack of strong reflection and extreme relativistic smearing does not necessarily imply the absence of an inner disk , because the reflection signature from the disk can be masked by complex ionization effects . for completely ionized gas , there is no photo electric absorption opacity left , so the reflected spectrum is simply due to electron scattering in the disk atmosphere .
there is then no characteristic iron line or edge feature from the inner disk , so the extreme relativistically smeared components of the line are not present . however , illumination of a disk with density given by the ss models results in a power law distribution of ionization state with radius ( ross & fabian 1993 ; matt , fabian & ross 1993 ) .
this does not fit the data when the inner radii have a high enough ionization to mask their reflected signature ( et al .
, 1998 ; 1999 ; done & 1999 ) .
the key problem is that the observed reflected spectra from the low / hard state gbhc are relatively neutral ( et al . , 1998 ; 1999 ; done & 1999 ) . yet
if the inner disk is extremely ionized , then there should exist a transition region in which the disk is cool enough for h and he like iron to exist , before getting to the low to moderate ionization material at larger radii .
these species give an unmistakable reflected signature , with very strong , highly ionized line and edge features ( lightman & white 1988 ; ross & fabian 1993 ; matt , fabian & ross 1993ab ; et al . , 1994 ; matt , fabian & ross 1996 ; ross , fabian & brandt 1996 ) , which are simply not present in the data .
this leads to the conclusion that either the disk really does subtend a small solid angle to the x ray source , or that its ionization state drops dramatically ( as a step function rather than a power law ) from extreme ionization to relatively neutral ( done & 1999 ) , or / and that the x ray source does not illuminate the inner disk ( e.g. through the anisotropic illumination pattern resulting from trans relativistic outflow from magnetic reconnection : beloborodov 1999a , b ) .
this argument is based on reflection models which neglect the energy change which can result from compton scattering below @xmath6 kev , and assume that the ionization state is constant with depth in the material .
in practice , at high ionization states the electron temperature in the disk can be high enough to cause appreciable comptonization smearing of the resulting reflected spectrum and line .
also , there is vertical ionization structure of the illuminated material , the details of which depend on the vertical density structure of the disk ( ross & fabian 1993 ; et al 1994 ; matt , fabian & ross 1996 ; ross , fabian & young 1999 ) .
ross , fabian & young ( 1999 ) argue that these effects alone are sufficient to hide the otherwise unmistakable signature of highly ionized reflection from h and he like iron , but this has yet to be tested against real data .
however , all these works use an _ assumed _ vertical density structure .
this is a very crude simplification , which can severely misrepresent the reflection signature from the disk ( nayakshin 1999b ; ross , fabian & young 1999 ) .
x ray illumination _
changes _ the density structure of the material .
material at the top of the disk is heated by the illumination , and so can expand , lowering its density .
deeper into the disk there is less x ray heating , so the material is cooler , and hence denser .
the self consistent density structure is especially important as there is a thermal ionization instability which affects x ray illuminated material in pressure balance ( field 1965 ; krolik , mckee & tarter 1981 ; kallman & white 1989 ; ko & kallman 1994 ; r@xmath7aska & czerny 1996 ; r@xmath7aska 1999 ; nayakshin , kazanas & kallman 2000 , hereafter nkk ) .
this can result in the ionization state of the material changing very rapidly , with a highly ionized skin forming on top of a mainly neutral disk .
if the ionization state of the skin is high enough to completely strip iron then it is almost completely reflective , and forms no spectral features .
the observed reflected signature is then dominated by photons reflected from deeper in the disk , where the material is mainly neutral .
although the material is still expected to be less ionized at larger radii , the meaning of less ionized is very different from that in the constant density models the skin remains completely ionized with a negligible amount of h and he like iron , but its thomson depth decreases with radius ( see nayakshin 2000 ) .
the instability gives a physical mechanism for a discontinuous transition in ionization state , as required by the data .
the small observed solid angle subtended by the disk could then be an artifact of fitting single zone ionization models to data from systems with more complex ionization structure .
ideally we would want to reanalyze the data , fitting these x ray illuminated disk reflection models to derive the solid angle subtended by the reflecting material . if these models give a worse fit to the data than the single ionization models , or if they give a derived solid angle of the reflector which is much less than @xmath3 then the `` hole in the disk '' geometry inferred from the single zone ionization models is still favored . if instead the complex ionization reflector models are a better fit to the data and give a solid angle which is compatible with @xmath3 then the ss disk with magnetic coronal x ray source is probably correct , and the previous results which ruled out this geometry are a model dependent interpretation of the data . however , direct data fitting is prohibitively cpu intensive . the models require simultaneous computation of the hydrostatic , ionization and energy balance for the illuminated gas plus optically thick radiation transfer and relativistic effects to determine the reflected spectrum .
a single model calculation requires @xmath8 day of cpu on a 500 mhz workstation .
given this , we have chosen to tackle the inverse problem . instead of fitting real data with the complex ionization reflector code
, we use the complex ionization reflector code to synthesize a spectrum with the same signal to noise and spectral resolution as given from typical _ ginga _ observations of the low / hard state spectra from galactic black hole candidates .
we then fit this model spectrum with the standard ( and computationally very fast ) single zone ionization reflection models which are widely used in x ray spectral analysis .
we can then quantify the effect of using the single zone ionization reflection models to describe a more complex ionization disk structure .
we show that the resulting spectra from a complex ionization structure disk subtending a solid angle of @xmath3 are well fit over the 220 kev range by the single zone reflection models , but that the presence of the ionized skin leads to an underestimate of the covering fraction of the reflector .
the ss disk models then can not be ruled out by the fact that single zone ionization reflector models derive a solid angle of @xmath9 .
the complex ionization structure disks can also produce a correlation between the amount of reflection as measured by the single zone ionization models and the spectral index ( as observed e.g. by zdziarski et al . , 1999 ) _ without _ changing the solid angle subtended by the disk .
we end by discussing how we can observationally distinguish between the truncated disk models and complex ionization structure .
we use the code of nkk to calculate the reflected spectrum from an x ray illuminated disk atmosphere .
this code computes the vertical structure assuming hydrostatic equilibrium , solving simultaneously for the density , ionization structure and energy balance of the illuminated gas , as well as doing the radiative transfer for the continuum and emission lines .
locally , the amount of photo ionization can be described by the parameter , @xmath10 , which is the ratio of the illuminating radiation pressure ( @xmath11 , where @xmath12 is the illuminating hard x ray flux ) to the gas pressure ( @xmath13 acting outwards , where @xmath14 is the hydrogen density and t is the temperature ) .
this relates to the more usual density ionization parameter @xmath15 by @xmath16 . as results of nkk show ,
the vertical density structure of the illuminated layer is rather complex ( see also nayakshin & kallman 2000 ) , as the gas in radiative equilibrium with the illuminating x
rays is subject to a thermal pressure ionization instability ( krolik , mckee & tarter 1981 ) . in the context of the given problem
, the instability can be described as following .
the x ray heating clearly depends on height , with the material at the top of the layer being heated the most . because the material is in pressure balance ,
the gas pressure is smallest at the top of the disk , so @xmath17 is at its maximum here .
this corresponds to the uppermost stable branch of the ionization equilibrium s curve ( see fig . 1 in nkk ) , with low density material which is highly or completely ionized .
atomic cooling is negligible so the temperature is a fraction of the local compton temperature ( see krolik et al . , 1981 and nayakshin 2000a ) ; compton heating is balanced by bremsstrahlung and compton cooling .
going deeper down into the disk , the x ray heating decreases with optical depth to electron scattering , @xmath18 , and the gas pressure increases .
hence , both the temperature and ionization parameter @xmath17 decrease .
however , there is then a turning point on the s curve ( point c in fig . 1 of nkk ) .
at this point the temperature and density is such that the rapidly increasing bremsstrahlung cooling causes the temperature to dramatically decrease .
this pulls the gas pressure down , but hydrostatic equilibrium requires that the gas pressure must monotonically increase as a function of depth into the disk .
the only way for the pressure to increase in a region with rapidly decreasing temperature is to rapidly increase the density , but this pushes up the cooling still further , and so decreases the temperature .
ionic and atomic species can now exist , so line transitions give a yet further increase in the cooling .
eventually this stabilizes onto the bottom part of the s curve , where the x ray heating is balanced primarily by line cooling from low temperature , high density material .
this results in an almost discontinuous transition from a highly ionized , hot and relatively tenuous skin to mainly neutral , cool and relatively dense material over a very small ( @xmath19 see appendix in nkk ) change in optical depth at @xmath20 few .
the optical depth where this instability occurs plus the compton temperature ( which depends on the illuminating power law as well as the local thermal disk spectrum ) are the main determinants of the resulting reflected spectrum .
the majority of the reflected flux arises from the hot layer if its optical depth @xmath21 .
the reflected spectrum then contains the imprints of the ionized material which can be either highly ionized ( where iron is mainly he or h like ) , or extremely ionized ( where even iron is completely stripped of bound electrons ) depending on the equilibrium compton temperature ( nayakshin & kallman 2000 ) . for smaller optical depths , @xmath22 ,
most of the reflection occurs in material below the ionization instability point , so the reflected spectrum contains the signatures of the cool material at low ionization states .
nkk define a dimensionless parameter @xmath23 , which is essentially the ratio between the gravity force and that from the x ray radiation pressure ( @xmath24 ) . where gravity is dominant ( @xmath25 ) then the vertical disk structure is close to that of an un illuminated disk ( assumed here for simplicity to be the gaussian density profile expected from a gas pressure dominated disk : shakura & sunyaev 1973 ) and the optical depth of the hot , ionized layer is very small .
conversely , where @xmath26 then the illumination determines much of the vertical structure , and the optical depth of the hot layer is large .
nkk give a simple analytic approximation to this hot layer optical depth of @xmath27 ( see also kallman & white 1989 ; nayakshin 2000 for more accurate analytical estimates of @xmath18 ) .
given that the observations of flat spectrum agn and low / hard state gbhc seem to show small but not negligible amounts of low to moderate ionization reflection , then the parameters which are most likely to reproduce this are @xmath28 , which corresponds to @xmath29 for hard spectra .
if the compton temperature in the upper , ionized skin is high enough then the reflected spectrum from this will be featureless , and so not detected in x ray spectral fitting .
the observable reflection signature will then be given only by the material below the ionization instability point , where the incoming flux has been decreased by scattering in the hot layer .
this reflected continuum then has to escape out through the hot layer , so scattering again reduces its amplitude and introduces some broadening to the reflected features
. a truly self consistent model should uniquely determine the illuminating flux at the surface of the disk as a function of the system parameters ( mass , radius and mass accretion rate through the disk ) .
since the x ray energy generation mechanism for magnetic flares is not yet understood in detail then such a model is not yet available .
instead we _ assume _ a given x
ray illuminating flux in order to demonstrate the physical behavior of a disk atmosphere irradiated by magnetic flares .
we calculate the angle averaged upward reflected spectra ( corresponding to an inclination of @xmath30 ) resulting from a disk at @xmath31 ( @xmath32 , the schwarzschild radius ) , with an accretion rate of @xmath33 onto a super massive black hole of @xmath34 .
the large black hole mass means that we avoid disk photons potentially contributing to the observed bandpass .
we assume a power law illuminating continuum with exponential cutoff at 200 kev .
the illumination parameter is fixed at @xmath35 , and the resulting angle averaged upward reflected spectra ( so corresponding to an inclination of @xmath30 ) are calculated for photon indices of @xmath36 and @xmath37 .
we do not include any general or special relativistic shifts for simplicity .
figure [ fig : spectr_tempr ] shows the reflected spectra ( without the illuminating power law ) and the gas temperature profiles for the 6 different values of @xmath38 .
we simulate a 10ks observation with the ginga x ray satellite , using the fakeit command in xspec ( arnaud 1998 ) , adding 0.5 per cent systematic errors to the results since it seems unlikely that any satellite will be calibrated to better than this limit .
we compare these simulated spectra with the angle dependent reflection code of magdziarz & zdziarski ( 1995 ) , with ion populations calculated as in done et al .
( 1992 ) , implemented in the pexriv model in xspec .
this calculates the continuum reflection for an input spectrum which is a power law with exponential cutoff . it does not include the associated iron line , so this is put in as a narrow ( intrinsic width fixed at @xmath39 kev ) gaussian line of free normalization , and energy ( except that this latter is constrained to be between 5.5 and 7.5 kev , as expected for iron with some shifts allowed for doppler and gravitational effects ) .
we assume solar abundances ( morrison & mccammon 1983 ) and an inclination of @xmath40 .
we also include a free absorbing column in our fits so as to match as closely as possible the uncertainties which arise in fitting real data .
figure [ fig:2 ] shows the spectral index and amount of reflection inferred from fitting the simulated datasets with the pexriv model , with full results given in table 1 .
there is a clear correlation of the amount of reflection and its ionization state with spectral index . for hard spectra
the ionization of the detected reflected spectrum is low , and the amount of reflection is much less than unity .
this is because the compton temperature of the upper layer of the disk is high , its ionization is extreme and it is thomson thick ( @xmath41 , see fig .
[ fig : spectr_tempr ] ) .
this produces a reflected spectrum which is mostly from electron scattering in the skin , which has few spectral features and can not easily be identified as reflected flux .
the ionization drops dramatically at the depth where the ionization instability sets in , so there is also a small reflected component from low ionization material which is fit by the pexriv model .
however , this itself has to escape out through the skin , and so some fraction is compton scattered , reducing the line and edge features from the low ionization reflected spectrum , but keeping its overall shape . for steeper spectra the compton temperature of the upper layers drops , so that its ionization is not so extreme , leading to strong spectral features from the ionized upper layer , plus the skin is actually thinner ( see fig .
1 ) , and hence most of the reflected flux is easily identified .
figure [ fig:2 ] also shows the detected line equivalent width .
it is generally low , but consistent with the theoretically expected equivalent width for the given ionization state ( assuming that resonant trapping is not important ) and spectral index , scaled by the observed solid angle ( & czerny 1994 ; george & fabian 1991 ) .
lcccccccc @xmath42 & @xmath43 & @xmath44 & @xmath45 & @xmath46 & @xmath47 & @xmath48 & @xmath49 & @xmath50 @xmath51 & @xmath52 & @xmath53 & @xmath54 & @xmath55 & @xmath56 & @xmath57 & @xmath58 & @xmath59 @xmath60 & @xmath61 & @xmath53 & @xmath62 & @xmath63 & @xmath64 & @xmath65 & @xmath66 & @xmath67 @xmath68 & @xmath69 & @xmath70 & @xmath71 & @xmath72 & @xmath73 & @xmath74 & @xmath75 & @xmath76@xmath45 & @xmath77 & @xmath70 & @xmath78 & @xmath79 & @xmath80 & @xmath81 & @xmath82 & @xmath76@xmath37 & @xmath83 & @xmath84 & @xmath85 & @xmath86 & @xmath87 & @xmath74 & @xmath88 & @xmath89 the high energy cutoff is never detected ( see table 1 ) , despite the fact that an exponential roll over temperature at 200 kev gives a 10 per cent effect at 20 kev .
this is probably due to the weak line and edge features compared to the compton hump in the complex models due to compton scattering of the low ionization reflected signature in the skin .
therefore , for a given reflected _ continuum _ , the models of nkk always have less of the absorption edge than the single ionization zone models ( see also ross , fabian & young 1999 who came to similar conclusions for exponential density profiles ) .
hence , when using the single zone models , by fixing the depth of the photo absorption edge ( with the edge energy fixing the ionization state ) , one in fact picks a particular value for @xmath90 already .
this value is too small to reproduce the continuum reflection rise of our models shown in fig .
[ fig : spectr_tempr ] , but the fitting compensates for this by increasing the exponential cutoff energy .
however , we caution that an exponential cutoff is much less sharp than the roll over in a true comptonized spectrum , so this aspect of the modeling need not be reproduced in detail by real illuminated disks .
all the fits are statistically very good , with @xmath91 for 22 degrees of freedom
. however , there should be some observable features from these disk models with structured ionization which distinguish them from simple single ionization state models .
figure 3 shows the total spectrum and reprocessed components from the @xmath92 flat spectrum simulation , together with the best fit model using the pexriv simple ( single zone ) ionization reflected spectrum .
plainly the absorbed power law and its reflected ( pexriv ) component give an excellent fit to the total spectrum derived from the complex ionization state models , but the individual components are not at all well modeled .
the true reflected flux is dramatically underestimated and the derived `` intrinsic '' spectrum is somewhat steeper . these two together would give a rather steeper total spectrum , but this is compensated for by the lack of high energy cutoff in the fitted spectrum . with broad band data , extending significantly above 20 kev then there should be significant residuals from fitting true complex ionization spectra with the pexriv code . even in the 220 kev band
it is plain that there is excess reflected flux at low energies in the stratified ionization model which is not matched by the single zone ionization state code .
this excess is at the 10 per cent level , and should easily be detectable .
we include a broad soft excess component in the fit to mimic this , using the compst ( sunyaev and titarchuk 1980 ) model in xspec .
this leads to a significant reduction in @xmath93 for the flat spectra ( see table 2 ) .
the resulting reduced @xmath93 values are typically very much less than unity because the model is sometimes an adequate fit to the simulated data _ without _ systematic errors ( photon counting errors are of the order of 0.05 per cent ! ) .
error ranges are then derived from @xmath94 , rather than the standard @xmath95 which assumes that the fit has @xmath96 .
several of the parameters are tightly correlated , so error ranges on a single parameter are somewhat misleading .
figure 4 shows the resulting fit to the simulated @xmath92 spectrum .
comparison with figure 3 shows that the broad soft excess component is used to describe the highly ionized reflected flux at low energies , while the fitted reflected flux is still dramatically underestimated .
lcccccccc @xmath42 & @xmath97 & @xmath98 & @xmath99 & @xmath100 & @xmath101 & @xmath102 & @xmath103 & @xmath104 @xmath51 & @xmath105 & @xmath106 & @xmath107 & @xmath108 & @xmath109 & @xmath110 & @xmath111 & @xmath112 @xmath60 & @xmath113 & @xmath114 & @xmath115 & @xmath116 & @xmath117 & @xmath118 & @xmath119 & @xmath120@xmath68 & @xmath121 & @xmath54 & @xmath122 & @xmath123 & @xmath124 & @xmath125 & @xmath126 & @xmath127 @xmath45 & @xmath128 & @xmath129 & @xmath130 & @xmath131 & @xmath132 & @xmath133 & @xmath134 & @xmath135 @xmath37 & @xmath136 & @xmath137 & @xmath138 & @xmath139 & @xmath140 & @xmath141 & @xmath142 & @xmath143
the classic ionization instability can cause a dramatic transition in the vertical structure of an x ray illuminated accretion disk , with an extremely ionized skin overlying low ionization material . with flat spectrum illumination
the compton temperature of the skin is high enough that even iron is completely stripped , so it only scatters the incident power law radiation and does not create any atomic features in the reflected spectrum ( such as emission lines and photo absorption edges ) .
the only observable reflected signature then comes from the material below the ionization instability , which has low to moderate ionization .
the instability suppresses any signature of h and he like iron , which was the key argument used by done & ( 1999 ) against substantial ionization of the disk .
where the optical depth in the ionized skin is @xmath144 then the observed reflected spectrum and line derived from single zone ionization models is suppressed by about a factor 34 , as observed in the flattest agn and gbhc .
thus , the geometry in which the disk covers half the sky as seen from the x ray source ( e.g. , magnetic flares above a disk ) _ can _ be consistent with the observed data because the reflected spectra computed with hydrostatic balance models _ appear _ to have a small reflection fraction when fitted with the current single zone models .
however , more detailed studies of this issue , with radial stratification of the skin properties , and with effects of possible x ray induced evaporation ( which is important for magnetic flares , see nkk ) is needed to resolve the issue truly self consistently .
we estimate that thomson depth of the skin should be in the range of unity to few to explain spectra typical of hard state of gbhcs . at first glance
, this seems to be a rather narrow range in parameter space , making the model appear contrived .
however , the dependence of @xmath18 on @xmath12 becomes very weak once @xmath145 , because the incident x
rays are efficiently down scattered in the skin . as discussed by kallman & white ( 1989 ) and nayakshin ( 2000a )
, it is then very difficult to produce @xmath146 a few even with arbitrarily large values of @xmath12 .
therefore , the range of `` acceptable '' values of thomson depths , @xmath147few , may actually correspond to a change in the x ray ionizing flux by several orders of magnitude , and hence the model may be quite robust in the parameter space .
we find a strong apparent correlation between the inferred intrinsic spectral index and the amount of reflection and its ionization ( see fig
. 2 and table 1 ) . the gas pressure at which the transition from the hot to the cold material occurs , @xmath148 , scales with the compton temperature as @xmath149 ( see eq . 4 in nayakshin 2000a ) .
therefore , the cold layer lies closer to the top of the skin for larger @xmath38 for a given illuminating flux . in our numerical simulations the thomson depth of the skin changes from @xmath150 for @xmath151 to @xmath152 for @xmath153 ( see fig .
the thicker the skin , the more it masks the cold material below it ( e.g. , see nkk ) , and so the correlation between @xmath38 and @xmath154 that we found in this paper is to be expected .
the correlation of the inferred ionization state ( @xmath155 ) with @xmath38 comes about as a consequence of changes in @xmath18 and the actual ionization state of the skin with changes in the spectral index . in particular , when @xmath38 increases , the compton temperature , and hence the skin s temperature , decrease ( see nayakshin & kallman 2000 on this ) . for @xmath156 ,
the skin is highly ionized , rather than extremely ionized , and so it produces strong features from h and he like iron . however , its optical depth is not very large for the parameters chosen in this paper ( see fig .
1 ) . hence
, there is a substantial contribution to the reflected spectrum from the low to moderate ionization material under the ionized skin .
the `` observed '' ionization state is a compromise between the optically thin , ionized skin and the low ionization material below , which then leads to the inferred correlation of @xmath155 and @xmath38 .
it is also interesting to note that a larger thomson depth of the ionized skin leads to larger integrated x ray albedo , @xmath157 .
the thermal flux resulting from reprocessing of the incident x ray flux is @xmath158 .
thus , the thicker the ionized skin , the less reprocessed radiation will emerge , which means that compton cooling of the magnetic active regions above the disk will be reduced .
this is why nayakshin ( 1999b ) and nayakshin & dove ( 2000 ) find that larger values of @xmath18 lead to hotter coronae above the disk , i.e. , smaller values of @xmath38 if the optical depth of the corona is fixed .
in other words , a complete self consistent calculation , where @xmath38 is not a fixed parameter , but it is rather found via balancing energy heating and cooling in the corona would produce _ even stronger _
apparent correlation between @xmath38 and @xmath154 than the one we found here .
these findings appear to be very relevant for the recently found correlations between @xmath38 and @xmath154 for a sample of agn and gbhcs , as well as for some individual sources ( e.g. , zdziarski , et al .
1999 ) .
note also that the correlation between @xmath155 and @xmath38 inferred in our models suggest that sources with steeper x ray spectra should produce spectra that _ appear _ to be more ionized than those of sources with harder spectra .
this statement is only correct in the statistical sense , though .
namely , if for every @xmath38 there is a range of illumination parameters @xmath23 , then at a given @xmath23 , there will be a correlation of @xmath155 and @xmath38 similar to that shown in fig .
1 . folded with a range of @xmath23 for every @xmath38 , one still expects to see a similar correlation , probably somewhat weaker than the one we found here .
this theoretical expectation does indeed appear to match the behavior of the ( steep spectrum ) narrow line seyfert 1 s , which have a large amount of ionized reflection ( nls1 pounds , done & osborne 1995 ; vaughan et al .
this supports the picture where the difference between nls1s and normal seyfert 1s is that nls1s accrete at a high fraction of the eddington limit ( pounds et al .
1995 ) , as the steep spectrum gives a low compton temperature and so reflection from highly ionized as opposed to extremely ionized material .
however , this correlation does _ not _ hold for mcg63015 , which has a steep spectrum together with a large amount of apparently _ neutral _ reflection ( lee et al 1999 ; zdziarski et al 1999 ) , although the variability of the line may indicate that ionization effects are important ( reynolds 2000 ) .
clearly the models used here are incomplete .
the optical depth of the ionized layer changes as a function of radius , and relativistic smearing effects should also be included .
however , it is clearly necessary to develop a basic understanding of how the complex ionization models differ from simplified single zone ones for a specific radius with no relativistic effects .
having understood that , one can move to the reflected spectra for complete disks , which we plan to do in future work . based on physical intuition and some preliminary work
, we can speculate on the full disk spectra .
nkk and nayakshin & kallman ( 2000 ) show that if the illuminating x ray flux , @xmath159 , exceeds that of the locally produced disk thermal emission , @xmath160 , and the x ray spectrum is relatively hard , i.e. , @xmath161 , then the skin is nearly completely ionized . the depth of the completely ionized skin decreases with increasing radius in the disk ( see nayakshin 2000a ) .
hence , larger radii contribute relatively more of the atomic features ( e.g. , k@xmath0 line and iron edges ) to the overall spectrum , while smaller radii contribute less . if @xmath162 in the inner disk then the extremely relativistic reflected signature will be absent from the data . a low ionization reflection signature is only seen from radii larger than the point at which @xmath163 .
this would suppress the broadest components of the iron line , so could make the derived line profile look _ qualitatively _ like that from a truncated disk ( see fig .
2b in nayakshin 2000b ) .
one still hopes that detailed and direct spectral fitting , combined with future more sensitive observations , will allow us to discriminate between the physically truncated and completely ionized disks .
the key element in the complex ionization models is that for hard x ray spectra the ionization instability results in a sharp discontinuity between the extremely ionized ( invisible ) skin and the low ionization material which gives a clearly identifiable reflection signature . given further complications that can arise due to relativistic broadening in the disk , it is quite likely that the _ full _ x ray illuminated disks have a reflected continuum which can mimic that of a truncated disk .
can we then distinguish between these two models ?
we have shown that there are _ observational _ tests .
firstly broad bandpass , high signal to noise data should show the deficiencies in the pexriv model fit ( see figure 3 ) .
such data currently exist only for the galactic black hole candidates , with cyg x1 having the best determined spectrum ( gierlinski et al .
1997 ; di salvo et al . 2000 ) .
these clearly show that the reflected spectrum from single ionization state material illuminated by a comptonized continuum is insufficient to describe the observed spectral curvature .
further , even in the 220 kev bandpass there is an observable difference between the single zone and complex ionization reflection models .
the complex ionization disks predict the existence of a highly ionized reflected skin which contributes to the spectrum at low energies .
physically , since there is no photo absorption opacity left in the skin , photons of all energies are scattered ( reflected ) rather than being absorbed in the skin .
the low ionization reflection component has negligible flux at low photon energies ( see , e.g. , basko et al . 1974 , lightman & white 1988 ) , so that spectrum from complex ionization models yields a soft excess as compared with the single
zone models , which can be modeled by a broad spectral component at low energies .
an additional and significant source of the soft x
ray excess is the bremsstrahlung emission in the skin , and compton down scattered reflected flux . according to the analytical theory of the compton heated completely ionized skin ( krolik et al .
1981 ; nayakshin 2000 ) , cooling due to bremsstrahlung emission constitutes between @xmath164 zero ( at the top of the skin ) and @xmath165 of the total cooling rate ( on the bottom ) .
the characteristic temperature at which the compton heated branch ceases to exist is of order a few hundred ev to a few kev ( see fig .
1 ) , so that the soft x ray excess will be rolling over at similar energies , consistent with our results here . there is a considerable evidence for a similar soft excess in the galactic black hole candidates .
cyg x1 plainly shows such a component in asca ( ebisawa et al . , 1996 ) and
sax ( di salvo et al . , 2000 ) data , and this is probably the reason why the rxte results include a blackbody at @xmath144 kev rather than a reflected component ( dove et al 1997 ) .
figure 5a shows the pca rxte data of cyg x1 ( dataset 10241 , as used by dove et al
. , 1997 ) fit to a simple single zone ionization model ( @xmath166 ) , while figure 5b shows the resulting fit including soft component ( @xmath167 ) .
the spectrum is much better described with the inclusion of the soft component .
the resulting spectral curvature is detected at @xmath168 per cent confidence on an @xmath169 test , and these real data look remarkably like the synthesized model spectra shown in figure 4 .
clearly this can be interpreted as strong evidence for the existence of complex ionization structure in the accretion disk .
however , this is not unambiguous as the gbhc have other ways to produce spectral curvature since the seed photons from the disk are so close to the energy range observed .
a truncated disk with hot inner flow produces a comptonized spectrum where the first few scattering orders are anisotropic , since the illuminating disk photons are anisotropic .
this produces a soft excess of about 10 per cent of the isotropic spectrum , i.e. similar in size to that expected from the complex ionization reflected spectrum ( p.t .
, private communication ) .
additionally , the transition between the cool truncated disk and hot inner flow is unlikely to be completely sharp , and there may be a further soft component from heated disk material .
these would both affect the gbhc spectra more than those from agn , so perhaps the true test of whether the disk has complex ionization structure or is truncated is whether the agn with flat spectra are significantly better fit by including a soft excess component
. however , even with the agn there are ambiguities .
one of the most promising models for the origin of the variability behavior of agn and gbhc has spectral evolution of the comptonized component from individual flares .
this results in a time averaged spectrum which is significantly curved , similar to the curvature seen here in cyg x1 ( poutanen & fabian 1999 ) .
we caution that deconvolving the reflected spectrum from more realistic , complex continuum models will have to be done before we can observationally test whether the truncated disk or complex ionization models give a better description of the data .
in this paper , we have shown that reflected x ray spectra from models which include hydrostatic balance for the illuminated gas have very different properties than single zone ionization models . in the former models , due to the thermal ionization instability , the vertical disk structure is almost discontinuous , with a dramatic transition from an extremely ionized skin on the top to a low to moderate ionization material below it .
the reflection signature is then a composite of a power law like spectrum from the extremely ionized material , and a low to moderate ionization reflection siganture from the material below the skin .
the constant density reflection models do not have such a discontinuous behavior . as a result , when the spectra of hydrostatic balance models are fit with simpler , single ionization zone models ( done et al .
1992 ; magdziarz & zdziarski 1995 ) , the physical parameters of the reflector can be miscalculated by a large factor .
the single zone models fit to the low to moderate reflection component , which is weak even with respect to the incident power law due to scattering in the ionized skin .
but the ionized skin reflection can not be reproduced in the single zone models , so this power law like reflected component is misidentified with the incident power law .
since the reflected fraction @xmath170 is proportional to the ratio of the inferred reflected spectrum to that of the illuminating , then this further dilutes the `` observed '' reflected fraction .
this implies that the values of @xmath171 often seen in gbhc and agn may be an artifact of the constant density reflection models used , and need not necessarily mean that the cold ss disk is truncated at some distance @xmath172 .
further , our results also imply that a low value of ionization parameter ( and `` neutral '' line at 6.4 kev ) , as found by the single zone ionization models , _ does not _
prove that the material is actually cold .
on the contrary , the upper layer of the disk may be completely ionized and obscure the cold material below .
these findings are in agreement with earlier suggestions of nayakshin ( 1999b ) and ross , fabian & young ( 1999 ) . in addition
, we find an apparent correlation between the power law index @xmath38 of the x ray continuum and the reflection fraction .
this correlation is similar to the one reported by zdziarski et al . , ( 1999 ) for agn and gbhc .
we produce the correlation using only the physical nature of the ionization instability for x ray illuminated gas in hydrostatic balance .
this should operate in _ all _ models where the accretion disk is illuminated by an x ray source .
in particular , other ways to produce the correlation of the amount of reflection with @xmath38 such as a moving inner disk edge ( e.g. poutanen 1998 ) or with anisotropic illumination by magnetic flares ( beloborodov 1999a , b ) will produce even stronger correlations with the inclusion of the ionization physics .
the _ only _ situation where the instability is unimportant for the observed reflected spectrum is if the optical depth of the hot layer is always very much less than unity .
however , one clear prediction of the ionization instability is that the steep spectrum agn should have strongly ionized reflected spectra , which does indeed seem to be the case for the narrow line seyfert 1 class of agn ( although not for the archetypal extreme relativistic disk object mcg63015 ) .
thus it seems likely that the ionization instability _ is _ important in determining the observed reflected spectrum , although it may operate in conjunction with anisotropic illumination or a moving disk edge .
we show that reflection from the ionized skin produces a soft x
ray continuum which can be modeled as a broad soft x
ray excess , similar to that seen in data from cyg x1 . taken together ,
our findings revive models for magnetic flares above an untruncated accretion disk in agn and gbhcs which seemed to be ruled out by observations of the small solid angle of the reflector ( see , e.g. , the discussion in done & 1999 ) .
it seems likely that radial gradients in the ionization structure should be able to suppress the extreme relativistic iron line profile from the inner disk , and so lead to the observed apparent truncation of the disk in the low / hard state gbhc and in some agn .
however , this should be confirmed by further detailed modeling involving complete disk spectra and relativistic smearing .
finally , another difficulty in unambiguously determining between complex ionization structure on a disk which extends down to the last stable orbit , and a truncated disk , is that the underlying illuminating continuum shape is not well understood .
continuum curvature can mimic the `` soft excess '' signature of complex ionization structure , and is predicted in the truncated disk geometry , and by models of the variability .
sadly , the observational constraints on the geometry of the accretion flow around the black hole in cyg x1 still can not distinguish between disk / coronae and adaf models for the x ray emission .
on the other hand , new data and complete disk calculations may present better constraints , allowing these very different geometries ( and accretion theories ) to be distinguished .
sn acknowledges support from nrc research associateship .
the authors thank demos kazanas , piotr and andrei beloborodov for useful discussions . | reflection of x rays from cool material around a black hole is one of the few observational diagnostics of the accretion flow geometry .
models of this reflected spectrum generally assume that the accretion disk can be characterized by material in a single ionization state .
however , several authors have recently stressed the importance of the classic ionization instability for x ray irradiated gas in hydrostatic balance .
this instability leads to a _ discontinuous _ transition in the vertical structure of the disk , resulting in a hot ionized skin above much cooler material .
if the compton temperature of the skin is high then even iron is completely ionized , and the skin does not produce any spectral features .
these new models , where the ionization structure of the disk is calculated self consistently , require an excessive amount of computing power and so are difficult to use in directly fitting observed x ray spectra .
instead , we invert the problem by simulating x ray spectra produced by the new reflection models , and then fit these with the old , single zone reflection models , to assess the extent to which the derived accretion geometry depends on the reflection model used .
we find that the single zone ionization models can severely underestimate the covering fraction of the `` cold '' material as seen from the x ray source if the optical depth in the ionized skin is of order unity , and that this can produce an apparent correlation between the covering fraction and the x ray spectral index similar in nature to that reported by zdziarski , lubiski and smith ( 1999 ) .
accretion , accretion disks radiative transfer line : formation x rays : general radiation mechanisms : non - thermal |
the dwarf spheroidal galaxies ( dsphs ) are the least luminous galaxies known .
they display remarkably high mass - to - light ratios ( @xmath4100 to 1000 ) , and the stars in each system are known to reside at the centre of a massive dark halo ( @xmath5 ) which extends far beyond the observed limiting radii @xcite . in terms of stellar population , these systems can have surprisingly complex star formation histories ( sfhs ) . all dsphs contain a population of ancient stars @xcite , however some ( such as fornax ) have been able to maintain multiple epochs of star formation and chemical enrichment over a hubble time .
these systems are relatively simple environments compared to larger galaxies , and are therefore a starting point in the study of star formation and enrichment .
simulations suggest a cyclical process , in which the gas collapses to form stars , is then chemically enriched and blown out by pockets of massive star formation , and then collapses again to repeat the cycle . @xcite
propose a time frame of @xmath4250 myr for a single cycle .
there are , however , open questions regarding star formation in dwarf galaxies .
population gradients suggest that the most recent bursts of star formation in each dsph were concentrated towards the object s centre @xcite .
also , the milky way halo contains a population of extremely metal - poor stars , with abundances of [ fe / h ] @xmath6 @xcite , whereas the dsphs do not contain a population with metallicities below [ fe / h ] @xmath7 @xcite .
this suggests that the gas sourcing the first generation of stars in these systems was pre - enriched .
moreover , @xcite have argued that the variety of sfhs in dsphs is not the result of reionization , and hence ` local processes ' have influenced star formation in each object .
these include the regulation of gas dynamics due to internal feedback @xcite while ram pressure stripping and tidal interaction with the milky way are also important factors @xcite . with an integrated absolute magnitude of @xmath8 @xcite , fornax is the brightest dsph associated with the milky way ( excluding the tidally dirsupting sagittarius system ) .
it lies at a distance of 138 kpc , and proper motion measurements indicate its orbit around the milky way is roughly circular with an eccentricity of @xmath9 and an orbital period of @xmath10 gyr @xcite .
@xcite have completed a large kinematic survey of this system , collecting radial velocities for 206 member stars over the entire surface of fornax .
they measured a flat velocity dispersion profile , indicating a significant dark component , and find a mass - to - light ratio of @xmath11 within 1.5 kpc ( approximately half the tidal radius ) .
compared to other dsphs , the star formation history of fornax is unusually complex .
all dsphs contain some number of ancient stars @xcite , however early studies of fornax revealed an extended giant branch , including carbon stars , suggesting a strong intermediate - age component @xcite .
additionally , fornax contains a significant young stellar component .
originally discovered by @xcite , the photometry of @xcite and @xcite later identified main sequence stars with ages as young as 100200 myr .
the analysis of @xcite indicates a burst of star formation in the centre of fornax 12 gyr ago , which has continued almost to the present day .
indeed , more than half the stars on the rgb are thought to be younger than 4 gyr @xcite .
of all the dsphs , fornax has experienced the most recent star formation .
fornax also contains population gradients , in which the young stars are preferentially located towards the centre @xcite .
this property is common in the dsph population @xcite and suggests that the gas required for subsequent star formation episodes was more successfully retained in the core of the dark halo than in the outer regions . in fornax
, the young component is not aligned with the main body and is highly structured , including a shell - like feature which may indicate an accretion event @xmath42 gyr ago @xcite .
in addition to an age spread , the stars in this system cover a significant range in metallicity .
@xcite , @xcite and @xcite have examined the chemical abundances of fornax red giants using spectra of the ca ii triplet features , and found a metallicity range of @xmath12 [ fe / h ] @xmath13 .
the data set of battaglia et al . contained enough fornax members ( 562 ) such that stellar metallicities could be accurately related to kinematics .
they found that the metal - rich stars have a colder velocity dispersion , and the metal - poor component shows signs of non - equilibrium kinematics towards the centre of fornax ( @xmath14 ) .
additionally , high resolution spectra of 81 red giants in the centre of fornax indicate that s - process elements are unusually strong , hence stellar winds ( such as those from agb stars ) have dominated the chemical enrichment of fornax in the last 24 gyr @xcite . despite the progress described above
, fornax has been lacking a homogeneous study of its sfh over the entire surface of the system .
the _ hst _
photometry of @xcite confirmed the presence of young stars at the centre of fornax and included evidence of seperate bursts of star formation . by combining these results with ca ii triplet metallicities
, @xcite created _ schematic _ star formation and chemical enrichment histories .
the preliminary sfh produced by @xcite was based on vlt / fors1 photometry ( depth @xmath15 ) also located at the centre of fornax , and another field approximately one core radius from the centre .
young stars were present in both fields , however the authors noted a significant difference in sfh between the two . in summary , although the general trend of star formation at the _ centre _ of fornax is known , the aggregate history is yet to be determined .
hence , we present the first results of a deep , homogeneous photometric data set over the entire surface of fornax .
we have extracted the sfh from this photometry using the cmd - fitting techniques developed by @xcite .
not only has this allowed us to derive a global sfh for this system , we have also examined the sfh as a function of position to search for population gradients .
our data have a limiting magnitude of @xmath16 , thus we are sensitive to main sequence stars in fornax with an age of 3 gyr and less . also , the red giant branch , red clump and horizontal branch stars allowed us to track the ages and metallicities of the populations with ages @xmath17 gyr .
this is the first complete sfh of this system derived from deep photometry .
images encompassing the surface of the fornax dsph were obtained using the eso / mpg 2.2 m telescope equipped with the wide - field imager .
this instrument provides a @xmath18 field of view using a @xmath19 mosaic of @xmath20 pix@xmath21 ccds and a pixel resolution of @xmath22 pix@xmath23 .
the survey aim is to obtain photometry to 23rd magnitude over the entire body of fornax .
the first stage presented here contains fornax itself and the outer shell noted by @xcite . thus far , 21 pointings have been obtained , covering a sky area of @xmath24 deg@xmath21 .
an overlap region of @xmath4@xmath25 between each field was chosen to ensure the photometric zeropoint was constant across the survey . a schematic diagram of the fields is shown in fig . [ obsmap ] .
we obtained three 600s dithered exposures in both @xmath26 and @xmath27 for all fields , allowing us to reach magnitudes of @xmath28 and @xmath29 ( @xmath30 completeness limits ) in all fields .
the images were taken during 11 nights in oct / nov 2006 in median seeing conditions of @xmath31 ( range @xmath32 ) .
the data were reduced using standard procedures in the _ mscred _ package in iraf : the overscan region and the bias frames were used to subtract the pedastal current from each science image , which were then trimmed .
twilight flat fields were combined to produce a master flat frame in @xmath26 and @xmath27 for each night , which were then used to flat - field the science images .
three dithered images were taken in each field , and we used routines in the _ mscred _ package to combine the images and remove gaps between the ccds @xcite .
an astrometric solution was constructed for all images by matching them to the first usno ccd astrograph catalogue ( ucac1 ) , which has an average precision of 31 mas in the magnitude range @xmath33 @xcite .
rms _ of our solution was @xmath34@xmath35 in all fields .
the individual ccd images were then combined using the _ mscimage _ routine to produce three single ( i.e. non - mosaic ) images for each field in both filters .
these three images were then median combined with the _ mscstack _ routine , which matches images based on their astrometry and removes the ccd gaps . finally ,
the combined image for each field was then corrected for zeropoint gradients across the field of view using the _ mscskysub _ routine .
photometry was derived using daophot @xcite .
we measured the background level in each field to calculate a standard deviation of the sky , @xmath36 .
each image was then searched for all sources @xmath37 above the background level , and aperture photometry was used to estimate their brightnesses . in a crowded field such as the centre of fornax ,
the psf - fitting technique in daophot provides a more accurate measure of stellar magnitudes compared to aperture photometry , as it allows the signal of adjacent stars to be disentangled .
hence , we examined the brightest 60 stars in each image and used those with no apparent neighbours and a well - defined gaussian shape to construct a master psf for each science frame .
this was then fitted to every source in the image using a fitting radius of 8 pixels ( @xmath38 ) or @xmath39 half - width , half - maxima of the psf ( depending on seeing ) . to measure the completeness and photometric accuracy of our photometry
, we performed artificial star tests on all science frames in both filters ( i.e. 42 science images ) .
we placed 1600 artificial stars in the image and attempted to recover them with daophot , where the photometric uncertainty was then determined as the dispersion of the returned magnitudes about the mean ( that is , not the input ) .
this was repeated for artificial stars at every 0.25 magnitudes in the @xmath26 and @xmath27 frames . to ensure a constant zeropoint across the survey , we matched stars in the overlap regions between fields using their astrometry and measured the mean inter - field difference in @xmath26 and @xmath27 .
this is the same technique used in our previous fornax survey @xcite .
the inter - field corrections were accurate to @xmath4@xmath40 mag in both @xmath26 and @xmath27 .
as an example , the final match between the fields f6 and f11 is shown in fig .
[ f11f6 ] .
a final zeropoint correction was made by matching our data to the catalogue of @xcite , which contains photometry of the core of fornax in @xmath26 and @xmath27 to a similar limiting magnitude as our survey .
the @xmath27 filter attached to the 2.2m / wfi is a standard cousins filter , and the @xmath27-band photometry was well matched between the two datasets .
in contrast , the @xmath26 filter ( bb#b/123_eso878 ) covers a larger wavelength range than the standard johnson filter , and therefore requires a colour correction to match standard @xmath26 magnitudes .
the eso website provides a correction for the @xmath26-band photometry using the @xmath41 colour ; we determined a @xmath42 colour correction by comparing our data to the stetson et al .
catalogue , yielding the following result , @xmath43 with a bootstrap error of 0.02 mag .
the overall photometric zeropoints are accurate to 0.03 mag .
the colour magnitude diagram ( cmd ) of fornax reinforces the complex star formation history ( sfh ) of this object : it contains multiple stellar populations with a vast range in age and chemical abundance .
a full description of the stellar populations at the centre of fornax is given by @xcite and @xcite .
however , the outer regions of fornax are not well known , and in fig .
[ cmd ] we present the first deep cmds for the entirety of fornax .
it has been theorised that fornax may contain strong radial population gradients ( e.g. @xcite ) , hence we have divided the data set into the four elliptical regions shown in fig .
[ obsmap ] .
the regional boundaries are at radii of @xmath44 , @xmath45 , @xmath46 and @xmath47 , where @xmath48 and @xmath49 are the core and tidal radii listed by @xcite .
the distribution of young stars in fornax is not aligned with the system s major axis and it is also known to contain strong asymmetries ( e.g. @xcite ) , however the majority of these stars are contained in the core region ( this will be shown in the next paper in this series ; coleman , in prep . ) and hence the assumption of elliptical regions is not vital for this sub - population .
region 1 ( the core region ) contains approximately 28,000 sources . moving outwards ,
another 41,000 stars were selected in region 2 , 31,000 in region 3 , and 51,000 in region 4 . for convenience
, we have also performed a background subtraction on each cmd , removing the field population using the ` signal - to - noise ' technique provided by @xcite .
a full description of this technique as applied to the current data set will be given in the next paper in this series .
in summary , we have divided the cmd into a grid of cells , counted the number of fornax and field stars in each cell , and then used this to remove the field contamination ( mostly foreground milky way stars ) .
the results are shown in the lower panels of fig .
[ cmd ] .
the central cmd ( upper left panel ) shows all sources in region 1 , drawn from within the inner red ellipse shown in fig .
[ obsmap ] .
immediately visible is the red giant branch extending downwards from @xmath50 , which contains old and intermediate age stars ( age @xmath51 gyr ) and is thickened due to the age and metallicity spread .
the most densely packed feature is the red clump ( rc ) , centred at @xmath52 , @xmath53 .
this is made up of core helium burning stars , and is essentially a young - to - intermediate age , metal - rich horizontal branch ( hb ) .
a hint of a hb extending towards the blue is also visible .
we also note that the hb appears to extend redward from the rc , however these objects are artefacts of the observational dithering pattern : a cmd - selection indicated that they lie pre - dominantly in regions in which only a single frame of @xmath26-band data was available .
this effect is most prominent in the hb star - rich central four fields , where a few of the hb stars display large photometry errors in the @xmath26 filter ( and hence an artificial colour spread in the hb itself is created ) .
although it is a small artefact ( it exists in less than @xmath54 of the surveyed area ) , its effect on our star formation history results is discussed further in [ sec : abvar ] .
returning to our discussion of the central region cmd , a slight overdensity lies approximately @xmath55 mag above the rc , identified as the colour - magitude clumping at the start of the agb cycle @xcite . below the rc
lies the sub - giant branch , which is also thickened by the age and metallicity spread .
finally , the blue column of stars extending downwards from @xmath56 , @xmath57 is the young main sequence .
@xcite discovered this feature in fornax , and @xcite subsequently identified main sequence stars as young as 200 myr .
indeed , of all the milky way dsphs , fornax has the most recent star formation . at first glance ,
the thick red giant and sub - giant branches are common to all four regions , indicating that the wide range of stellar ages and metallicities continues well beyond the core radius .
however , the differences encountered in the stellar population when moving outwards from the centre of fornax are remarkable .
we see a decrease in the prominence of the young main sequence , possibly indicating that recent star formation ( i.e. less than 4 gyr ago ) was preferentially located towards the centre of fornax .
another clear difference between the four regions lies in the morphology of the hb .
the red clump is present in all four regions , however the hb extends further into the blue region as we move outwards ; it terminates at @xmath58 in the region 4 cmd .
this indicates that the outer regions of fornax contain a significant ( if not dominant ) population of old , metal - poor stars .
further emphasis of this point is provided by the red giant branch ( rgb ) : the lower panels of fig .
[ cmd ] show that as we move outwards , the rgb shifts towards the blue , thus indicating a decrease in mean metallicity with increasing radius .
overall , a comparison of the four cmds shown in fig . [ cmd ] would suggest that later bursts of star formation and chemical enrichment were preferentially located towards the centre of fornax .
numerical fitting of cmds allows a study of the sfh of a dwarf galaxy ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
moreover , the large mosaic presented in this paper provides an opportunity to study the spatial variation of the sfh within fornax . to obtain a detailed picture of the sfh we use the cmd - fitting software match @xcite , which applies maximum - likelihood methods to fit photometric data with simple model cmds . by converting data and models to so - called hess - diagrams ( 2-d histograms of the stellar density as a function of colour and magnitude ; @xcite ) a direct pixel - by - pixel comparison is possible .
the model cmds are based on theoretical isochrones from @xcite and include realistic photometric errors and completeness , which are obtained from the artificial star tests described earlier . by determining the best - fitting linear combination of model cmds for different age and metallicity bins ,
the sfh and metallicity evolution are then constrained .
the accuracy of the recovered metallicities depends not only on the data quality , but also on the quality of the isochrones and the stellar evolution tracks on which they are based .
@xcite tested match on a set of six globular clusters with varying metallicities using isochrones based on the same stellar evolution tracks from @xcite .
they show that the recovered metallicities are always within 0.2 dex of the spectroscopic values . in all results presented in this paper
we therefore include an additional contribution to the metallicity uncertainties of 0.2 dex .
since there are variations in seeing and sky brightness between the different fields in the mosaic , the sfh fits are done separately for each field .
furthermore , the four radial bins are treated separately to enable an analysis of the radial variation of the stellar populations in fornax .
the main free parameters in cmd fitting are distance , age , metallicity and extinction , although the binary fraction and the assumed initial mass function ( imf ) also play a role . to limit the number of free parameters to the age and metallicity
, we assume reasonable priors on the others parameters . for all our fits we assume a binary fraction of 0.5 and a salpeter imf @xcite , which to the cmd depth probed here is practically equal to , for example , a kroupa imf @xcite . during the past decade
, several different studies have all found consistent distances to fornax of @xmath59 kpc @xcite , justifying a prior on the distance in our sfh fits . to account for the small uncertainty in the distance
, we perform all fits for three fixed distances , namely 135 , 138 and 141 kpc . according to the dust extinction maps from @xcite the foreground reddening towards fornax varies between @xmath60 and @xmath61 mag .
fits are performed for different combinations of foreground and internal extinctions , with the former fixed at @xmath60 , 0.02 , 0.025 and 0.03 mag , and the latter at @xmath62 , 0.1 , 0.2 , 0.3 and 0.4 mag . for each individual fit , the distance and
two kinds of extinction are fixed , while the star formation rates ( sfrs ) for the age and metallicity bins serve as the free parameters .
since the isochrones are spaced more evenly in log(@xmath63 ) than in @xmath63 , the age bins are defined in log(@xmath63 ) .
they have bin - widths of @xmath64 , with the oldest bin corresponding to ages between 11 and 16 gyr and the youngest to 10 and 16 myr , for a total of 21 age bins . the full metallicity range spanned by the isochrones , [ fe / h ] @xmath65 to 0.0 is covered with 16 bins of 0.15 dex width . to account for contamination by foreground stars and faint background galaxies , a control field cmd
is created from all star - like sources outside the limiting radius of fornax , and used as an additional model cmd in the fits .
all stars with @xmath66 and @xmath67 and @xmath68 are fit , using hess - diagram bin sizes of 0.16 in magnitude and 0.08 in colour . figs .
[ fig : f15_bin1_fit ] and [ fig : f15_bin3_fit ] show the hess diagrams of regions 1 and 3 recovered from field 15 with the corresponding best - fit models and residuals .
in general the fits are good , but some systematic problems arise in reproducing the exact shape of the hb / rc and its extent towards the red .
this imperfect modeling of the hb / rc region is a general problem for theoretical isochrones , due to the complicated processes taking place during this phase in stellar evolution .
newer isochrone sets than the one currrently used by match @xcite should improve this situation , but our current results are not strongly affected by these problems . however , as the details of the theoretical isochrones used do influence the exact values of metallicities and ages we recover , some extra uncertainty should be taken into account when interpreting the fit results .
after running the 60 fits ( 3 distances , 4 foreground and 5 internal extinction values ) for each region , the values of the maximum - likelihood measure of goodness - of - fit @xmath69 ( see * ? ? ?
* ; * ? ? ?
* ) of all fits are compared . the expected random ( i.e. not due to actual sfh differences ) variance in this @xmath69 parameter , @xmath36 , when fitting a specific cmd ,
is calculated using poisson statistics and from monte carlo simulations using random drawings from the best model .
all fits that have a value of @xmath69 within 1@xmath36 of the best fit are considered as ` good ' fits and used to construct the sfhs presented in the remainder of this paper .
as a first pass , we examine the aggregate sfh of the fornax system . combining the sfhs obtained from all regions and fields
gives the total sfh of fornax , presented graphically in fig .
[ fig : totalsfh ] and in tabular form in table [ tab : totalsfh ] .
clearly , fornax has a complex history and has been forming stars continuously over most of the age of the universe , with non - zero sfrs being found in most age bins .
after the first stars formed more than 10 gyr ago , a slow decline is detected until approximately 4 gyr ago , when the star formation rate sharply increased for a period of @xmath41 gyr .
after this sudden intensified star formation episode , a very low level of star formation has continued until the present day . for comparison
, we also show in fig .
[ fig : totalsfh ] the schematic sfh derived by @xcite based on photometry and ca ii triplet results .
we have shifted the peak of their relative star formation rate to match our peak sfr .
the results are well matched for the last few gyr , however our cmd - fitting code is better able to extract detailed star formation rates from the early history of fornax .
we also show the total stellar mass formed during each bin in table [ tab : totalsfh ] and fig .
[ masshistory ] . from this
, we calculate the total stellar mass formed in fornax to be @xmath70 @xmath71 , where we have integrated the salpeter imf down to a mass of @xmath72 .
initially , the average metallicity of the stars shows little evolution .
a large fraction of the ancient stars seems to have formed from pre - enriched gas , as the mean metallicity in the oldest age bin ( @xmath73 gyr ) is already [ fe / h ] @xmath74 .
the metallicity only starts to increase significantly from this value at an age of around 5 gyr .
overall , these results are in good agreement with previous studies of the sfh and metallicity of fornax .
we show in the lower panel of fig .
[ fig : totalsfh ] the approximate age - metallicity relation derived by @xcite based on spectra of 562 rgb stars extending from the centre of fornax to the tidal radius .
this line represents our estimate of the evolution of the average iron abundance from fig .
23 of @xcite , however we note that each age bin encompasses a wide range of stellar metallicities and thus this line should be regarded as a guide only .
although they are well within the uncertainties , our results are slightly more metal - rich than the spectroscopic results .
such a small offset in metallicity might be caused by slight inaccuracies in the isochrones , the photometric calibration , or the distance to fornax .
population gradients are a common feature of dsphs @xcite .
generally , more recent star formation events are more centrally concentreated and produce stars with a greater metallicity compared to the older stellar population . in fig .
[ fig : radialsfh ] we show the total sfhs ( left ) and metallicity evolutions ( right ) for all four radial regions ( e.g. for region 1 the four sfhs of the portions in fields 10 , 11 , 15 , and 16 have been combined ) .
overplotted in each panel are the sfhs and metallicities for four individual regions , to give an idea of the field - to - field variation .
since the panels on the right only show the _ average _ metallicities and provide no insight into the metallicity distribution , the combined fit results for the four radial regions with the complete age - metallicity grid are shown in fig .
[ fig : radialfit ] .
there is a striking difference in the sfh when moving from region 1 ( top panels ) to region 4 ( bottom panels ) . in the center of fornax ,
the burst of star formation that occurred @xmath1 gyr ago stands out strongly .
moving outwards , the proportion of stars produced during this burst decreases and in the outskirts are not found at all .
stars younger than 1 gyr are also mostly found in the center , with hardly any significant star formation at these ages in the outermost radial bin .
this kind of radial dependence , with young stars more centrally concentrated than old stars is a common characteristic of dsphs , and evidence for this was found before in fornax @xcite .
at all radii , the metallicity evolution follows the same pattern .
it starts off at [ fe / h ] @xmath75 and remains constant until @xmath45 gyr ago .
then the metallicity starts to increase rapidly to @xmath76 , after which there is little indication of further enrichment .
this rapid increase in metallicity coincides with the strong burst of star formation . comparing the sfhs of individual fields within a radial bin ( i.e. the coloured lines in fig .
[ fig : radialsfh ] ) shows no significant difference in the central bin . since in the central part the dynamical timescales are shortest , the populations are well - mixed . in the outermost region
there is also no sign of spatial variation , which is probably because there are mostly ancient stars which have had sufficient time to diffuse throughout the galaxy . in regions 2 and 3 , however , significant field - to - field variation is visible in the strength of the 4 gyr burst .
the detailed structural properties of the stellar populations in fornax is the subject of the next paper in this series ( coleman , in prep . ) .
although fornax is found to be relatively metal - rich , fig .
[ fig : radialfit ] shows that the spread in metallicities is significant . at all ages
the metallicity spread is at least @xmath41 dex , and especially the oldest stars ( @xmath7710 gyr ) show a very large spread , with metallicities ranging between [ fe / h ] @xmath78 to our metallicity cut - off at @xmath79 .
note that in all regions there is a non - zero sfr in the most metal - rich two age bins .
this is caused by match attempting to fit the redward extension of the hb ( see fig .
[ cmd ] ) with an additional , very red rc .
this redward extension is an artefact of the observational dithering pattern in which some ( less than @xmath54 ) of the hb stars have poor @xmath26-band photometry , hence this detection of very metal - rich , ancient stars is also an artefact , and they are excluded from our star formation history .
based on spectroscopy of 562 rgb stars in fornax , @xcite found two distinct populations : a metal - rich ( [ fe / h ] @xmath80 ) component , and a metal - poor ( [ fe / h ] @xmath81 ) component with a large metallicity spread .
consistent with observations of other dsphs , they found the metal - rich stars to be more centrally concentrated than the metal - poor population . to examine the metallicities of the oldest stars , we show histograms of the metallicity distribution of the stars in the oldest age bin in fig .
[ fig : ancient_feh ] . in region 1 , the metallicity distribution of these ancient stars peaks at [ fe / h ] @xmath82 , but also contains many of stars with lower metallicities .
the peak at [ fe / h ] @xmath82 is also apparent in region 2 , but becomes less strong in region 3 and is practically absent in region 4 .
this outer bin appears to harbour a broader peak at [ fe / h ] @xmath83 , and a third peak at [ fe / h ] @xmath84 .
it should be noted that this third peak is most likely an accumulation of more metal - poor stars outside our metallicity range .
however , the histograms of the inner three regions are consistent with the presence of three peaks at metallicities of [ fe / h ] @xmath85 , @xmath86 , and @xmath87 dex . to summarise ,
we have recovered the metal - rich component ( centred at [ fe / h ] @xmath88 ) of fornax , however our results support the presence of _ three _ distinct peaks in the metallicity distribution of the ancient stars rather than the two proposed by @xcite . to test the statistical significance of the three peaks
, we created ten monte carlo realisations of the fornax stellar population using our sfh and age - metallicity relation .
a comparison of the resulting metallicity distributions indicated that , despite some slight differences , the three peaked function was present in every synthetic population .
we therefore argue that the metal - poor component of @xcite can be sub - divided into two separate populations .
this suggests that fornax experienced three main star formation events in the period @xmath7710 gyr ago , a hypothesis to be tested with further spectroscopic data . in a previous , wide - area survey
, @xcite noticed a shell - like feature near the center of fornax in our field 15 ( see fig .
[ obsmap ] ) .
subsequently , @xcite confirmed the presence of the feature based on deep photometry obtained with magellan , and determined an age of 1.4 gyr and a metallicity of [ fe / h ] @xmath89 for its stars . because of the relatively small number of stars in this overdensity and the large contamination by the overall fornax stars , obtaining a sfh is difficult , as also shown by @xcite .
therefore , we opt for constraining the properties of the stars in the feature using a single component ( sc ) fitting strategy , described in detail in @xcite . in short , simple stellar population models with a narrow age and metallicity range are fit to the observed cmd and their goodness - of - fit is compared to that of the best - fitting single component model .
all stars in a 25 wide elliptical annulus containing the inner shell were extracted from the survey .
this was then divided into two regions : the @xmath90 arc containing the shell , while the remainder of the annulus was used as a control field .
the control field - subtracted hess diagram of the shell is shown in panel ( a ) of fig .
[ fig : shell_scfit ] , where the msto and rc of the shell population clearly stand out .
panel ( b ) of the same figure shows the areas in the age - metallicity plane that lie within 1 , 2 , and 3@xmath36 of the best - fit values , indicated with an asterisk .
we find the age of the stars in the shell to be @xmath91 gyr and the metallicity [ fe / h ] @xmath92 dex .
our sfh results indicate a very low sfr for this age and the average age - metallicity relation predicts a higher metallicity of [ fe / h ] @xmath93 for stars of the general fornax population of the same age .
thus , these results seem consistent with the interpretation of the shell resulting from an accretion event @xcite , rather than being part of the underlying stellar populations .
this accretion hypothesis will be discussed further in a later publication . with the sfh in hand ,
it is possible to construct the cmd of fornax as it would have looked at some time in the past . in this way
, the total luminosity of the system can be traced as a function of time . combining the sfh fits of all fields , the overall sfr as function of age and metallicity
was used to construct artificial cmds for fornax at various points in the last 10 gyr , assuming in all cases a binary fraction of 0.5 and a salpeter imf . by extending the cmds far enough down the lf
, the flux of all stars can be used to calculate the total luminosity .
although several systematic effects hamper a very precise measurement of @xmath94 , the value we obtain for the present day , @xmath95 , is very close to the literature value of @xmath8 @xcite .
the evolution of the v - band luminosity of fornax during the past 10 gyr is shown in fig .
[ fig : lumevol ] .
this figure is for illustrative purposes only , and any uncertainties given would be estimates at best .
however , we can see that while fornax has experienced a general trend towards increasing brightness as more gas is converted to stars , there are significant variations caused by bursts of star formation .
we find that the total luminosity of fornax has a range of at least 1 magnitude over a hubble time , hence at some point it was less than @xmath30 of its current brightness .
the mechanisms of star formation and chemical enrichment in low - mass galaxies are not well understood .
qualitatively , the process is simple to envision . initially , gas collapses at the centre of a dark halo to form the first generation of stars .
the heaviest of these quickly evolve and blowout material , injecting chemicals and energy into the surrounding gas cloud .
this stellar feedback causes the enriched gas cloud to expand and , if the potential well is deep enough , then collapse back to the centre of the dark halo to create a new generation of chemically enriched stars .
thus , a cyclical process is established in which multiple generations of stars form with a progressive increase in chemical abundance . however , there are still many unresolved questions regarding dsph star formation . @xcite
have shown that , in contrast to the galactic halo , the dsphs are conspicuously lacking stars with [ fe / h ] @xmath96 .
this suggests that the gas sourcing the oldest stars in dsphs was pre - enriched .
given that all dsphs are thought to contain some fraction of ancient stars ( i.e. with ages @xmath97 gyr ; @xcite ) the initial enrichment must have been an extremely rapid process , via some mechanism which is unclear .
furthermore , the sfhs for each dsph are remarkably different .
some dsphs contain purely old stars ( ages @xmath2 gyr ) and are characterised by a simple sfh ( e.g. draco ) , whereas others are dominated by intermediate - age stars and have been able to maintain multiple epochs of star formation and chemical enrichment ( e.g. fornax ) .
this is the leading question concerning star formation in dwarf galaxies : why do draco and fornax reside at the centres of dark halos with similar masses @xcite , yet they differ in brightness by a factor of more than ten ?
@xcite discussed this point , and described two possible scenarios : ( i ) either fornax and draco initially contained the same amount of gas , hence local effects have allowed fornax to produce stars with a greater efficiency , or , ( ii ) fornax had access to a tenfold greater reservoir of gas , and draco experienced a similar ( but scaled down ) initial star formation .
we note that these are extreme scenarios , and something in between is not precluded .
@xcite found that the reionization of the universe did not cause the expected reduction in dsph star formation , hence ` local effects ' are thought to be the dominant factor producing the variety of dsph sfhs .
differences in the level of tidal distortion , mechanical feedback and gas infall experienced by each system are cited , however the precise nature of these local effects is uncertain ( e.g. @xcite ) . in this
regard , @xcite have noted that a dsph s dark matter fraction will influence its sfh .
external forces ( e.g. tidal and ram pressure stripping ) can remove blown - out gas from a dsph , however a massive dark halo will allow the satellite to retain its gas .
fornax contains a large number of stars and has a complex star formation history , hence it is an ideal object to compare to simulations .
firstly , our best fits show that the ancient stars ( age @xmath2 gyr ) in this system have a _
metallicity of [ fe / h ] @xmath98 ( fig .
[ fig : totalsfh ] ) .
this indicates that the first few gyr contained an intense period of star formation and chemical enrichment . at some level
, this enrichment appears to have occurred throughout the whole body of fornax : fig .
[ fig : radialfit ] indicates that the oldest stars in every radial bin contain a number of [ fe / h ] @xmath88 stars .
however , we also note a metallicity gradient in this ancient population , such that the central stars display a mean iron abundance approximately 0.3 dex greater than those in the outer regions .
this is consistent with the spectroscopic results of @xcite .
the three peaks in the metallicity distribution function ( fig .
[ fig : ancient_feh ] ) are possibly evidence for three main bursts for star formation in the early universe . in summary , while our results indicate that the first few gyrs saw a swift chemical enrichment process in fornax , they also show that this enrichment was enhanced towards the centre of the system .
in general , our results for the first few gyr of fornax are well reproduced by models .
@xcite constructed a 3d hydrodynamic model describing gas dynamics and chemical enrichment in a dwarf galaxy including the contribution of supernovae .
the metalicity distribution function we find for the ancient fornax stars is similar to that produced by the marcolini et al . model .
furthermore , they find that stars located towards the centre of a dsph are the product of a more efficient chemical enrichment @xcite , as is seen in fornax @xcite . @xcite presented a semi - analytic cosmological model following star formation in a dsph galaxy in a milky way - type environment , pre - enriched to an abundance of [ fe / h ] @xmath7 .
they showed that a dsph experiences intense star formation in the first few hundred myr , with multiple cycles of star bursts followed by gas blowout and infall , with accompanying chemical enrichment .
their metallicity distribution function and mean metallicity are roughly equivalent to those we measured for the old stars in fornax .
it is clear that simulations are able to accurately reproduce the first epoch of star formation in a dsph environment .
@xcite find a rapidly decreasing star formation rate with time , such that approximately 2.5 gyr after virialization the rate has fallen well below @xmath99
@xmath71 yr@xmath100 .
this corresponds to an age of @xmath49 gyr , or the second data point in our global sfh for fornax ( fig .
[ fig : totalsfh ] ) , where we measure a star formation rate of approximately @xmath101 @xmath71 yr@xmath100 .
hence , although the @xcite model provides an excellent reproduction of star formation in a dsph such as sculptor , they note that a more complex sfh ( such as that seen in fornax ) requires a different set of conditions . in this context
, we find the star formation rate in fornax to be an approximately constant value250 myr .
] of @xmath101 @xmath71 yr@xmath100 in the period from 9 to 4 gyr ago .
this period also witnessed a slow , monotonic increase in iron abundance .
however , star formation in fornax experienced a sudden increase approximately 3 to 4 gyr ago , jumping threefold to @xmath4@xmath102 @xmath71 yr@xmath100 with an accompying spike in chemical enrichment .
these results all agree with the ca ii triplet results of @xcite .
our results also suggest that this epoch of star formation was relatively short - lived , lasting @xmath39 gyr , and was confined to the central @xmath4@xmath103 ( 1500 pc ) of fornax ( fig .
[ fig : radialsfh ] ) .
there are a variety of explanations for this continued star formation .
@xcite state that a refinement of the reionization criterion would allow massive dsphs with a lower initial gas - to - dark ratio in their models , thereby leading to less efficient mechanical feedback and more regular star formation activity .
this could explain the steady star formation seen in the period from 9 to 4 gyr ago , however it can not account for the subsequent burst . as an alternative scenario , fornax may have experienced an injection of new gas to fuel this next generation of stars .
we have already presented evidence for a merger in fornax , proposing that a gas - rich dwarf galaxy merged with this system to fuel strong star formation activity @xcite .
however , the timing is problematic : the original scenario requires that this merger occured approximately 2 gyr ago , or 2 gyr _
after _ the suddent burst of star formation found here .
furthermore , our chemical enrichment history shows that the new burst of stars was accompanied by a sudden increase in abundance , hence this would imply that , prior to star formation , the metal abundance of the gas was at least that of fornax and therfore unlikely to be of foreign origin .
we would therefore argue that the 4 gyr burst was fueled by gas originating in fornax .
this scenario requires gas blown away by star formation to have resided in the outer regions of fornax for at least 5 gyr before collapsing back to the dsph .
gas expelled to the outer regions ( or even the halo ) of a satellite system is generally expected to be removed by ram pressure stripping and tidal distortion .
however , @xcite presented evidence that hi clouds exist in many local group dsphs , situated up to 10 kpc ( or , approximately three tidal radii in the case of fornax ) from the centre of the system .
this is supported by the strong evidence for hi associated with sculptor @xcite .
indeed , @xcite discovered an hi cloud located to the north of fornax . whether this is associated with the dsph is not certain ( the radial velocity of fornax is inconveniently close to that of the milky way in this direction ) , however @xcite state a minimum mass of @xmath104 @xmath71 at the distance of fornax .
@xcite noted that an object following an orbit with a low ellipticity ( such as fornax ; @xcite ) will be better able to retain its gas .
it therefore seems possible that the massive , extended dark halo of fornax could have allowed gas in the outer regions to remain bound to the system for an extended period .
an external influence on star formation in a satellite system are tidal interactions with the larger host galaxy @xcite .
tidal forces experienced by a satellite system as it orbits its host can induce burts of star formation @xcite as the interstellar gas clouds are compressed @xcite . using _ hst _
images over a four year epoch , @xcite measured the proper motion of fornax and derived an orbital period of @xmath10 gyr with an eccentricity of @xmath9 .
tidal forces scale as @xmath105 , hence this orbit implies that fornax will experience a distortion force change of at least @xmath30 as it moves from pericentre to apocentre . a pericentric passage
approximately 4 gyr ago is possible within the current solution ( we are grateful to s. piatek for sharing his orbital code ) , however the uncertainties are currently too large for a fair comparison between orbit and sfh .
finally , we examine the recent activity in fornax . following the 4 gyr burst
, there was another star forming event @xmath106 myr ago , and a more recent event approximately 100 myr ago ( fig .
[ fig : totalsfh ] ) .
these 100 myr old stars have been previously noted by @xcite and @xcite , and are the youngest stars yet observed in a dsph . however , as a new result , we have also detected a small number of younger stars .
table [ tab : totalsfh ] indicates that in the period @xmath107 myr ago , @xmath41500 @xmath71 were formed in fornax . this is minute compared to fornax as a whole
( see fig .
[ masshistory ] ) , yet it is tentative evidence that _ fornax may have been forming stars almost to the present day_. we ignore the abundance measurements for this youngest population , given that they are based on only a few stars in each age bin and have significant errors . to demonstrate the difficulty in detecting this small , age @xmath108 myr population , we show an artificial cmd of the fornax young stars in fig .
[ youngcmd ] .
the left panel is a ` typical ' prediction for the young stars based on our sfh .
extinction , photometric errors and completeness for the survey have been included .
note that the distribution of the blue points ( age @xmath108 myr ) is almost identical to that of the red points ( age @xmath109 myr ) , thus these populations are degenerate and it is not possible to distinguish between them based only on photometry .
however , it shows that such a very young population could easily have evaded detection up to now .
given the small number of young stars predicted by our sfh , and the difficulty in separating them from slightly older stars , we classify this as tentative evidence for an ultra - young population in fornax .
nonetheless , we expect very little foreground contamination blueward of @xmath110 , hence the brightest blue stars shown in the right panel would argue that fornax did indeed experience star formation less than 100 myr ago .
high resolution spectra of these young stars is required to accurately determine their age .
lithium is an important age diagnostic for late - type stars as it is easily destroyed in stellar interiors , hence a follow - up survey could target the resonance doublet of li i at 6708 ( e.g. @xcite ) in these young stars to demarcate the age of the most recent star forming event in fornax .
based on two - filter photometry to a magnitude of @xmath0 , we used a cmd - fitting technique to derive the star formation history for the fornax dsph .
all dsphs contain some number of ancient stars , however these systems are known to display a wide variety of sfhs .
fornax formed a significant number of its stars in the early universe ( age @xmath2 gyr ) and subsequently experienced a constant star formation rate .
this behaviour can be reproduced by the simulations . however , in the period @xmath1 gyr ago , fornax experienced a sudden burst of star formation , approximately three times the rate for the previous 5 gyr .
the cause of this activity is unclear .
the star formation rate has been decreasing ever since , with smaller events @xmath106 myr and 100 myr ago .
we also find tentative evidence for a small number of stars ( total mass @xmath4@xmath111 ) which have formed in the past 100 myr .
fornax contains the most recent star formation activity of any local group dsph .
strong radial gradients in the sfh are also evident .
as noted by previous authors @xcite , recent star formation has been concentrated towards the centre of fornax .
this trend is seen in other dsphs @xcite .
furthermore , we have found that chemical enrichment was more efficient at the dsph s centre .
even the oldest stars display a metallicity gradient , such that the inner stars have a mean iron abundance approximately @xmath112 dex greater than the outer stars .
indeed , the first few gyr in fornax appear to have been a time of intense star formation and chemical enrichment : the age @xmath2 gyr stars display a mean abundance of [ fe / h ] @xmath113 .
the oldest stars in fornax also show three peaks in the metallicity distribution , possibly evidence for three main bursts of star formation in the period @xmath7710 gyr ago .
we find the metallicity to have increased monotonically in the period @xmath114 gyr ago , and then experienced a sharp increase in conjunction with the intense burst of star formation described above .
thus , while the first few gyr of star formation in fornax can be reproduced by models , the cause of the burst 4 gyr ago is unclear .
we have previously proposed that a gas - rich dwarf galaxy merged with fornax approximately 2 gyr ago to produce sub - structure @xcite , however we can not reconcile the timing of this event with the observed peak in the sfh .
therefore , we suggest that gas enriched and blown out by earlier star forming events hi clouds settled in the outer regions of fornax , and then re - collapsed to fuel an intense period of star formation at the centre of the dsph .
this event may have been caused by tidal interactions with the milky way during a pericentric passage .
the authors thank the anonymous referee for their comments which have improved the manuscript .
the authors also thank s. piatek and c. pryor for their helpful advice regarding the orbit of fornax in different potentials , and we acknowledge the use of their orbital algorithms .
we are indebted to h .- w .
rix for his helpful discussions , and the mpia observing team who obtained the mpg / eso 2.2 m data described in this paper .
aaronson , m. , & mould , j. 1980 , , 240 , 804 aaronson , m. , & mould , j. 1985 , , 290 , 191 aparicio , a. , gallart , c. & bertelli , g. 1997 , , 114 , 680 barton , e. j. , geller , m. j. , & kenyon , s. j. 2000 , , 530 , 660 battaglia , g. , et al .
2006 , , 459 , 423 beauchamp , d. , hardy , e. , suntzeff , n. b. , & zinn , r. 1995 , , 109 , 1628 bersier , d. 2000 , , 543 , l23 blitz , l. , & robishaw , t. 2000 , , 541 , 675 bouchard , a. , carignan , c. , & mashchenko , s. 2003 , , 126 , 1295 bouchard , a. , carignan , c. , & staveley - smith , l. 2006 , , 131 , 2913 buonanno , r. , corsi , c. e. , castellani , m. , marconi , g. , fusi pecci , f. , & zinn , r. 1999 , , 118 , 1671 carignan , c. , beaulieu , s. , ct , s. , demers , s. , & mateo , m. 1998 , , 116 , 1690 christlieb , n. , et al .
2002 , , 419 , 904 coleman , m. , da costa , g. s. , bland - hawthorn , j. , martnez - delgado , d. , freeman , k. c. , & malin , d. 2004 , , 127 , 832 coleman , m. g. , da costa , g. s. , bland - hawthorn , j. , & freeman , k. c. 2005 , , 129 , 1443 de jong , j. t. a. , rix , h - w . , martin , n. f. , zucker , d. b. , dolphin , a. e. , belokurov , v. , evans , w. , 2008 , , 135 , 1361 dekel , a. , & silk , j. 1986 , , 303 , 39 demers , s. , & kunkel , w. e. 1979 , , 91 , 761 dolphin , a. e. 1997 , , 2 , 397 dolphin , a. e. 2002 , , 332 , 91 ferrara , a. , & tolstoy , e. 2000 , , 313 , 291 frebel , a. , et al .
2005 , , 434 , 871 gallart , c. , aparicio , a. , bertelli , g. , & chiosi , c. 1996 , , 112 , 1950 gallart , c. , aparicio , a. , zinn , r. , buonanno , r. , hardy , e. , & marconi , g. 2005 , iau colloq . 198 : near - fields cosmology with dwarf elliptical galaxies , 25 girardi , l. , bertelli , g. , bressan , a. , chiosi , c. , groenewegen , m. a. t. , marigo , p. , salasnich , b. , & weiss , a. 2002 , , 391 , 195 grebel , e. k. , gallagher , j. s. , iii , & harbeck , d. 2003 , , 125 , 1926 grebel , e. k. , & gallagher , j. s. , iii 2004 , , 610 , l89 grillmair , c. j. , freeman , k. c. , irwin , m. , & quinn , p. j. 1995 , , 109 , 2553 harbeck , d. , et al . 2001 , , 122 , 3092 harris , j. & zaritsky , d. 2001 , , 136 , 25 held , e. v. , saviane , i. , momany , y. , & carraro , g. 2000 , , 530 , l85 helmi , a. , et al . 2006 , , 651 , l121 hernandez , x. , gilmore , g. & valls - gabaud , d. 2000 , , 317 , 831 holtzman , j. a. et al . 1999 , , 118 , 2262 johnston , k. v. , choi , p. i. , & guhathakurta , p. 2002 , , 124 , 127 kroupa , p. , tout , c. a. , & gilmore , g. 1993 , , 262 , 545 letarte , b. 2007 , phd thesis , `` chemical analysis of the fornax dwarf galaxy '' mackey , a. d. , & gilmore , g. f. 2003 , , 340 , 175 marcolini , a. , dercole , a. , brighenti , f. , & recchi , s. 2006 , , 371 , 643 marcolini , a. , dercole , a. , battaglia , g. , & gibson , b. k. 2008 , , accepted , arxiv:0803.0385 mateo , m. 1998 , , 36 , 435 mayer , l. 2005 , iau colloq . 198 : near - fields cosmology with dwarf elliptical galaxies , 220 mayer , l. , mastropietro , c. , wadsley , j. , stadel , j. , & moore , b. 2006 , , 369 , 1021 mihos , j. c. , & hernquist , l. 1996 , , 464 , 641 montes , d. , lpez - santiago , j. , fernndez - figueroa , m. j. , & glvez , m. c. 2001 , , 379 , 976 navarro , j. f. , frenk , c. s. , & white , s. d. m. 1997 , , 490 , 493 olsen , k. a. g. 1999 , , 117 , 2244 olszewski , e. w. , mateo , m. , harris , j. , walker , m. g. , coleman , m. g. , & da costa , g. s. 2006 , , 131 , 912 piatek , s. , pryor , c. , bristow , p. , olszewski , e. w. , harris , h. c. , mateo , m. , minniti , d. , & tinney , c. g. 2007 , , 133 , 818 pont , f. , zinn , r. , gallart , c. , hardy , e. , & winnick , r. 2004 , , 127 , 840 salvadori , s. , ferrara , a. , & schneider , r. 2008 , , accepted , arxiv:0802.0462 saviane , i. , held , e. v. , & bertelli , g. 2000 , , 355 , 56 schlegel , d. , finkbeiner , d. & davis , m. 1998 , , 500 , 525 simon , j. d. , & geha , m. 2007 , apj , submitted , arxiv:0706.0516 stetson , p. b. 1987 , , 99 , 191 stetson , p. b. , hesser , j. e. , & smecker - hane , t. a. 1998 , , 110 , 533 tolstoy , e. & saha , a. 1996 , , 462 , 672 tolstoy , e. , irwin , m. j. , cole , a. a. , pasquini , l. , gilmozzi , r. , & gallagher , j. s. 2001 , , 327 , 918 tolstoy , e. , et al .
2004 , , 617 , l119 valdes , f. g. 2002 , automated data analysis in astronomy , 309 walker , m. g. , mateo , m. , olszewski , e. w. , bernstein , r. , wang , x. , & woodroofe , m. 2006 , , 131 , 2114 walker , m. g. , mateo , m. , olszewski , e. w. , gnedin , o. y. , wang , x. , sen , b. , & woodroofe , m. 2007 , , 667 , l53 zacharias , n. , et al .
2000 , , 120 , 2131 lccc age bin & sfr & [ fe / h ] & mass + ( log - yr ) & ( @xmath11510@xmath116 yr@xmath100 ) & & ( @xmath117 ) + @xmath118 & @xmath119 & @xmath120 & @xmath121 + @xmath122 & @xmath123 & @xmath124 & @xmath125 + @xmath126 & @xmath127 & @xmath128 & @xmath129 + @xmath130 & @xmath131 & & @xmath131 + @xmath132 & @xmath131 & & @xmath131 + @xmath133 & @xmath131 & & @xmath131 + @xmath134 & @xmath135 & @xmath136 & @xmath137 + @xmath138 & @xmath131 & & @xmath131 + @xmath139 & @xmath140 & @xmath141 & @xmath142 + @xmath143 & @xmath144 & @xmath145 & @xmath146 + @xmath147 & @xmath148 & @xmath149 & @xmath150 + @xmath151 & @xmath152 & @xmath153 & @xmath154 + @xmath155 & @xmath156 & @xmath157 & @xmath158 + @xmath159 & @xmath160 & @xmath161 & @xmath162 + @xmath163 & @xmath164 & @xmath165 & @xmath166 + @xmath167 & @xmath168 & @xmath169 & @xmath170 + @xmath171 & @xmath172 & @xmath173 & @xmath174 + @xmath175 & @xmath176 & @xmath177 & @xmath178 + @xmath179 & @xmath180 & @xmath181 & @xmath182 + @xmath183 & @xmath184 & @xmath185 & @xmath186 + @xmath187 & @xmath188 & @xmath189 & @xmath190 + | based on a deep imaging survey , we present the first homogeneous star formation history ( sfh ) of the fornax dwarf spheroidal ( dsph ) galaxy .
we have obtained two - filter photometry to a depth of @xmath0 over the entire surface of fornax , the brightest dsph associated with the milky way , and derived its sfh using a cmd - fitting technique .
we show that fornax has produced the most complex star formation and chemical enrichment histories of all the milky way dsphs .
this system has supported multiple epochs of star formation .
a significant number of stars were formed in the early universe , however the most dominant population are the intermediate age stars .
this includes a strong burst of star formation approximately @xmath1 gyr ago .
significant population gradients are also evident .
similar to other dsphs , we have found that recent star formation was concentrated towards the centre of the system .
furthermore , we show that the central region harboured a faster rate of chemical enrichment than the outer parts of fornax . at the centre , the ancient stars ( age @xmath2 gyr )
display a mean metallicity of [ fe / h ] @xmath3 , with evidence for three peaks in the metallicity distribution .
overall , enrichment in fornax has been highly efficient : the most recent star formation burst has produced stars with close to solar metallicity .
our results support a scenario in which fornax experienced an early phase of rapid chemical enrichment producing a wide range of abundances .
star formation gradually decreased until @xmath44 gyr ago , when fornax experienced a sudden burst of strong star formation activity accompanied by substantial chemical enrichment .
weaker star forming events followed , and we have found tentative evidence for stars with ages less than 100 myr . |
motion camouflage is a stealth strategy employed by various visual insects and animals to achieve prey capture , mating or territorial combat . in one type of motion camouflage ,
the predator camouflages itself against a fixed background object so that the prey observes no relative motion between the predator and the fixed object . in the other type of motion camouflage ,
the predator approaches the prey such that from the point of view of the prey , the predator always appears to be at the same bearing .
( in this case , we say that the object against which the predator is camouflaged is the point at infinity . ) assuming that the prey can readily observe optical flow , but only poorly sense looming , this type of motion by the predator is then difficult to detect by the prey .
for example , insects with compound eyes are quite sensitive to optical flow ( which arises from the transverse component of the relative velocity between the predator and the prey ) , but are far less sensitive to slight changes in the size of images ( which arise from the component of the relative velocity between the predator and prey along the line between them )
. more broadly such interactions may also apply in settings of mating activity or territorial maneuvers as well . in the work , @xcite of srinivasan and davey ,
it was suggested that the data on visually mediated interactions between two hoverflies , _ syritta pipiens _ obtained earlier by collett and land @xcite , supports a motion camouflage hypothesis .
later , mizutani , chahl and srinivasan @xcite , observing territorial aerial maneuvers of dragonflies _ hemianax papuensis _ , concluded that the flight pattern is motivated by motion camouflage ( see figure 1 in their paper ) .
see also @xcite for a review of related themes in insect vision and flight control .
motion camouflage can be used by a predator to stealthily pursue prey , but a motion camouflage strategy can also be used by the prey to evade a predator .
the only difference between the strategy of the predator and the strategy of the evader is that the predator seeks to approach the prey while maintaining motion camouflage , whereas the evader seeks to move away from the predator while maintaining motion camouflage . besides explaining certain biological pursuit strategies
, motion camouflage may also be quite useful in certain military scenarios ( although the `` predator '' and `` prey '' labels may not be descriptive ) . in some settings , as is the case in @xcite , @xcite ,
@xcite it is more appropriate to substitute the labels `` shadower '' and `` shadowee '' for the predator - prey terminology . in this work ,
we take a structured approach to deriving feedback laws for motion camouflage , which incorporate biologically plausible ( vision ) sensor measurements .
we model the predator and prey as point particles moving at constant ( but different ) speeds , and subject to steering ( curvature ) control .
for an appropriate choice of feedback control law for one of the particles ( as the other follows a prescribed trajectory ) , a state of motion camouflage is then approached as the system evolves .
( in the situation where the predator follows a motion - camouflage law , and the speed of the predator exceeds the speed of the prey , the predator is able to pass `` close '' to the prey in finite time . in practice ,
once the predator is sufficiently close to the prey , it would change its strategy from a pursuit strategy to an intercept strategy . )
what distinguishes this work from earlier study of motion - camouflage trajectories in @xcite is that we present _ biologically plausible feedback laws _ leading to motion camouflage .
furthermore , unlike the neural - network approach used in @xcite to achieve motion camouflage using biologically - plausible sensor data , our approach gives an explicit form for the feedback law which has a straightforward physical interpretation .
the study of motion camouflage problems also naturally extends earlier work on interacting systems of particles , using the language of curves and moving frames @xcite-@xcite .
for concreteness , we consider the problem of motion camouflage in which the predator ( which we refer to as the `` pursuer '' ) attempts to intercept the prey ( which we refer to as the `` evader '' ) while appearing to the prey as though it is always at the same bearing ( i.e. , motion camouflaged against the point at infinity ) . in the model we consider , the pursuer moves at unit speed in the plane , while the evader moves at a constant speed @xmath0 .
the dynamics of the pursuer are given by @xmath1 where @xmath2 is the position of the pursuer , @xmath3 is the unit tangent vector to the trajectory of the pursuer , @xmath4 is the corresponding unit normal vector ( which completes a right - handed orthonormal basis with @xmath3 ) , and the plane curvature @xmath5 is the steering control for the pursuer .
similarly , the dynamics of the evader are @xmath6 where @xmath7 is the position of the evader , @xmath8 is the unit tangent vector to the trajectory of the evader , @xmath9 is the corresponding unit normal vector , and @xmath10 is the steering control for the evader .
figure [ framefig2d ] illustrates equations ( [ pursuer2d ] ) and ( [ evader2d ] ) .
note that @xmath11 and @xmath12 are planar natural frenet frames for the trajectories of the pursuer and evader , respectively .
we model the pursuer and evader as point particles ( confined to the plane ) , and use natural frames and curvature controls to describe their motion , because this is a simple model for which we can derive both physical intuition and concrete control laws .
( furthermore , although we save the details for a future paper , this approach generalizes nicely for three - dimensional motion . )
flying insects and animals ( also unmanned aerial vehicles ) have limited maneuverability and must maintain sufficient airspeed to stay aloft , so treating their motion as constant - speed with steering control is physically reasonable , at least for some range of flight conditions .
( note that the steering control directly drives the angular velocity of the particle , and hence is actually an acceleration input .
however , this acceleration is constrained to be perpendicular to the instantaneous direction of motion , and therefore the speed remains unchanged . )
we refer to ( [ pursuer2d ] ) and ( [ evader2d ] ) as the `` pursuit - evader system . '' in what follows , we assume that the pursuer follows a _ feedback _ strategy to drive the system toward a state of motion camouflage , and close in on the evader .
the evader , on the other hand , follows an open - loop strategy .
the analysis we present for the pursuer feedback strategy also suggests ( with a sign change in the control law ) how the evader could use feedback and a motion - camouflage strategy to conceal its flight from the pursuer . ultimately , it would be interesting to address the game - theoretic problem in which both the pursuer and evader follow feedback strategies , so that the system would truly be a pursuit - evader system .
( what we address in this work would be more properly described as a pursuer - pursuee system .
however , we keep the pursuer - evader terminology , because it sets the stage for analyzing the true pursuer - evader system , which we plan to address in a future paper . )
motion camouflage with respect to the point at infinity is given by @xmath13 where @xmath14 is a fixed unit vector and @xmath15 is a time - dependent scalar ( see also section 5 of @xcite ) .
let @xmath16 be the vector from the evader to the pursuer .
we refer to @xmath17 as the `` baseline vector , '' and @xmath18 as the `` baseline length . ''
we restrict attention to non - collision states , i.e. , @xmath19 . in that case
, the component of the pursuer velocity @xmath20 transverse to the base line is @xmath21 and similarly , that of the evader is @xmath22 the _ relative _ transverse component is @xmath23 * lemma * ( infinitesimal characterization of motion camouflage ) : the pursuit - evasion system ( [ pursuer2d ] ) , ( [ evader2d ] ) is in a state of motion camouflage without collision on an interval iff @xmath24 on that interval .
* proof * : @xmath25 suppose motion camouflage holds .
thus @xmath26.\ ] ] differentiating , @xmath27 . hence , @xmath28.\end{aligned}\ ] ] @xmath29 suppose @xmath30 on @xmath31 $ ] .
thus @xmath32 so that @xmath33 where @xmath34 and @xmath35 .
@xmath36 it follows from the * lemma * that the set of all motion camouflage states constitutes a 5-dimensional smooth manifold with two connected components , each diffeomorphic to @xmath37 in the 6-dimensional state space @xmath38 of the problem . in practice
we are interested in how far the pursuit - evasion system is from a state of motion camouflage . in what follows ,
we offer a measure of this .
consider the ratio @xmath39 which compares the rate of change of the baseline length to the absolute rate of change of the baseline vector . if the baseline experiences pure lengthening , then the ratio assumes its maximum value , @xmath40 .
if the baseline experiences pure shortening , then the ratio assumes its minimum value , @xmath41 . if the baseline experiences pure rotation , but remains the same length , then @xmath42 . noting that @xmath43 we see that @xmath44 may alternatively be written as @xmath45 thus , @xmath44 is the dot product of two unit vectors : one in the direction of @xmath17 , and the other in the direction of @xmath46
note that @xmath47 is well - defined except at @xmath48 , since @xmath49 for convenience , we define the notation @xmath50 to represent the vector @xmath51 rotated counter - clockwise in the plane by an angle @xmath52 .
thus , for example , @xmath53 .
the transverse component @xmath54 of relative velocity , expression ( [ wdefn ] ) , then becomes @xmath55 \left(\frac{\bf r}{|{\bf r}|}\right)^{\perp } \nonumber \\ \hspace{-.2 cm } & = & \hspace{-.2 cm } - \left(\frac{\bf r}{|{\bf r}| } \cdot \dot{\bf r}^{\perp}\right ) \left(\frac{\bf
r}{|{\bf r}|}\right)^{\perp}.\end{aligned}\ ] ] for convenience , we define @xmath56 to be the ( signed ) magnitude of @xmath54 , i.e. , @xmath57 and refer also to @xmath56 as the transverse component of the relative velocity . from the orthogonal decomposition @xmath58
\left(\frac{\dot{\bf r}}{|\dot{\bf r}|}\right)^\perp,\ ] ] it follows that @xmath59 ^ 2 = \gamma^2 + \frac{|w|^2}{|\dot{\bf r}|^2}.\ ] ] thus @xmath60 is a measure of the distance from motion camouflage .
differentiating @xmath47 along trajectories of ( [ pursuer2d ] ) and ( [ evader2d ] ) gives @xmath61 \nonumber \\ & & + \frac{1}{|\dot{\bf r}|}\left [ \frac{\bf r}{|{\bf r}| } - \left(\frac{\bf r}{|{\bf r}|}\cdot \frac{\dot{\bf r}}{|\dot{\bf r}|}\right ) \frac{\dot{\bf r}}{|\dot{\bf r}| } \right ] \cdot \ddot{\bf r}.\end{aligned}\ ] ] from ( [ rdefnplanar ] ) we obtain @xmath62 and @xmath63 also , @xmath64
\hspace{-.2 cm } & = & \hspace{-.2 cm } \left[\frac{\bf r}{|{\bf r}| } \cdot \left(\frac{\dot{\bf r}}{|\dot{\bf r}|}\right)^{\perp } \right ] \left(\frac{\dot{\bf r}}{|\dot{\bf r}|}\right)^{\perp } \nonumber \\
\hspace{-.2 cm } & = & \hspace{-.2 cm } \frac{1}{|\dot{\bf r}|^2 } \left(\frac{\bf r}{|{\bf r}| } \cdot \dot{\bf r}^{\perp } \right ) \dot{\bf r}^{\perp}.\end{aligned}\ ] ] then from ( [ dotgamma ] ) we obtain @xmath65 \nonumber \\ & & + \frac{1}{|\dot{\bf r}|}\left [ \frac{\bf r}{|{\bf r}| } - \left(\frac{\bf r}{|{\bf r}|}\cdot \frac{\dot{\bf r}}{|\dot{\bf r}|}\right ) \frac{\dot{\bf r}}{|\dot{\bf r}| } \right ] \cdot \big ( { \bf y}_p u_p - \nu^2 { \bf y}_e u_e \big ) \nonumber \\ \hspace{-.2 cm } & = & \hspace{-.2 cm } \frac{|\dot{\bf r}|}{|{\bf r}|}\left [ \frac{1}{|\dot{\bf r}|^2 } \left(\frac{\bf r}{|{\bf r}| } \cdot \dot{\bf r}^{\perp } \right)^2 \right ] \nonumber \\ & & + \frac{1}{|\dot{\bf r}|}\left[\frac{1}{|\dot{\bf r}|^2 } \left(\frac{\bf r}{|{\bf r}| } \cdot \dot{\bf r}^{\perp } \right ) \dot{\bf r}^{\perp}\right ] \cdot \big ( { \bf y}_p u_p - \nu^2 { \bf y}_e u_e \big ) .
\nonumber \\\end{aligned}\ ] ] noting that @xmath66 and @xmath67 we obtain @xmath68 \nonumber \\ & & + \frac{1}{|\dot{\bf r}|}\left[\frac{1}{|\dot{\bf r}|^2 } \left(\frac{\bf r}{|{\bf r}| } \cdot \dot{\bf r}^{\perp } \right ) \right ] \big(1 - \nu ( { \bf x}_p \cdot { \bf x}_e ) \big ) u_p \nonumber \\ & & + \frac{1}{|\dot{\bf r}|}\left[\frac{1}{|\dot{\bf r}|^2 } \left(\frac{\bf r}{|{\bf r}| } \cdot \dot{\bf r}^{\perp } \right ) \right ] \big(\nu - ( { \bf x}_p \cdot { \bf x}_e ) \big ) \nu^2 u_e.\end{aligned}\ ] ] suppose that we take @xmath69 \nu^2 u_e,\ ] ] where @xmath70 , so that the steering control for the pursuer consists of two terms : one involving the motion of the evader , and one involving the transverse component of the relative velocity
. then @xmath71 \left [ \frac{1}{|\dot{\bf r}| } \left(\frac{\bf r}{|{\bf r}| } \cdot \dot{\bf r}^{\perp } \right ) \right]^2,\ ] ] and for any choice of @xmath70 , there exists @xmath72 such that @xmath73 for all @xmath17 such that @xmath74 .
thus , for control law ( [ planarup ] ) , @xmath75
control law ( [ planarup ] ) has the nice property that for any value of the gain @xmath76 , there is a disc of radius @xmath77 ( depending on @xmath78 ) such that @xmath79 outside the disc .
however , the problem with ( [ planarup ] ) is that the pursuer needs to know ( i.e. , sense and estimate ) the evader s steering program @xmath10 .
here we show that by taking @xmath78 sufficiently large , motion camouflage can be achieved ( in a sense we will make precise ) using a control law depending only on the transverse relative velocity : @xmath80 in place of ( [ planarup ] ) , provided @xmath81 is bounded . comparing ( [ planarupgain ] ) to ( [ transrelvel3 ] )
, we see that , indeed , @xmath5 is proportional to the signed length of the relative transverse velocity vector . we will designate this as the _ motion camouflage proportional guidance _ ( mcpg ) law for future reference ( see section v below ) . as is further discussed in section v , ( [ planarupgain ] ) requires range information as well as pure optical flow sensing . however , the range information can be coarse , since range errors ( within appropriate bounds ) have the same effect in ( [ planarupgain ] ) as gain variations .
we say that ( [ planarupgain ] ) is _ biologically plausible _ because the only critical sensor measurement required is optical flow sensing .
optical flow sensing does not yield the relative transverse velocity directly , but rather the angular speed of the image of the evader across the pursuer s eye .
in fact , it is the sign of the optical flow that is most critical to measure correctly , since errors in the magnitude of the optical flow , like range errors , only serve to modulate the gain in ( [ planarupgain ] ) . for biological systems , the capabilities of the sensors
_ vis - a - vis _ the sensing requirements for implementing ( [ planarupgain ] ) constrain the range of conditions for which ( [ planarupgain ] ) represents a feasible control strategy . in the high - gain limit we focus on below , sensor noise ( which is amplified by the high gain )
would be expected to have significant impact .
however , to illustrate the essential behavior , here we neglect both sensor limitations and noise .
let us consider control law ( [ planarupgain ] ) , and the resulting behavior of @xmath47 as a function of time . from ( [ dotgammaupue ] ) , we obtain the inequality @xmath82
\left [ \frac{1}{|\dot{\bf r}| } \left(\frac{\bf r}{|{\bf r}| } \cdot \dot{\bf r}^{\perp } \right ) \right]^2 \nonumber \\ & & + \frac{1}{|\dot{\bf r}|}\left[\frac{1}{|\dot{\bf r}|^2 } \left(\frac{\bf r}{|{\bf r}| } \cdot \dot{\bf r}^{\perp } \right ) \right ] \big(\nu - ( { \bf x}_p \cdot { \bf x}_e ) \big ) \nu^2 u_e \nonumber \\
\hspace{-.2 cm } & \le & \hspace{-.2 cm } - \left(1-\gamma^2 \right ) \left[\frac{\mu}{|\dot{\bf r}|}\big(1-\nu ( { \bf x}_p \cdot { \bf x}_e)\big ) - \frac{|\dot{\bf r}|}{|{\bf r}| } \right ] \nonumber \\ & & + \frac{1}{|\dot{\bf r}|^2}\sqrt{1-\gamma^2 } \big| \big(\nu - ( { \bf x}_p \cdot { \bf x}_e ) \big ) \nu^2 u_e \big| \nonumber \\ \hspace{-.2 cm } & \le & \hspace{-.2 cm } - \left(1-\gamma^2 \right ) \left[\frac{\mu}{|\dot{\bf r}|}(1-\nu ) - \frac{|\dot{\bf r}|}{|{\bf r}| } \right ] \nonumber \\ & & + \left(\sqrt{1-\gamma^2}\right)\frac{\nu^2(1+\nu)(\max |u_e| ) } { |\dot{\bf r}|^2 } \nonumber \\
\hspace{-.2 cm } & \le & \hspace{-.2 cm } - \left(1-\gamma^2 \right ) \left[\mu\left(\frac{1-\nu}{1+\nu}\right ) - \frac{1+\nu}{|{\bf r}| } \right ] \nonumber \\ & & + \left(\sqrt{1-\gamma^2}\right)\left[\frac{\nu^2(1+\nu)(\max |u_e| ) } { ( 1-\nu)^2}\right],\end{aligned}\ ] ] where we have used ( [ dotrbound ] ) . for convenience , we define the constant @xmath83 as @xmath84 for any @xmath70 , we can define @xmath72 and @xmath85 such that @xmath86 ( and it is clear that many such choices of @xmath77 and @xmath87 exist ) .
note that ( [ mudecomp ] ) implies @xmath88 thus , for @xmath89 , ( [ dotgammabound ] ) becomes @xmath90 \nonumber \\ & & + \left(\sqrt{1-\gamma^2}\right)c_1 \nonumber \\
\hspace{-.2 cm } & = & \hspace{-.2 cm } - \left(1-\gamma^2 \right ) c_o + \left(\sqrt{1-\gamma^2 } \right)c_1.\end{aligned}\ ] ] suppose that given @xmath91 , we take @xmath92 . then for @xmath93 , @xmath94 where @xmath95 * remark * : there are two possibilities for @xmath96 the state we seek to drive the system toward has @xmath97 ; however , ( [ oneminusgammasqreps ] ) can also be satisfied for @xmath98 .
( recall that @xmath99 . )
there is always a set of initial conditions such that ( [ oneminusgammasqreps ] ) is satisfied with @xmath98 .
we can address this issue as follows : let @xmath100 denote how close to @xmath101 we wish to drive @xmath47 , and let @xmath102 denote the initial value of @xmath47 .
take @xmath103 so that ( [ dotgammabound2 ] ) with ( [ c2defn ] ) applies from time @xmath104 .
@xmath36 from ( [ dotgammabound2 ] ) , we can write @xmath105 which , integrating both sides , leads to @xmath106 where @xmath107 .
noting that @xmath108 we see that for @xmath89 , ( [ dotgammabound2 ] ) implies @xmath109 where we have used the fact that @xmath110 is a monotone increasing function .
now we consider estimating how long @xmath89 , which in turn determines how large @xmath111 can become in inequality ( [ gammabound ] ) , and hence how close to @xmath112 will @xmath44 be driven . from ( [ gammadotprod ] ) we have @xmath113 which from ( [ dotrbound ] ) and @xmath114 , @xmath115 , implies @xmath116 from ( [ dotnormrbound ] ) , we conclude that @xmath117 and , more to the point , @xmath118 ) to be meaningful for the problem at hand , we assume that @xmath119 . then defining @xmath120 to be the minimum interval of time over which we can guarantee that @xmath79 , we conclude that @xmath121 from ( [ gammafinal ] ) , we see that by choosing @xmath122 sufficiently large ( which can be accomplished by choosing @xmath92 sufficiently large ) , we can force @xmath123 . noting that @xmath124 for @xmath125 , we see that @xmath126 thus , if @xmath92 is taken to be sufficiently large that @xmath127 then we are guaranteed ( under the conditions mentioned in the above calculations ) to achieve @xmath128 at some finite time @xmath129 . * definition * : given the system ( [ pursuer2d ] ) , ( [ evader2d ] ) with @xmath47 defined by ( [ gammadotprod ] ) , we say that `` motion camouflage is accessible in finite time '' if for any @xmath130 there exists a time @xmath131 such that @xmath128 .
@xmath36 * proposition * : consider the system ( [ pursuer2d ] ) , ( [ evader2d ] ) with @xmath47 defined by ( [ gammadotprod ] ) , and control law ( [ planarupgain ] ) , with the following hypotheses : * @xmath132 ( and @xmath133 is constant ) , * @xmath10 is continuous and @xmath81 is bounded , * @xmath134 , and * @xmath135 .
motion camouflage is accessible in finite time using high - gain feedback ( i.e. , by choosing @xmath70 sufficiently large ) .
* proof * : choose @xmath72 such that @xmath136 .
choose @xmath137 sufficiently large so as to satisfy ( [ c2bound ] ) , and choose @xmath87 accordingly to ensure that ( [ dotgammabound2 ] ) holds for @xmath138 .
then defining @xmath78 according to ( [ mudecomp ] ) ensures that @xmath139 , where @xmath140 is defined by ( [ bigtdefn ] ) . @xmath36 * remark * : assumption @xmath141 above can be generalized to @xmath142 .
( the @xmath143 case corresponds to a stationary `` evader , '' so that the natural frenet frame ( [ evader2d ] ) and steering control @xmath10 for the evader are not defined . )
the following simulation results illustrate the behavior of the pursuit - evasion system ( [ pursuer2d ] ) , ( [ evader2d ] ) , under the control law ( [ planarupgain ] ) for the pursuer and various open - loop controls for the evader .
the simulations also confirm the analytical results presented above .
figure [ straight_traj ] shows the behavior of the system for the simplest evader behavior , @xmath144 , which corresponds to straight - line motion . because control law ( [ planarupgain ] ) is the same as ( [ planarup ] ) when @xmath144 , @xmath47 tends monotonically toward @xmath101 ( for the initial conditions and choice of gain @xmath78 used in the simulation shown ) . in figure
[ straight_traj ] , as in the subsequent figures showing pursuer and evader trajectories , the solid light lines connect the pursuer and evader positions at evenly - spaced time instants . for a pursuit - evasion system in a state of motion camouflage
, these lines would all be parallel to one another .
also , each simulation is run for finite time , at the end of which the pursuer and evader are in close proximity .
( the ratio of speeds is @xmath145 in all of the simulations shown . )
figure [ sine_traj ] illustrates the behavior of the pursuer for a sinsusoidally - varying steering control @xmath10 of the evader , and figure [ sine_gamma ] shows the corresponding behavior of @xmath44 . in figure
[ sine_gamma ] , increasing the value of the feedback gain @xmath78 by a factor of three is observed to decrease the peak difference between @xmath47 and @xmath101 by a factor of about @xmath146 .
this is consistent with the calculations in the proof of the * proposition*. figure [ random_traj ] illustrates the behavior of the pursuer for a randomly - varying steering control @xmath10 of the evader , and figures [ random_gamma ] and [ random_gamma_initial ] show the corresponding behavior of @xmath44 .
similarly to figure [ sine_gamma ] , figure [ random_gamma ] shows that increasing the feedback gain @xmath78 by a factor of three decreases the peak difference between @xmath47 and @xmath101 by a factor of about @xmath146 .
figure [ random_gamma_initial ] shows the initial transient in @xmath44 for @xmath111 small .
as would be expected , increasing the gain @xmath78 increases the convergence rate .
( the time axes for figures [ random_gamma ] and [ random_gamma_initial ] differ by a factor of 200 , which is why the initial transient can not be seen in figure [ random_gamma ] . ) finally , figure [ circle_traj ] illustrates the behavior of the pursuer for a constant steering control @xmath10 , resulting in circling motion by the evader .
there is a vast literature on the subject of missile guidance in which the problem of pursuit of an ( evasively ) maneuvering target by a tactical missile is of central interest .
a particular class of feedback laws , known as _ pure proportional navigation guidance _ ( ppng ) occupies a prominent place @xcite . for planar missile - target engagements , the ppng law determining the steering control for the missile / pursuer is @xmath147 where @xmath148 denotes the rate of rotation ( in the plane ) of the line - of - sight ( los ) vector from the pursuer to the evader . here
the gain @xmath149 is a dimensionless positive constant known as the navigation constant .
notice that our motion camouflage guidance law ( mcpg ) given by ( [ planarupgain ] ) has a gain @xmath78 which has the dimensions of @xmath150^{-1 } $ ] .
also , it is easy to see that @xmath151 so , to make a proper comparison we let @xmath77 as in section iii be a length scale for the problem and define the dimensionless gain @xmath152 thus , our mpcg law takes the form @xmath153 it follows that motion camouflage uses range information to support a high gain in the initial phase of the engagement , ramping down to a lower value in the terminal phase @xmath154 . in nature
this extra freedom of gain control is particularly relevant for echolocating bats ( see @xcite ) , which have remarkable ranging ability .
analysis of the performance of the ppng law is carried out in @xcite , using arguments similar to ours ( although our sufficient conditions appear to be weaker ) . while motion camouflage as a strategy is discussed in @xcite , under `` parallel navigation , '' to the best of our knowledge ,
the current work is the first to present and analyze a feedback law for motion camouflage .
in work under preparation , we have generalized the analysis to the three - dimensional setting , and to planar motion camouflage with respect to a finite point .
the three - dimensional analysis is made possible by the use of natural frenet frames , analogously to the three - dimensional unit - speed particle interaction laws described in @xcite .
because we are able to treat the motion camouflage problem within the same framework as our earlier formation control and obstacle - avoidance work @xcite-@xcite , we would like to understand how teams of vehicles can make use of motion camouflage , and whether we can determine the convergence behavior of such systems .
various biologically - inspired scenarios for motion camouflage with teams have been described in @xcite .
considering additional military applications without biological analogs , there are thus a variety of team motion camouflage problems to study .
the authors would like to thank m.v .
srinivasan of the research school of biological sciences at the australian national university for valuable discussions and helpful comments on an earlier draft of this paper .
this research was supported in part by the naval research laboratory under grants no . n00173 - 02 - 1g002 ,
n00173 - 03 - 1g001 , n00173 - 03 - 1g019 , and n00173 - 04 - 1g014 ; by the air force office of scientific research under afosr grants no .
f49620 - 01 - 0415 and fa95500410130 ; by the army research office under oddr&e muri01 program grant no .
daad19 - 01 - 1 - 0465 to the center for communicating networked control systems ( through boston university ) ; and by nih - nibib grant 1 r01 eb004750 - 01 , as part of the nsf / nih collaborative research in computational neuroscience program . | motion camouflage is a stealth strategy observed in nature .
we formulate the problem as a feedback system for particles moving at constant speed , and define what it means for the system to be in a state of motion camouflage .
( here we focus on the planar setting , although the results can be generalized to three - dimensional motion . )
we propose a biologically plausible feedback law , and use a high - gain limit to prove accessibility of a motion camouflage state in finite time .
we discuss connections to work in missile guidance .
we also present simulation results to explore the performance of the motion camouflage feedback law for a variety of settings . |
in the last few years band offset engineering " ( _ i.e. _ the possibility of tuning the electronic and transport properties of semiconductor heterojunctions through modifications of their valence band offset ( vbo ) induced by strain , growth process , number of layers etc . )
has attracted great attention for both scientific and technological reasons @xcite . only recently , however , have lattice mismatched heterostructures begun to realize their potential , due to new developments in preparation techniques which finally allowed pseudomorphic crystal growth without misfit dislocations @xcite . in a parallel way ,
some theoretical works @xcite focused on strained heterojunctions , but still much effort is needed to understand what mainly affects the band line - up at the interface . in this work , we consider homopolar isovalent heterostructures .
in particular , we examine iii - v superlattices ( sls ) , namely the common - anion system gasb / insb ( lattice mismatch of 5.7 @xmath0 ) and the common - cation system inas / insb ( lattice mismatch of 6.4 @xmath0 ) , concentrating on the effect of ordering direction and strain conditions determined by sl pseudomorphic growth on a given substrate . _
ab - initio _ self - consistent full potential linearized augmented plane wave flapw @xcite local density calculations were performed for ( gasb)@xmath1/(insb)@xmath1 and ( inas)@xmath1/(insb)@xmath1 sls , grown along the [ 001 ] ( tetragonal symmetry ) and [ 111 ] ( trigonal symmetry ) directions , in which the interface was represented by three alternating layers of each binary constituent ( 12 atoms in each unit cell ) . in what follows ,
it is important to keep in mind that if strain is not taken into account , the valence bands in unstrained gasb and insb are predicted to line - up ( therefore giving zero vbo ) , whereas the unstrained insb topmost valence band is expected to be 0.51 ev _ higher _ in energy than that in unstrained inas ( see ref.@xcite and references therein ) .
as pointed out by mailhot and smith @xcite , [ 111 ] ordered strained - layer sls show large polarization fields oriented along the growth direction , which lead to a shift in the electronic energy levels .
however , this effect is negligible for ultrathin sls , such as those examined in the present work .
moreover , the effect completely vanishes for [ 001 ] ordered sls , due to the symmetry properties of the strain tensor .
finally , we neglect interdiffusion processes which lead to interfacial composition changes ( i.e. we consider an atomically abrupt geometry ) and relaxations at the interface of the anion - cation distance ( bulk bond length away from the interface are considered equal to those immediately next to it ) : this , in fact , is expected @xcite to introduce little modification of the charge rearrangement at the junction and hence of the vbo .
most of the computational parameters are common to those used previously for [ 001 ] and [ 111 ] ordered ( 1x1 ) sls @xcite , except for the wave function cut - off ( @xmath2 = 2.7 a.u.).[17 ] tests performed by increasing the @xmath2 up to 3.1 showed a change in the vbo of less than 0.01 ev .
there is , however , an important difference with respect to the calculations performed for ultrathin ( @xmath3 = 1 ) sls regarding the cation @xmath4-shell : in the present work , the ga 3@xmath4 and in 4@xmath4 electrons are treated as part of the core and not as valence electrons ( previous works @xcite demonstrated a slight dependence ( about 0.03 ev ) of the vbo in common - cation systems on the @xmath4-shell treatment ) - which results in a strong reduction of the computational effort .
furthermore , the core charge spilling out of the ga and in muffin tin spheres was treated using an exact overlapping charge method , thus minimizing the error introduced by the treatment of semicore states .
the structural parameters ( reported elsewhere @xcite ) are determined according to the macroscopic theory of elasticity ( mte ) , whose validity in predicting the correct structure for the determination of the vbo is established @xcite .
our choice is also justified by the results obtained in the case of ultrathin sls @xcite , which were found to be in good agreement with those obtained from total energy minimization . in order to study the dependence of the vbo on the strain
, we examined different strain conditions for ( ac)@xmath1/(bc)@xmath1-type [ 001 ] ordered sls ( the common cation case ( ab)@xmath1/(ac)@xmath1 is treated in an analogous way ) : ( i ) pseudomorphic growth of a bc epilayer on an ac substrate ; ( ii ) free standing mode " , equivalent to a system grown on an a@xmath5b@xmath5c substrate ( denoted in the following as av . subs . " ) ; and ( iii ) pseudomorphic growth of an ac epilayer on a bc substrate . on the other hand
, the dependence of the vbo on the ordering direction is studied through a comparison of the [ 001 ] and [ 111 ] ordered sls grown on a fixed substrate with average lattice constant , but different crystallographic orientations .
this choice of the substrate and the consequent lattice relaxation , leads to a small difference ( about 1.4 @xmath0 ) between the lattice constants of the binary constituents along the [ 001 ] and [ 111 ] growth direction .
in analogy with the common experimental approach followed in photoemission measurements , we have evaluated the vbo using core electron binding energies as reference levels @xcite .
we have chosen the @xmath6-levels of the common atom c ( _ i.e. _ sb in the common - anion system and in in the common - cation system ) .
note that other choices of core levels for different atoms ( _ i.e. _ ga and in in the common - anion system , as and sb in the common - cation system ) would produce a vbo value differing from those reported here by at most 0.06 ev , which has thus to be considered as our numerical uncertainty .
the calculation of the vbo , @xmath7 , is done according to the following expression : @xmath8 where the interface term @xmath9 indicates the relative core level alignment of the two c atoms at opposite sides of the interface ( one belonging to the ac side and the other to the bc side ) , while @xmath10 indicates the binding energy difference ( relative to the valence band maximum ( vbm ) ) of the same core levels evaluated in the binary constituents , opportunely strained to reproduce the elastic conditions of the sl .
first of all , we focus our attention on the elastically relaxed av . subs . "
[ 001 ] and [ 111 ] ordered sls .
we should now notice that there are two inequivalent interfaces along the [ 111 ] growth axis @xcite ; for example , in the common - anion sl , we can have the ordering direction parallel either to the insb interface bond or to the gasb interface bond .
our results indicate that the @xmath9 term is essentially the same ( within 0.02 ev ) in the two different situations , suggesting that the effect due to the particular geometry at the interface is very small .
table [ vbo1 ] lists the contributions due to the interface ( @xmath9 ) and to the strained bulks ( @xmath10 ) and the resulting values of the vbos ( @xmath7 ) as a function of the ordering direction ; the superscripts @xmath11 and @xmath12 indicate respectively the non - relativistic and relativistic ( _ i.e. _ spin - orbit coupling treated in a perturbative approach ) calculations .
we note that the interface term has a positive sign , indicating that the sb - core levels are deeper at the gasb ( inas ) side of the common - anion ( common - cation ) interface , compared to the corresponding levels at the insb side .
further , the first contribution ( @xmath13 is seen to be sensitive to the crystallographic ordering ( the two values for [ 001 ] and [ 111 ] growth axis differ by 0.1 ev both in the common - anion and the common - cation systems ) , while the second contribution to the vbo ( @xmath10 ) is almost uninfluenced by the ordering direction . on the whole , we do not observe a marked dependence of the vbo on the crystallographic ordering at the interface . since it is well - known for lattice matched structures @xcite that the band line - up is independent of interface orientation , the vbo change we find in going from the [ 001 ] to the [ 111 ] ordered sls has to be related only to the appreciable mismatch which causes a different relaxation of the interface bond - lengths .
let us now look at the role of the mismatch in determining the band line - up .
our results for [ 001 ] systems with different pseudomorphic growth conditions are shown in table [ vbo2 ] , where the notation is analogous to that of table [ vbo1 ] .
note that the interface term @xmath9 is very similar in the same sl grown on the three different substrates , implying that the charge readjustment at the interface is almost independent of the strain conditions .
this is confirmed by the results obtained with a small change ( by as much as 0.6 @xmath0 ) of the bond length at the interface : the calculated @xmath9 is consistent with the one obtained for mte structures within 0.03 ev .
on the other hand , the @xmath10 term ( _ i.e. _ the bulk contribution to the vbo ) varies dramatically , showing that the core level binding energies in the strained binary suffers an appreciable change when growing the sl on different substrates .
in fact , the energy of the topmost valence level ( and hence the binding energy @xmath14 ) is determined by the interplay of the spin - orbit coupling and the non - cubic crystal field " @xcite .
in particular , the second of these two effects is critically dependent on strain conditions , and is thus the origin of the large difference between the @xmath15 in table [ vbo2 ] .
furthermore , what is remarkable about table [ vbo2 ] is the clear trend shown by the vbo as a function of the substrate lattice constant : the smaller the @xmath16 , the more the insb topmost valence level is raised with respect to the vbm of the other sl constituent . in order to understand more fully the action of the strain on the vbo
, we have also estimated the band line - up with respect to unstrained binaries for the [ 001 ] interfaces .
thus , we substituted in eq.([equazione ] ) the @xmath9 value obtained for [ 001 ] interfaces ( which was found to be almost independent of strain effects ) and the @xmath10 , evaluated starting from the zincblende bulk unstrained constituents ( i.e. disregarding the effect of strain on the binary s vbm ) .
taking into account the spin - orbit coupling , we obtain @xmath17 = 0.03 ev and @xmath17 = 0.49 ev for the gasb / insb and inas / insb heterojunctions , respectively .
these results match perfectly those reported in ref .
@xcite and the above mentioned valence bands perfect alignment for the common anion system and @xmath17 = 0.51 ev for the common cation case , showing that , if strain is not correctly taken into account , completely different results are found .
it is important to notice that the strain acting on the energy of the topmost valence band level is also responsible for the spatial localization of this state .
we find , in fact , that in all of the common anion ( common cation ) structures , the hole carriers are mainly localized on the insb side of the heterojunction , while in the gasb / insb system grown on an insb substrate we find a complementary situation .
this is clear from the decreasing trend of the vbo as the lattice constant is increased and from the sign change in the insb substrate case ( showing that the vmb in gasb is higher in energy than in insb ) .
figures [ fig1 ] and [ fig2 ] , for common - anion and common - cation interfaces respectively , illustrate the linear dependence ( see the solid line in the figures ) of the vbo on the lattice parameter which determines the sl pseudomorphic growth . thus , the two figures show that the gasb / insb and inas / insb sls provide a good opportunity for tuning their vbo : a range of about 0.5 ev for common - anion and of 0.7 ev for common - cation systems is covered by varying the strain conditions determined by the substrate .
let us now compare our results with other theoretical predictions , obtained from model @xcite , semi - empirical @xcite and _ ab - initio _ @xcite calculations , as illustrated in figures [ fig1 ] and [ fig2 ] .
note that all the predicted values agree with those of the present work ( except those of ref .
@xcite ) , within their uncertainty of a few hundredths of an ev @xcite and our error bars , respectively .
incidentally , we observe that a similar disagreement between _ ab initio _ results and those obtained by cardona and christensen @xcite was also found in other iii - v isovalent heterojunctions , such as gap / inp @xcite and gaas / inas @xcite . furthermore , the linear trend of the band offset as a function of the strain found in the present work is in excellent agreement with the predictions of other theoretical work @xcite and is reasonably expected to reproduce the real situation .
so far , we have completely omitted a discussion of the conduction band offset ( @xmath18 ) , due to well known failures of lda in predicting the correct band gap energies .
however , we should now point out that , using an approximate estimate of empirical band gaps ( see ref .
@xcite for details ) , we can obtain information on the different kinds of band line - ups as growth conditions are changed .
in fact , we find a type i alignment for all the [ 001 ] common anion interfaces , while [ 111 ] gasb / insb shows a type ii staggered alignment ( partial overlap of the band gaps ) . on the other hand , for [ 001 ] common
cation grown both on an inas and on an average substrate , we find a type ii broken gap line - up , with the inas conduction band minimum lower in energy than the insb vbm . finally , the [ 001 ] inas / insb grown on insb and
the [ 111 ] inas / insb heterojunctions contain a semimetallic compound ( inas ) , thus leading to a type iii alignment .
starting from the results obtained from _ ab initio _
calculations @xcite for different [ 001 ] oriented strained layer interfaces ( with similar lattice mismatch and grown on a substrate having lattice constant averaged over those of the constituents ) , it is interesting to discuss the vbo trend as a function of the atomic species involved .
we compare the value obtained for the common anion gasb / insb ( lattice mismatch of 5.7 @xmath0 and @xmath7 = 0.07 ev , in good agreement with @xmath7 = 0.04 ev reported in ref .
@xcite ) , with those obtained for gap / inp @xcite ( lattice mismatch of 7.4 @xmath0 and @xmath7 = 0.01 ev ) and for gaas / inas @xcite ( lattice mismatch of 5.7 @xmath0 and @xmath7 = 0.00 ev ) . taking into account that these results are obtained by different computational methods and are therefore affected by different error bars , it appears that , under similar strain conditions , the vbo is almost uninfluenced by the change of both anions in the common anion systems
. therefore , since the @xmath9 term is expected to be constant ( due to similar ionicity difference of the constituents ) , these results seem to suggest that also the @xmath10 term has to be similar in all the gax / inx ( x = p , as , sb ) structures
. a similar @xmath7 is also found , if we compare our results for inas / insb ( lattice mismatch of 6.4 @xmath0 and @xmath7 = 0.54 ev ) with gaas / gasb ( lattice mismatch of 7.2 @xmath0 ) , where both cations are changed ( @xmath7 = 0.65 @xmath19 0.1 ev @xcite ) .
this observation can be explained , considering the anionic character ( see ref .
@xcite , ref . @xcite and references therein ) of the topmost valence level in the binary constituents , which determines the binding energy contribution ( @xmath10 ) to the valence band offset : for example , provided that the state of strain is similar in the constituent materials grown on an average substrate , we do nt expect a strong difference in the band line - up if we change the cation at both sides of the common cation interface . from the experimental point of view , the high mismatch ( about 6 @xmath0 ) between the lattice constants of the two sl constituents results in great difficulty to grow the sl systems , without misfits and dislocations .
only recently , an insb quantum - well has been realized in gasb @xcite ; starting from photoluminescence peak emission energy data and from calculations based on a standard finite square - well model @xcite ( taking into account strain @xcite ) , a vbo of 0.16 ev was obtained that is quite different from the one , 0.34 ev , reported in table [ vbo2 ] .
unfortunately , a similar disagreement between theoretical and experimental data , obtained using different techniques , was found also for other homopolar isovalent iii - v interfaces , such as the gaas / inas ( see ref . @xcite and references therein ) and gap / inp ( @xmath7 = 0.01 ev , the theoretical result @xcite against @xmath7 = 0.60 ev , the experimental results from photoluminescence @xcite ) systems , that are reasonably close to the common - anion one studied here .
in particular , in a recent work focused on gaas / inas sls , ohler _ et al . _
@xcite obtained ( from ultraviolet photoelectron spectroscopy measurements of the cation @xmath4-core levels ) a @xmath7 value in disagreement ( by as much as 0.3 ev ) with theoretical predictions @xcite .
notwithstanding some differences between the experimental and theoretical vbo values , many important observations discussed above are confirmed by ohler _
et al . _ experimental work @xcite .
for example , the linear trend found @xcite for @xmath20 ( in gaas / inas sls ) as a function of @xmath16 agrees with theoretical predictions @xcite and with our results ; further , the independence of the @xmath21 term on the strain conditions discussed above , is experimentally confirmed @xcite by the trend of the in 4@xmath22 and ga 3@xmath23 core - level binding energy difference , which is almost unaffected by the different substrate used in growing the heterostructure . as a last comment
, we think that it worthwhile to remark that the linear trend found in ref .
@xcite for the gaas / inas vbo s as a function of the growth conditions leads to a slope that is almost equal to that of fig .
[ fig1 ] for gasb / insb systems .
starting from the band offset transitivity rule - which is well established ( to within 0.02 ev ) @xcite for [ 001 ] common - atom superlattices - we can derive the vbo for the inas / gasb system , an almost lattice matched interface ( the two lattice constants differ only by 0.6 @xmath0 ) .
this system is attracting more and more attention recently for its unusual type ii broken - gap band line - up . through a linear interpolation of the common - anion band offsets as a function of the substrate lattice constant ,
we have calculated the vbo for the gasb / insb system as if grown on an inas substrate . as evidenced above ( see fig .
[ fig1 ] ) , the linear approximation is expected to be reasonably valid ; furthermore , in this case the extrapolation is obtained for a lattice constant ( @xmath24 ) which differs by only 0.7 @xmath0 from one of our self - consistent results ( @xmath25 ) .
we thus report in fig .
[ fig3 ] our calculated vbo for the common - cation sl on an inas substrate and the extrapolated value for the ideal common - anion sl grown on an inas substrate . using the transitivity rule ,
we obtain : @xmath26 so that @xmath27 = 0.88 - 0.40 = 0.48 ev .
this result is in good agreement with the available experimental values ( 0.46 ev @xcite , 0.51 ev @xcite ) .
in summary , we have studied the valence band offsets in [ 001 ] and [ 111 ] gasb / insb and inas / insb interfaces by means of _ ab - initio _ flapw calculations , focusing our attention on its dependence on ordering direction and strain conditions .
our results indicate that , under the same strain conditions , the former has quite a small effect on the band line - up mainly due to the different structural relaxation of the interface atoms at the [ 001 ] and [ 111 ] heterojunctions . on the other hand ,
a much more important effect is due to pseudomorphic growth on different substrates : the high tunability of the vbo ( about 0.5 ev and 0.7 ev for common - anion and for common - cation sls , respectively ) is evidenced by its linear decreasing trend as the substrate lattice constant is increased , mainly due to the bulk contribution to the band line - up .
finally , the transitivity rule was used to determine the inas / gasb valence band offset and good agreement between theory and experiment was obtained .
we thank b. w. wessels and m. peressi for stimulating discussions and a careful reading of the manuscript .
useful discussions with s. massidda are also acknowledged .
work at northwestern university supported by the mrl program of the national science foundation , at the materials research center of northwestern university , under award no .
dmr-9120521 , and by a grant of computer time at the nsf supported pittsburgh supercomputing center .
partial support by a supercomputing grant at cineca ( bologna , italy ) through the consiglio nazionale delle ricerche ( cnr ) is also acknowledged .
.interface term ( @xmath9 ) , strained bulk term ( @xmath10 ) and valence band offset ( @xmath7 ) for elastically relaxed ( gasb)@xmath1/(insb)@xmath1 and ( inas)@xmath1/(insb)@xmath1 -average substrate- superlattices as a function of the ordering direction ( @xmath28 and @xmath29 ) and including ( @xmath30 and @xmath31 ) spin - orbit effects ) .
energy differences ( in ev ) are considered positive if the level relative to the insb layer is higher in energy with respect to the gasb ( inas ) layer in the common - anion ( common - cation ) system . [ cols="^,<,>,^,^,^,^",options="header " , ] s. baroni , m. peressi , r. resta , a. baldereschi , _ theory of band offsets at semiconductor heterojunctions _ , in _ proceedings of the 21@xmath32 international conference on the physics of semiconductors _ , edited by ping jiang and hou - zhi zheng ( world scientific , singapore , 1993 ) , p. 689 . | first - principles full potential linearized augmented plane wave ( flapw ) calculations have been performed for lattice - mismatched common - atom iii - v interfaces . in particular , we have examined the effects of epitaxial strain and ordering direction on the valence band offset in [ 001 ] and [ 111 ] gasb / insb and inas / insb superlattices , and found that the valence band maximum is always _ higher _ at the insb side of the heterojunction , except for the common - anion system grown on an insb substrate .
the comparison between equivalent structures having the same substrate lattice constant , but different growth axis , shows that for comparable strain conditions , the ordering direction slightly influences the band line - up , due to small differences of the charge readjustment at the [ 001 ] and [ 111 ] interfaces . on the other hand
, strain is shown to strongly affect the vbo ; in particular , as the pseudomorphic growth conditions are varied , the bulk contribution to the band line - up changes markedly , whereas the interface term is almost constant . on the whole ,
our calculations yield a band line - up that decreases linearly as the substrate lattice constant is increased , showing its high tunability as a function of different pseudomorphic growth conditions .
finally , the band line - up at the lattice matched inas / gasb interface determined using the transitivity rule gave perfect agreement between predicted and experimental results . |
we provide a short review of the first seven years of observing with the taurus tunable filter at the aat ; for the first three years , the ttf was also used at the wht .
the instrument has had most of its use on the aat where it gets between 1015% of the telescope time scheduled by the patt ( uk ) and atac ( australia ) committees . in a nutshell
, the ttf allows for wide - field ( 10@xmath2 ) spectrophotometric imaging from 370 nm to 1000 nm with resolving powers generally in the range 100 to 1000 .
an important feature of the instrument is the use of charge shuffling synchronized to band switching in order to greatly suppress systematic errors associated with conventional imaging .
another aspect of the ttf is time series readout coupled to band switching which has lead to important new work on compact variable sources .
technical accounts of the instrument , and its related charge shuffling modes , can be found in appendices a and b. there have been several recent reviews on tunable filters including bland - hawthorn ( 2000@xmath3 ) , jones ( 2001@xmath4 ) and jones ( 2001 ) .
recent scientific reviews can be found in veilleux ( 2002 ) and veilleux ( 2002@xmath3 ) .
a list of largely unexplored science areas with tunable filters is given in bland - hawthorn ( 2001 ) .
the purpose of this review is to stress the versatility of tunable filters in observational astronomy .
the taurus focal reducer has also produced important results in its other modes of operation ( see appendix a ) : polarimetry , fabry - perot ` staring ' and emission line scanning , and most recently multi - object spectroscopy ( e.g. riddick , lucas & roche 2003 ) .
but here we concentrate specifically on the ttf mode of operation .
all observations discussed here were undertaken on 4 m telescopes , often in non - ideal observing conditions .
these authors eagerly anticipate the tunable filters which are planned or under way for the new generation optical / ir telescopes on superb sites .
these include , inter alia , the osiris tunable filter on the grantecan 10.2 m telescope ( cepa 2002 ) , a possible tunable filter within fors on the vlt , the proposed maryland magellan tunable filter ( mmtf ) , and the tunable filter within the goodman spectrograph on the soar 4 m telescope .
when one reviews the results below , it is important to keep in mind that the new generation of tunable filters need only improve on one of the following site conditions , instrument performance , field of view , pixel sampling , telescope aperture to achieve a major gain .
tunable filters are ideally suited to surveys of star forming galaxies in different environments .
the object selection is based on the property we are trying to measure , the star formation rate via the line .
this was the principle behind the ttf field galaxy survey which was the basis of d.h .
jones thesis ( 2001 ) .
he obtained photometric data and restricted data on 4 clusters and 9 field positions .
all observations were highly successful in identifying line emitting galaxies , typically @xmath5 objects above @xmath6 per pointing ( although rather fewer in ) , finding many more objects than comparable studies with conventional imaging techniques ( hu & mcmahon 1996 ; thommes 1998 ) .
the first part of ttf field galaxy survey was published in jones & bland - hawthorn ( 2001 ) .
recent evidence suggests a decline in the volume - averaged star formation rate ( sfr ) with the advance of cosmic time since @xmath7 .
the survey set out to derive luminosity functions in three discrete wavelength intervals at @xmath8 , 0.24 , and 0.39 .
one of the interesting surprises was the discovery of a population of compact sources found to have moderate amounts of emission : some of these sources would have been dismissed as stars in earlier photographically selected surveys .
the ttf field galaxy survey measures volume - averaged star formation rates at intermediate redshift that are a factor of @xmath9 times higher than those measured from multi - slit or multi - fibre surveys .
since these results have been largely confirmed by the cadis emission line survey ( hippelein et al 2003 ) and the subaru emission line survey ( fujita et al 2003 ) , it is likely that the restricted apertures of fibres and slits have underestimated the total h@xmath10 line flux at low redshift . in a new study of galaxies in the canada - france redshift survey ( cfrs ) , tresse ( 2002 )
combine the ttf field galaxy survey measurements with their isaac / vlt luminosity function at @xmath7 ( see fig . 1 ) .
they find that the comoving luminosity density increases by a factor of 12 from @xmath11 to @xmath12 .
their results confirm a strong rise of the star formation rate at @xmath13 proportional to @xmath14 .
tober ( 2003 ; see also glazebrook 1997 ) carried out a deep survey of 20 square arcminutes in the hubble deep field north ( hdf - n ) using the taurus tunable filter and william herschel telescope ( wht ) to observe in contiguous sequences of narrow band slices in the 710@xmath15910 nm region of the spectrum .
these authors sampled almost 70 narrowband images in four redshift intervals . in this way
, one gets a _ multiplicity effect _ in that the same images can be used to find different sources at different redshifts , thereby allowing for a derivation of the star formation rate over a wide range in redshift .
the data , which were taken during four long dark nights at the wht , have excellent image quality with psf @xmath16 0.7 fwhm after combining the full data set .
cross - matching with keck spectroscopy of the hdf - n results in a line - luminosity limited sample , with very highly complete redshift identification , containing oii , and emitters over the @xmath17 . from this direct star - formation rate selected sample
they reconstruct , via maximum likelihood techniques , the star - formation history of the universe to @xmath18 .
their study finds no evidence for any new population of low - luminosity strong - lined objects to an approximate flux limit of @xmath19 w m@xmath20 and covering a cosmological volume of 1000 mpc@xmath21 at @xmath18 . from their complete narrow - band sample of a single sight line ,
tober find a star - formation history consistent with earlier estimates from broad - band selected samples .
2.7 in 2.7 in until now , identifying the majority of cluster members from a combination of kinematics and a colour - magnitude diagram has been difficult , particularly for blue objects , due to contamination of sightline galaxies . but ttf observations can be used to identify faint cluster members unambiguously , in particular those with line emission .
these objects are typically blue so the ttf is an ideally suited to unscrambling the region of contamination ( see fig .
2(_right _ ) ) .
we find that k@xmath22a galaxies ( dressler & gunn 1983 ) can also be identified from absorption ( jones & bland - hawthorn 1999 , fig . 1 ) . in fig
, we demonstrate the power of ttf to find line emitters even in poor seeing ( 2@xmath23 ) .
dalcanton ( 1996 ) and zaritsky ( 1997 ) use ccd drift scanned images to identify clusters through enhanced surface brightness fluctuations out to @xmath24 .
for one cluster at @xmath25 , in just 2 hours , we identified 10 cluster members at the aat in 2 seeing , compared to 6 members in 1.5 nights using the keck lris spectrograph .
for the rich cluster a3665 ( ac 106 ) , first studied by couch & newell ( 1984 ) , zaritsky & jones identify 40 emitting candidates above 3@xmath26 , with many more candidates at lower thresholds ( down to 0.03 solar masses per year at @xmath27 ) , in just four hours .
example ttf scans are shown in fig .
3(_right _ ) , with the line profiles in fig .
3(_left _ ) . to date
, average star formation rates in @xmath28 clusters appear to be quite low compared to the field ( cf .
couch 2001 ) .
baker ( 2001 ) used the [ oii]372.7 nm line to detect starforming galaxies in the vicinity of mrc b0450 - 221 .
many emission line galaxies can be visible over the whole field ( down to a few solar masses per year ) .
their projected separations from the quasar range from 200 to 700 kpc .
a nice demonstration of the power of ttf imaging is the detection of very faint extended [ oii ] emission along the radio axis of the quasar .
francis ( 2001@xmath4,@xmath29 ) have continued their study of the ` francis cluster ' , an extended ensemble of galaxies at @xmath30 that was originally identified from the damped ly@xmath10 absorption occurring at the same redshift in a paired quasar sightline . in this study , they look for ly@xmath10 emission from star forming galaxies within the ensemble .
it is clear that the future holds great promise for using quasars to identify gas - rich starforming environments at high redshift .
there have been claims of euv excesses in clusters which , if true , would suggest that 10@xmath31k gas accounts for a large fraction of missing baryons ( lieu 1999 ; bowyer 1999 ) .
euv emission is exceedingly difficult to detect at earth : it is too hot to be detected optically , too cool to be seen by x - ray satellites , and undergoes molecular line absorption as it propagates through the ism . using the euve satellite , lieu ( 1999 ) claim a direct euv detection of the cluster abell 1795 .
maloney & bland - hawthorn ( 2001 ) found that such a strong euv field would ionize the molecular and hi disks of all spirals in the cluster to such an extent that they should all be bright .
ttf observations of the cluster were kindly taken by a. edge in ` straddle shuffle ' ( maloney & bland - hawthorn 2001 ) mode in order to subtract the continuum light of the cluster very accurately .
only very faint emission was found in the cluster spirals .
jaffe & bremer ( 2000 ) used ` on - off ' charge shuffling in order to detect very faint levels of emission in cooling flow clusters . in the case of abell 2597
, they find a nebula extending over 50@xmath1560 kpc from the cluster centre .
a. edge and collaborators have looked at a large sample of cooling flow clusters and detect optical line emission in all cases .
tunable filters are ideally suited to observe extended sources , in particular , irregularly shaped sources with no axis of symmetry .
earlier work with d.h .
jones on star formation in clusters occasionally turned up spectacular gravitational lenses throughout the cluster .
as seen through the ttf , these sources can have high contrast because the background object is commonly a high - redshift starforming galaxy which emits strongly in emission lines .
hewett ( 2000 ) used the ttf to detect extended emission in a galaxy - galaxy gravitational lens due to an intermediate redshift elliptical lensing a @xmath32 star forming galaxy .
they were following up a suggestion by miralde - escud & lehar ( 1992 ) that the surface density of high - redshift starforming galaxies is so high , galaxy - galaxy lensing should be relatively common .
the tunable filter has the major advantage of providing detailed spectrophotometric information over a larger field of view ( fov ) than that of other 3d instruments .
it would be very difficult and very expensive to build an integral field spectrograph which could provide the same field and the same photometric integrity at low light levels .
the tunable filter is therefore ideally suited to study nearby galaxies where the line - emitting gas extends over several arcminutes .
this ionized material is an excellent probe of the phenomena taking place in the core of starburst and active galaxies , and can be used to quantify the impact of nuclear and star - formation activity on the environment and vice - versa . once again , the reader is encouraged to look at recent reviews by veilleux ( 2002@xmath3 ) .
tadhunter and collaborators have undertaken a comprehensive survey of extended ionized haloes around powerful radio galaxies ( tadhunter2000 ; solorzano - inarrea 2002 ; solorzano - inarrea & tadhunter 2003 ) .
it is well known that optical line emission often aligns with the radio axis , as these authors find in all cases .
however more surprising was the frequency of optical emission perpendicular to the radio axis . in fig . 4
, the optical emission in coma a extends over more than 150 kpc in an egg - shaped nebula ( tadhunter 2000 ) .
more remarkably , the faint outer nebula emission matches the extended outer radio lobe over these same scales .
the frequent match - up between optical , radio and x - ray emission in radio galaxies , seyfert and starburst galaxies is a topic of great interest ( see below ) . in the case of low power radio jets ,
the emission is thought to arise from sychrotron , whereas in high power radio jets , the x - rays appear to arise from inverse compton emission ( birkinshaw 2001 ) .
shopbell ( 1999 ) recently published a remarkable extended nebula surrounding the x - ray selected quasar , mr 2251 - 178 .
the summed image in fig .
5 is one of the faintest extended emission - line images ever published .
the nebula , which extends over more than 200 kpc , shows pronounced tidal arm structure .
these authors chose two narrow bands spaced by half a bandwidth in order to study the kinematics .
the entire structure appears to be in rotation about the quasar .
the gas is not obviously connected with any other companions in the field .
uv radiation from the quasar would have to be escaping almost isotropically to account for the ionization .
both the source of the gas and the source of ionization remain a mystery .
seyfert galaxies often show the impact of the nuclear activity on the very large scale disk of a spatially resolved galaxy in the near field .
two of the most recent and spectacular examples are ngc 1068 and ngc 7213 .
can we use the ism as a screen to interpret what is essentially unresolved and unseen at the core ?
ngc 1068 is one of the most remarkable of seyfert galaxies in the near field .
it is well known that the core harbours a concealed seyfert 1 ( ` low power quasar ' ) nucleus , but just how energetic is the source ?
the radio jet axis exhibits a wide range of activity ( cecil 2002 ) but can we infer the true energetics and nuclear spectrum from multi - wavelength studies ( alexander 2000 ; pier & krolik 1993 ) ?
shopbell ( 2001 ; see also veilleux 2001 ) find that the well known ionization cone ( pogge 1988 ) extends up to and _ beyond _ the hi edge of the early - type spiral .
the observed nebula requires rather special conditions to be detectable over such a huge extent .
the most likely interpretation is that a multi - phase , vertically distended ism is being blasted by a very energetic central source ( @xmath33 erg s@xmath34 ) .
indeed , the chandra observations appear to show the most spectacular example to date of an ` x - ray ionization cone ' ( young 2001 ) .
these galaxy - scale ionization cones are not unique to ngc 1068 .
another has been detected with the ttf in ngc 7213 ( rupke 2002 ; veilleux2001 ) to rival the spectacular system in ngc 5252 studied by tadhunter & tsvetanov ( 1989 ) .
ionization cones are often associated with nuclear jets .
recent examples from ttf observations include ic 5063 ( cecil 2001 ) and circinus ( veilleux & bland - hawthorn 1997 ) .
circinus , a large spiral close to the galactic plane , is particularly noteworthy : the system shows evidence for a whole network of ` artillery shells ' blasting away from the nucleus .
one such filament shows a spectacular ` herbig haro'-like morphology .
extended ionization cones are seen in a variety of sources , including the radio galaxies studied by tadhunter ( see above ) .
one of us has obtained a short exposure [ oiii ] image of cen a using the straddle shuffle mode in order to achieve a clean subtraction of the elliptical galaxy . in fig .
6 , we find a very highly ionized halo of [ oiii ] emission above and below the dust lane .
the component was first discovered by phillips ( 1984 ) who found that the faint emission appeared to rotate slowly compared to the dust lane , and typical line dispersions are hundreds of fwhm .
we suspect that this emission arises from a highly ionized wind which encompasses the radio jet over a much larger solid angle .
cecil ( 2000 ) show that the famous braided jets in ngc 4258 , which can be traced over 10 kpc or more , emanate from the nuclear regions .
these jets have been seen in radio continuum , rosat / chandra x - rays , and optical emission .
they obtained deep observations in the outer hi disk which gave the first detailed picture of faint tendrils of the jets .
a re - analysis of the radio continuum data has revealed that these faint features have radio counterparts .
the ngc 1068 and ngc 4258 observations were able to detect faint emission unseen previously with less sensitive imaging .
it is important to realize that both observations were taken within 20of the full moon , _ which demonstrates the effectiveness of the ttf in any phase of the lunar cycle .
_ a number of powerful starburst galaxies show evidence for large - scale winds along their minor axis ( see reviews by veilleux 2001 and heckman 2002 ) .
this phenomenon appears to be common at low and high redshift ( lehnert & heckman 1996 ; veilleux 2001 ; pettini 2001 ) .
the most detailed studies have concentrated on objects like m82 , ngc 253 and ngc 3079 .
it remains unclear just how much energy , mass and metals these objects contribute to the intergalactic medium ( igm ) . in some cases ,
it is not clear if the wind is driven by a central starburst , an agn or a combination of these .
however there is mounting evidence that the optical diagnostics may greatly underestimate the true wind energetics ( strickland & stevens 2000 ) .
veilleux & rupke ( 2002 ) provide spectacular evidence for a large - scale wind in the edge on , early type galaxy ngc 1482 .
they obtain a clear separation of disk material from the outflowing gas .
the kinematics show all the hallmarks of a biconal outflow .
more impressively , the entire emission line complex has diagnostics which are entirely consistent with fast shocks . in other wind systems
, there is often a large contribution from the central ionizing stellar radiation field .
more recently , veilleux ( 2003 ) have presented results of a pilot imaging study of 10 nearby starburst and active galaxies conducted with the ttf on the anglo - australian telescope .
an important goal of this study was to search for warm emission - line gas on scales larger than @xmath35 10 kpc to constrain the zone of influence of these galaxies .
large - scale structures are discovered or confirmed in four of the galaxies ( ngc 1068 , ngc 1482 , ngc 4388 , ngc 6240 , and ngc 7213 ) .
unsuspected structures are also seen for the first time in the galactic winds of ngc 1365 , ngc 1705 , circinus , and eso484-g036 .
the ttf data are combined with new optical long - slit spectra as well as published and archived radio and x - ray maps to constrain the origin and source of ionization of these filaments .
a broad range of phenomena is observed , including large - scale ionization cones and galactic winds , tidal interaction , and ram - pressure stripping by an intracluster medium . for his thesis work ,
miller ( 2002 ) has undertaken an emission line survey of edge - on spiral galaxies in order to trace the connection of disk star formation with activity in galactic haloes .
miller & veilleux ( 1999 ) and veilleux ( 2001 ) show spectacular examples of diffuse ionized gas several kiloparsecs off the plane of normal spirals .
the ratio maps show enhanced [ nii ] emission with respect to as we move away from the plane which is most easily explained as a higher electron temperature in the halo gas ( cf .
sokolowski 1993 ; reynolds 1999 ; collins & rand 2001 ) .
the source of the temperature increase is a topic of debate in contemporary astrophysics .
miller & veilleux ( 2003@xmath3 ) confirm a strong correlation between the mass of ionized halo gas and the star formation rate per unit disk area .
very little is known about the frequency and nature of star formation in elliptical galaxies , even though we have known for a long time that a large fraction contain cold gas ( van gorkom 1997 ; knapp 1999 ) .
ferguson and collaborators ( 2001 ) have been making observations with ttf of a large sample of hi - selected ellipticals .
essentially all of these systems ( see fig . 7 ) show evidence of star formation to date . for her thesis work ,
cianci ( 2003@xmath3 ) is undertaking a detailed comparison of 20 face - on spirals observed at and , and uv ( 150 nm ) images from the ultraviolet imaging telescope ( see fig .
her particular interest is to understand asymmetries in spiral galaxies , and the connection of the diffuse ionized gas with the hii regions and dust distribution ( cf .
zurita 2001 ; zurita 2000 ) . while spiral galaxies are commonly assumed to be axisymmetric , it is surprising how few galaxies approach this ideal .
the origin of asymmetries may have profound implications for their properties .
an interesting possibility is that the spiral galaxy disk ` sloshes around ' within the dark matter potential ( e.g. syer & tremaine 1996 ; schoenmakers , franx & de zeeuw 1997 ) or that the dark halo is lopsided ( jog 1999 ) .
simulations show that the effects can survive for one or two rotation periods ( @xmath36 yr ) . to date
, studies of lopsidedness have concentrated on gas morphology and dynamics .
but if the simulations are right , asymmetric variations should show up in the young stellar population of any lopsided spiral .
cianci ( 2003 ) may now have detected an @xmath37 asymmetry in the spiral arms of the interacting galaxies m83 and m51 .
the uv continuum is enhanced relative to h@xmath10 along spurs that fall on the outside of the western arm .
simulations with starburst99 indicate that these regions are older than 20 myr , compared to the h - alpha bright regions which appear to be younger than 7 myr .
ngc 2915 is a dark - matter dominated , blue compact dwarf galaxy with an hi disk extending to a radius of 15 kpc . in optical continuum , it is a relatively nondescript with an optical radius of only 3 kpc .
no stars were known to exist beyond this radius .
meurer ( 1999 ) obtained deep imaging with the ttf revealing , for the first time , faint hii regions at projected radii of 3 to 6 kpc .
these have since been confirmed with the advanced camera for surveys on the hubble space telescope ( meurer 2003 ) .
higdon ( 1997 ) have undertaken a multi - line study of the cartwheel galaxy .
this is the most spectacular of the class of ` ring galaxies ' which are excited by the central impact of an interloper .
they identify up to 100 star forming regions throughout the disk which they will model as propagating star formation .
there is a related class of galaxies excited by off - centre impacts .
one of these , ngc 1512 , has been looked at in detail with the ttf by f. briggs .
moreover , there is a wider class of objects under the general heading ` event driven star formation ' ( strong mergers , jet induced star formation , etc . ) which are relatively unexplored with the ttf
. 2.7 in one of the most important areas where very little work has been done is imaging in stellar absorption lines .
beauchamp & hardy ( 1997 ) have demonstrated that this is possible even with small aperture telescopes .
molla , hardy & beauchamp ( 1999 ) emphasize the importance of spatially resolved , stellar absorption line mapping of face - on spiral galaxies .
ryder , fenner & gibson ( 2003 ) have now mapped two spirals , ngc 6221 and ngc 7213 , using the mg_2 @xmath38 and fe @xmath39 features , two of the most prominent lick / ids indices ( worthey et al .
the lick spectral indices do not measure abundances per se , but can be transformed to quantities like [ fe / h ] via the method of spectral synthesis .
the late - type spiral ngc 6221 shows a steep increase in both indices at the core , before dropping back to the nuclear values further out .
this is in contrast to the early - type spiral ngc 7213 which has relatively constant indices in the core , followed by a smooth rise in the outer disk .
bland - hawthorn ( 2001 ) provide a list of stellar absorption line projects which should be considered for tunable filter work .
jones ( 2002@xmath4 ) discovered an extended nebula associated with a pulsar moving rapidly through the interstellar medium .
several of these sources are now known , but in the case of pulsar b0740 - 28 , the conic nebula is pinched perpendicular to the long axis .
possible interpretations are variations in the pulsar wind or in the external ism .
2.7 in schuberth & burton ( 2000 ) used the ttf to observe arguably the most exotic line to date in the optical , the [ ci]872.7 nm line buried deep within the oh forest .
this line is extremely important in photodissociation regions ( pdr ) .
ngc 2023 is an hii region on the surface of a molecular cloud .
these authors postulated that carbon atoms emitting at 872.7 nm might delineate the pdr .
their [ ci ] image ( see fig . 9 ) is the first true image of extended carbon emission ever obtained .
they show that the emission correlates very closely to the so - called mysterious extremely red emission ( ere ) at about 650 nm .
[ ci ] is clearly a powerful tool for probing galactic star forming regions .
there is presently a lot of interest in detecting variability from brown dwarfs .
if the atmospheres and environs are cool enough , models suggest that there should be dust condensations which swirl around the brown dwarf producing variability in the light curve .
this was detected for the first time by tinney & tolley ( 1999 ) in the m - type brown dwarf lp 944 - 20 .
they used ttf in a novel set up which required the charge to step on the ccd in synchrony with the ttf being tuned between different bands .
since then , variability has been found in another brown dwarf kelu-1 ( clarke 2002 ; clarke , tinney & covey 2002 ) .
these sources do indeed appear to have complex weather patterns .
deutsch ( 1998 ) have used ttf in time series mode to study the [ oi]844.6 nm emission line from the x - ray binary star v2116 oph , the optical counterpart of gc1 + 4 .
the symbiotic - like optical spectrum of v2116 oph shows the presence of a red giant .
the x - ray source is highly variable , so the [ oi ] line was studied in the hope of detecting variability .
this was rejected to a high level of confidence .
it is thought that the system is being seen at a special time , and is probably an x - ray pulsar undergoing rapid evolution .
ryder ( 2001 ) was able to image mars in four narrow bands , a difficult experiment since the source is so bright . with four bands set at 390 , 500 , 668 and 707 nm , he produced high quality images of the ice caps , the maria and terae , and various cloud formations .
jeremy bailey ( 2003 , private communication ) has recently emphasized the importance of tunable filters in monitoring weather patterns on mars and venus in narrow spectral bands .
he compares the relative merits of a scanning slit , a tunable filter and an integral field spectrograph on an 8@xmath1510 m telescope .
the demands on planetary science are ( i ) spatial stability , ( ii ) spatial integrity , ( iii ) spatial resolution of 25 mas or better , ( iv ) ability to switch rapidly between bands to maximize the time resolution , and to combat the planet s rotation . all of these requirements are met with a near - infrared tunable filter ( resolving power @xmath40 ) used in conjunction with adaptive optics on a large telescope .
there are numerous ways in which tunable filters can find new and important uses in astronomical programs
. one possible advance for tunable filters are devices which can operate at cryogenic temperatures .
problems like piezo creep and slow gain at cryo temperatures can now be overcome .
the largest tunable filter that queensgate can make ( 150 mm aperture ) is not well suited to 810 m telescopes .
but there is no reason why a 300 mm aperture etalon , for example , can not be made .
we have an initial design for such a system that uses an additional control stack at the centre of the plates , well matched to the central obstruction of most telescopes .
this design requires a slightly reconfigured cs100 control system .
it is commonly thought that tunable filters are restricted to switching between bands which fall within the restricted bandpass of the order sorting filter .
in fact , a multiband order sorter ( offer & bland - hawthorn 1998 ) allows spectral bands at opposite ends of the optical spectrum to be observed in sequence ( cianci 2000 ) .
there exists a wide class of sources which have never benefitted from detailed narrowband imaging largely because even through conventional narrow bands , the field stars saturate the detector .
one of many examples was an attempt to measure gas metallicities in the vicinity of the trapezium .
but rapid switching through differential ultranarrow bands makes this entirely feasible now .
the straddle shuffle demonstration by maloney & bland - hawthorn ( 2001 ) has been used on galaxies with very bright disks and cores , leading to perfect cancellation of the continuum light and therefore revealing very weak levels of line emission .
this same method should work well on young globular clusters in order to search for emission , but this has not been attempted to date .
tunable filters have major applications for adaptive optics imaging and for coronographic imaging .
the use of differential narrowband imaging would greatly help to suppress any stray light around the central or nearby bright stars .
b. woodgate has achieved some spectacular observations of emission line nebulae around symbiotic stars with the hubble space telescope , but the central stars are so bright that diffraction spikes are a major problem .
there are numerous other applications , some of which require minor modifications to the existing hardware .
these include new time series modes , nod & shuffle imaging , mask & shuffle imaging , broad - narrow switching , tunable polarimetric imaging ( all discussed in bland - hawthorn 2000 ) and tunable echelle imaging ( baldry & bland - hawthorn 2000 ) .
there are web sites which describe the basic observing modes of taurus : * taurus homepage : * www.aao.gov.au/taurus * taurus tunable filter : * www.aao.gov.au/ttf * taurus++ : * www.aao.gov.au / taurus++ * taurus polarimetry : * www.aao.gov.au/taurus_pol
* general review on tunable filters : * bland - hawthorn ( 2000@xmath3 ) , bland - hawthorn ( 2001 ) * ttf instrument summary : * bland - hawthorn & jones ( 1998 ; 1999 ) , jones & bland - hawthorn ( 1998 ) , jones ( 2001@xmath4 ) .
* charge shuffle modes : * bland - hawthorn & barton ( 1995 ) , glazebrook & bland - hawthorn ( 2001 ) , maloney & bland - hawthorn ( 2001 ) , jones ( 2001@xmath29 ) . * ttf data analysis : * jones ( 2002@xmath29 ) .
alexander , t. 2000 , apj , 536 , 710 baker , j.c .
2001 , aj , 121 , 1821 baldry , i.k . , bland - hawthorn , j. 2000 , pasp , 112 , 1112 beauchamp , d. & hardy , e. 1997 , aj , 113 , 1666 birkinshaw , m. , worrall , d.m . ,
hardcastle , m.j .
2001 . in new visions of the x - ray universe , estec , in press bland - hawthorn , j. , barton , j. 1995 , aao newsletter , 75 , 8 bland - hawthorn , j. , jones , d. h. 1998 , pasa , 15 , 44 bland - hawthorn , j. , jones , d. h. 1998 , spie , 3355 , 855 bland - hawthorn , j. 2000@xmath4 , in encyclopaedia of astronomy & astrophysics , ( iop publ .
& macmillan publ .
: bristol ) , ed .
murdin ( see http://nedwww.ipac.caltech.edu/level5/hawthorn2/frames.html ) bland - hawthorn , j. 2000@xmath29 , in imaging the universe in three dimensions , eds .
w. van breugel , j. bland - hawthorn , asp conf ser , 195 , 34 bland - hawthorn , j. 2001 , apj , 563 , 611 bowyer , s. , berghfer , t.w . , korpela , e.j . 1999 ,
apj , 526 , 592 cecil , g. 2000 , spie , 4008 , 73 cecil , g. 2000 , apj , 536 , 675 cecil , g. 2001 , baas , 199 , 5011 cecil , g. 2002 , apj , 568 , 627 cepa , j. 2002 , in galaxies : the third dimension , eds m rosada , l binette , l arias , in press cianci , s. 2000 , aao newsletter , 94 , 3 cianci , s. 2000 , in imaging the universe in three dimensions , eds
. w. van breugel , j. bland - hawthorn , asp conf ser , 195 , 601 cianci , s. 2003@xmath4 , aao newsletter , 102 , 4 cianci , s. 2003@xmath29 , phd , univ . of sydney cianci , s. 2003 , apj , submitted clarke , f. 2002 , phd , univ . of cambridge clarke , f. , tinney , c.g . ,
covey , k.r .
2002 , mnras , in press ( astro - ph/0201162 ) collins , j. , rand , r. 2001 , apj , 551 , 57 couch , w.j . , newell , e.b
. 1984 , apjs , 56 , 143 couch , w.j .
2001 , apj , 549 , 820 dalcanton , j.j .
1996 , apj , 466 , 92 deutsch , e.w .
1998 , pasp , 110 , 912 dressler , a. , gunn , j.e . 1983 , apj , 270 , 7 ferguson , a. , van der hulst , t. , van gorkom , j. 2001 , aao newsletter , 96,4 francis , p. , wilson , g.m .
& woodgate , b.e .
2001@xmath4 , pasa , 18 , 64 francis , p. 2001@xmath29
, apj , 554 , 1001 fujita , s.s .
2003 , apjl , in press ( astro - ph/0302473 ) glazebrook , k. 1997 , in extragalactic astronomy in the infrared , moriond , eds .
ga mamon , tx thuan , p. 467 glazebrook , k. , bland - hawthorn , j. 2001 , pasp , 113 , 197 heckman , t. 2002 , in extragalactic gas at low redshift , eds js mulchaey and jt stocke , asp conf ser , 254 , 292 hewett , p.c .
2000 , in imaging the universe in three dimensions , eds .
w. van breugel , j. bland - hawthorn , asp conf ser , 195 , 94 higdon , j.l , cecil , g. & bland - hawthorn , j. 1997 , aas , 191 , 8907 hippelein , h. 2003 , a&a , accepted ( astro - ph/0302116 ) hu , e.m . , mcmahon , r.g .
1996 , nature , 382 , 231 jaffe , w. , bremer , m. 2000 , aao newsletter , 93 , 3 jog , c.j . 1999 ,
apj , 522 , 661 jones , d.h . , bland - hawthorn , j. 1998 , pasp , 110 , 1059 jones , d.h . , bland - hawthorn , j. 1999 , in looking deep in the southern sky , eds r morganti , wj couch , p. 320
jones , d.h .
2001 , phd , australia national university jones , d.h .
2001@xmath4 , pasp , 113 , 255 jones , d.h .
2001@xmath29 , in scientific drivers for eso future vlt / vlti instrumentation , eds . j. bergeron and g. monnet jones , d.h . , bland - hawthorn , j. 2001 , apj , 550 , 593 jones , d.h . , renzini , a. , rosati , p. , seifert , w. 2001 , eso mess , 103 , 10 jones , d.h .
, stappers , b. & gaensler , b. 2002@xmath4 , a&a , in press jones , d.h . 2002@xmath29 , mnras , 329 , 759 knapp , j. 1999 , in star formation in early - type galaxies , asp conf .
ser . , 163 , 119 lehnert , m.d . & heckman , t. m. 1996 , apj , 462 , 651 lieu , r. 1999 , apj , 527 , l27 maloney , p.r . &
bland - hawthorn , j. 2001 , apj , 553 , l129 meurer , g. 1999 , baas , 194 , 501 meurer , g. 2003 , iau symp .
220 , dark matter in galaxies miller , s.t . ,
veilleux , s. 1999 , aao newsletter , 88 , 4 miller , s.t .
2002 , phd , univ . of maryland miller , s.t . & veilleux , s. 2003@xmath4 , apjs , in press ( astro - ph/0305026 ) miller , s.t . & veilleux , s. 2003@xmath29 , apj , in press ( astro - ph/0304471 ) miralde - escud , j. , lehar , j. 1992 , mnras , 259 , 31p molla , m. , hardy , e. & beauchamp , d. 1999 , apj , 513 , 695 offer , a. , bland - hawthorn , j. 1998 , mnras , 299 , 176 pettini , m. 2001 , apj , 554 , 981 phillips , m.m .
1984 , nature , 310 , 554 pier , e.a . , krolik , j.h .
1993 , apj , 418 , 673 pogge , r.w .
1988 , apj , 328 , 519 reynolds , r. , haffner , m. , tufte , s. 1999 , apj , 525 , l21 riddick , f. , lucas , p. & roche , p. 2003
, aao newsletter , 102 , 10 rupke , d. 2002 , in prep ryder , s. 2001 , aao newsletter , 98 , 11 ryder , s. , fenner , y. & gibson , b.k .
2003 , aao newsletter , 102 , 6 schoenmakers , a.p . , franx , m. & de zeeuw , p.t .
1997 , mnras , 292 , 349 schuberth , y. , burton , m.g .
2000 , aao newsletter , 92 , 3 shopbell , p.l . 1999 ,
apj , 524 , l83 shopbell , p.l .
2001 , baas , 198 , 5706 smail , i. 1998 , mnras , 293 , 124 sokolowski , j. 1993 , phd , rice university solorzano - inarrea , c. 2002 , mnras , 331 , 673 solorzano - inarrea , c. & tadhunter , c.n .
2003 , mnras , accepted ( astro - ph/0301157 ) strickland , d.k . ,
stevens , i. r. 2000 , mnras , 314 , 511 syer , d. & tremaine , s. 1996 , mnras , 281 , 925 tadhunter , c.n . , tsvetanov , z. 1989 , nature , 341 , 422 tadhunter , c.n .
2000 , mnras , 314 , 849 thommes , e. 1998 , mnras , 293 , l6 tinney , c.g . ,
tolley , a.j . 1999 , mnras , 304 , 119 tober , j. , glazebrook , k. , thomson s. , bland - hawthorn , j. & abraham , r.g .
2003 , apj , submitted tresse , l. , maddox , s.j . , le fevre , o. , cuby , j .-
2001 , mnras , 337 , 369 van gorkom , j. 1997 , in the nature of elliptical galaxies , asp conf ser , 116 , 363 veilleux , s. , bland - hawthorn , j. 1997 , apj , 479 , l105 veilleux , s. 2002@xmath4 , in extragalactic gas at low redshift , eds js mulchaey and jt stocke , asp conf ser , 254 , 313 veilleux , s. 2002@xmath29 , in galaxies : the third dimension , eds m rosada , l binette , l arias , in press veilleux , s. , rupke , d. 2002 , apj , 565 , l63 veilleux , s. , cecil , g. , bland - hawthorn , j. & shopbell , p.l .
2002 , in line emission from jet flows , eds .
w. henney , w. steffen , l. binette & a. raga ( isla mujeres ) , _ rev .
_ , 13 , 222229 veilleux , s. , shopbell , p.l . ,
rupke , d. & bland - hawthorn , j. 2003 , apj , submitted veilleux , s. & miller , s.t .
2003 , aao newsletter , 102 , 8 young , a.j . ,
wilson , a.s . , shopbell , p.l .
2001 , apj , 556 , 6 zaritsky , d. 1997 , apj , 480 , l91 zurita , a. 2001 , phd thesis , instituto astrofisica , la laguna zurita , a. 2000 , a&a , 363 , 9 | the taurus tunable filter ( ttf ) has now been in regular use for seven years on the anglo - australian telescope ( aat ) .
the instrument was also used for three years ( 19961999 ) on the william herschel telescope ( wht ) .
we present a brief review of the different applications in order to illustrate the versatility of tunable filters in optical / ir spectrophotometric imaging .
tunable filters are now planned or under development for 6 - 10 m class telescopes which ensures their use for years to come .
@xmath0 anglo - australian observatory , po box 296 , epping , nsw 2121 + [email protected] + @xmath1 dept . of astrophysics & optics ,
school of physics , university of new south wales , nsw 2052 + [email protected] + |
among the fundamental parameters of the standard model of particle physics , the angle @xmath5 of the unitarity triangle formed from elements of the cabibbo - kobayashi - maskawa quark mixing matrix @xcite has a particular importance .
not only is it one of the least well determined fundamental parameters of the standard model , it is also the only @xmath4-violating parameter that can be measured using only tree - level decays @xcite .
the precise determination of @xmath0 is thus a critical element of the programme towards understanding the baryon asymmetry of the universe , and is one of the main objectives of planned future @xmath6 physics experiments ( see , for example , @xcite ) .
a method to measure the ckm phase @xmath0 in @xmath1decays by comparing the dalitz plot distribution obtained when the neutral @xmath3 meson is reconstructed in @xmath4 eigenstates to that when flavour - specific states are used has been proposed recently @xcite .
the method builds on the original proposal of gronau , london and wyler ( glw ) @xcite for a similar analysis using @xmath7decays , and exploits the fact that @xmath4 violation effects are expected to be enhanced in @xmath8 decays since the interfering amplitudes are of comparable magnitude @xcite .
since the charge of the kaon in @xmath1decays unambiguously tags the flavour of the decaying @xmath6 meson , there is no need for time - dependent analysis @xcite .
compared to previous proposals for similar analyses @xcite this method has several attractive features : 1 ) the amplitudes can be measured directly in the dalitz plot analysis , relative to the flavour - specific @xmath9 amplitude ; 2 ) there is no need to normalise to decays with another final state ; 3 ) there is sufficient information to extract @xmath0 using only @xmath3 decays to @xmath4-even states ;-
odd @xmath3 decays with large branching fractions such as @xmath10 are difficult to reconstruct , particularly in experiments at hadron colliders . ]
4 ) the sensitivity to @xmath0 does not depend strongly on the values of @xmath0 and strong phases involved ; 5 ) @xmath0 is determined with only a single unresolvable ambiguity ( @xmath11 ) .
these advantages have much in common with those of the @xmath7 , @xmath2method , which is currently providing the most constraining measurements of @xmath0 @xcite . moreover
, this technique is especially promising for the lhcb experiment since it does not require the reconstruction of neutral particles in the final state .
there is , however , a potential drawback of the method , in that the fit to the dalitz plot distribution is inherently model - dependent .
the magnitude of the model - dependence has been estimated to be small @xcite , but in view of the aim to achieve an @xmath12 measurement of @xmath0 at future experiments , it is certainly desirable to explore possibilities to remove such systematic uncertainties . a model - independent analysis procedure has been proposed for the extraction of @xmath0 from @xmath7 , @xmath2decays @xcite .
the method uses data from charm factory experiments specifically , @xmath4-tagged charm mesons to obtain information about the strong phase variation across the @xmath3 decay dalitz plot . a similar approach to probe the strong phases present in the @xmath1dalitz plot is not immediately available since there is no realistic way to obtain sufficient samples of @xmath4-tagged @xmath6 decays . in this paper
we show that the study of dalitz plot distributions in the decay chain @xmath1 , @xmath2allows to translate the model - independent measurement of the @xmath2amplitude to the @xmath1mode .
this approach allows not only to measure @xmath0 model - independently in this decay chain , but also to use the obtained constraints on the @xmath1amplitude in the @xmath13dalitz plot analysis . for clarity ,
our method is independent of model assumptions in both @xmath1and @xmath2decays .
the more straightforward cases which are either model - dependent in @xmath1but model - independent in @xmath2 , or vice - versa , are also possible but are not discussed in detail .
a fully model - dependent analysis is , of course , also possible .
the remainder of the paper is organised as follows . in section [ sec : idea ] we outline the basic idea and the formalism of the analysis . in section [ sec : dflav ]
we discuss how possible complications arising from the use of @xmath14 decays as quasi - flavour - specific states can be resolved . in section [ sec : models ] we describe the dalitz plot models that we use in our feasibility study .
we finally present the results of the study in section [ sec : sensitivity ] , before concluding in section [ sec : conclusion ] .
we begin by recalling the essentials of the @xmath7 , @xmath2model - independent method @xcite .
the amplitude of the @xmath7 , @xmath2decay can be written @xmath15 where @xmath16 is the amplitude of the @xmath17decay , @xmath18 is the amplitude of the @xmath19decay , @xmath20 is the ratio of the absolute values of the interfering @xmath21 and @xmath22 amplitudes , and @xmath23 is the strong phase difference between these amplitudes . assuming no @xmath4 violation in @xmath3 decay , @xmath24 .
the density of the @xmath3 decay dalitz plot from @xmath7 decay is given by the absolute value squared of the amplitude @xmath25 where @xmath26 the functions @xmath27 and @xmath28 are the cosine and sine of the strong phase difference @xmath29 between the @xmath19and @xmath17 amplitudes decay amplitudes used by cleo @xcite .
note that alternative conventions are used in the literature , _
e.g. _ ref .
@xcite uses a convention where the sign of @xmath30 is opposite . ] : @xmath31 the equations for the charge - conjugate mode @xmath32 are obtained with the substitution @xmath33 .
considering both @xmath6 charges , one can obtain @xmath0 and @xmath23 separately .
once the dalitz plot is divided into @xmath34 bins symmetrically to the exchange @xmath35 , the expected number of events in the @xmath36 bin of the dalitz plot of @xmath3 decay from @xmath7 is @xmath37 \ , , \label{n_b}\ ] ] where @xmath38 is the number of events in the corresponding bin of the dalitz plot where the @xmath3 meson is in a flavour eigenstate ( conveniently obtained using @xmath39 samples ) and @xmath40 is a normalisation constant .
the bin index @xmath41 ranges from @xmath42 to @xmath43 ( excluding 0 ) ; the exchange @xmath44 corresponds to the exchange @xmath45 .
the coefficients @xmath46 and @xmath47 , which include information about the cosine and sine of the phase difference , are given by @xmath48 where the dependences on dalitz plot position of @xmath49 , @xmath50 and @xmath30 have been suppressed .
the parameters @xmath47 are defined similarly with cosine substituted by sine . here
@xmath51 represents the dalitz plot phase space and @xmath52 is the bin region over which the integration is performed .
the symmetry under @xmath53 requires @xmath54 and @xmath55 , thus for @xmath34 bins there are @xmath56 unknowns . for each bin
there are two observables in @xmath57 decays , being the numbers of events in @xmath58 and @xmath59 samples , and thus there are a total of @xmath60 observables .
the system can consequently be solved for @xmath61 . however , due to the small value of @xmath20 in @xmath7decays , there is very little sensitivity to the @xmath46 and @xmath47 parameters , leading to a reduction in the precision on @xmath0 that can be obtained .
use of external constraints on @xmath46 and @xmath47 , that can be provided by charm - factory experiments @xcite , leads to much improved sensitivity to @xmath0 .
turning now to @xmath1decays , we find that the simplest approach towards performing a model - independent analysis is not viable .
using the convention @xmath62 , where @xmath63 and @xmath64 correspond to @xmath4-even and @xmath4-odd final states , respectively , the amplitude for @xmath13decay is given by @xmath65 where @xmath66 is the amplitude of the @xmath67decay and @xmath68 is the amplitude of the @xmath69decay . in these expressions
we have factored out the @xmath4-violating phase @xmath0 so that the amplitudes @xmath70and @xmath71are @xmath4-conserving . both the strong phase difference between @xmath67and @xmath69decays and the ratio of the suppressed and allowed amplitudes @xmath20 are now functions of @xmath6 decay dalitz plot variables and enter the expressions for amplitudes .
let us now consider dividing the @xmath6 decay dalitz plot into bins ( denoted by the index @xmath72 , @xmath73 ) .
the number of expected events in each @xmath13dalitz plot bin is @xmath74 \ , .
\end{array } \label{num_bin_cp}\ ] ] in this expression the factors @xmath75 and @xmath76 give the numbers of events in the corresponding bin in @xmath67and @xmath69decays , respectively ( complications arising due to the difficulty to reconstruct @xmath3 mesons in pure flavour eigenstates are discussed in section [ sec : dflav ] below ) .
the factor @xmath77 is a global normalisation constant . in eq .
( [ num_bin_cp ] ) we have introduced the parameters @xmath78 and @xmath79 that describe the average strong phase difference and amplitude difference between @xmath67and @xmath69decays in each bin , and that are defined by : @xmath80 where the dependence on the @xmath6 decay dalitz plot position of @xmath81 , @xmath82 and @xmath83 have been suppressed , @xmath51 again represents the phase space ( now of the @xmath6 decay dalitz plot ) and @xmath84 is the bin region over which the integration is performed .
the parameters @xmath85 are given by similar expressions with cosine substituted by sine .
until this stage the discussion for @xmath1decays has run in parallel to that for @xmath7 , @xmath2 .
however , unlike the @xmath7 , @xmath2case , the two amplitudes @xmath86 and @xmath87 are inherently different and thus there is no symmetry in the dalitz plot to be exploited . in the case of
@xmath4-eigenstate @xmath3 decays , we therefore have @xmath88 unknowns ( even if the normalisation is fixed externally ) , that can not be solved for with @xmath89 observables .
hence external constraints on @xmath78 and @xmath79 are necessary @xcite .
let us now consider both dalitz plots simultaneously in the @xmath1 , @xmath2 decay chain .
the amplitude can be written as @xmath90 the decay probability is proportional to the absolute value squared of the amplitude @xmath91 : @xmath92 \ , . }
\end{array } \label{ampl_sq}\ ] ] to implement the model - independent analysis both @xmath6 and @xmath3 plots have to be binned .
corresponding bin indices are denoted here , as before , by greek and latin characters , respectively .
the expected number of events in the bins of @xmath1 , @xmath2double dalitz plot is then @xmath93 \big\ } \ , , } \end{array } \label{num_bin}\ ] ] where the definition of the event numbers @xmath76 , @xmath75 , @xmath38 and phase terms @xmath94 is the same as for the processes mentioned above , and @xmath95 is a normalisation constant . if the number of bins in the @xmath1dalitz plot is @xmath96 and the number of bins in the @xmath2dalitz plot is @xmath34 , then after exploiting the symmetry of the @xmath2decay as before , the number of equations represented by eq .
( [ num_bin ] ) is @xmath97 .
the number of unknowns ( including the normalisation factor @xmath95 ) is @xmath98 .
naturally , information on @xmath0 can not be extracted from the decays of only one @xmath6 flavour , as is apparent since eq .
( [ num_bin ] ) is invariant under the rotation of @xmath0 with the simultaneous rotation of all @xmath99 pairs . to determine the @xmath4 violation parameter it is , of course , necessary to add the opposite @xmath6 flavour ( with the substitution @xmath100 ) .
the system can now be fully resolved : in total , there are @xmath101 observables and @xmath98 unknowns , _
i.e. _ it is solvable for @xmath102 . since the coefficients @xmath46 and @xmath47 are the same as in the model - independent @xmath57 analysis , and can be obtained at a charm factory , the number of unknowns can be reduced even further .
as a by - product of the @xmath1 , @xmath2analysis , the coefficients @xmath103 and @xmath85 will be determined .
the values of these parameters can then be used to enable a model - independent analysis of @xmath13decays .
technically the combination of @xmath1 , @xmath2and @xmath13modes is conveniently done by using a combined likelihood fit that includes both the expressions eq .
( [ num_bin ] ) and eq .
( [ num_bin_cp ] ) .
in practise it is hard to reconstruct secondary @xmath3 mesons from @xmath6 decays in pure flavour eigenstates . (
the possibility to exploit the charge of the associated pion in @xmath39 is not available . )
semileptonic @xmath104 decays in principle provide a source of pure flavour - specific states , but owing to small branching fractions and large backgrounds these are experimentally difficult to deal with , especially at hadron experiments such as lhcb .
it is more convenient to use hadronic final states such as @xmath105 , which are not pure flavour tags due to the presence of a small admixture of doubly - cabibbo - suppressed amplitude .
this contribution can significantly affect a precision measurement . in the proposed analysis , hadronic final states can be used after some correction of the procedure .
the amplitude of a @xmath1 , @xmath106decay is @xmath107 where @xmath108 and @xmath109 are the ratio of the magnitudes of the suppressed and favoured @xmath3 decay amplitudes and the strong phase between them , respectively .
that is consistent with that used in the majority of the literature ( see , _ e.g. _ ref .
@xcite ) , but is opposite in sign to that used for @xmath30 .
another different convention is used by cleo @xcite . ]
the number of events in the @xmath1dalitz plot bins with @xmath3 detected in a @xmath110 state is : @xmath111\ , .
\label{bin_num_kp } \end{array}\ ] ] similarly for the suppressed decay @xmath1 , @xmath112 , the amplitude is @xmath113 and the number of events is @xmath114\ , .
\label{bin_num_kpbar } \end{array}\ ] ] the expressions for @xmath115 decays are obtained by the substitution @xmath116 .
the relations of eq .
( [ bin_num_kp ] ) and eq .
( [ bin_num_kpbar ] ) add another @xmath117 equations to the system of equations to be used in the analysis ( @xmath89 equations for each @xmath6 flavour ) , but since the constants @xmath108 and @xmath109 are known from charm analyses @xcite , only @xmath89 new unknowns are added ( the numbers @xmath76 and @xmath75 ) .
hence , the inclusion of the suppressed final states into the analysis can help to improve the overall sensitivity to @xmath0 @xcite .
recent studies suggest further improvement could be gained from including also the @xmath118 decays @xcite .
in order to perform a feasibility study of the proposed method it is necessary to define a @xmath6 decay dalitz plot model .
the sensitivity depends crucially on the structure of the @xmath69and @xmath67amplitudes , specifically on the interference between them .
if the strong phase difference @xmath23 is nearly constant ( _ e.g. _ in the case when both amplitudes are dominated by @xmath119 decay ) , the double dalitz analysis reduces to a conventional dalitz plot analysis technique similar that used for @xmath120 decays . note that in this type of analysis the possible contribution from non-@xmath121 states can be taken into account model - independently by introducing the coherence factor @xmath122 to the interference term @xcite ( analyses using this approach have been performed , albeit with low statistics @xcite ) .
however , if the resonance structure in @xmath1decays is rich with multiple overlapping states in different channels ( @xmath121 , @xmath123 and @xmath124 ) , the proposed technique should provide a significant benefit compared to analysis of the @xmath2 dalitz plot with the coherence factor . [ cols="<,^,^,^,^,^,^ " , ] & @xmath125 & @xmath126 & @xmath127 & @xmath128 & @xmath129 + different strategies for using @xmath1decays to measure @xmath0 are studied using mc simulation .
the results are shown in table [ sens_table ] .
the uncertainties are obtained from the spread of results in the samples of 200 pseudo - experiments .
an unbinned fit of @xmath13decays with free parameters @xmath20 , @xmath0 and @xmath23 ( option 1 ) gives an estimate of the statistical error for a model - dependent dalitz plot analysis of this decay @xcite .
a binned fit of the same mode ( option 2 ) with fixed values of @xmath130 and @xmath131 shows a drop of sensitivity due to binning of the @xmath1dalitz plot . for most amplitude models ,
the loss in sensitivity is about 20% .
we next consider selecting only events in the @xmath121 region of @xmath1phase space , introducing a coherence factor @xcite , and using @xmath2decays . a conventional single dalitz plot analysis ( option 3 )
gives comparable sensitivity to the unbinned @xmath13analysis ( option 1 ) .
however , only a fraction of all @xmath1events are used indicating that further gains are potentially possible . changing from unbinned to a binned model - independent approach ( option 4 )
reduces the sensitivity by about @xmath132% . relaxing the @xmath46 and @xmath47 coefficients in the binned approach is also possible ( option 5 ) @xcite .
although the sensitivity in this case reduces significantly , the reduction is far less than in the @xmath120 mode @xcite due to the larger interference term in neutral @xmath6 decays and also due to the optimal binning used in this study .
the double dalitz plot analysis ( option 6 ) uses all available events in the @xmath1phase space , providing better @xmath0 accuracy compared to option 4 , and also giving as a by - product the values of @xmath1amplitude coefficients @xmath103 and @xmath85 .
the values of @xmath103 and @xmath85 returned from the fit and a comparison with their calculated values is shown in figure [ bdkp_bins ] ( b ) .
relaxing the @xmath2amplitude coefficients @xmath46 , @xmath47 ( option 7 ) results in a much smaller reduction of the sensitivity than that with single dalitz plot analysis selecting only the @xmath121 region .
the fitted values of the @xmath46 , @xmath47 coefficients for this option are shown in figure [ dkpp_bins ] ( b ) . combined with @xmath13decays ( option 8)
, the double dalitz analysis allows to further improve the statistical precision .
however the gain in sensitivity for most models is about @xmath133 worse than the weighted error for options 6 and 2 , which means that the precision of @xmath103 and @xmath85 terms obtained from the dalitz analysis affects the @xmath13mode . with very large statistics it may be possible to mitigate this effect by using a finer binning of the @xmath1dalitz plot .
an example of the residual distribution for @xmath0 using option 6 ( double dalitz plot fit ) with amplitude model c is shown in figure [ dbldlz_resid](a ) .
no systematic bias is seen in any of the fits .
we emphasise that , although the fit procedure uses a binning that depends on the decay amplitude , the method does not generate systematic bias even if the amplitude used to define the binning is wrong or the model - blind binning is used .
to demonstrate this , we perform the fit of the data generated using model c , but using model - blind rectangular dalitz plot bins ( the bin boundaries in @xmath134 are @xmath135 and 7 mev@xmath136/@xmath137 ; those in @xmath138 are @xmath139 and 12 mev@xmath136/@xmath140 , the bins above @xmath141 mev@xmath136/@xmath137 are combined so that the number of bins is 7 as in the other fits ) .
the quality factor for this binning is @xmath142 .
the residual distribution for @xmath0 using option 6 ( double dalitz plot fit ) is shown in figure [ dbldlz_resid](b ) .
the fit is unbiased , and the statistical accuracy compared to the fit with binning based on the correct underlying model is roughly proportional to the ratio of the @xmath143 values : the @xmath0 resolution is @xmath144 .
the statistical accuracy of the method does not depend significantly on the value of @xmath0 .
the precision on @xmath0 as a function of @xmath0 is shown in figure [ dbldlz_resid](c ) for model c , options 6 ( double dalitz plot fit only ) and 8 ( double dalitz plot fit combined with @xmath13 ) .
we expect that the dependence on the strong phases in @xmath1decays , which were taken arbitrarily in our study , should also not be large .
( -340,130 ) ( -185,130 ) ( -30,130 )
we have proposed a technique to obtain the angle @xmath0 of the unitarity triangle in a model - independent way by using the decay chain @xmath1 , @xmath2 .
our method is independent of model assumptions in both @xmath1and @xmath2amplitudes .
the proposed approach allows not only to measure @xmath0 in this decay chain , but also to use the constraints on the @xmath1amplitude obtained in this measurement for a model - independent @xmath13dalitz plot analysis . the precision of the @xmath0 measurement using our technique
will depend strongly on the structure of @xmath1amplitude , which is as - yet unknown .
however , mc simulation results show that if interference effects between @xmath69and @xmath67amplitudes are strong enough to allow a meaningful model - dependent measurement of @xmath0 using @xmath13 decays @xcite , and the number of reconstructed @xmath1 , @xmath2decays is comparable to the number of @xmath13decays , the model independent technique proposed here should enable a significant improvement in the sensitivity .
we are grateful to our colleagues from the lhcb experiment , and would particularly like to thank alex bondar and mike williams for discussions .
we would like to acknowledge support from the university of warwick research development fund , the science and technology facilities council ( united kingdom ) and from the ministry of science and education of the russian federation .
n. cabibbo , phys .
* 10 * ( 1963 ) 531 .
m. kobayashi and t. maskawa , prog . theor .
* 49 * ( 1973 ) 652 .
m. antonelli _ et al .
_ , arxiv:0907.5386 [ hep - ph ] . t.e .
browder _ et al . _ , jhep 02 ( 2008 ) 110 [ arxiv:0710.3799 [ hep - ph ] ]
. m. artuso _ et al .
_ , eur .
j. c * 57 * ( 2008 ) 309 [ arxiv:0801.1833 [ hep - ph ] ] .
browder , t. gershon , d. pirjol , a. soni and j. zupan , arxiv:0802.3201 [ hep - ph ] . t. gershon , phys . rev .
d * 79 * ( 2009 ) 051301(r ) [ arxiv:0810.2706 [ hep - ph ] ] .
m. gronau and d. london , phys .
b * 253 * ( 1991 ) 483 .
m. gronau and d. wyler , phys .
b * 265 * ( 1991 ) 172 .
i. i. y. bigi and a. i. sanda , phys .
b * 211 * ( 1988 ) 213 .
i. dunietz , phys .
b * 270 * ( 1991 ) 75 .
m. gronau , phys .
b * 557 * ( 2003 ) 198 [ arxiv : hep - ph/0211282 ] .
s. pruvot , m. h. schune , v. sordini and a. stocchi , arxiv : hep - ph/0703292 .
r. aleksan , t. c. petersen and a. soffer , phys .
d * 67 * ( 2003 ) 096002 [ arxiv : hep - ph/0209194 ] .
b. aubert _ et al . _ [ babar collaboration ] , phys .
d * 78 * ( 2008 ) 034023 [ arxiv:0804.2089 [ hep - ex ] ] .
k. abe _ et al . _ [ belle collaboration ] , arxiv:0803.3375 [ hep - ex ] .
t. gershon and m. williams , arxiv:0909.1495 [ hep - ph ] .
a. giri , y. grossman , a. soffer and j. zupan , phys .
d * 68 * ( 2003 ) 054018 [ arxiv : hep - ph/0303187 ] .
a. bondar and a. poluektov , eur .
j. c * 55 * ( 2008 ) 51 [ arxiv:0801.0840 [ hep - ex ] ] .
r. a. briere _ et al .
_ [ cleo collaboration ] , arxiv:0903.1681 [ hep - ex ] .
j. l. rosner _
[ cleo collaboration ] , phys .
lett . * 100 * ( 2008 ) 221801 [ arxiv:0802.2264 [ hep - ex ] ] .
d. m. asner _
[ cleo collaboration ] , phys .
d * 78 * ( 2008 ) 012001 [ arxiv:0802.2268 [ hep - ex ] ] .
d. atwood , i. dunietz and a. soni , phys .
* 78 * ( 1997 ) 3257 [ arxiv : hep - ph/9612433 ] .
d. atwood , i. dunietz and a. soni , phys .
d * 63 * ( 2001 ) 036005 [ arxiv : hep - ph/0008090 ] .
n. lowrey _ et al . _
[ cleo collaboration ] , phys .
d * 80 * ( 2009 ) 031105 [ arxiv:0903.4853 [ hep - ex ] ] .
b. aubert _ et al .
_ [ babar collaboration ] , phys .
d * 79 * ( 2009 ) 072003 arxiv:0805.2001 [ hep - ex ] .
b. aubert _ et al . _ [ babar collaboration ] , phys .
d * 80 * ( 2009 ) 031102 [ arxiv:0904.2112 [ hep - ex ] ] .
b. aubert _ et al .
_ [ babar collaboration ] , phys .
* 96 * ( 2006 ) 011803 [ arxiv : hep - ex/0509036 ] .
b. aubert _ et al .
_ [ babar collaboration ] , phys .
d * 77 * ( 2008 ) 071102 [ arxiv:0712.3469 [ hep - ex ] ] .
j. brodzicka _ et al .
_ [ belle collaboration ] , phys .
* 100 * ( 2008 ) 092001 [ arxiv:0707.3491 [ hep - ex ] ] .
k. abe _ et al . _ [ belle collaboration ] , arxiv:0803.3375 [ hep - ex ] .
k. akiba _ et al .
_ , cern - lhcb-2008 - 031 .
a. bondar and a. poluektov , eur .
j. c * 47 * ( 2006 ) 347 [ arxiv : hep - ph/0510246 ] . | it is shown that it is possible to perform a model - independent extraction of the ckm unitarity triangle angle @xmath0 using only the decays @xmath1with
@xmath2and @xmath1with flavour - specific @xmath3 decays .
the proposed method can also utilise the @xmath1data with @xmath4-eigenstate decays of the @xmath3 meson in a model - independent way . |
the disintegrating sagittarius ( sgr ) dwarf galaxy remains a riddle , wrapped in a mystery , inside an enigma .
large scale photometric surveys , such as the two micron all - sky survey ( 2mass ) and the sloan digital sky survey ( sdss ) have now revealed the structure of the tidal tails of the sgr over more than @xmath2 radians on the sky @xcite . by tallying all the stellar debris in the streams and remnant , we now know that the progenitor galaxy had a luminosity of @xmath3 , comparable to the present day small magellanic cloud @xcite .
the ingestation of such a large progenitor , together with its dismantling under the actions of the galactic tides , can provide us with a wealth of information about both the galaxy and the sgr , if we can only decode it .
radial velocities , and sometimes metallicities and chemical abundances , are now known for many hundreds of stars in the sgr tails ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
there are also @xmath0 10 globular clusters associated with the sgr tails @xcite .
this rich mosaic of positions and velocities of sgr tracers has proved surprisingly difficult to understand .
although there is no shortage of sgr disruption models in the literature ( see e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , they all have significant shortcomings , and fail to reproduce a substantial portion of the datasets .
the most successful recent attempt is by @xcite , though they do not explain the striking two stream morphology seen in the sdss data @xcite .
additionally they advocate the use of a triaxial halo for the galaxy with minor axis contained with the galactic plane , which is unattractive on other grounds ( e.g. * ? ? ?
given the impasse , it is natural to look to proper motions of the sgr stream as so as to obtain a clearer picture of its space motion .
@xcite have provided the first measurements of the proper motion of the sgr trailing tail .
they took advantage of archival photographic plate data in some of kapteyn s selected areas which provides a 90 year baseline .
they derive proper motions for four @xmath4 fields covering locations on the trailing tail between @xmath5 and @xmath6 from the sgr core .
however , the number of stars in each field remains modest ( @xmath7 ) , and so the precision of the proper motion measurement is still quite low ( @xmath8 mas yr@xmath1 ) . here
, we will pursue a different tack to obtain proper motions of the trailing stream in roughly the same area of sky . as part of a project to detect supernovae , the sloan digital sky survey scanned a @xmath9 square degree region on the celestial equator , known as stripe 82 ( e.g. , * ? ? ?
proper motions can be derived by matching objects between the @xmath10 epochs @xcite obtained over time period of @xmath0 7 years , whilst the co - added optical data is roughly 2 magnitudes deeper than a single epoch sdss measurement . although the baseline is small so the precision of a measurement of proper motion of a single star is still low , we can take advantage of the large number of sgr tracers to get a high precision ( @xmath11masyr@xmath1 ) measurement for the proper motion of the ensemble .
the paper is arranged as follows .
section 2 describes the extraction of proper motions for stars from the stripe 82 data using background quasars to provide an absolute reference frame .
section 3 shows how to identify the sgr stars in stripe 82 , where they occupy a distinctive niche in magnitude and radial velocity space .
section 4 discusses our modelling of the proper motions using both photometric and spectroscopic samples .
we extract the proper motion for both the bright and faint sgr streams identified by @xcite . in section 5
, we compare our proper motions both with the earlier work of @xcite and with the simulation data .
stripe 82 has already been the subject of numerous studies .
its multi - epoch and multi - band imaging allows study of the variable sky and identification of many kinds of transient phenomena ( see e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the stripe 82 dataset has also been used to derive proper motions by @xcite .
this light - motion catalogue was subsequently exploited to build reduced proper motion diagrams @xcite and analyse kinematical properties of galactic disk and halo populations @xcite .
even so , the proper motion measurements pioneered by @xcite can be improved . for the bulk proper motion of sgr
, we are interested in the statistical properties of a large ensemble of faint tracers , so proper motions with the smallest possible systematic errors are highly desirable .
the original catalogue by bramich does not provide proper motions for stars fainter than @xmath12 20.5 and is known to have some noticeable systematics . since the sgr stream has a very large number of tracers in stripe 82 @xcite , we do not require a proper motion measurement for every star , but rather need small systematic errors and well understood error - bars for an ensemble . for this purpose , it makes sense to measure the proper motions relative to quasars ( qsos ) .
stripe 82 has a number of both spectroscopically and photometrically identified qsos . in this work
, we have used the catalogue of spectroscopic qsos from @xcite and the sample of photometrically identified qsos from @xcite to extract denizens of stripe 82 .
the purity of the spectroscopic catalogue of qsos is guaranteed all the objects are qsos and must have zero proper motion .
however , the photometric catalogue is known to have some contamination by stars . in order to minimize contaminants ,
we use the cut good@xmath13 , as recommended by @xcite .
this guarantees a small stellar contamination , certainly @xmath145% .
given a sample of qsos each with zero proper motion , then for each star in the vicinity of the qso , we may determine proper motion relative to the quasar .
as an input catalog for the stars , we took the stripe 82 co - add dataset @xcite , from which we select primary objects , classified by the sdss pipeline as stars .
the individual source detections are taken from the stripe 82 portion of the sdss dr7 database @xcite using only those fields having acceptable and good data quality flags .
matching co - added sources to detections at individual epochs is done with the 0.5 arcsec radius using the q3c module for the postgresql database @xcite .
this procedure makes the catalogue incomplete for high proper motion objects ( with proper motions @xmath15 100 mas yr@xmath1 ) , but we are not interested in such objects in our current study .
then , for each pair ( star , qso ) observed multiple times by sdss within one field , we analyse the positional offsets and errors . for the right ascension ,
these are defined via @xmath16 with similar equations for the declination .
the model for the positional offsets is @xmath17 where @xmath18 is the proper motion of the star , @xmath19 is the date of the observation , @xmath20 is the fraction of outliers , and @xmath21 is the scatter around the linear relation , whilst @xmath22 is the rectangular function to account for the outliers . the resulting likelihood
is then minimized with respect to the 4 parameters @xmath18 , @xmath23 , @xmath20 , @xmath21 with the error - bars determined from the hessian at the minimum .
repeating this procedure for the offsets in declination gives us the proper motions and their errors , @xmath24 , @xmath25 , @xmath26 , @xmath27 for all the sources with a spectroscopic or a photometric qso nearby .
after performing the computation of the proper motion for individual stars , and before trying to measure the statistical proper motions for ensembles of stars , it is important to check for the presence of possible systematics , as well as to examine the accuracy of the error bars .
[ fig : qsocheck ] presents such an assessment . in this paper
, we are focusing on the particular part of stripe 82 , which intersects with the sgr stream . as systematic effects
may depend on the right ascension , fig .
[ fig : qsocheck ] uses only the proper motions in the right ascension range @xmath28 .
the left panel of fig .
[ fig : qsocheck ] shows the histogram of proper motions of the spectroscopic qsos measured relative to the photometric qsos and normalized by the error bar provided by our fitting procedure .
the red overplotted curve shows a gaussian with the center at zero and unity dispersion .
the excellent match of the histogram with the gaussian curve shows that that the proper motions do not possess noticeable systematic offsets , and that the error bars on the proper motions are a faithful description of the precision .
it is also quite clear from the plot that this is not true for proper motion measurements by @xcite , which have noticeable systematics and error underestimation .
the middle panel of fig .
[ fig : qsocheck ] shows the the proper motion of the spectroscopic qsos relative to the photometric qsos versus the @xmath29 color difference .
the points with error bars show the median proper motion in bins of @xmath29 , while the error bars are 1.48 times the median absolute deviation of proper motions within a corresponding @xmath29 bin . since all the points lie within 1 @xmath21 of zero , we conclude that the color - dependent terms in the proper motions are negligible down to the precision [email protected] mas yr@xmath1 .
this holds true at least within the color range @xmath31 , which is the color - range applicable to 99% percent of the photometric qsos .
later we will see that the proper motion of the sgr stream determined from the photometric sample with @xmath32 , @xmath33 and the spectroscopic sample with @xmath34 , @xmath35 agree each other , further confirming the small level of color - related systematic errors . and
last , the right panel of fig .
[ fig : qsocheck ] shows the precision of our proper motion measurements as a function of @xmath36-band magnitude . there is a considerable scatter , but the median proper motion precision of @xmath37mas yr@xmath1 is significantly better than that of @xcite for large range of magnitudes .
ccc 87.5 & @xmath38 & @xmath39 + 92.5 & @xmath40 & @xmath41 + 97.5 & @xmath42 & @xmath43 + 102.5 & @xmath44 & @xmath45 + 107.5 & @xmath46 & @xmath47 + 112.5 & @xmath48 & @xmath49 + 117.5 & @xmath50 & @xmath51 + 125.0 & @xmath52 & @xmath53 + 135.0 & @xmath54 & @xmath55 [ tab : sgr_vel ]
the sgr stream in the southern galactic hemisphere is known to have a complex structure .
@xcite used main - sequence turn - off ( msto ) stars extracted from sdss data release 8 ( dr8 ) to demonstrate the existence of two streams a thicker , brighter stream and a thinner , fainter stream displaced by @xmath56 .
the brighter stream has multiple turn - offs as well as a prominent red clump , whereas the fainter stream does not .
@xcite argued that the brighter stream was composed of more than one stellar population , including some metal - rich stars , whereas the fainter stream is dominated by a single metal - poor population . here ,
we are primarily interested in the intersection of the sgr stream with the sdss stripe 82 . fig .
[ fig : sag_map ] shows the density of msto stars extracted via the cuts @xmath57 and @xmath58 in the southern hemisphere , together with stripe 82 demarcated by the red lines .
again , two distinct streams structures are clearly visible a brighter one crossing the equator and stripe 82 at a right ascension of @xmath59 and a dimmer one crossing the equator at a right ascension of @xmath60 .
the magnitude distribution of msto stars along stripe 82 is shown in the middle panel of fig .
[ fig : sag_map ] .
the sgr streams are clearly evident , with the brighter stream visible at @xmath61 and fainter one at @xmath62 . since the distances to the streams are not constant with @xmath63 , and since stripe 82 crosses the stream at an angle , we observe a clear distance gradient with right ascension .
the right panel of fig .
[ fig : sag_map ] shows the radial velocity as a function of right ascension for giant stars , extracted by the cut @xmath64 using the sdss spectroscopic measurements .
this time not only are stars within stripe 82 included , but also those satisfying @xmath65 and lying near the stream ( @xmath66 ) .
as already found by @xcite , the sgr stream is visible at the @xmath67kms@xmath68=235 km@xmath1 @xcite and a solar peculiar velocity of ( u , v , w)=(-8.5,13.38,6.49)kms@xmath1 from @xcite ] .
the figure also shows the variation of radial velocities as the stream stars return from the apocenter of the sgr orbit . as they will be needed later , we extract the radial velocities along the stream , together with the velocity dispersion , by fitting a gaussian to the sgr signal . although figure [ fig : sag_map ] shows the radial velocities as a function of right ascension , it is important to realise that we perform these fits in bins along @xmath69 , which is the angle along the stream from the remnant , measured positive in the trailing direction , as defined in @xcite ) .
the resulting measurements as a function of @xmath69 are shown in table [ tab : sgr_vel ] .
importantly , fig .
[ fig : sag_map ] demonstrates that the sgr stream stars are visible as a distinct bright feature in magnitude and velocity space .
therefore , this information can be used to select sgr member stars and statistically measure their properties such as proper motion .
kms@xmath1 and peculiar velocity from @xcite . ] ccccccccccc fp1 & 5.00 & 22.00 & 0.05 & -2.51 & 0.11 & 0.11 & 0.10 & -2.51 & 0.11 & 0.12 + fp2 & 25.00 & 40.00 & -0.50 & -2.24 & 0.09 & 0.09 & 0.81 & -2.15 & 0.09 & 0.11 + fp3 & 42.00 & 52.00 & 0.19 & -2.36 & 0.14 & 0.14 & 1.87 & -1.45 & 0.11 & 0.15 + fs4 & 14.00 & 50.00 & -0.42 & -2.65 & 0.11 & 0.13 & 1.07 & -2.46 & 0.11 & 0.11 [ tab : pm_tab ] ccccccccccccc fp1 & 89.5 & 10.0 & -15.0 & 9.6 & -22.6 & 25.4 & 194 & 14 & -46 & 14 & 23 & 8 + fp2 & 106.1 & 0.6 & -23.5 & 5.2 & -24.4 & 29.1 & 290 & 12 & 33 & 14 & -20 & 9 + fp3 & 118.6 & -6.7 & -31.4 & 0.4 & -25.0 & 33.9 & 270 & 17 & -57 & 24 & -48 & 16 + fs4 & 105.6 & 0.8 & -23.2 & 5.3 & -24.3 & 28.9 & 308 & 15 & -16 & 18 & -40 & 11 [ tab:3dvel_tab ]
as demonstrated above , in the southern galactic hemisphere , it is possible to achieve clean separation of the sgr trailing tail stars and the galactic foreground in both apparent magnitude and the radial velocity space .
however , even for high - confidence stream members , the uncertainty of individual proper motion measurements ( @xmath70mas yr@xmath1 ; see section [ sec : pms ] ) is comparable or higher than the expected tangential velocity of the stream ( @xmath71mas yr@xmath1 ; * ? ? ?
it is therefore crucial to combine the signal from as many stream members as possible to beat down the noise . @xcite
have shown that , for the regions of apparent magnitude and radial velocity space dominated by the stream , an accurate measurement of the systemic proper motion of an ensemble of stars belonging to the stream can be obtained through simple background subtraction .
alternatively , the overall stellar distribution in the space of observables can be modeled , yielding the direct contributions of the galaxy and the stream .
if such models can be cast in the 3d space of proper motion and magnitude or radial velocity , it is feasible that a superior measurement of the proper motion can be achieved as the contribution of the background to both systematic and random noise will be reduced .
unfortunately , adequate analytical models of the sgr stream and the galaxy are not readily available .
therefore , we choose to approximate these distributions by a sum of multi - dimensional gaussians .
this so - called gaussian mixture is a well known semi - parametric technique widely used to model multi - dimensional datasets @xcite .
gaussian mixtures have several key properties that make the model - fitting straightforward and fast .
first , the uncertainties associated with the measurements ( as well as the missing data ) are naturally incorporated into the model .
secondly , there exists a guaranteed fast convergence procedure the expectation minimisation ( em ) algorithm @xcite . in this work ,
we have used the extreme - deconvolution package , the open - source gaussian mixture implementation by @xcite . as evident from the dissection of the sdss dataset shown in fig . [
fig : sag_map ] , main sequence turn - off stars ( msto ) are the most numerous sgr stream specimens available to carry out the proper motion analysis .
the typical density of the msto stars in the stream is 150 stars per square degree , compared to the foreground density of around 50 .
stripe 82 slices through both the bright and the faint streams at an angle , resulting in a slightly tilted bi - modal distribution in the plane of right ascension and @xmath36 band magnitude .
a faint eastern wing to the bright stream can also be discerned at higher right ascension ( see middle panel of fig . [
fig : sag_map ] ) .
motivated by the evolution of the distance and the structure of the stream along the stripe , we split the msto sample into three parts : the faint stream at @xmath72 , the bright stream at @xmath73 , and the eastern wing of the bright stream at @xmath74 ( we label those fields as fp1 , fp2 , fp3 respectively ; they are shown by red error - bars on fig .
[ fig : sag_map ] ) . in each of the three right ascension bins , the overall 3d distribution of msto stars in the space of ( @xmath75 )
was modeled with a mixture of 5 gaussians , the initial guess for the free parameters was obtained by running the k - means algorithm @xcite .
the choice of number of gaussians ( @xmath76 ) was motivated by looking at the cross - validated log - likelihood @xcite as a function of @xmath76 .
this initially rises as a function of @xmath76 , peaks at @xmath77 5 and stays roughly constant for @xmath785 , demonstrating the goodness of fit and the absence of overfitting .
we have also made sure that the objects in the sgr stream are represented by a single gaussian .
when applying the extreme deconvolution , we force the covariance matrix for the sgr component to be diagonal , e.g. assuming zero covariance between the proper motion and the apparent magnitude .
the proper motions for the stream stars were not corrected for the solar reflex motion .
the errors in the proper motion measurement , i.e. the uncertainties in determining the centre of the gaussian representing the sgr stream , were determined either from bootstrap procedure or from the hessian matrix of the likelihood function .
[ fig : pm_hist ] shows the data analysed as well as the best fit gaussian mixture model of the main sgr stream @xmath73 .
it is reassuring to see that the gaussian mixture model was able to describe the data distribution adequately .
the resulting measurements of the proper motion together with the uncertainties are given in table [ tab : pm_tab ] . among @xmath0 7000 stars analyzed , according to the model ,
@xmath0 4000 stars belong to sgr , while @xmath0 3000 stars belong to the background / foreground population .
the number of members of the sgr stream with sdss spectroscopic measurements is significantly smaller than the number of msto stars used in the previous section . despite this , knowledge of the radial velocities and surface gravities allows us to have much purer samples of the sgr stream members . in this section ,
we perform a complementary measurement of the stream s proper motion using spectroscopic members only . to this end , we select all stars with spectra in stripe 82 lying at @xmath79 ( we label this area fs4 and show it with the blue error - bar on fig . [
fig : sag_map ] ) and classified by the sdss spectroscopic pipeline as giants @xmath64 . despite very wide range of selected right ascensions , most of the sgr members with spectroscopy in the stripe 82 region located at the center of the bright stream , at @xmath80 . according to the right panel of fig .
[ fig : sag_map ] , the sgr stream s radial velocity changes along the stripe . therefore , to ease the modeling
, we subtract the variation of the radial velocity centroid from the data using the measurements presented in table [ tab : sgr_vel ] .
the resulting radial velocity is approximately constant as a function of ra , and the measurements of proper motions and radial velocities @xmath81 can now be represented by a mixture of gaussians .
we run the extreme deconvolution on the sample of @xmath01500 stars and find that 3 gaussian components are sufficient to describe the dataset .
we follow the strategy outlined in section [ sec : phot_sample ] , with the difference that the covariance matrix of the gaussian representing the stream component is now set free . fig .
[ fig : vel_pm_hist ] shows the density distribution of the data , together with the best - fit gaussian mixture model for the stripe 82 stars with spectra . by comparing the grey - scale density with the red contours
, we can confirm that the model has captured the properties of the dataset reasonably well .
the last line of table [ tab : pm_tab ] reports the values of the proper motion for the spectroscopic sample , and confirms that the measurements for the two independent stream samples agree within 2@xmath21 .
it is also particularly reassuring since sgr members in the spectroscopic sample have very different colors ( g@xmath30i @xmath0 0.75 ) from the sgr members in the photometric sample ( g@xmath30i @xmath0 0.3 ) of the sgr giants .
therefore the agreement of the proper motions from two samples is a proof of a small level of color - related systematic effects .
the proper motion signals we have measured have a very large contribution of the solar reflex motion . since we have distance estimates to the sgr stream in the south measured elsewhere ( e.g. * ? ? ?
* ) , we can correct for this and check whether the proper motions are actually aligned with the streams .
figure [ fig : arrows_plot ] shows the proper motion vectors after applying the solar reflex corrections . as we can clearly see , the proper motions are indeed properly aligned with the streams and are consistent with each other .
for this calculation we have still used the distance to the faint stream as given by @xcite , although , there is evidence in @xcite that the stream the faint stream is 3@xmath305 kpc closer .
there are many models of the sgr stream in the literature , all purporting to provide the distances , velocities and proper motions as a function of position on the sky ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
although quite detailed , all the models fail to reproduce at least some of the features that we see on the sky ( e.g. , * ? ? ?
* ; * ? ? ?
but , it still instructive to see where our data measurements lie relative to the existing models .
we have chosen the @xcite model as a comparison benchmark , because it is arguably the most comprehensive and up - to - date .
there are several caveats to be borne in mind .
the first is related to the fact that while our central measurement in stripe 82 at @xmath82 or @xmath83 corresponds directly to the center of the trailing tail in the simulation , the fainter stream which crosses stripe 82 at @xmath84 or @xmath85 does nt have a counterpart in the simulation by @xcite .
the second is related to the choice of rotation velocity of the local standard of rest . as shown in @xcite ,
the observed proper motion signal is sensitive to the adopted v@xmath86 .
the models of @xcite have been computed and fitted under the assumption of the iau standard value of the @xmath87=220kms@xmath1 , while the current best estimates are slightly higher at @xmath88kms@xmath1@xcite . throughout the paper
, we adopted the value of 235kms@xmath1 .
[ fig : comparison ] shows our data points overplotted onto the distribution of tracers from @xcite model , where we have selected tracers from the trailing tail only ( @xmath89 and @xmath90 ) , and lying within 20 degrees of stripe 82 ( @xmath91 ) .
the @xmath69 values of our measurements correspond to the centers of the rectangular areas used to perform the measurements , and the error - bars indicate the extent of those areas .
the agreement for the central field ( @xmath92 ) , as well as for the edge of the bright stream ( @xmath93 ) , is quite good . at the location of the fainter sgr stream ( empty circle on the plot )
, the model does nt have many particles , as expected . in order to facilitate further comparisons , table [ tab:3dvel_tab ]
gives the positions and velocities corresponding to our measurements . here , ( @xmath94 ) and ( @xmath95 ) are in a right handed galactocentric coordinate system with the sun located at @xmath968.5kpc , and @xmath97kms@xmath1 .
the distances taken from the work by @xcite , and the error on the velocities does not take into account any possible systematic error distance of @xmath98 .
readers wishing to compare sgr disruption models with our observations should use table [ tab:3dvel_tab ] only for rough checks .
a proper comparison entails measuring the average proper motions within the same area of the sky as we do and comparing these numbers directly with our table [ tab : pm_tab ] .
we also compare our measurements with the data from @xcite . out of 6 fields analysed by @xcite ,
4 have measurable sgr signal and only 2 of those ( sa 93 and sa 94 ) lie near the stripe 82 .
the other 2 fields with detectable sgr signal are @xmath0 15 degrees away from the stripe 82 . as figure [
fig : comparison ] shows , the error - bars of our measurements are much tighter than @xcite , and the proper motions themselves agree approximately within the combined errors .
the comparison of the measured stripe 82 proper motions ( red with error bars ) with both the earlier measurements of ( * ? ? ? * grey with error bars ) and the simulations of ( * ? ? ?
* blue , green , yellow dots ) .
our measurement of the proper motion of the fainter of the two trailing tails is identified by an empty red circle .
note that the datapoints from @xcite are selected from trailing tail only and from the region near stripe 82 @xmath99 .
the points are colored according to @xmath100 such that points with @xmath101 are dark blue , points with @xmath102 are orange , and points near stripe 82 are light - green .
two of the fields from @xcite lie within stripe 82 , while two other fields are @xmath0 15 degrees away from the stripe 82 . ]
we have measured the proper motion of the sgr stream using stripe 82 data . by tying the astrometry to the known qsos and combining the measurements for large samples of stars , we have been able to achieve high precision proper motions .
our measurements have been performed for two distinct groups spectroscopically selected red giants / subgiants with sdss radial velocities and photometrically selected msto stars .
the proper motions of those agree very well and after correcting for the solar reflex motion are tightly aligned with the sgr stream .
our results are in agreement with earlier measurements by @xcite , but , by virtue of the large numbers of stars in our samples , our statistical error bars are substantially smaller , typically about @xmath103 mas yr@xmath1 .
there are three fields in stripe 82 for which the proper motion of the photometric sample has been measured , and one field for the spectroscopic sample .
combining this information with distances from @xcite gives us the full six dimensional phase space coordinates of the sgr trailing stream at four locations along stripe 82 .
we provide a table of three dimensional positions and velocities of sgr stream stars , in which the contributions of the motion of the sun and lsr have been removed . to complement the work on the sgr streams in the south , it would be particularly useful to carry out a corresponding kinematical analysis for the streams in the north .
this is a subject to which we plan to return in a later contribution .
the combination of the proper motions , radial velocities and distances in both galactic hemispheres should allow us to make further progress in the understanding of the complicated structure of the sgr streams and solve some of the riddles posed by their existence . | using multi - epoch observations of the stripe 82 region done by sloan digital sky survey , we measure precise statistical proper motions of the stars in the sagittarius stellar stream .
the multi - band photometry and sdss radial velocities allow us to efficiently select sgr members and thus enhance the proper motion precision to @xmath0 0.1masyr@xmath1 .
we measure separately the proper motion of a photometrically selected sample of the main sequence turn - off stars , as well as of a spectroscopically selected sgr giants .
the data allow us to determine the proper motion separately for the two sgr streams in the south found in @xcite .
together with the precise velocities from sdss , our proper motion provide exquisite constraints of the 3-d motions of the stars in the sgr streams . |
in a recent paper , lieu , mittaz , zhang 2006 ( lmz06 ) published the hitherto most comprehensive direct correlation of the _ wilkinson microwave anisotropy probe _ first year ( wmap1 ) data with the x - ray data of rosat and asca , in search for the sze in the temperature of the cosmic microwave background ( cmb ) along the directions to 31 randomly chosen rich clusters located above the galactic plane .
the wmap1 passbands being analyzed cover the frequency range of 41 94 ghz .
this investigation led to the astonishing finding that on average the level of sze detected by wmap1 is no deeper than the intrinsic cmb primary anisotropy as seen by wmap in directions of blank sky ( i.e. away from rich clusters and groups ) , and in any case accounts only for 1/3 to 1/4 of the level expected from the x - ray measurements of the sample clusters , moreover , lmz06 explored and excluded a variety of possible reasons for the discrepancy , including emission by radio point sources in the clusters , which fail by a large margin to deliver sufficient flux to explain the apparent lack of sze in the wmap1 w - band of 94 ghz . the results of lmz06 were corroborated recently by bielby & shanks 2007 , which presented a similar correlation study between wmap3 and x - ray observations by chandra and rosat , and likewise reported a substantially less than expected sze in the wmap3 data for a much larger sample size than the 31 clusters of lmz06 .
the additional information provided by bielby & shanks 2007 also included a ` truncation test ' where the authors abruptly cut off the x - ray gas profiles of the chandra cluster sample at a radius as small as 2 arcmin , and still found a sze discrepancy of a factor of two along the central line - of - sight .
thus the anomaly is real and must be resolved .
a separate and superficially unrelated phenomenon of clusters is the discovery of excess euv and soft x - ray emission in the energy range 0.1 1.0 kev which rises above the level expected from the spectrum of the hot virialized cluster gas .
this property has been known for more than one decade to exist in some clusters ( e.g. lieu et al 1996 , kaastra et al 1999 , nevalainen et al 2003 ) . in this paper
we demonstrate that it is possible to connect the non - thermal inverse compton interpretation of the cluster soft excess ( hwang 1997 , ensslin & biermann 1998 , sarazin & lieu 1998 ) with the sze anomaly in wmap by extending the dynamic range off this power - law distribution of electrons , so that the number density of electrons causing the sze scattering is actually reduced .
we also examine the viability of attributing any remaining sze flux discrepancy ( after the above effect is taken into account ) , to synchrotron radiation from an unmapped intracluster population of relativistic electrons . apart from the three 1997 - 98 papers ,
there have been numerous suggestions of a general , non - thermal intracluster environment , an early example being jaffe ( 1977 ) and , more recently , quenby et al ( 1999 ) . in particular , a model of the acceleration via alfven waves driven by major cluster mergers is given by brunetti et al ( 2004 ) .
of further interest is the idea that relativistic jets carry a significant portion of the total energy output of radio galaxies , causing x - ray emission up to mpc distance scales ( ghiellini & celotti 2001 ) .
celotti , ghisellini , & chiaberge ( 2001 ) provided an analysis of such a scenario for pks 0637 - 752 . unless a separate physical mechanism is proposed ( and indeed there is at least one serious paper on the prospect of neutralino dark matter decay as a cluster s non - thermal reservoir , see colafranceso , profumo , & ullio 2006 ) , the possibility of cosmic rays as the key to solving the soft excess and s - z puzzles hinges upon the manner in which energy from agn is distributed widely throughout a cluster , and the speed in which re - acceleration can compensate for losses .
it is well known that bhm diffusion happens too slowly for this purpose ( see section 3 ) .
our model invokes alfven waves as the spreading agent which is also responsible for rapid fermi statistical acceleration .
we assume a continuous non - thermal injection rate into the cluster environment of 10@xmath2 ergs s@xmath3 in both electron output and poynting flux .
this is based upon the analysis of observations by ghisellini and celotti ( 2001 ) and the the numbers arising from jet injection in the hydrodynamic models of zanni et al ( 2005 ) .
the jet may be of cluster size in one dimension ( nulsen et al 2005 ) .
the cluster radius is assumed to be @xmath4 1 mpc and the intracluster magnetic field in the range 1 to 10 @xmath1 g .
( e.g. govoni et al 2001 ; medvedev , silva , & kamionkowski 2005 ) .
ambient gas density is @xmath4 10@xmath5 to @xmath6 @xmath7 ( e.g. brunetti et al 2004 ) .
we also took a typical cluster distance of 400 mpc , an appropriate number for the clusters of the lmz06 sample , since the mean redshift of the sample is @xmath8 0.1 , corresponding to a distance of 428.57 mpc in a @xmath9 0.3 , @xmath10 0.7 , and @xmath11 0.7 cosmology ( bennett et al 2003 , spergel et al 2006 ) . while an energetic jet existing a reasonable fraction of the lifetime of a large , relaxed cluster can transport energy in one dimension on a cluster scale of 1.0 mpc in @xmath12 year for a jet propagating at 1000 km s@xmath3 , a more general three dimensional dispersion mechanism is necessary to account for the distributed acceleration required here .
we adopt a b=6 @xmath1 g magnetic field , tangled on a @xmath13 kpc scale over the central 0.5 mpc region and a density of 10@xmath14 @xmath7 ( eg coma cluster , fretti et al . , 1995 ) as representative parameters .
alfven waves then move with velocity @xmath15 410 km / s , allowing the transport of non - thermal energy over a cluster volume , radius 0.4 mpc , in @xmath16 years .
instabilities within the jet and at the edges may produce the alfven waves which continue to spread the energy after the bulk flow is dissipated .
rayleigh - taylor and kelvin - helmholz instabilities in particular were mentioned by zanni et al ( 2005 ) .
these authors show how the disturbance in density and entropy could have spread after the jet and shock were switched off , the computations being followed up to @xmath12 years .
medvedev , silva , & kamionkowski ( 2005 ) computed a model of cluster - wide field amplification caused by the non - relativistic weibel instability , although only 0.1 % of the total energy ends up in the magnetic field .
they claim that the short wavelength turbulence initially generated gives rise to much longer wavelengths on a cosmological time scale .
nishikawa et al , 2003 , provide an example of this instabilty generating turbulence in a relativistic jet .
merger shocks and intense stellar winds are alternative sources of heat input to the cluster medium which may add to the wave energy .
fretti et al ( 2004 ) developed this idea and fitted synchrotron and inverse compton models to observed spectra , based on giant radio halos .
a non - thermal component to the electron spectrum is indicated by measurements of a cluster soft excess from regions where thermal gas at a temperature considerably lower than virial is not likely to exist .
sarazin and lieu ( 1998 ) estimated the intracluster cosmic ray electron energy necessary to account for the observed euv luminosity of some clusters as due to inverse compton ( ic ) interaction between the electrons and the cmb .
the ic radiation power is predominantly in the euv and soft x - rays waveband , when the cmb scatters off lower energy ( 300 @xmath17 1,000 ) electrons .
they quoted @xmath18 and demanded @xmath19 for a1795 @xmath20 may be as high as 10@xmath2 ergs s@xmath3 ( mittaz , lieu , and lockman 1998 ) , while for the coma cluster it is @xmath4 a few @xmath21 10@xmath22 ergs s@xmath3 ( lieu et al 1999 ) .
thus @xmath23 ranges from 10@xmath24 to 10@xmath25 ergs s@xmath3 .
we will take the larger of the two values of cosmic ray energy content , viz .
@xmath26 erg , as requiring explanation , further , we map the ic output photons to electrons in the energy range just above 100 mev , and assume that the electron spectrum cuts off at energies much below this value . recently , the report of strong soft x - ray excess in the cores of the clusters as1101 ( werner et al 2007 ) and a3112 ( bonamente et al 2007 ) bolsters the non - thermal interpretation of the soft excess from cluster centers . the alternative model which invokes a warm thermal component to account for
the excess must explain how this component can co - exist with the hot virialized cluster medium without being clumped into rapdily cooling high density clouds .
moreover , stringent upper limits on the amount of intracluster o vii were derived from these observations , due to the absence of detectable signatures of o vii emission , the most sensitive tracer of warm gas hitherto available ( see also lieu & mittaz 2005 ) .
the 2 - 7 kev luminosity of the power - law population of emitted photons required to fit the soft excess data is an appreciable fraction of the total x - ray luminosity in this passband , reaching 30 % in the case of a3112 ( bonamente et al 2007 ) . to understand how the presence of non - thermal electrons could mean a significant reduction in the total density of electrons in the region of concern of a cluster ,
let us compare the radiative losses of a single electron by thermal bremstrahlung and inverse compton scattering . in an electron - proton plasma ,
the thermal loss at energy @xmath27 and velocity @xmath28 is @xmath29 for plasma density @xmath30 , gaunt factor @xmath31 , electron radius @xmath32 and fine structure constant @xmath33 . for inverse compton losses at electron lorentz factor @xmath34 , @xmath35 where @xmath36 is the cmb energy density .
hence the ratio of thermal to non - thermal power is @xmath37 at a virialized gas temperature of kt=9.5 kev .
hence the required number density of @xmath38 electrons is down by a factor @xmath39 relative to the number density of 10 kev electrons .
the energetic , cosmic ray electrons in the cluster should not exceed equipartition with the magnetic field . to see that there is no danger here ,
take an x - ray luminosity of @xmath40 ergs s@xmath3 for the central 0.5 mpc cluster region and , using the ic loss rate of @xmath41 one obtains a cosmic ray number density of @xmath42 @xmath7 and an energy density @xmath43 ergs @xmath7 ; this is to be compared with an equivalent equipartition field , which is as high as 10@xmath1 g. evidence for cross - check the determination of a cluster s mass by attributing all its x - ray luminosity to a hot virialized gas came from weak gravitational lensing .
smail et al .
( 1997 ) find a @xmath44 efficiency correction is needed to the lens shear strengths to bring x - ray determined masses into agreement .
a lower x - ray cluster mass resulting from some of the emitted x - rays being of non - thermal origin would imply a higher efficiency of shear measurement , i.e. the correction factor of 75 % may not be necessary .
adoption of the reduced number density in the thermal electron cluster component will also lead to an increase in the heavy element abundances inferred from line intensities because the continuum levels are reduced .
it is not useful to invoke diffusive shock acceleration as the distributed cosmic ray electron source since the larmor radius , @xmath45 , even at 10@xmath46 ev electron energy is only @xmath4 0.1 pc in a 1 @xmath1 g field , i.e. diffusive propagation over cluster scales is impossible ( see also the second half of this section ) .
instead we appeal to a general fermi acceleration phenomenon throughout the medium . however , the existence of shock accelerated cosmic rays at the boundaries of the large scale agn jet driving the additional cluster heating is important for the generation of the necesaary alfven wave spectrum .
the model assumes a continuous injection of energy into at least short wavelength alfven waves from the agn injection , continuing to take place well after switch off of the jet as the large scale pressure wave spreads . as the alfven wave front propagates across the cluster , the broad spectrum cosmic ray population streams ahead into a low scattering medium .
quasi - linear theory predicts a diffusive scattering coefficient ( bell 1978 ) , @xmath47 where @xmath48 is the energy density of the waves in resonance with particles of momentum p per unit logarithmic bandwidth .
scattering and hence acceleration and wave growth depends on the relation @xmath49 between wave number and larmor radius .
bell ( 1987 ) shows also that an approximation to quasi - linear theory for the growth rate of waves due to streaming is @xmath50 @xmath51 is the particle distribution function .
the quantity @xmath52 is the pressure per unit logarithmic bandwidth of particles of momentum p. assuming equipartition between particle and wave energy at resonance in each energy band within the turbulent alfven wave front driven by the region of shock acceleration , the growth rate in the region originally of lower scattering power is simply @xmath53 where @xmath54 is the scale size for the gradient in @xmath51 .
a @xmath55 pc produces @xmath56 years , the time constant we will show is needed for acceleration of @xmath57 ev electrons , the energy obtained in the most extreme version of the model developed .
this may be compared with a minimum structural size determined by the larmor radius at this energy when the jet is far out of @xmath58 cm .
the strong turbulence resulting from the streaming instability allows a diffusion mean free path within an order of magnitude of the bhm limit throughout the accelerated particle energy range ( zank et al 2000 ) .
this treatment of scattering is in contrast to that of brunetti et .
2004 ) who used an assumed initial kolmogorov or kraichman turbulence spectrum determined by a minimum k value to specify the quasilinear diffusion coefficent and then allow the wave particle interactions to cause decay of the wave s spectral intensity . the fermi acceleration time constant is @xmath59 where @xmath60 is the particle velocity , and @xmath61 is the diffusion mean free path .
experience in measuring @xmath61 in the turbulent interplanetary medium suggests @xmath62 , rather than the bhm value .
the lifetime of an electron of energy @xmath63 against inverse compton interactions ( losses ) on the cosmic microwave background ( cmb ) of energy density @xmath64 , is @xmath65 by equating inverse compton and fermi time constants after applying our adopted parameters , we obtain a maximum permitted electron energy of 1.9 @xmath21 10@xmath46 ev . and a life of @xmath66 years .
this limit is important to our ensuing calculation of a possible upper limit to the electron spectrum . with negligible reacceleration
, the loss time at @xmath38 is @xmath67 years .
vlk , aharonian and breitschwerdt ( 1996 ) reviewed the non - thermal energy content of galaxies and discussed the contribution of shocks to the cosmic ray content .
we may consider two alternatives to the alfvenic heating model for the supply of high energy electrons .
the first employs shocks at collisions between clusters and hence parameters for the medium inside clusters , so @xmath68 cm / s and @xmath69 g. if the model to explain the sze anomaly requires distributed synchrotron radiation at 41 ghz , the electron @xmath34 given by @xmath70 needs to reach @xmath71 , where the ic loss time is @xmath72 years . now using as the approximate shock acceleration time @xmath73 where @xmath74 is the diffusion coefficient
, one finds that @xmath75 years .
however , the time to diffuse over a cluster scale of @xmath76 kpc , @xmath77 is @xmath78 years .
the second alternative concerns relativistic agn jets near their source . we take
@xmath79 and @xmath69 g. these parameters yield @xmath80 year .
the timescales relating to shock acceleration indicate that there should be no lack of localised sources of relativistic electrons , especially in an agn environment where a relativistic jet of lorentz factor @xmath81 produces the @xmath82 factor increase of acceleration first noted quenby and lieu ( 1989 ) .
however , the long diffusion timescale spells the impossible task of filling the entire cluster with a non - thermal population using isolated diffusive shock acceleration sources as input , i.e. the energetic particles so produced are expected to emit synchrotron radiation only locally .
vlk et al had similarly found extremely long times for the escape of particles from clusters .
as an alternative , or even an addition to the model explaning of the sze , the extension of the cluster non - themal population to energies sufficient to provide synchrotron radiation in the wamp response range is expolored . without a detailed , non - linear model of the cosmic ray cluster electron spectrum , which is beyond the scope of this work , we can not predict the overall spectral shape .
models without particle - wave back reaction and relying only on synchrotron loss produce a pile up at high energies before a cutoff ( borovsky and eilek , 1986 ) .
a typical model where turbulent driving of waves occurs , in this case from eddies inherent in jet turbulence radiating waves , is studied by eilek and henriksen ( 1984 ) .
they find power a power law electron and synchrotron spectrum can arise . here
we will assume a power law , but only need it to relate low energy electrons yielding inverse compton photons and high energy electrons yielding synchrotron photons without need of detailed knowledge of the shape in between .
take the intracluster cosmic ray electron spectrum as @xmath83 the power output is @xmath84 now the peak frequency of the synchrotron spectrum is given by @xmath85 if the lower limit to @xmath34 could correspond to @xmath86 200 , or electron energy @xmath87 100 mev , below the peak of the cosmic ray electron flux in our galaxy , the minimum emitted frequency is @xmath88 1.7 @xmath21 10@xmath89 hz . the upper emitted frequency , corresponding to the electron cutoff energy of 1.9 @xmath21 10@xmath46 ev as explained after eq . (
11 ) , is @xmath90 3.4 @xmath21 10@xmath91 hz . the cluster volume assumed is @xmath92 @xmath93 . to explain the anomalies in the s - z observations , a synchrotron power output @xmath87 20,000 times less than that of the cmb ( i.e. @xmath94 ergs @xmath95 s@xmath3 sr@xmath3 hz@xmath3 ) is required at 41 ghz .
thus the power at this frequency is defined for equation ( 16 ) and represents the measurement at one energy of the assumed cosmic ray electron spectrum .
the second measurement the electron specrum is afforded by the cluster soft excess .
more precisely , by taking @xmath26 ergs as the cosmic ray energy content and linking the ic photons to the electron energy range just above 100 mev for an electron spectrum that cuts off below this value , a second point on the electron spectrum is defined .
the electron spectrum may be recast into a more recognizable form , as @xmath96 where @xmath97 a typical set of spectral parameters satisfying both the sze anomaly and the euv excess are : @xmath98 @xmath7 sr@xmath3 and @xmath994.0 @xmath21 10@xmath100 @xmath7 ev@xmath101 , applicable to the energies between 100 mev and 2 @xmath21 10@xmath102 gev , with an ensuing intracluster relativistic electron number density of @xmath4 2.0 @xmath21 10@xmath103 @xmath7 . for comparison
, our galaxy contains a cosmic ray proton number density of 6 @xmath21 10@xmath103 @xmath7 .
there is thus _ no difference _ between the requirements on acceleration in the intracluster medium containing a central agn and the cosmic ray flux in a normal galaxy . to ensure that the model works self - consistently , we emphasize that the proposed intracluster non - thermal electrons will not by themselves produce any obvious sze signal above the normal level from the intracluster hot gas . moreover
, their synchrotron radiation at radio frequencies is below normal radio astronomy sensitivity limits .
a typical cluster observation has its lowest contour at 1 mjy over a 43 arcsec beam resolution .
this corresponds to a background from the cluster of 10@xmath104 ergs @xmath95 s@xmath3 sr@xmath3 hz@xmath3 . thus , unless we contemplate a larger cosmic ray population than depicted by eq .
( 16 ) using maximum associated parameters , it would require an exceptionally low noise observation to pick out the aspect of cosmic ray signal suggested here .
there are indeed some large clusters where a radio halo appears .
giovannini et al ( 1999 ) found @xmath4 5 % of a complete x - ray selected cluster sample to have diffuse radio emission .
synchrotron radial profiles more extended than that of x - rays can produce 2 2.5 mpc source sizes , while mini - halos confined to central regions and radio relics on the cluster periphery were also reported ( brunetti 2004 ) .
there are some indications of non - thermal , hard x - ray cluster emission , but a typical egret gamma ray upper limit ( on the coma cluster , sreekumar et al , 1996 ) is 3 @xmath21 10@xmath103 photon @xmath95 s@xmath3 at 100 mev
. the higher @xmath34 end of our proposed electron power - law will produce gamma rays , also by the ic effect on the cmb . for a particular energy of the emerging photon
the required electron @xmath34 is given by @xmath105 and using the equation for the rate of ic energy loss for an electron , we find an electron density @xmath106 electrons @xmath7 above the recoil threshold for ic induced 100 mev gamma emission . assuming most of the emission is concentrated at around the threshold of @xmath107 6 @xmath21 10@xmath89 , and a cluster distance of 100 mpc , the flux at earth is @xmath4 1 @xmath21 10@xmath108 photon @xmath95 s@xmath3 , i.e. even under the scenario of maximal cosmic ray pressure the egret limit is not violated by our model .
although the basic ideas of utilizing agn activity and alfvenic heating as a prime energy source for the intracluster medium have been proposed , we focussed upon a number of specific non - thermal cluster processes .
our model started with the limited spatial extent of the agn related jet feeding .
alfven waves then distribute the energy within a cluster time frame , resulting in a cluster - wide population of relativistic electrons with a power - law spectrum .
the chief constraint on electron lifetime , and hence spectral hardness , is inverse compton loss on the cmb .
provided a reasonable fraction of the observed cluster agn power can spread outwards , there will be enough energy to explain the wmap1 sze anomaly of lmz06 and the cluster soft excess phenomenon .
the proposed mechanism is in - line with previous ideas in interpreting giant radio halos , although we treat the generation of the intracluster wave spectrum in a different manner . among the two models offered ,
the simplest explanation of the wmap sze anomaly is that the cluster electron population has hitherto been overestimated , because all the emitted x - rays from a cluster were attributed to the virialized hot thermal medium when as many as half of these thermal electrons should in fact be replaced by a much smaller population ( in terms of number density ) of cosmic ray electrons accelerated to modest lorentz factors . after correcting for this effect ,
the result is a significant reduction of the predicted sze . in the alternative ( more indirect , and hence perhaps less likely ) approach to solving the observational problems , we invoked synchrotron radiation in the microwave frequency range from the higher energy end of the same power - law distribution of cluster cosmic ray electrons as that responsible for the sze anomaly .
such electrons can exist only if a continuous input from alfvenic acceleration at high lorentz factors ( to combat ic losses ) is available .
the salient features of our proposed model to reduce the thermal electron content of clusters are summarized as follows .
( a ) for non - thermally active clusters the wmap sze anomaly would not implicate negatively upon the cosmological origin of the cmb .
rather , it probes the properties of the intracluster medium , which ( for some clusters at least ) harbor a cosmic ray energy density approaching that of our own galaxy .
( b ) contribution to cluster mass from the thermal medium are significantly overestimated in clusters with anomalous sze .
cosmic abundances may be underestimated .
( c ) anomalous sze could arise from clusters with one or more of the following special characteristics : powerful agns , cluster scale x - ray jets , long radio jets , radio ` ridges ' , radio halos , or other evidence for recent cluster merger .
( d ) similarly , soft excess emissions are more likely where there is evidence of large - scale cluster turbulence , as above .
( e ) extended cluster radio emission identified in some massive clusters could be explained by an enhanced version of the same alfvenic heating and ic loss balance presented herein .
( f ) relatively higher fluxes of non - thermal x - rays are emitted by clusters with anomalous sze .
( g ) the full sze is expected to be found in relaxed clusters that do not exhibit any of the characteristics outlined in section 3 .
sreekumar , p. , bertsch , d. l. , dingus , b. l. , esposito , j. a. , fichtel , c. e. , + fierro , j. , hartman , r. c. , hunter , s. d. , kanbach , g. , kniffen , d. a. , lin , y. c. , + mayer - hasselwander , h. a. , mattox , j. r. , michelson , p. f. , von montigny , c. , + mukherjee , r. , nolan , p. l. , schneid , e. , thompson , d. j. , and willis , t. d. , 1996 , + apj .
464 , 628 . | wmap s detection of the sunyaev - zeldovich effect ( sze ) at a much reduced level among several large samples of rich clusters is interpreted in terms of conventional physics .
it is by now widely believed that the central soft x - ray and euv excess found in some clusters can not be of thermal origin , due to problems with rapid gas cooling and the persistent non - detection of the o vii line , but may arise from inverse - compton scattering between intracluster relativistic electrons and the cosmic microwave background ( cmb ) .
in fact , recently chandra and xmm observations of the soft x - rays from abell 3112 are well fitted by the sum of a power law and a thermal virialized gas component of comparable luminosities .
therefore the missing sze flux could simply be due to an overestimate of the central density of virialized electrons which scatter the cmb .
we also considered if higher energy electrons drawn from the same power - law population as those responsible for the soft excess may synchrotron radiate in the intracluster magnetic field of strength b @xmath0 a few @xmath1 g to produce cluster microwave emissions in the wmap passbands that account for the missing sze flux .
either explanation of the wmap anomaly would bolster the current model of central cluster soft excesses , viz .
non - thermal activities prevail in the core of at least some clusters
. the energetic electrons may originate from agn jet injection , then distributed cluster - wide with accompanying _ in situ _ fermi acceleration , by alfven waves .
however the missing thermal electron explanation provides a less demanding model . |
Subsets and Splits
Data Mining Abstracts
This query searches for and retrieves 2000 documents containing the phrase 'data mining' in the abstract, providing limited but specific document samples.