article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
freezing is a basic phenomenon , the most inevitable of all phase changes .
when a liquid freezes into a crystalline solid its continuous symmetry of translation and rotation is broken into one of the bravais lattices .
a crystalline solid has a discrete set of vectors @xmath1 such that any function of position such as one particle density @xmath2 satisfies @xmath3 for all @xmath4 [ 1 ] . because of localization of particles on lattice sites , a crystal is a system of extreme inhomogeneities where values of @xmath2 show extreme differences between its values on the lattice sites and in the interstitial regions .
the density functional formalism of classical statistical mechanics has been used to develop theories for liquid- solid transitions [ 2 , 3 ] .
this kind of approach was initiated in 1979 by ramakrishnan and yussouff [ 4 ] .
the central quantity in this formulation is the excess reduced helmholtz free energy due to interparticle interactions of both the crystal @xmath5 $ ] and the liquid @xmath6 $ ] [ 4,5 ] . for the crystal @xmath5 $ ] is a unique functional of @xmath7 whereas for the liquid @xmath8 is simply a function of liquid density @xmath9 .
the density functional formalism is used to write an expression for @xmath5 $ ] in terms of one and two - particle distribution functions of the solid [ 2 - 5 ] .
the direct pair correlation function ( dpcf ) that appears in this expression is a functional of @xmath10 [ 2 ] . when this functional dependence is ignored and the dpcf is replaced by that of the coexisting uniform liquid [ 4,5 ] or by that of an effective uniform fluid " [ 6,7 ] the free energy functional becomes approximate and fails to provide an accurate description of freezing transitions for a large class of intermolecular potentials .
attempts to include a term involving the three - body direct correlation function of the coexisting liquid in the free energy functional failed to improve the situation [ 8 , 9 ] .
it has recently been emphasized [ 10 - 12 ] that at the freezing point a qualitatively new contribution to the correlations in distribution of particles emerges due to spontaneous symmetry breaking .
this fact has been used to write the dpcf of a frozen phase as a sum of two qualitatively different contributions ; one that preserves the continuous symmetry of uniform liquid and the other that breaks it and vanishes in the liquid . the double functional integration in density space of a relation that connects @xmath11 $ ] to the dpcf led to an exact expression for @xmath11 $ ] . the freezing transitions in three - dimensions have been investigated using this new free energy functional .
the results found for the isotropic - nematic transition [ 10 ] , crystallization of power - law fluids [ 11 ] and the freezing of fluids of hard spheres into crystalline and glassy phases [ 12 ] are very encouraging .
in this paper we apply the free energy functional to investigate the liquid - solid transition in two - dimensions .
it may , however , be noted that in contrast to three - dimensional solid , a two - dimensional solid melts in two - steps ; the intermediate phase known as hexatic has a very narrow stability region in between liquid and crystal [ 13 - 15 ] . since inclusion of the hexatic phase in the density functional formalism
has not so for been possible , we neglect its presence and focus on the freezing of a fluid into the crystalline phase .
similar approach has been taken by others [ 16 - 21 ] . here
our motivation is to examine how well this new free energy functional ( described briefly in the following section ) compares with other free energy functionals in describing the crystallization of two - dimensional fluids .
the paper is organized as follow : in sec .
ii we give a brief description of the free - energy functional for a symmetry broken phase that contains both the symmetry - conserving and symmetry - broken parts of the dpcf . in sec .
iii we describe methods to calculate these correlation functions for a two - dimensional system .
the theory is applied in sec .
iv to investigate the freezing of power - law fluids into a crystalline solid of hexagonal lattice .
the paper ends with a brief summary and perspectives in sec . v.
the reduced free energy functional @xmath12 $ ] of an inhomogeneous system is functional of @xmath13 and is written as [ 2 ] @xmath14= a_{id}[\rho ] + a_{ex } [ \rho ] .
\nonumber \hspace*{5.5 cm } ( 2.1)\end{aligned}\ ] ] the ideal gas part @xmath15 is exactly known and is written in terms of @xmath16 as @xmath17 = \int { d \br \rho ( \br ) [ ln(\rho ( \br)\lambda)- 1 ] } , \nonumber \hspace*{4.cm } ( 2.2 ) \end{aligned}\ ] ] where @xmath18 is a cube of the thermal wavelength associated with a molecule .
the second functional derivative of the excess part @xmath19 $ ] with respect to @xmath16 defines the dpcf @xmath20 of the system [ 2 ] , @xmath21 } { \delta \rho(\br_{1})\delta\rho(\br_{2 } ) } = -c ( \br_{1},\br_{2 } , [ \rho ] ) .
\nonumber \hspace*{5.cm } ( 2.3)\end{aligned}\ ] ] the function @xmath22 that appears in this equation is related to the total correlation function @xmath23 through the ornstein -zernike(oz ) equation , @xmath24 both functions h and c are functional of @xmath16 since breaking of continuous symmetry of a uniform liquid at the freezing point gives rise to a qualitatively new contribution to correlations in the distribution of particles [ 10 - 12 ] , the dpcf of the frozen phase is written as a sum of two contributions ; @xmath25 ) = c^{(0)}(|\br_{2}-\br_{1}|,\rho_{0 } ) + c^{(b)}(\br_{1},\br_{2 } ; [ \rho ] ) , \nonumber \hspace*{3 cm } ( 2.5)\end{aligned}\ ] ] where @xmath26 is symmetry - conserving and @xmath27 symmetry - broken parts of the dpcf . while @xmath26 depends on magnitude of interparticle separation r and is function of average density @xmath28 ,
@xmath27 is invariant only under discrete set of translations and rotations and is functional of @xmath29 .
+ using eq .
@xmath30 we rewrite eq . @xmath31 as @xmath32}{\delta \rho(\br_{1 } ) \delta\rho(\br_{2 } ) } = - c^{(0 ) } ( |\br_{2}- \br_{1}|,\rho_{0 } ) , \nonumber \hspace*{5.5 cm } ( 2.6)\end{aligned}\ ] ] @xmath33}{\delta \rho(\br_{1 } ) \delta\rho(\br_{2 } ) } = - c^{(b ) } ( \br_{1},\br_{2},[\rho ] ) , \nonumber \hspace*{6.cm } ( 2.7)\end{aligned}\ ] ] where @xmath34 + { a^{(b)}}_{ex } [ \rho ] = { a}_{ex } [ \rho ] $ ] , the expressions for @xmath34 $ ] and @xmath35 $ ] are found from functional integrations of eqs . @xmath36 and @xmath37 , respectively . in this integration , as described elsewhere [ 10 - 12 ] , the system is taken from some initial density to the final density along a path in the density space ; the result is independent of the path of integration .
these integrations give , @xmath38 = { a}_{ex } ( \rho_{l } ) + \beta \mu - \ln { ( \rho_{l } \lambda ) } -{\frac{1}{2 } } \int { d\br_{1 } \int d \br_{2 } ( \rho(\br_{1 } ) -\rho_{l } ) ( \rho(\br_{2 } ) - \rho_{l } ) \overline{c}^{(0)}(|\br_{2}-\br_{1}| ) } \nonumber , \hspace*{.5 cm } ( 2.8 ) \end{aligned}\ ] ] and @xmath39 = -\frac{1}{2 } \int d\br_{1 } \int d\br_{2 } ( \rho(\br_{1 } ) - \rho_{0 } ) ( \rho(\br_{2 } ) - \rho_{0 } ) \overline{c}^{(b ) } ( \br_{1},\br_{2 } ) \nonumber , \hspace*{4.cm } ( 2.9)\end{aligned}\ ] ] where @xmath40 @xmath41 here @xmath42 is excess reduced free energy of the coexisting uniform liquid of density @xmath43 and chemical potential @xmath44 , @xmath45 is average density of the solid and @xmath46 , @xmath47 being the boltzmann constant and t is the temperature and @xmath48 is an order parameter arising due to breaking of symmetry .
the expression for @xmath49 $ ] given by eq .
@xmath50 is found from functional integrations when density @xmath51 of the coexisting fluid is taken as a reference .
the expression for @xmath52 $ ] given by eq .
@xmath53 is found by performing double functional integrations in the density space corresponding to the symmetry broken phase .
the path of integration in this space is characterised by two parameters @xmath54 and @xmath55 .
these parameters vary from @xmath56 to @xmath57 .
the parameter @xmath54 raises the density from zero to final value @xmath58 as it varies from @xmath56 to @xmath57 , whereas the parameter @xmath55 raises the order parameters from zero to their final values @xmath59 .
the result is independent of the order of integration .
+ the free energy functional for the symmetry broken phase is sum of @xmath60 $ ] , @xmath61 $ ] and @xmath62 $ ] thus @xmath63= \int d\br \rho(\br ) [ ln ( \rho ( \br)\lambda ) - 1 ] + { a_{ex}}(\rho_{l } ) + \beta ( \mu - ln(\rho_{l } \lambda ) ) \int d\br ( \rho ( \br ) -\rho_{l } ) \nonumber \\
-\frac{1}{2 } \int d \br_{1 } \int d \br_{2 } ( \rho ( \br_{1 } ) -\rho_{l } ) ( \rho ( \br_{2 } ) -\rho_{l } ) \overline{c}^{(0 ) } ( |\br_{2 } -\br_{1}| ) \nonumber \hspace*{3.cm } \\ - \frac{1}{2 } \int d \br_{1 } \int d \br_{2 } ( \rho ( \br_{1 } ) -\rho_{0 } ) ( \rho ( \br_{2 } ) -\rho_{0 } ) \overline{c}^{(b ) } ( |\br_{1},\br_{2}| ) \nonumber \hspace*{2 cm } ( 2.12)\end{aligned}\ ] ] where @xmath64 and @xmath65 are given , respectively by @xmath66 and @xmath67 . in deriving eq .
@xmath68 no approximation has been introduced . in the ramakrishnan - yussouff free energy functional @xmath69
is neglected and @xmath64 is replaced by @xmath70 . + in locating transition the grand thermodynamic potential defied as @xmath71 is generally used as it ensures that the pressure and the chemical potential of the two phases remain equal at the transition .
the transition point is determined by the condition @xmath72 , where @xmath73 is the grand thermodynamic potential of the coexisting liquid . from above expressions
one gets the following expression for @xmath74 ; @xmath75 -\frac{1}{2 } \int d \br_{1 } d \br_{2 } ( \rho ( \br_{1})-\rho_{l } ) ( \rho ( \br_{2})-\rho_{l } ) { \bf \overline{c}^{(0 ) } } ( |\br_{2}-\br_{1}| ) \nonumber \\ -\frac{1}{2 } \int d \br_{1 } d \br_{2 } ( \rho ( \br_{1})-\rho_{0})(\rho ( \br_{2})-\rho_{0 } ) { \bf { \overline{c}}^{(b ) } } ( |\br_{1},\br_{2}| ) .
\nonumber \hspace*{4.cm } ( 2.14 ) \end{aligned}\ ] ] minimization of @xmath74 with respect to @xmath76 subjected to the perfect crystal constraint leads to @xmath77 where @xmath78 and @xmath79 the value of lagrange multiplier @xmath80 in eq .
@xmath81 is found from the condition @xmath82 one needs , in principle , the values of @xmath70 and @xmath83 to calculate self consistently the value of @xmath29 that minimizes w. in practice however , one finds it convenient to do minimization with an assumed form of @xmath16 .
the ideal part is calculated using a form for @xmath16 which is a superposition of normalized gaussians centred around the lattice sites ; @xmath84 , \nonumber \hspace*{3.5 cm } ( 2.17 ) \end{aligned}\ ] ] where @xmath85 is the localization parameter .
for the interaction part it is convenient to use the fourier expression , @xmath86 where @xmath87 are reciprocal lattice vectors ( rlv s ) of the lattice and @xmath88 are order parameters , taking fourier transform of eq .
@xmath89 one finds @xmath90 .
we consider model fluids interacting via inverse power pair potentials @xmath91 ; where @xmath92 and n are potential parameters and r is molecular separation .
the parameter n measures softness of the repulsion ; @xmath93 corresponds to the hard disk and @xmath94 to the one component plasma .
such repulsive potentials can be realised in colloidal suspensions .
one such systems in two - dimensions has been provided by paramagnetic colloidal particles in a pendant water droplet , which are confined to the air- water interface [ 13 ] . by applying an external magnetic field perpendicular to the interface , a magnetic moment
is introduced in the particles resulting into a tunable mutual dipolar repulsion between them .
the pair interaction thus created is repulsive and proportional to @xmath95 .
the crystallization of this system has been investigated by van teeffelen et.al .
[ 20,21 ] using several versions of density functional theory ( dft ) .
other example where short range repulsion between molecules is found is microgel spheres whose diameter could be temperature tuned [ 15 ] .
most computer simulation studies on these systems suffer from the finite - size effects . in case of hard disks
recently a large scale monte- carlo simulation , large enough to access the thermodynamic regime , has been performed [ 22 ] .
the result confirms two - steps transitions from liquid to solid with the intermediate haxatic phase[23,24 ] .
however , the liquid- hexatic transition , in contrast to the prediction of kosterlitz - thouless - halperin - nelson - young(kthny ) theory [ 23,24 ] is found to be first - order while the hexatic - solid transition is second - order .
the density functional theory predicts the liquid - solid transition to be first - order .
in addition to being a pair potential that can be realized in a real system , it has a well known scaling property according to which the reduced excess thermodynamic properties depend on a single variable ( or coupling constant ) which for a two - dimensional system is defined as @xmath96 using this scaling the potential is written as @xmath97 where @xmath98 and r is measured in units of @xmath99 .
the pair correlation functions of a classical system can be found in any spatial dimensions as a simultaneous solution of the oz equation ( eq.@xmath102 ) and a closure relation that relates functions h , c and the potential u(r ) .
several closure relations including the percus- yevick ( py ) relation , the hypernetted chain ( hnc ) relation , modified hypernetted chain ( mhnc ) relation etc . have been used to describe structure of a uniform fluid [ 25 ] .
we may however note that while the oz equation is general and connects the total and direct pair correlation functions of liquids as well as of symmetry broken phases , the closure relations that exist in the literature have been derived assuming translational invariance [ 25 ] .
they are therefore valid only for normal fluids .
we use the integral equation theory involving suitable closure relations to find symmetry - conserving part of pair correlation functions @xmath103 and @xmath104 and their derivatives with respect to density , the symmetry - broken part of the dpcf is calculated using a method described in ref.[11 ] .
+ the oz equation for a uniform system of density @xmath105 reduces to @xmath106 the hnc closure relation and a closure relation proposed by roger and young [ 26 ] by mixing the py and the hnc relations in such a way that at @xmath107 it reduces to the py and for @xmath108 it reduces to the hnc relation , can be written together as @xmath109 -1 \nonumber , \hspace*{3.cm } ( 3.2)\end{aligned}\ ] ] where @xmath110 and @xmath111 is a mixing function with an adjustable parameter @xmath112 . for @xmath113 or , f(r)= @xmath57 , eq .
@xmath114 reduces to the hnc closure relation . in the roger- young relation
, @xmath115 is chosen to guarantee thermodynamic consistency between the virial and compressibility routes to the equation of state .
+ the differentiation of eqs . @xmath116 and @xmath114 with respect to density @xmath105 yields following two relations @xmath117 and @xmath118 \frac{\partial \chi ( r ) } { \partial \rho } { f^{-1}(r ) } \nonumber \hspace*{4.cm } ( 3.4)\end{aligned}\ ] ] the closed set of coupled equations
@xmath119 have been solved for four unknowns @xmath120 , @xmath70 , @xmath121 and @xmath122 .
the method can be extended to include higher order derivatives . in fig .
1 we plot the fourier transform of @xmath123 defied as @xmath124 for @xmath125= @xmath126 , @xmath127 and @xmath128 . the values given in fig .
@xmath129 ( a ) for n=3 are in good agreement with values found by van teeffelen et.al [ 20,21 ] ( see fig.1 of their paper ) . as has been reported in ref .
[ 21 ] the hnc closure underestimates values of @xmath130 whereas the roger - young ( ry ) closure gives relatively better but not very accurate values . in fig .
1 we also give values found from an approach proposed by kang and ree ( kr ) [ 27 ] .
+ the exact closure relation which one finds from the liquid state theory [ 25 ] can be written as @xmath131 \nonumber , \hspace*{3.cm } ( 3.6 ) \end{aligned}\ ] ] where @xmath132 is the bridge function . in the hnc closure relation @xmath132
is taken equal to zero . in the kr approach the bridge function calculated for a reference potential and denoted as @xmath133
is used for @xmath132 in eq . @xmath134 .
the evaluation of @xmath133 is done prior to and separated from the main integral equation by solving the martynov - sarkisov [ 28 ] integral equation .
we briefly summarise here the way this is done for soft repulsive potentials in two - dimensions .
+ the potential @xmath135 is first divided into a reference @xmath136 and a perturbation part @xmath137 .
@xmath138 where @xmath139 @xmath140 here @xmath141 and @xmath142 is nearest neighbour distance for hexagonal lattice at given density @xmath105 .
the @xmath143 for the reference potential is evaluated using the oz equation @xmath144 and closure relation @xmath145 \nonumber .
\hspace*{3.cm } ( 3.11)\end{aligned}\ ] ] for @xmath146 the mortynov- sarkisov [ 28 ] relation @xmath147^{{1}/{s } } -1 -\chi_{0 } ( r ) \nonumber , \hspace*{4.cm } ( 3.12)\end{aligned}\ ] ] with @xmath148=2 is used .
the values of @xmath146 are found by solving eqs.(3.10)-(3.12 ) self - consistently .
the value of @xmath143 as a function of r for n= 3 is plotted in fig .
the nature of @xmath143 is same as was found in case of three - dimensions [ 27 ] .
this value of @xmath143 found for the reference potentials is used in relation ( 3.6 ) which is used to solve the oz equation self - consistently to get values of @xmath149 and @xmath150 .
the values of @xmath151 found by this method are shown in fig . 1 by full lines
. these values are close to the simulation values given by van teeffelen et.al [ 20,21 ] for n=3 . for n=6 and 12 values found from the ry closure and values found from the kr method
are close showing that for short - range repulsive potentials the ry closure yields good values of pair correlation functions . for a crystal @xmath153
is invariant only under discrete set of translations corresponding to lattice vectors @xmath154 .
if one chooses a center of mass variable @xmath155 and difference variable @xmath156 , the @xmath83 can be written as [ 11,12 ] @xmath157 ) \nonumber , \hspace*{4.cm } ( 3.13 ) \end{aligned}\ ] ] where g are rlv s . since @xmath83 is real and symmetric with respect to interchange of @xmath158 and @xmath159 , @xmath160 and @xmath161 .
the function @xmath162 can be expanded in terms of higher body direct correlation functions of uniform liquid [ 2 ] ; @xmath163)= \int d \br_{3 } { c_{3}}^{(0 ) } ( \br_{1},\br_{2},\br_{3};\rho_{0 } ) ( \rho(\br_{3})- \rho_{0 } ) \nonumber \hspace*{6.5 cm } \\ + { \frac{1}{2 } } \int d \br_{3 } d \br_{4 } { c_{4}}^{(0 ) } ( \br_{1},\br_{2 } , \br_{3 } , \br_{4};\rho_{0 } ) ( \rho(\br_{3})- \rho_{0 } ) ( \rho(\br_{4})- \rho_{0 } ) + \cdots \nonumber , \hspace*{1.cm } ( 3.14 ) \end{aligned}\ ] ] where @xmath164 , and @xmath165 are the n - body direct correlation functions of a uniform liquid of density @xmath166 .
these correlation functions are related to derivatives of @xmath167 with respect to density @xmath166 as follows [ 2 ] ; @xmath168 @xmath169 etc .
+ the values of derivatives of @xmath104 appearing on the left hand side of above equations can be found using the integral equation theory described above .
the usefulness of this method to find @xmath170 depends on convergence of series eq .
@xmath171 which is a series in ascending powers of order parameters , and our ability of finding values of n - body @xmath172 direct correlation functions from eq . @xmath173 .
barrat et.al [ 8 ] have shown that @xmath174 can be factored as @xmath175 and the function t(r ) can be determined from relation ( see eq .
@xmath173 ) @xmath176 this method can be extended for higher @xmath165 [ 29 ] . since @xmath177 is averaged over density @xmath105 and over order parameters @xmath59 the contributions made by successive terms of eq .
@xmath178 in @xmath52 $ ] is expected to decrease rapidly [ 11 ] . in the case of three dimensions
it was found that it is only first term of the series @xmath179 which needs to be considered to describe accurately the fluid -solid transition [ 11 , 29 ] . in two - dimensions
the convergence is expected to be faster as number of nearest neighbours is less compared to the three - dimensions and therefore the higher body correlation functions expected to be less important . in view of this , we consider here the first term of the series @xmath178 only and examine its contribution in stabilizing the hexagonal lattice at the transition point . from known values of @xmath180 we solved numerically eq .
@xmath181 to find values of t(r ) for different values of @xmath182 . in fig .
3 , we plot values of t(r ) for n = @xmath183 and @xmath184 at values of @xmath182 close to the freezing point .
+ taking only the first term of eq . @xmath179 the expression for @xmath162 in terms of t(r ) can be written as @xmath185 using the relation @xmath186 eq .
@xmath187 reduces to eq .
@xmath188 , i.e. @xmath189 where @xmath190 using the relation @xmath191 where @xmath192 is the bessel function of the first kind of integral order m we find following expression for @xmath193 ; @xmath194 where @xmath195 and @xmath196 for hexagonal lattice @xmath197 .
the value of @xmath198 depends on values of order parameters @xmath199 and on the values of rlv s . in figs . 4 - 6
, we plot harmonic coefficients @xmath200 and @xmath201 for @xmath202 and @xmath184 for rlv s of first four sets , respectively . for different set of rlv s
@xmath198 varies with r in different way ; the values in all cases become negligible for r ( measured in units of @xmath203)@xmath204 . for any given value of g ,
the values of @xmath200 is about an order of magnitude larger than @xmath205 at their maxima and minima .
as magnitude of g increases value of @xmath206 decreases and after the sixth set of rlv s values of @xmath206 become negligible for all values of n.
substituting expressions of @xmath207 given by eqs . @xmath89 and @xmath208 and of @xmath209 given by eq .
@xmath210 in eq .
@xmath211 we find .
@xmath212 where @xmath213 + 1 \nonumber , \hspace*{1.cm } ( 4.2 ) \\
\frac{\delta w_{0}}{n } = - \frac{1}{2 } \delta { \gamma } c^{(0)}(0 ) - \frac{1}{2 } \sum_{g \neq 0 } { |\mu_{g}|}^{2 } \widehat{c}^{(0 ) } ( g ) \nonumber , \hspace*{1.5 cm } ( 4.3)\\ \frac{\delta w_{b}}{n } = - \frac{1}{2 } \sum_{g } { \sum_{g_{1}}}^ { ' } \mu_{g_{1 } } \mu_{-g - g_{1 } } \hat{\overline{c } } ( { \bf{g}_{1 } + \frac{1}{2 } \bf{g } } ) \nonumber .
\hspace*{1.5 cm } ( 4.4)\end{aligned}\ ] ] where @xmath214 ; the subscripts s and l represent solid and liquid respectively .
hear @xmath215 and @xmath216 are respectively , the ideal , symmetry - conserving and symmetry - broken contributions to @xmath74 .
the prime on a summation in eq @xmath217 indicates the condition @xmath218 and @xmath219 , and @xmath220 we used above expressions to locate the liquid - crystal ( hexagonal lattice ) transition by varying values of @xmath221 and @xmath85 .
the results given in table 1 for n= @xmath183 and @xmath184 correspond to the ry closure relation .
we note that the contribution arising due to symmetry broken part of the dpcf is far from negligible and its importance increases with the softness of potential .
while it is about 7.3 @xmath222 to the symmetry conserving term for n=12 , it increases about 44 @xmath222 for n=3 .
this explains why the ramakrishnan - yussouff theory gives good results for hard core potentials but fails for potentials that have soft core and/or attractive tail .
as the contribution of @xmath223 is negative , it stabilizes the solid phase . without
it the theory strongly overestimates the stability of fluid phase specially for softer potentials [ 20,21 ] .
the contribution made by the symmetry broken part of the dpcf is , as expected , small compered to that in three - dimensions ( 3d ) at the freezing point for the same potential .
for example , the contribution in 3d [ 11 ] for n=12 is 22.2 @xmath224 compared to 7.3 @xmath224 in 2d whereas for n=6 the contribution is 37 @xmath224 in 3d and 18 @xmath224 in 2d . in table 2
we compare results of the present calculation using both the ry closure and the kr procedure to calculate pair correlation functions for n=3 with the results found from other free - energy functionals as reported in ref @xmath225 $ ] .
the experimental results obtained from real - space microscopy measurements of magnetic colloids confined to an air - water interface [ 13 ] and values found from numerical simulations [ 30,31 ] are also given in the table .
while the ry closure gives slightly higher values of @xmath226 and @xmath227 compared to experimental values , the kr closure gives slightly lower values .
but these values along with the values of other parameters , particularly the values of @xmath228 , are in better agreement with the experimental values compared to any other versions of the dft .
although the extended modified weighted - density approximation ( ema ) [ 32 ] with verlet closure [ 33 ] gives values of @xmath226 which is close to the one found by us using the kr closure , but the values of @xmath229 is significantly lower ; @xmath230 compared to the values found by us @xmath231 , and the experimental value , @xmath232 .
the real - space experimental data are not available for other systems .
the computer simulation results [ 34 , 35 ] show the liquid - solid transition at @xmath233 and @xmath234 respectively for @xmath235 and @xmath184 .
these values are close to the one given in table 1 .
we used a free energy functional that contains both the symmetry - conserving part of the dpcf @xmath104 and the symmetry - broken part @xmath236 to investigate the freezing of a two - dimensional fluid into a two - dimensional crystal of hexagonal lattice .
the values of @xmath104 and its derivatives with respect to density @xmath105 as a function of interparticle separation r have been determined using an integral equation theory comprising the oz equation and the closure relations of roger and young [ 26 ] and of kang and ree [ 27 ] . for soft potential ( n=3 ) the two results
are found to differ ; the kr closure seems to give better result . for more repulsive potentials the two results are close as shown in fig .
1 . for @xmath236 which is functional of @xmath29 and is invariant only under discrete set of translations and rotations , we used an expansion in ascending powers of order parameters .
this expansion involves higher body direct correlation functions of isotropic phase , which in turn were found from the density derivatives of @xmath104 using a method proposed by barrat et.al [ 8 ] .
the contribution of symmetry - broken part of dpcf to the free energy is found to depend on nature of pair potentials ; the contribution increases with softness of potentials .
this result is in agreement with that found in three - dimensions and explains why the ramakrishnan - yussouff free - energy functional was found to give a reasonably good description of the freezing transition of hard core potentials but failed for potentials that have soft core and/or attractive tail .
the results found here and the results reported for 3d indicate that the theory described here can be used to investigate the freezing transitions of all kinds of fluids .
since our free energy functional takes into account the spontaneous symmetry breaking it can be used to study various phenomena of ordered phases .
the results indicate that the density - functional approach provides an effective frame work for theoretical study of a large variety of problems involving inhomogeneities . however , the question not adequately addressed yet is the size of fluctuations effect which play important role in two - dimensional systems .
the other important question is the inclusion of hexatic phase in the theory .
99 p. m. chaikin and t. c. lubensky , * principles of condensed matter physics * ( cambridge university press , 1995 ) y. singh , phys .
, 351 ( 1991 ) h. lowen , phys . rep .
, 249 ( 1994 ) t. v. ramakrishnan and m. yussouff , phys .
b * 19 * , 2775 ( 1979 ) a. d. j. haymet and d. w. oxtoby , j. chem .
phys . * 74 * , 2559 ( 1981 ) a. r. denton and n. w. ashcroft , phys .
a * 39 * , 4701 ( 1989 ) a. khein and n. w. ashcroft , phys .
* 78 * , 3346 ( 1997 ) j. l. barrat , j. p. hansen and g. pastore , mol . phys . * 63 * , 747 ( 1988 ) w. a. curtin , j. chem . phys . * 88 * , 7050 ( 1988 ) p. mishra and y. singh , phys .
* 97 * , 177801 ( 2006 ) ; + p. mishra , s. l. singh , j. ram and y. singh , j. chem
. phys . * 127 * , 044905 ( 2007 ) s. l. singh and y. singh , europhys .
lett . * 88 * , 16005 ( 2009 ) s. l. singh , a. s. bharadwaj and y. singh , phys .
e * 83 * , 051506 ( 2011 ) k. zahn , r. lenke and g. maret , phys .
lett . * 82 * , 2721 ( 1999 ) ; + h. h. von grunberg , p. keim , k. zahn and g. maret , phys .
lett . * 93 * , 255703 ( 2004 ) s. z. lin , b. zheng and s. trimper , phys . rev .
e * 73 * , 066106 ( 2006 ) y. han , n. y. ha , a. m. alsayed and a. g. yodh , phys .
e * 77 * , 041406 ( 2008 ) t. v. ramakrishnan , phys .
rev . lett . * 48 * , 541 ( 1982 ) x. c. zeng and d. w. oxtoby , j. chem . phys . * 93 * , 2692 ( 1990 ) j. c. barrat , h. xu , j. p. hansen and m. baus , j. phys . c.
* 21 * , 3165 ( 1988 ) v. n. ryzhov and e. e. tareyeva , phys .
b * 51 * , 8789 ( 1995 ) s. van teeffelen , c. n. likos , n. hoffmann and h. lowen europhys . lett .
* 75 * , 583 ( 2006 ) s. van teeffelen , h. lowen and c. n. likos , j. phys
. condens .
matter * 20 * , 404217 ( 2008 ) e. p. bernard and w. krauth , phys .
lett . * 107 * , 155704 ( 2011 ) . b. i. halperin and d. r. nelson , phys .
lett . * 41*,121 ( 1978 ) ; + d. r. nelson and b. i. halperin , phys .
b * 19 * , 2457 ( 1979 ) a. p.
young , phys . rev .
b * 19 * , 1855 ( 1979 ) .
j. p. hansen and i. r. mcdonald , * theory of simple liquids , 3rd ed * ( academic press , boston , 2006 ) .
f. j. rogers and d. a. young , phys .
a * 30 * , 999 ( 1984 ) .
h. s. kang and f. h. ree , j. chem . phys . * 103 * , 3629 ( 1995 ) g. a. martynov and g. sarkisov , mol . phys .
* 49 * , 1495 ( 1983 ) a. s. bharadwaj and y. singh , unpublised h lowen , phys .
e * 53 * , r29 ( 1996 ) r. haghgooie and p. s. doyle , phys .
e * 72 * , 011405 ( 2005 ) c. n. likos and n. w. ashcroft , phys .
rev . lett . * 69 * , 316 ( 1992 ) ; j. chem .
99 , 9090 ( 1993 ) l. verlet , phys .
, 201 ( 1968 ) j. q. broughton , g. h. gilmer and j. d. weeks , phys .
b * 25 * , 4651 ( 1982 ) m. p. allen , d. frenkel , w. gignac and j. p. mctague , j. chem . phys . * 78 * , 4206 ( 1983 ) table 1 .
freezing parameters @xmath237 and the pressure p at coexistence along with the contributions of ideal , symmetry - conserving and symmetry - broken parts of @xmath238 .
these results correspond to the roger - young closure [ 26 ] . of the direct pair correlation function of @xmath239
plotted against @xmath240 shown are data found from the integral equation theory using the ry closure ( dashed line ) , hnc closure ( dotted line ) and kr closure ( full line ) at values of @xmath182 shown in figure ( a ) , ( b ) and ( c ) for n=3 , 6 and 12 , respectively.,width=384,height=672 ] for rlv s of first four sets for n=3 , @xmath246 .
notations are as follow : full line represents values of the first set , the dotted line of the second set , dashed line of the third set and dashed - dotted line of the fourth set .
the distance r is expressed in unit @xmath247 , where @xmath248.,width=384,height=345 ] | a free - energy functional for a crystal proposed by singh and singh ( europhys .
lett . * 88 * , 16005 ( 2009 ) ) and which contains both the symmetry conserved and symmetry broken parts of the direct pair correlation function has been used to investigate the crystallization of a two - dimensional fluid .
the results found for fluids interacting via the inverse power potential @xmath0 for n= 3 , 6 and 12 are in good agreement with experimental and simulation results . the contribution made by the symmetry broken part to the grand thermodynamic potential at the freezing point
is found to increase with the softness of the potential .
our results explain why the ramakrishnan - yussouff ( phys . rev .
b * 19 * , 2775 ( 1979 ) ) free - energy functional gave good account of freezing transitions of hard - core potentials but failed for potentials that have soft core and/or attractive tail . |
the large scale structure of large parts of a brown dwarf atmosphere is influenced by the life cycle of dust ( @xcite ) which eventually governs their spectral appearance .
this life cycle of dust is a consequence of an interplay of convection , dust formation ( nucleation and growth ) , evaporation and gravitational settling ( @xcite ) thereby providing a likely cause for the observed non - periodic atmospheric variability ( e.g. @xcite ; + @xcite ) .
dust is the most active chemical component in the stellar atmosphere fluid because it is _
i ) _ the main opacity carrier in substellar atmospheres once it has formed , _ ii ) _ the most active chemical component which consumes condensable elements from the gas phase , and hence , _
iii ) _ it changes the atmospheric @xmath0 structure
. every convectively unstable gas will be turbulent providing a source of additional small scale variability . because a complete three - dimensional and time - dependent solution of the radiative transfer and the fluid dynamics including a complex chemistry is simply not possible yet
, the turbulent closure problem must be tackled in connection with dust formation and element conservation in substellar objects .
a numeric approach to such challenges is the large eddy simulation ( les ) method in which the range of resolvable scales is numerically treated and the unresolved , or residual scales are represented by a subgrid model .
anticipating for instance the ir spritzer observations with a spectral range up to @xmath1 m ( see @xcite ) which luckily is typical for dust , we followed two paths in our investigations : 1 .
the theoretical background of gravitational settling ( rain out ) is studied on the base of extended dust moment equations which allow for a simultaneous modelling of dust nucleation , growth / evaporation and size dependent drift ( sect .
[ sec : lsr ] ) .
we have investigated the small scale regime in 1d and in 2d to identify the governing processes , and to understand the interaction and the feedback mechanisms of turbulence and dust formation ( sect .
[ sec : ssr ] ) .
@xcite have extended the theory of dust formation developed by @xcite into the substellar regime .
the dust formation can be treated as a two step process because of time - scale arguments : first seed particles form ( nucleation ) which subsequently grow to macroscopic sizes .
@xcite developed a system of dust moment equations which have the character of conservation equations with source terms due to nucleation and growth / evaporation . by taking into account additionally the size dependent drift
particularly important in substellar atmospheres
, simultaneous with nucleation and growth / evaporation , a time - dependent system of extended dust moment equation results which has to be solved consistently with the element conservation equations .
a handsome set of static ( @xmath2 ) , stationary ( @xmath3 ) dust moment equations including auxiliary conditions for element conservation has been derived in @xcite .
its solution for tio@xmath4-grains on top of a prescribed substellar model structure @xmath5 has been demonstrated ( for model details please consult @xcite ) .
figure [ fig1 ] ( top ) shows the calculated stratification which results from the interplay between convective up - mixing , dust formation , and drift .
the up - mixing is modelled as a convective overshoot with exponentially decreasing efficiency with hight above the convective active zone . the large - scale quasi - static cloud distribution in a brown dwarf atmosphere results from two counteracting processes : the upward mixing of condensable elements by convective overshoot and the downward rain out . the distribution of dust inside and close to the cloud is merely determined by the sensitive dependence of the dust formation processes ( nucleation , growth / evaporation ) on the local temperature ( @xmath6 ) , the local gas density ( @xmath7 ) , the local element abundances ( @xmath8 ) , and the local mixing time scale .
the bottom sketch in fig .
[ fig1 ] visualises the close resemblance of terrestrial expectations of clouds and rain .
the study of the dimensionless dust moment equations for typical substellar atmosphere conditions reveals a hierarchical appearance of the source terms nucleation @xmath9 growth / evaporation @xmath9 drift ( @xcite ) .
nucleation ( seed particle formation ) is the dominant source term ( region i in fig .
[ fig1 ] ) if the local thermodynamic conditions allow for it .
only if nucleation becomes inefficient or impossible , the growth / evaporation term dominates ( regions ii and iv in fig .
[ fig1 ] ) .
only if nucleation and growth are both inefficient , the drift term will govern the equations solution ( region iii and partly also regions ii , iv in fig .
[ fig1 ] ) .
[ [ section ] ] high above the convection zone , the mixing time - scale is large and the elemental replenishment of the gas is too slow to allow for considerable amounts of dust to be present in the atmosphere .
the few particles forming here are very small which is , however , sufficient to make @xmath10 , which causes these atmospheric layers to become dust - poor . the gas phase is strongly depleted in condensible elements .
[ [ section-1 ] ] nucleation takes placemainly in the upper parts of the cloud layer , where the temperatures are sufficiently low and the elemental replenishment by mixing is sufficiently effective .
although the gas is strongly depleted in heavy elements in these layers , it is nevertheless highly supersaturated ( @xmath11 ) such that homogeneous nucleation can take place efficiently .
the dust grains remain small @xmath12 m . the mean drift velocities are even smaller than in region 0 because of the higher gas densities .
[ [ section-2 ] ] with the inward increasing temperature , the supersaturation ratio @xmath13 decreases exponentially which leads to a drastic decrease of the nucleation rate @xmath14 .
consequently , nucleation becomes unimportant at some point , . here , the condensible elements mixed up by convective overshoot are mainly consumed by the growth of already existing particles , which have formed in region
i and have drifted into region ii .
the gas is still strongly supersaturated @xmath15 , indicating that the growth process remains incomplete , the condensible elements provided by the mixing are not exhaustively consumed by growth .
[ [ section-3 ] ] with the increasing mean size of the dust particles , the total surface decreases which makes the consumption of condensible elements from the gas phase less effective .
at the same time the drift velocities increase .
when the dust particles have reached a certain critical size , @xmath16 m to @xmath17 m , the drift becomes more important than the growth , and the qualitative behaviour of the dust component changes ( _ rain edge _ ) .
although the gas is still supersaturated @xmath18 , the in - situ formation of dust grains is as ineffective as in region ii .
the depletion of the gas phase vanishes in region iii , the gas abundances approach about the solar values .
the grains reach their maximum radii at the lower boundary of region iii ( _ cloud base _ ) : @xmath19 m to @xmath20 m .
[ [ section-4 ] ] the gravitationally settling dust particles finally cross the cloud base and sink into the under - saturated gas situated below , where @xmath21 and @xmath22 . here , the dust grains raining in from above evaporate thermally . the evaporation of these dust particles , however , does not take place instantaneously , but produces a spatially extended evaporation region iv with a thickness of about 1 km for the model atmosphere shown here . beside the general cloud structure ( figs .
[ fig1 ] ) and details about the atmospheric dust component like rate of nucleation @xmath23 ( solid line , 2nd panel fig .
[ fig1 ] , top ) , mean particle sizes and mean drift velocities ( both 4th panel fig .
[ fig1 ] , top ) , we find also detailed information about the molecule abundance in the gas phase ( 5th panel fig .
[ fig1 ] , top ) .
the remaining gas composition depends on the dust formation processes and follows the atmosphere structuring by the dust as suggested in sect .
[ ss : dust ] .
[ [ section-5 ] ] the gas phase isstrongly depleted by those elements involved into the dust formation process ( merely ti in the case depicted in fig
. [ fig1 ] ) .
this metal deficiency amounts for instance to about 6 orders of magnitude in the case of tio@xmath4 in the visible outer region 0
. this metal deficiency decreases with decreasing height and its observational relevance will depend on where the @xmath24level is reached .
the model atmosphere adopted here would appear strongly metal deficient .
[ [ section-6 ] ] the gas phase has approximately solar abundances since it is strongly mixed with gas from the deep interior and the depletion by the growth of the few large particles drifting through is has only little influence .
[ [ section-7 ] ] the size dependent evaporation of the inward drifting dust grains causes a height dependent element enrichment of the inner , hot and dense atmospheric regions .
it might , however , lay below the @xmath24level and may therefore be only visible for rather warm substellar objects .
if the inertia of the atmospheric gas is larger than its friction , large reynolds number , turbulence will develop if a driving mechanism is present .
substellar atmospheres seem to provide both : high reynolds numbers and convection as driving mechanism .
consequently , simulations of brown dwarf and planetary atmospheres have to deal with the multi - scale problem as it is characteristic for turbulent systems . in classic static model
atmospheres the mixing length theory is used to describe the multi - scale problem of turbulent convection . only one the most energy containing scale of the system @xmath25 is used to model the whole scale spectrum ( fig .
[ fig2 ] ) . in simulations of dynamic model atmospheres more advanced methods have been used , like the artificial viscosity ( fig .
[ fig2 ] ) which damps fluctuations such that no energy is contained in the smallest but still resolvable scale ( @xmath26 ) .
the large eddy simulation tries to use information from the edge of the resolvable regime to model the residual , unresolved scales ( fig .
[ fig2 ] ) .
obviously , knowledge needs to be accumulated about this small - scale regime . a hierarchical investigation from small to large scale
has been started by @xcite as sketched by in fig .
the basic idea is that convective elements do collide and thereby create a whole spectrum of disturbances , turbulent fluctuations .
+ fluctuations carry expansion waves which can initiate dust formation by a local drop of the temperature below the nucleation threshold .
if thermal stability is assured , these seed particles grow to macroscopic sizes and start to radiate thermally .
the result is a reinforced cooling of the gas / dust mixture in the small scale regime .
a feedback loop sets in which assures more dust to form until all condensible material is consumed from the gas phase .
the dust complex has then reached a steady state ( compare fig . 3 in @xcite ) .
relying on these results , @xcite have adopted the concept of driven turbulence where a whole spectrum of superimposed waves constantly enters and thereby disturbs an initially dust free gas inside a test volume .
figure [ fig3 ] demonstrates the principle findings : the dust formation process is initiated on the smallest length scales resulting in a very inhomogeneous distribution of dust .
similar results were obtained based on asymptotic analysis of the nucleation rate in fig . 2 in @xcite . the smallest observable dust structure in the present fig .
[ fig3 ] ( top ) is @xmath27 m after @xmath28s with the smallest resolved scale being @xmath29 m . these small scale heterogeneity lasts not very long .
ongoing hydrodynamic disturbances cause the dust structures to gather in larger and larger cloud - like structures causing a ( mesoscopic ) large - scale heterogeneity . lanes and curls appear and disappear during these dynamic process of gathering .
structures of the size of @xmath30 m have developed in the presented simulation after @xmath31s ( fig .
[ fig3 ] , bottom ) .
note that the appearance of curls indicates the development of vorticity which is typical for turbulent systems . during the time of active dust formation
, sites with high vorticity appear and the question may be raised if the formation of dust can cause turbulence .
the answer seems to be positive since the appearance of dust in a system has a strong feedback onto the system s temperature and the velocity field , the thermal and the dynamic structure of the atmosphere .
this idea is supported by the results of 1d simulations of which one example is shown in fig .
[ fig4 ] where means with standard deviations are depicted .
the apparent standard deviation is largest in the period of most active dust formation , between 0.05s and 3s , independent of the initial reference temperature .
almost no variations are present in the dust quantities if the dust complex has reached its steady state which is much in contrast with the hydrodynamic quantities ( top four panels in fig .
[ fig4 ] ) which are continuously perturbed by the turbulence driving mechanism .
[ [ optical - depth ] ] optical depth : + + + + + + + + + + + + + + adopting rosseland mean gas opacity typical for the hot , inner layers of a brown dwarf atmosphere and the rosseland mean dust opacity for astronomical silicates , the variability of the total opacity and the space mean optical depth for the size of our test volume ( 500 m ) has been investigated . for the simulation depicted in fig .
[ fig4 ] the dust and the gas opacities differ by about 1.5 orders of magnitude .
the dust opacities vary by about 0.5 magnitudes , whereas the gas opacity by only about 0.2 order of magnitudes .
this is independent of the characteristic time scale of the system , the large - scale mach number of the initial configuration .
the optical depth increases by a factor of 10 when the dust forms , in accordance with the opacity increase .
there is , however , a delay between the onset of dust formation and the time when the maximum optical depth is reached because the dust formation sets in earlier than a cloud becomes optically thick .
this delay is sensitive to the characteristic time scale of the system .
the gas / dust mixture will only be optically thick for the case of non - transparent dust particles depending also on the wavelength considered .
glassy grains efficiently absorb in the far ir ( @xmath32 m ) and one might , depending on the observed wavelength , detect typical dust features and possibly even the rapid formation events which , however , may need to evolve on macroscopic scales to be observable .
turbulent fluctuations do not seem to produce considerable variations in view of the space mean of the optical depth in the long term ( here @xmath33s ) if astronomical silicates are assumed as typical dust opacity carriers , though turbulent fluctuations are essential for the initiation of dust formation and eventually for the origin of large scale structures .
[ [ dependence - on - mach - number ] ] dependence on mach number : + + + + + + + + + + + + + + + + + + + + + + + + + + the use of dimensionless equations allows to describe the character of the model equations and their solution by a set of characteristic numbers .
for instance , the mach number = 1 suggest a sound - like flow while mach number @xmath34 implies a quasi - static fluid field .
this concerns only the mean behaviour of the reactive fluid
. deviations from this can occur and a local rather than a global mach number may occasionally be considered .
the mach number of the local turbulent fluctuations is always sub - sonic due to the underlying model of driven turbulence ( see @xcite ) in our simulations , m@xmath35 .
the global mach number describes the turbulent energy input on the largest resolvable scale and has an indirect influence on m@xmath36 .
however , the dust formation is a local process , and its onset is delayed in case of low global mach number .
the overall pattern formation remains very much alike but needs longer times to evolve for smaller mach numbers .
presently , two schools exist for modelling the large - scale structures of substellar atmospheres
. both aim on the understanding of the physiochemical processes being responsible for the detected spectral appearance and the photometric variability .
\a ) the classic static model atmosphere includes beside the assumption of hydrostatic equilibrium
an elaborate frequency - dependent ( static ) radiative transfer treatment assuring flux conservation also with regard to the convective flux treated in the mixing length approximation .
the solution of the chemical equilibrium for hundreds of gaseous - and if needed also for solid - species closes the system of model equations .
\b ) multi - dimensional modelling originating from solar and engineering research .
the complete time - dependent hydrodynamic equations are solved including energy conservation .
the time - dependence and multi - dimensionality is payed by either a simplified radiative transfer treatment ( @xcite ) .
in addition , the treatment of the gas phase chemistry is very limited .
only @xcite did include a time - dependent model for dust formation ( see sect . [
sec : lsr ] ) .
it seems tempting to combine the advantage of the one ( e.g. frequency - dependent radiative transfer for spectrum interpretation ) with the advantage of the other ( e.g. hydrodynamic cloud formation ) . what do we need to achieve such a goal ? [ [ section-8 ] ] a complete hydrodynamic modelling of a substellar atmosphere will need to solve the multi - scale problem inherent to any turbulent media .
hence , we need a model able to describe the large scale hydrodynamic structures commonly ascribed to convection
. the model must simultaneously describe the small - scale turbulent fluid field .
one way to do so is to apply the method of the large eddy simulation ( les ) where the spatial and the associated time scale are resolved down to a cut - off length .
the scales smaller than that must be represented by some appropriate subgrid model .
a further challenge is the simulation of the physical system close to a quasi - static situation .
numeric truncation errors may cause unwanted spurious effects which can accumulate in the course of the simulation .
[ [ section-9 ] ] brown dwarfs are shown to be fast rotators ( see scholz & eislffel this conference ) and we know from jupiter that rotation shapes its initially intermittent atmosphere .
rotation will introduce a day / night behaviour and tides on the scale of the entire globe .
it is , hence , a large - scale process and a globe - sized simulation will be needed .
so far , only rotation free simulations are available for brown dwarfs .
[ [ section-10 ] ] for spectrum interpretation and for a correct atmospheric temperature - density structure , the photon - matter interaction has to be modelled .
huge data bases with millions of lines for the various molecules of interest ( h@xmath4o , tio , tic , tih , co , sio , vo , ch@xmath37 , cah , crh , feh , h@xmath38o , nh@xmath38 , ph@xmath38 , cia h@xmath4 ) and others , which influence the thermal structure but not show up in spectral features , are available and under construction . obeying observational evidence , atomic data are needed as well as good models for line profiles ( k i , na i , ca ii ; rb i , cs i ) .
multi - dimensional and time - dependent simulations are not able to handle such amounts of data which makes the 1d models the best and only tool for detailed spectrum interpretation also in the far future . however , the radiative transfer in spatially inhomogeneous media may need further attention also in substellar objects since it can support structure formation in dusty media where e.g. shadow casting is a major effects ( see question by steinacker ) .
[ [ section-11 ] ] so far , time - dependent models were presented which include only a very simple model for the chemical gas phase composition of the dust forming gas .
our work in progress allows to calculate the chemical equilibrium among 33 molecules and the conservation in space and time of 10 elements which is sufficient for the study of heterogeneous dust formation .
it may be necessary to study chemical non - equilibrium effects influencing the composition of the gas phase .
such effects may be important if the chemical reaction time is considerably smaller than the time on which the local temperature and density changes , which require that convection is much faster than chemistry .
chemical equilibrium , in contrast , holds if the chemical reactions are much faster than the hydrodynamic changes .
since photochemical processes are irrelevant in inner substellar atmospheric regions , the above ideas are concerned with the local thermo- and hydrodynamic condition , or in other words , it can be considered in lagrangian coordinates .
hence , observation of co and ch@xmath37 for instance in gl 229b do not necessarily mean that the convective time scale is smaller than the chemical reaction time for these molecules .
it may very well be that a convective cell of hot gas with its respective chemical equilibrium composition travels upward and is detected in a considerable cooler surrounding . however , the local conditions inside this cell are appropriate for the presence of co and the other gas compounds in chemical equilibrium since they are not influence by the surrounding nor by photo - processes yet . hence , there are at least two reasons not to expect chemical non - equilibrium : 1 ) the atmospheres are almost entirely convective thereby providing a efficient mixing mechanism , and 2 ) substellar atmospheres are very dense , hence , the chemical time scales are very small due to high collision rates .
[ [ section-12 ] ] first steps for understanding dust formation and for providing a consistent description for substellar atmospheres are made ( see sect . [ sec : lsr ] ) .
so far , the extended model equation for dust formation which include the drift effect have not be solved consistently inside a model atmosphere .
note that the formation process of dust is a non - equilibrium process since it takes place far away from thermodynamic equilibrium .
the gas needs to be substantially over - saturated , only then the phase transition is kinetically possible . the next crucial step may be the consideration of coagulation processes .
coagulation is likely to play a role in the inner , drift dominated atmospheric layers . here the drift may cause grain - grain collision and thereby an non - continuous , jump - like dust growth process leading to much larger dust grains than we presently do predict .
we have presented work on two entirely different scale regimes of substellar atmospheres , one dealing with the observational large - scale regime and the other dealing with the small - scale regime where structure formations can be initiated .
this allows us to cautiously draw a picture which sketches processes and interactions most relevant to rain and cloud formation in substellar atmospheres .
figure [ fig5 ] summarises the expected large scale structure as well as the small scale structure formation due to turbulence seen as a zoom in into the large scale structures of the atmosphere .
first , the turbulent character of a substellar atmosphere does allow the formation of grains in else - wise dust - hostile regions , i.e. further inside the atmosphere .
the constantly occurring wave input from e.g. colliding convection cells causes the dust to form and later to gather into larger and larger structure ( clouds ) which can block radiation and cause backwarming due to the high opacity of dirty dust grains .
second , the dust formation does very likely occur inside elemental rich , upstreaming gaseous material .
once first seed particles have formed , they immediately start to sink into deeper , much denser atmospheric layer . during this travel
the particle grow to @xmath39 m sized grains and speed up their fall as they get larger .
the region of seed formation is characterised by very small grains and a high metal deficiency .
the dust particles each their maximum size in deep layers , just at the borderline of thermal stability .
what we know from earth rain we find also for brown dwarfs , the rain drops simply path through some atmospheric regions before they hit the ground or evaporate as it is in the case of brown dwarfs and giant gas planets .
hence , there is a distinct upper and lower boundary where the dust clouds are to be expected .
the upper borderline is related to th condition that enough condensable elements must be mixed upwards , the lower borderline is given by the condition of thermal stability .
it is clear from thermodynamic arguments , that the cloud layer sinks deeper into the atmosphere as cooler the object is and visa versa .
it is , hence , not surprising that cooler substellar objects do appear more metal deficient .
this work has been supported by the _ dfg _ ( grants se 420/19 - 1&2 , kl 611/7 - 1 , kl 611/9 - 1 ) and the _ berliner programm zur frderung der chancengleichheit fr frauen in forschung und lehre_. most of the literature search has been performed with the ads system .
* burgasser : * are the structures you show at later times an analogous `` picture '' of what you see on the surface of a dust - forming star / brown dwarfs ? have you considered rotation effects ?
great talk \1 ) in case of true self - similarity , this could be true .
however , i would hesitate to simply scale up the present results since they are obtained for a very different scale regime where major large scale processes like for instance convection have no direct influence on the local fluid dynamics , though they indirectly are present as driving mechanism for turbulence .
\2 ) the local transition time scale is very small and it merely depends on the dust formation time scale .
the global time scale will be much longer in a turbulent medium since areas can remain dust - free longer than others . to give a rough estimate , duration from @xmath41s are possible , depending also on the mach number of the fluid .
\1 ) non of the dust simulations presented here need to include such reaction energy effects as is has been discussed in ( woitke & helling 2003 a&a 399 , 297 ) .
we have shown that this effect can only amount to a maximum heating of @xmath42k of any grain which we consider as negligible under typical brown dwarf atmosphere conditions .
2)the release of condensation heat during grain grow can be important for the gas , possibly driving the convection as in the earth atmosphere .
the abundance of high - temperature condensates in brown dwarf atmospheres is smaller than that of water in the earth atmosphere .
therefore , we expect smaller effects .
* ludwig : * which is the dominant spatial scale ( relative to the scale on which turbulence is driven ) of the dust resulting structures ? in how far do smaller scales influence the dominant dust structures ? which is the minimum level of fluctuations necessary to significantly influence dust formation ?
\3 ) the smallest global mach number studied so far is @xmath45 which determines the energy input at the largest scale .
the local turbulent fluctuation does , however , fall into a regime of smaller mach numbers .
the quantitative character of structure formation is very similar to the @xmath46 regime , but the time scales are longer
. proper radiative transfer is not included so far .
present dust cooling is treated by a relaxation ansatz towards a constant temperature .
multi - dimensional radiative transfer simulations do , however , suggest the support of such inhomogeneous media than a smoothing out of these structures ( see e.g. woitke el al . 2000 a&a 358 , 665 ) .
* dobler : * in your model the stratification quickly changes from adiabatic to isothermal , which is very different from what we have in the earth s troposphere .
what is the physics behind this difference ?
first , the optical properties of the considered particles are large in brown dwarf atmospheres .
we have to expect dirty , i.e. non - transparent , dust particles in brown dwarfs which is very different to what we have in the earth s atmospheric layers .
second , the ratio radiative / thermal energy of the gas is much larger in brown dwarf atmospheres since it scales with @xmath47 and brown dwarfs are considerably warmer than the earth . | the large scale structure of a brown dwarf atmosphere is determined by an interplay of convection , radiation , dust formation , and gravitational settling , which possibly provides an explanation for the observed variability .
the result is an element depletion of the dust forming regions and an element enrichment of the dust evaporating sites .
the formation of dust cloud structures in substellar atmospheres is demonstrated based on a consistent theoretical description of dust formation and destruction , gravitational settling , and element depletion including the effect of convective overshoot .
since the viscosity is small in brown dwarf atmospheres , the convection creates a turbulent environment with fluctuations of all thermodynamic state variables on a wide range of spatial scales .
hence , the classical turbulent closure problem needs to be tackled in connection with dust formation in substellar objects , because a complete three - dimensional and time - dependent solution of the model equations is simply not possible .
structure formation may be seeded on the smallest scales , when chemical processes are involved . in order to understand the interaction of turbulence and dust formation
, we have performed investigations of the smallest scale regimes in 1d and in 2d in order to identify the governing processes of the unresolved scale regime . |
many questions about spectra and eigenfunctions of elliptic operators are motivated by bohr s correspondence principle in quantum mechanics , asserting that a classical dynamical system manifests itself in the semiclassical ( as planck s constant @xmath0 ) limit of its quantization .
when the quantum system is given by the laplacian @xmath1 on the riemannian manifold @xmath2 ( describing a quantum particle on @xmath2 in the absence of electric and magnetic fields ) the corresponding classical system is the geodesic flow @xmath3 on @xmath2 , so in the high energy limit eigenfunctions should reflect the properties of the geodesic flow .
one of the most studied question concerns limits of eigenfunctions . to an eigenfunction @xmath4 with @xmath5 one can associate a measure @xmath6 on @xmath2 with the density @xmath7 ; its phase space counterpart is a distribution @xmath8 on the unit cosphere bundle @xmath9 , projecting to @xmath6 .
it can be defined as follows : given a smooth function @xmath10 on @xmath9 ( an observable ) , we choose a `` quantization , '' a pseudodifferential operator @xmath11 of order zero with principal symbol @xmath10 , and let @xmath12 the definition of the measures @xmath8 depends of course on the choice of the quantization map @xmath13 , but any two choices differ by an operator of a lower order , and that does not affect the asymptotic behaviour of @xmath14 .
the measures @xmath8 are sometimes called _ wigner measures . _
a natural problem is to study the set of weak * limit points of @xmath15-s ; it follows from egorov s theorem that any limit measure is invariant under the geodesic flow , but the limits are quite different for manifolds with integrable and ergodic geodesic flows . if @xmath2 has completely integrable geodesic flow @xmath3 , sequences of @xmath8-s concentrate on liouville tori in phase space satisfying the quantization condition .
let @xmath16 be the the counting function of the laplacian ( as usual , we enumerate eigenvalues taking into account their multiplicities ) .
recall that , by the weyl formula , @xmath17 if @xmath3 is ergodic , the following fundamental result ( sometimes called _ quantum ergodicity _ theorem ) holds .
[ thm : qerg ] let @xmath2 be a compact manifold with ergodic geodesic flow , and let @xmath18 be a zero order pseudodifferential operator with principal symbol @xmath19 . then @xmath20 or , equivalently , @xmath21 where @xmath22 is the normalized canonical measure on @xmath9 .
the theorem shows that for a subsequence @xmath23 of eigenfunctions of the full density , @xmath24 , and after projecting to @xmath2 we find that @xmath25 ( weak@xmath26 ) . in other words ,
almost all high energy eigenfunctions become equidistributed on the manifold and in phase space .
various versions of theorem [ thm : qerg ] were proved in @xcite by shnirelman , zelditch , colin de verdire and helffer martinez robert , as well as in other papers .
quantum ergodicity has been established for billiards in @xcite by grard
leichtnam and zelditch zworski .
important further questions concern the rate of convergence in , quantum analogues of mixing and entropy .
rudnick and sarnak conjectured that on negatively curved manifolds , the conclusion of theorem [ thm : qerg ] holds _ without averaging _ , or equivalently @xmath27 for _ all _ eigenfunctions ; this is sometimes called _ quantum unique ergodicity _ ( que ) .
this conjecture has been proved for some arithmetic hyperbolic manifolds by lindenstrauss , with further progress by soundararajan and holowinsky . on the other hand
, a. hassell ( @xcite ) has shown that on the bunimovich stadium billiard , there exist _ exceptional _ sequences of eigenfunctions concentrating on the `` bouncing ball '' orbits , so the analogue of the que conjecture does not hold for all billiards .
the aim of this paper is to identify the correct semiclassical dynamics corresponding to quantum systems with discontinuities .
more precisely , we are looking at manifolds ( possibly with boundary ) whose metrics are allowed to have jump discontinuities across codimension one hypersurfaces . such manifolds model situations in physics where waves propagate in matter that consists of different layers of materials . in the simplest case
we would have two isotropic materials touching at a hypersurface .
the metric in each layer is then given by @xmath28 , where @xmath29 is the refraction index in layer @xmath30 and @xmath31 is the euclidian metric on the tangent bundle .
wave propagation in these media is described by the wave equation @xmath32 where @xmath1 is the laplace operator with respect to the metric @xmath33 and transmissive boundary conditions are imposed on the solutions .
the high energy limit of such a systems shows properties that do not remind of classical mechanics : singularities of solutions travel on geodesics until they hit the discontinuity .
then they are reflected and refracted according to the laws of geometrical optics .
consequently , there is no classical flow on phase space that describes the high energy limit .
moreover , the naive generalization of egorov s theorem fails in this situation .
if @xmath18 is a pseudodifferential operator on the manifold supported away from the discontinuity then the quantum mechanical time - evolution @xmath34 of @xmath18 will in general fail to be a pseudodifferential operator .
thus , on the algebraic level of observables the quantum - classical correspondence fails .
we will show that after forming an average over the eigenstates the quantum dynamics relates to a certain probabilistic dynamics that takes into account the different branches of geodesics emerging in this way .
our main result establishes a quantum ergodicity theorem in the case where this classical dynamics is ergodic .
our proof relies on a precise symbolic calculus for fourier integral operators associated with canonical transformations ( see section [ s : fio ] ) and on a local weyl law for such operators ( theorem [ t : local - fio ] ) .
we construct a local parametrix for the wave kernel consisting of a sum of such fourier integral operators ( section [ t : parametrix ] ) and apply to them the above results .
the usual proof of quantum ergodicity is based on the consideration of the positive operator obtained by squaring the average of the time - evolution of a pseudodifferential operator .
egorov s theorem plays an important role in this construction .
since it does not hold in our setting , the standard proof can not be directly applied .
instead we use the local weyl law for an operator that is not necessarily positive but whose expectation value with respect to any eigenfunction is positive .
we then apply the symbolic calculus to obtain an explicit formula for the leading asymptotic coefficient .
the proof of the main result is presented in full detail in section [ s : ergodicity ] .
it should be mentioned that local weyl laws of the type given in theorem [ t : local - fio ] are very powerful and have many applications . in particular , a less explicit but similar result was recently stated and used in @xcite and @xcite to prove quantum ergodic restriction theorems .
ray splitting not only occurs in the quantum systems we consider but also happens in situations described by systems of partial differential equations and higher order equations .
moreover , ray - splitting occurs in a natural way in quantum graphs .
our results carry over in a straightforward manner to these situations .
ray - splitting billiards have been studied extensively in the physics literature , see e.g. @xcite and references therein .
the emphasis has been on spectral statistics , trace formulae , eigenfunction localization ( `` scarring '' ) , and the behaviour of periodic orbits . in the mathematical literature
, the emphasis has been on the propagation of singularities @xcite and spectral asymptotics @xcite .
quantum ergodicity has not previously been considered for ray - splitting ( or branching ) billiards .
our results fill an important gap in the semiclassical theory of such systems .
it is our hope that our results will serve as a motivation to further study a probabilistic dynamical system that we introduce .
following zelditch and zworksi @xcite , we say that @xmath2 is a compact manifold with piecewise smooth lipschitz boundary if @xmath2 is a compact subset of a smooth manifold @xmath35 such that there exists a finite collection of smooth functions @xmath36 such that 1 .
@xmath37 , 2 .
@xmath2 has lipschitz boundary and @xmath38 is an embedded submanifold of @xmath35 .
3 . @xmath39 .
a riemannian metric on @xmath2 is assumed to be the restriction of a riemannian metric on @xmath35 .
we will denote the regular part of the boundary @xmath40 of @xmath2 by @xmath41 and the singular part by @xmath42 .
suppose that @xmath43 is a compact manifold with piecewise smooth lipschitz boundary and @xmath44 is a co - dimensional one piecewise smooth closed hypersurface in @xmath43 such that @xmath45 .
we will also assume that the completion of @xmath46 with respect to the inherited uniform structure is again a manifold @xmath2 with piecewise smooth lipschitz boundary .
thus , cutting @xmath43 open along @xmath47 results in @xmath2 and we obtain a part @xmath48 of the boundary @xmath40 with a two - fold covering map @xmath49 . with boundary , title="fig:",width=188,height=151 ] with boundary , title="fig:",width=188,height=151 ] in the simplest case @xmath43
is oriented , @xmath47 is a closed hypersurface that separates @xmath43 into two parts @xmath50 and @xmath51
. then @xmath2 will be the disjoint union of @xmath52 and @xmath53 , @xmath48 is the disjoint union of two copies of @xmath47 , and the deck transformation simply interchanges these two copies of @xmath47 .
a smooth metric on @xmath2 defines a metric on @xmath46 .
since we do not require that the covering map @xmath54 be an isometry , this metric can in general not be continued to a metric on @xmath43 but has a jump discontinuity at @xmath47 .
manifolds @xmath43 with a metric of this form on @xmath46 can be thought of as riemannian manifolds with metric jump discontinuities at @xmath47 .
note that the construction ensures that the metric can be continued smoothly up to @xmath47 on either side of @xmath47 in local coordinates ( although the continuations from the left and from the right need not coincide ) .
the gluing construction defines a map @xmath55 which is again a two - fold covering map that lifts the original covering map .
note that the deck transformation of this cover does not in general preserve the length of covectors .
since @xmath48 has zero measure , functions in @xmath56 can also be understood as functions in @xmath57 .
moreover , functions in @xmath58 can be understood as functions in @xmath59 which are smooth away from @xmath47 and can have a jump discontinuity along @xmath47 . for the sake of notational simplicity , in the following we will not distinguish between the spaces @xmath56 and @xmath57 and understand @xmath58 as a subspace of @xmath59 .
we will assume that @xmath60 is either the sobolev space @xmath61 or the space @xmath62 of @xmath63-functions vanishing on @xmath64 .
we define the laplace operator @xmath1 on @xmath43 as the self - adjoint operator defined by the dirichlet quadratic form @xmath65 with domain @xmath60 , where @xmath66 is the standard riemannian density , @xmath67 is the volume element and @xmath68 is the riemannian norm of the covector @xmath69 .
contraction of the density @xmath70 with the outward pointing unit normal vector field results in a density @xmath71 on @xmath48 .
there then exists a positive function @xmath72 such that @xmath73 for all @xmath74 , where @xmath75 denotes the deck transformation of the density @xmath71 on @xmath48 .
the domain of the unique self - adjoint operator generated by this quadratic form coincides with @xmath76 in the case @xmath77 , and with @xmath78 in the case @xmath79 . in both formulae @xmath80 and @xmath81 are the outward pointing unit normal vector fields at @xmath82 and @xmath83 respectively , and @xmath84 is the deck transformation of @xmath85 .
thus , in the former case the operator is subject to dirichlet boundary conditions at @xmath64 and transmissive boundary condition at @xmath48 . in the latter case
the operator is subject to neumann boundary conditions at @xmath64 and transmissive boundary conditions at @xmath48 . to be more specific and for notational purposes we would like to describe the transmissive boundary condition in local coordinates near @xmath86 .
suppose that @xmath87 is a point .
locally the normal bundle of @xmath86 is trivial and , therefore , there exists a connected open neighbourhood @xmath88 of @xmath89 such that @xmath90 is the disjoint union of two connected components @xmath91 and @xmath92 .
after making such a choice , for each point @xmath93 there are two points @xmath94 and @xmath95 in @xmath48 that correspond to @xmath96 under the covering @xmath49 .
these points @xmath97 can be thought of as the different limits in @xmath2 that sequences have when converging to @xmath96 in @xmath98 and @xmath99 respectively . for any function @xmath100 on @xmath2
we can then define its two boundary values @xmath101 and @xmath102 .
the deck transformation group will act by interchanging the points @xmath94 and @xmath95 .
similarly , different metrics @xmath103 and @xmath104 on @xmath105 are obtained by passing to the limit in @xmath91 and @xmath92 respectively .
let @xmath106 be the corresponding quadratic forms , and let @xmath107 be the @xmath108-unit conormal vectors oriented into the @xmath108-sides . in the same way let @xmath109 be the induced riemannian metric on @xmath82 .
then @xmath110 and @xmath111 correspond to the different riemannian metrics on @xmath86 that are induced by the different limits from left and right hand side respectively .
the induced metric densities are then given by @xmath112 and the function @xmath113 is computed as @xmath114 . by construction @xmath115 .
elements @xmath116 in the domain of @xmath1 then satisfy the transmissive boundary conditions @xmath117 for all @xmath118 , where @xmath119 is the exterior normal derivative of @xmath116 at the point @xmath97 .
let @xmath120 be the sobolev space .
since these boundary conditions are elliptic , we have @xmath121 for any @xmath122 .
moreover , if @xmath123 then @xmath124 for technical reasons , it is more convenient to consider operators acting in the space of half - densities on @xmath43 rather than in the space of functions .
further on , if @xmath125 is a function space , we shall denote by @xmath126 the corresponding space of half - densities .
the operator @xmath127 with domain @xmath128 is said to be the _ laplacian in the space of half - densities .
_ clearly , it is a self - adjoint operator whose spectrum coincides with the spectrum of @xmath1 .
moreover , @xmath116 is an eigenfunction of @xmath1 if and only if the half - density @xmath129 is an eigenvector of @xmath127 corresponding to the same eigenvalue .
we shall denote the laplacian in the space of half - densities by the same letter @xmath1 specifying , when necessary , in what space we consider the operator .
as above let @xmath130 be the riemannian metric on @xmath105 and @xmath131 . for local statements near @xmath47 it will be enough to consider open sets @xmath88 as in the previous section and a choice @xmath132 as above so that we can use the above notation @xmath133 for the different boundary values of the metric .
if @xmath134 , where @xmath135 and @xmath136 , let denote by @xmath137 the hamiltonian trajectory generated by the hamiltonian @xmath138 and starting at @xmath139 .
its projection @xmath140 onto @xmath43 is the geodesic emanating from the point @xmath141 in the direction @xmath142 .
clearly , @xmath140 and @xmath143 are positively homogeneous functions of @xmath142 of degree 0 and 1 respectively .
furthermore , if @xmath144 then @xmath145 and the geodesic @xmath140 is parametrised by its length . the hamiltonian trajectory @xmath146 is well defined until the geodesic @xmath140 hits the boundary @xmath40 , i.e. , either the set @xmath47 or the boundary of @xmath43 . in the former case , according to the laws of geometrical optics , it splits into two geodesics .
one of them is obtained by reflection , and the other is the refracted trajectory which goes through @xmath47 but changes its direction . in the latter case
there is only the reflected trajectory . more precisely , there are the following possibilities . assume , for the sake of definiteness , that the trajectory @xmath147 approaches @xmath47 from the @xmath148-side and meets @xmath86 at the time @xmath149 .
let @xmath150 so that @xmath151 and @xmath152 .
denote by @xmath153 the @xmath108-orthogonal projections of @xmath154 onto the cotangent space @xmath155 .
clearly , @xmath156 where @xmath157 by definition , the _ reflected _ trajectory is the hamiltonian trajectory @xmath158 originating from the point @xmath159 at the time @xmath149 and going into the @xmath148-side of @xmath43 . if @xmath160 then the geodesic @xmath140 hits @xmath86 at zero angle . in this case
the reflected trajectory is not well defined .
assume that @xmath161 and denote @xmath162 obviously , @xmath163 .
the hamiltonian trajectory @xmath164 originating from the point @xmath165 at the time @xmath149 and going into the @xmath166-side is called the _ refracted _ trajectory .
note that in this case @xmath164 is the reflection of the trajectory @xmath167 coming from the @xmath166-side to the point @xmath168 .
the corresponding refracted trajectory coincides with @xmath158 , so that @xmath169 and @xmath167 have the same pair of reflected and refracted trajectories . if @xmath170 then there is no refraction . in this case
one says that @xmath171 is a point of _ total reflection .
_ if @xmath172 then the angle of refraction is zero and the refracted trajectory may not be well defined .
if the geodesic hits a point @xmath173 then @xmath146 is reflected in the same way as above .
namely , the reflected trajectory is the hamiltonian trajectory originating from @xmath174 where @xmath175 and @xmath176 are tangential and normal components of @xmath154 . the trajectory obtained by consecutive reflections and/or refractions
is called a billiard trajectory . in general
, there are infinitely many billiard trajectories originating from a given point @xmath134 ; moreover , the set of these trajectories is typically uncountable .
we shall denote them by @xmath177 , where @xmath178 is an index specifying the type of trajectory ( see below ) .
each billiard trajectory @xmath177 consists of a collection of geodesic segments @xmath179 , which are joined at @xmath47 and @xmath64 .
following @xcite , we shall suppose that @xmath178 is a ternary fraction @xmath180 where @xmath181 or @xmath182 for all @xmath183 , so that @xmath178 is a point of the cantor set in @xmath184 $ ] .
more precisely , we say that the trajectory has type @xmath178 if the following is true . *
if the @xmath185st segment of the trajectory is obtained by reflection and there exists the corresponding refracted ray then @xmath181 . * if the @xmath185st segment of the trajectory is obtained by refraction then @xmath186 , * if the trajectory has only @xmath183 segments or the @xmath183th segment ends at @xmath64 or at a point of total reflection then either @xmath181 or @xmath186 .
the last condition implies that a billiard trajectory may have different types @xmath178 . roughly speaking
, the equality @xmath186 means that @xmath185st segment is obtained by refraction whenever it is possible .
a billiard trajectory is not well - defined if * the trajectory hits @xmath40 infinitely many times in a finite time , * or the angle of incidence or the angle of refraction is equal to zero , * or the trajectory hits a point in @xmath42 .
trajectories of the first type are called _ dead - end _ , trajectories of the second type
are said to be _ grazing _ , and we call trajectories of the third kind _ singular_. let @xmath187 , @xmath188 , and @xmath189 be the sets of starting points of the dead - end , grazing and singular trajectories , respectively
. clearly , @xmath187 , @xmath188 , and @xmath189 are conic subsets of @xmath190 . throughout the paper
we shall suppose that the following assumption holds .
[ a : bad - points ] the conic set @xmath191 has measure zero in the cotangent bundle .
this assumption means the set of starting points of `` bad '' trajectories is of measure zero or , in other words , that the billiard trajectories @xmath192 are well - defined for all @xmath178 , all @xmath193 and almost all @xmath194 .
one can easily show that @xmath195 is a set of measure zero ( see , for instance @xcite , @xcite or @xcite ) . however , there are reasons to believe that @xmath196 may have a positive measure . in @xcite
the authors constructed such an example for a similar branching billiard .
let @xmath197 be the set of points @xmath198 such that all the billiard trajectories @xmath192 are well defined for @xmath199 $ ] and @xmath200 .
clearly , @xmath201 is an open conic subset of @xmath190 .
assumption [ a : bad - points ] implies that their intersection @xmath202 is a set of full measure in @xmath203 .
note that the mapping @xmath204 defined on @xmath201 is a homogeneous canonical transformation in the sense in symplectic geometry for each fixed @xmath205 $ ] and @xmath178 .
it preserves the canonical symplectic 1-form @xmath206 on @xmath203 and the standard measures on @xmath203 and @xmath9 ( see , for instance , @xcite or @xcite ) .
further on we shall abbreviate the words ` fourier integral operator ' and ` pseudodifferential operator ' to fio and @xmath207 , respectively .
we shall always be assuming that their symbols and amplitudes belong to hrmander s classes @xmath208 ( see , for instance , ( * ? ? ?
* chapter 18 ) ) . recall that the conic support @xmath209 of a function @xmath210 is defined as the closure of the union @xmath211 , where @xmath212 are the positively homogeneous functions appearing in the asymptotic expansion @xmath213 .
let @xmath214 be a smooth homogeneous canonical transformation in @xmath215 defined on an open conic set @xmath216 , and let @xmath217 be an operator with schwartz kernel @xmath218 ( that is , @xmath219 ) .
the operator @xmath220 is said to be a fio of order @xmath183 associated with @xmath221 if @xmath218 can be written as an oscillatory integral of the form @xmath222 modulo a half - density from @xmath223 .
here @xmath224 , @xmath225 and @xmath226 are smooth functions on @xmath227 satisfying the following conditions . 1 .
@xmath224 is an arbitrary cut - off function , which is positively homogeneous of degree 0 for large @xmath142 , is identically equal to 1 in a small neighbourhood of the set @xmath228 and vanishes outside another small neighbourhood of the set @xmath228 .
@xmath226 is an amplitude from a class @xmath208 with @xmath229 ; 3 .
@xmath225 is positively homogeneous in @xmath142 of degree 1 with @xmath230 .
4 . @xmath231 as @xmath232 . 5 .
@xmath233 for all @xmath234 and @xmath235 is the only solution of the equation @xmath236 on @xmath237 .
note that * @xmath238 because @xmath221 preserves the 1-form @xmath239 .
therefore the condition * ( a@xmath240 ) * is fulfilled whenever @xmath241 and @xmath237 is small enough . *
the right hand side of behaves as a half - density with respect to @xmath96 and @xmath141 and , consequently , the corresponding operator acts in the space of half - densities ; [ r : fio - def1 ] the above definition of a fio was introduced in @xcite ( see also ( * ? ? ?
* chapter 2 ) ) .
it is equivalent to the traditional one , which is given in terms of local real - valued phase functions parametrizing the lagrangian manifold @xmath242 ( see , for example , @xcite or @xcite ) .
[ r : fio - def2 ] by ( * ? ? ?
* theorem 1.8 ) , the schwartz kernel of a fio can be represented by an integral of the form with any phase function @xmath225 and cut - off function @xmath224 satisfying the above conditions .
[ r : fio - def3 ] one can define a fio using with an amplitude @xmath243 depending on @xmath96 instead of @xmath244 .
these two definitions are equivalent . indeed , since @xmath245 with some smooth matrix - function @xmath246
, one can always remove the dependence on @xmath96 by expanding @xmath247 into taylor s series at the point @xmath235 , replacing @xmath248 with @xmath249 and integrating by parts .
in particular , this procedure shows that with an @xmath96-dependent amplitude @xmath250 defines an infinitely smooth half - density whenever @xmath251 in a conic neighbourhood of the set @xmath252 .
one can find all homogeneous terms in the expansion of the amplitude @xmath244 by analysing asymptotic behaviour of the fourier transforms of localizations of the distribution .
this implies that @xmath244 is determined modulo a rapidly decreasing function by the fio and the phase function @xmath225 .
it is not difficult to show that the conic support @xmath253 does not depend on the choice of @xmath225 and is determined only by the fio @xmath220 itself ( see , for instance , ( * ? ? ?
* section 2.7.4 ) ) .
we shall denote it by @xmath254 .
[ r : fio - def4 ] by ( * ? ? ?
* corollary 2.4.5 ) , if a phase function satisfies * ( a@xmath255 ) * , * ( a@xmath256 ) * and the matrix @xmath257 is positive definite then @xmath258 .
one can deduce from this that the set of phase functions @xmath225 satisfying * ( a@xmath255)**(a@xmath240 ) * is connected and simply connected .
let @xmath259 be the @xmath260-principal bundle over @xmath261 on which the multivalued function @xmath262 becomes a single valued continuous function of @xmath263 depending continuously on @xmath225 .
the fibre at the point @xmath264 can be thought of as the set of equivalence classes of pairs @xmath265 , where @xmath225 is a phase function satisfying * ( a@xmath255)**(a@xmath240 ) * , @xmath10 is an integer , and the equivalence relation is @xmath266 iff @xmath267,\ ] ] where the branch of the argument in the right hand side is chosen to be continuous along any path in the set of phase functions satisfying * ( a@xmath255)**(a@xmath240)*. then @xmath262 can be defined as a continuous single valued function on the principal bundle @xmath259 determined by @xmath268 ) \in ( -\pi + a , \pi + a ] .\ ] ] locally real phase functions provide local trivializations of @xmath259 determining the topology on the total space . in the following
we will often suppress the argument @xmath269 $ ] and think of the function @xmath270 as a multivalued function on @xmath261 understood as a continuous single valued function on @xmath259 .
factoring out @xmath271 one obtains a @xmath272-principal bundle which we denote by @xmath273 . the complex line bundle associated with the representation @xmath274
is called the keller maslov line bundle .
our definition here is equivalent to the one given in the literature ( see , for instance , @xcite and @xcite ) because if @xmath275 and @xmath276 are real then @xmath266 is equivalent to @xmath277 sections of the keller maslov line bundle can be understood as functions @xmath278 on @xmath273 satisfying the equivariance condition @xmath279 where @xmath280 $ ] denotes the variable in the fibre on which the @xmath260-action is defined by @xmath281+n = [ ( \phi , a+n)]$ ] . for the purpose of this article we will think of them in this way .
the situation simplifies when the bundle @xmath259 is topologically trivial .
this is equivalent to the existence a branch of @xmath282 which is continuous on the set @xmath283 .
clearly , such a branch exists whenever @xmath284 is simply connected .
let @xmath286 be a symmetric @xmath287-matrix with a nonnegative ( in the sense of operator theory ) real part @xmath288 , and let @xmath289 be the orthogonal projection on @xmath290 .
we shall denote @xmath291 .
furthermore , we define the function @xmath292 in such a way that it is continuous with respect to @xmath293 on the set of matrices with a fixed kernel and is equal to zero when @xmath294 . in particular ,
if @xmath295 then @xmath296 where @xmath297 is the signature of @xmath298 ( see ( * ? ? ?
* section 3.4 ) ) . the following is ( * ? ? ?
* proposition 2.3 ) .
[ p : index ] the function @xmath299 as a function on @xmath259 does not depend on @xmath225 and local coordinates .
the function @xmath285 is continuous along any path on which @xmath300 is constant .
[ d : index ] if @xmath301 \to { { \mathcal d}}(\phi)$ ] is a path in @xmath284 let @xmath302 \to { { \mathcal d}}_{\mathbb{z}}(\phi)$ ] be any continuous lift .
@xmath303 is independent of the lift and called the maslov index of @xmath304 .
the above definition was introduced in @xcite ( the idea goes back to @xcite ) .
if the path is closed then it coincides with the maslov index defined in @xcite via a ech cohomology class associated with parametrizations of the lagrangian manifold by families of local real - valued phase functions .
the index function @xmath285 allows one to extend this notion to non - closed paths .
choosing @xmath224 with a sufficiently small support , we can rewrite in the form @xmath305 where @xmath306 is the section of the keller
maslov bundle obtained from the amplitude @xmath307 by the procedure described in remark [ r : fio - def3 ] .
clearly , @xmath306 belongs to the same class @xmath208 and @xmath308 where @xmath309 and @xmath310 are the leading homogeneous terms of the amplitudes @xmath306 and @xmath226 . by construction ,
@xmath311 is well defined for @xmath96 sufficiently closed to @xmath312 as a continuous function on @xmath273 satisfying the equivariance condition @xmath313 where @xmath314 is a variable for the fibre of @xmath273 .
therefore it is a section in the dual of the keller maslov bundle .
since the product of @xmath315 is single valued , @xmath316 is a section of the keller maslov line bundle
. we shall call it a _ full symbol _ of the corresponding fio .
the following is ( * ? ? ?
* theorem 2.7.11 ) .
[ t : symbol ] let @xmath317 be a fio whose schwartz kernel is given by .
then the leading homogeneous term @xmath309 of the amplitude @xmath318 is uniquely determined by the operator @xmath317 .
[ r : proof2 ] by theorem [ t : symbol ] , the leading homogeneous term @xmath309 does not depend on the choice of local coordinates and does not change when we change the phase function @xmath225 in the representation .
[ r : proof3 ] theorem [ t : symbol ] was proved in @xcite only for canonical transformations associated with billiards .
however , the same proof works in the general case . in the following
we will denote the leading homogeneous term @xmath309 of the symbol of a fio @xmath220 by @xmath319 and think of it as a multivalued function on @xmath261 or as a single valued function on @xmath273 respectively , that satisfies ( [ equivariance ] ) .
note that the product @xmath320 is single valued by construction as @xmath321 where @xmath310 is the leading homogeneous term of the amplitude @xmath226 from and @xmath322 is evaluated at @xmath323 .
[ e : symbol ] if @xmath221 is the identical transformation then @xmath324 and the corresponding fio @xmath317 is a @xmath207 ( see , for example , ( * ? ? ? * theorem 19.1 ) ) .
if we put @xmath325 then the principal symbol of @xmath317 ( in the sense of the theory of pseudodifferential operators ) coincides with the leading homogeneous term @xmath309 .
if the bundle @xmath273 is trivial we can globally fix a branch of @xmath326 for a fixed phase function .
note , that in general there will be no preferred branch . in case @xmath220 is a @xmath207 we shall however suppose that @xmath327 , so that in this case @xmath319 coincides with the traditional principal symbol . in what follows , in order to avoid ` boundary effects ' when considering compositions of various fios and @xmath207s on @xmath2 , we shall have to assume that supports of their schwartz kernels are separated from the boundary .
namely , we shall deal the following classes of operators .
* @xmath328 is the class of operators @xmath220 whose schwartz kernels @xmath218 vanish in a neighbourhood of the set @xmath329 .
* @xmath330 is the class of operators @xmath220 whose schwartz kernels @xmath218 vanish in a neighbourhood of the set @xmath331 .
* @xmath332 .
* @xmath333 is the class of operators that can be written in the form @xmath334 where @xmath335 .
the following results are in principle well - known .
however , we need explicit formulae for the principal symbols , which are not obvious ( partly , due to the fact that there are many possible definitions of the symbol ) .
therefore we have included a simple direct proof of theorem [ t : composition ] in appendix [ s : a ] . in the rest of this section @xmath336 or
@xmath337 will denote a map between the fibres of the @xmath272-bundles that depends only on the involved canonical transformations .
a construction of this map is implicitly contained in the proofs of the statements but is not needed for our purposes .
[ t : composition ] let @xmath338 be fios of order @xmath339 associated with canonical transformations @xmath340 , where @xmath341 .
assume that either @xmath342 or @xmath343 .
then the composition @xmath344 is a fio of order @xmath345 associated with the canonical transformation @xmath346 with principal symbol equal to @xmath347 such that @xmath348 recall that the inner product in the space of half - densities @xmath349 is invariantly defined , and so are the adjoint operators . taking @xmath350 in theorem [ t : composition ] , we obtain [ c : adjoint ] let @xmath220 be a fio of order @xmath183 associated with a canonical transformation @xmath221 .
if @xmath351 then the adjoint operator @xmath352 is a fio of order @xmath183 associated with the inverse transformation @xmath353 with principal symbol equal to @xmath354 such that @xmath355 theorem [ t : composition ] and corollary [ c : adjoint ] immediately imply [ c : composition ] let @xmath356 , @xmath357 be as in theorem [ t : composition ] .
assume that either @xmath358 or @xmath359 .
then @xmath360 is a fio of order @xmath345 associated with the transformation @xmath361 with principal symbol equal to @xmath362 such that @xmath363 [ r : simple ] in simple words , the above formulae for principle symbols mean that @xmath364 , @xmath365 and @xmath366 , where @xmath367 are integers which are uniquely determined by the canonical transformations .
it is well known that fios can be extended to the space @xmath368 of distributions with compact supports in @xmath369 ( see , for instance , @xcite or @xcite ) .
theorem [ t : composition ] and standard results on @xmath207s imply that a fio of order @xmath183 lying in @xmath370 is bounded from @xmath120 to @xmath371 .
let @xmath372 be @xmath207s , and let @xmath220 be a fio associated with a canonical transformation @xmath221 .
then , by corollary [ c : composition ] , @xmath373 is a fio associated with @xmath221 such that @xmath374 is a subset of @xmath375 the above inclusion implies that @xmath376 for all distributions @xmath377 , where @xmath378 denotes the wave front set . roughly speaking
, this means that singularities of a distribution are moved by the map @xmath221 under the action of the associated fio @xmath220 .
the following is a refined version of egorov s theorem .
[ t : egorov ] let @xmath356 and @xmath357 be fios of orders @xmath379 and @xmath380 associated with a canonical transformation @xmath221 . if @xmath381 is a @xmath207 of order @xmath183 then 1 .
the composition @xmath382 is a @xmath207 of order @xmath383 such that @xmath384 2 .
@xmath385 3 . if @xmath386 then @xmath387 .
the first two statements are obtained by applying corollary [ c : composition ] to the composition @xmath388 and then theorem [ t : composition ] to the composition of this operator with @xmath389 .
if @xmath390 then @xmath391 coincides with @xmath392 as the principal principal symbol of a fio , that is , modulo a factor @xmath393 with @xmath394 . if @xmath18 is a non - negative self - adjoint operator
then so is @xmath395 .
it follows that @xmath396 , which implies that @xmath397 . since @xmath398 continuously depends on @xmath18 , the same is true for all @xmath207s @xmath18
in this section and further on we shall denote @xmath400 since @xmath1 is self - adjoint , the operators @xmath401 form a strongly continuous unitary group in the space of half - densities in @xmath349 . in the following theorem
@xmath402 denotes the class of operators whose schwartz kernels are infinitely smooth on @xmath403\times m \times m$ ] up to the boundary .
[ t : parametrix ] let @xmath18 and @xmath246 be @xmath207s in the space of half - densities on @xmath2 such that @xmath404 , @xmath405 and @xmath406 . then ,
on the time interval @xmath403 $ ] , we have @xmath407 modulo @xmath402 , where the sum is finite and @xmath408 are parameter - dependent fios such that 1 .
each fio @xmath408 is associated with the canonical transformation @xmath409 , where @xmath410 and @xmath179 is a billiard segment defined in subsection [ s : billiards2 ] ; 2 .
each fios @xmath408 satisfies the equation @xmath411 modulo @xmath402 ; 3 .
the fio @xmath412 associated with the first segment ( which starts at @xmath413 ) satisfies the initial condition @xmath414 , for all other segments @xmath415 ; 4 .
the phase functions corresponding to the incoming , reflected and refracted trajectories coincide on the set @xmath416 , and so do the arguments @xmath417 ; 5 .
the principal symbols @xmath418 are ( locally ) independent of @xmath419 .
the theorem is proved using the standard technique , which goes back to @xcite . for a usual elliptic boundary value problem with branching billiards
, it is discussed in detail in @xcite . for the operator defined in section [
s : setting ] , a sketch of proof was given in @xcite .
[ r : negative - time ] the same arguments show that a similar result holds for negative times .
more precisely , theorem [ t : parametrix ] remains valid for @xmath420 $ ] under the assumption that @xmath421 , where @xmath422 is the set of starting points of billiard trajectories going in the reverse direction which are well defined on the time interval @xmath423 $ ] .
one can easily show that @xmath424 if and only if @xmath425 . following ( * ? ? ?
* section 2.6.3 ) , let us fix branches of @xmath417 for the fios @xmath408 assuming that * for the fio @xmath412 associated with the first segment of billiard trajectories , @xmath426 ; * the branches corresponding to the incoming , reflected and refracted trajectories coincide on the set @xmath427 .
in view of the parts ( 3 ) and ( 4 ) of theorem [ t : parametrix ] , these two conditions can be satisfied . clearly , they uniquely determine the branch of @xmath417 for all the fios @xmath408 for all times @xmath419 .
it follows that the bundles @xmath428 over @xmath429 associated with the transformations @xmath430 are trivial .
this allows us to consider the principal symbols @xmath431 of fios @xmath408 as single - valued functions on @xmath432 .
in particular , theorem [ t : parametrix](5 ) implies that the principal symbol @xmath433 associated with the first segment is identically equal to 1 .
let @xmath434 be the conic set of points @xmath435 such that all the billiard trajectories @xmath192 experience at most @xmath183 reflections and refractions for @xmath199 $ ] .
the sets @xmath434 are open , and their union over @xmath183 contains @xmath197 . since the intersection @xmath436 is compact , the conic support @xmath429 is covered by a finite collection of connected components of @xmath434 with a sufficiently large @xmath183 . assume , for the sake of simplicity , that @xmath429 lies in one connected component of @xmath434 , and let @xmath408 be one of the fios introduced in theorem [ t : parametrix ]
. since there are no dead - end , grazing or singular trajectories of length @xmath199 $ ] originating from @xmath434 , for each fixed @xmath178 and @xmath437 there are two possibilities : 1 .
all segments @xmath438 with @xmath410 end at the points of @xmath439 which do not belong to the set of total reflection ; 2 .
all segments @xmath438 with @xmath410 end at the points of total reflection in @xmath439 ; 3 .
all segments @xmath438 with @xmath410 end at @xmath440 . in the first case , in order to satisfy the boundary condition , one has to add to @xmath408 two fios @xmath441 and @xmath442 , corresponding to the reflected and refracted trajectories , respectively . in the second case ,
the fio @xmath441 corresponding to the reflected trajectories is chosen is such a way that the schwartz kernel of the sum @xmath443 has singularities only at the points @xmath444 with @xmath445 .
after that one can satisfy the boundary condition by adding a ` boundary layer term ' whose singularities are also located at the points @xmath444 with @xmath445 ( see ( * ? ? ?
* section 3.3.4 ) for details ) . since @xmath404
, the boundary layer terms do not appear in the sum representing the operator @xmath446 .
finally , in the third case , the dirichlet or neumann boundary condition is satisfied by adding only a fio @xmath441 corresponding to the reflected trajectories .
let @xmath431 , @xmath447 and @xmath448 be the principal symbols of the fios @xmath408 , @xmath441 and @xmath442 .
[ l : split ] assume that the trajectory @xmath449 approaches @xmath86 from the @xmath148-side and hits @xmath86 at the time @xmath450 at the point @xmath451 let @xmath452 be as in subsection [ s : billiards1 ] , and let @xmath453 if @xmath171 is a point of total reflection .
define @xmath454 then @xmath455 where 1 . in the first case @xmath456 and @xmath457 ; 2 . in the second case @xmath458 .
if the trajectory hits the boundary @xmath83 then we have with 1 .
@xmath459 for the neumann boundary condition , and + @xmath460 for the dirichlet boundary condition . clearly , if the geodesics approaches @xmath86 from the @xmath166-side , then the coefficients @xmath461 and @xmath462 are defined by the formulae which are obtained from the above equalities by swapping @xmath463 and @xmath464 .
[ r : energy ] note that @xmath465 in the case ( i ) , and @xmath466 in the cases ( ii ) , ( iii ) .
part ( iii ) of lemma [ l : split ] is a particular case of ( * ? ? ?
* corollary 3.4.7 ) . the formulae ( i ) and
( ii ) can be deduced from ( * ? ? ?
* proposition 3.3 ) , which states the same result but for principals symbols defined in a different way . however , the proof of proposition 3.3 in @xcite is very sketchy and is not easy to reconstruct
. therefore in appendix [ s : b ] we outline a direct proof , which uses the technique developed in @xcite . in this subsection
we shall briefly recall some results from ( * ? ? ?
* section 1.5 ) and ( * ? ? ?
* appendix d.6 ) .
strictly speaking , they were proved in @xcite only for billiards obtained by reflections .
however , the same technique of matching phase functions is applicable to refracted trajectories , and the proofs remain exactly the same .
let us denote the index functions of the transformations @xmath467 by @xmath468 .
theorem [ t : parametrix](4 ) implies that the index functions corresponding to two consecutive billiard segments coincide at the points of reflection and refraction .
this allows us to define the index function @xmath469 associated with the transformation @xmath204 .
consider the matrix of the first derivatives @xmath470 . since @xmath471 is positively homogeneous in @xmath142 of degree zero ,
@xmath472 is not greater than @xmath473 .
if @xmath474 for some @xmath122 then the point @xmath475 is said to be a conjugate point of the billiard trajectory @xmath476 , and the number @xmath477 is called its multiplicity .
at the points of reflection and refraction , @xmath472 is the same for the incoming , reflected and refracted trajectories .
thus the notions of a conjugate point and its multiplicity are well defined for all @xmath478 and all @xmath479 $ ] .
the following statement is an immediate corollary of ( * ? ? ?
* lemma 1.5.6 ) and ( * ? ? ?
* theorem d.6.8 ) .
[ p : theta ] if @xmath478 then every billiard trajectory @xmath476 , @xmath199 $ ] , has only finitely many conjugate points .
the number of its conjugate points counted with their multiplicities is equal to @xmath480 .
the operator @xmath1 is self - adjoint and has compact resolvent and therefore there exists an orthonormal basis @xmath481 in @xmath349 such that @xmath482 where @xmath483 . given an @xmath484-bounded operators @xmath18 , let us define @xmath485 the weyl asymptotic formula for @xmath486 implies that @xmath487 we shall denote @xmath488 whenever the limit exists .
formulae giving the value of @xmath489 in terms of other characteristics of the operator @xmath18 ( such as its symbol or schwartz kernel ) are usually called local weyl laws . [ calkin ] since @xmath490 whenever the operator @xmath491 is compact , @xmath489 depends only on the image of @xmath18 in the calkin algebra .
clearly , @xmath492 is a function with locally bounded variation , whose derivative @xmath493 is the sum of @xmath494-functions located at the points @xmath495 with coefficients @xmath496 . for every smooth rapidly decreasing function @xmath497 on @xmath498 we have @xmath499 where @xmath500 denotes the convolution . in many cases it is easier to investigate the asymptotic behaviour of @xmath501 .
after that , a local weyl law can be obtained by applying a suitable tauberian theorem . if @xmath381 and its schwartz kernel @xmath502 satisfies the condition 1 .
@xmath503 for all @xmath504 and @xmath505 then the trace @xmath506 exists as a distribution in @xmath419 .
indeed , the above condition implies that @xmath507 is an operator with a continuous kernel @xmath508 for all sufficiently large positive integers @xmath183 , so that we can define @xmath509 in this case @xmath510 where @xmath511 is the fourier transform of @xmath497 and @xmath512 is the inverse fourier transform .
let @xmath513 be a fio of order zero associated with canonical transformations @xmath221 . by ( * ? ?
* theorem 8.1.9 ) , the wave front of the schwartz kernel of @xmath220 is a subset of @xmath514 therefore @xmath220 satisfies the condition * ( c)*. further on * @xmath515 is the normalised measure on @xmath9 ; * @xmath516 is the set of fixed points of a transformation @xmath221 lying in @xmath9 .
the following two results are proved in appendix [ s : c ] .
[ t : local - fio ] if @xmath517 is a fio of order zero associated with a homogeneous canonical transformations @xmath221 then the limit @xmath518 exists and @xmath519 [ l : thetas ] if @xmath520 are fios of order zero associated with homogeneous canonical transformations @xmath340 and let @xmath521 and @xmath346 .
then @xmath522 almost everywhere on the set @xmath523 .
theorems [ t : composition ] , [ t : local - fio ] , and lemma [ l : thetas ] immediately imply [ c : local - fios ] if @xmath520 are fios of order zero associated with canonical transformations @xmath340 then the limit @xmath524 exists and coincides with @xmath525 where @xmath346 .
applying theorem [ t : local - fio ] with @xmath526 and the weyl formula for the counting function , we obtain the following well known result .
[ c : local - pdo ] if @xmath527 is a @xmath207 of order zero then the limit @xmath489 exists and is equal to @xmath528 .
a theorem similar to theorem [ t : local - fio ] was stated and proved under a clean intersection condition in @xcite .
note that the formula for @xmath518 given in @xcite differs from , as it does not contain the factor @xmath529 .
this is due to the fact that the author used an implicitly defined notion of `` scalar principal symbol '' .
an explicit definition of this object would have involved the index function or an analogue .
let @xmath201 be the conic subsets of the cotangent bundle @xmath190 defined at subsection [ s : billiards3 ] . by assumption [ a : bad - points ]
, @xmath201 is an open set of full measure for each @xmath530 .
if @xmath478 then all the billiard trajectories @xmath192 originating from @xmath264 are well defined for @xmath199 $ ] and experience only finitely many reflections and refractions .
it follows that , for each fixed @xmath478 and @xmath199 $ ] , the set of end points of the trajectories @xmath192 is finite .
let us denote it by @xmath531 .
[ r : dynamics1 ] note that @xmath532 is not uniquely defined if the trajectory hits the boundary at the time @xmath149 .
for the sake of definiteness , we shall be assuming that in this situation @xmath533 .
[ r : dynamics2 ] recall that the shifts along billiard trajectories @xmath204 are homogeneous canonical transformations in @xmath203 .
one can consider the branching billiard system as a family of multi - valued canonical transformations @xmath534 , mapping @xmath194 into the set @xmath535 .
suppose that a billiard trajectory @xmath192 is well defined and hits the boundary at the times @xmath536 for each @xmath537 we have the associated coefficient @xmath538 calculated in lemma [ l : split ] , where @xmath539 is the order number of the next segment .
more precisely , in the notation of lemma [ l : split ] , @xmath540 if the next segment is obtained by reflection , and @xmath541 if the next segment is obtained by retraction .
let us define @xmath542 in view of remark [ r : energy ] , @xmath543 for all @xmath478 and @xmath199 $ ] , where the sum is taken over all distinct billiard trajectories of ` length ' @xmath419 originating from @xmath264 .
[ r : weight ] we call the number @xmath544 the weight of the trajectory @xmath545 , @xmath546 $ ]
. it can be thought of as the proportion of energy transmitted along the billiard trajectory , or the probability for a particle to travel along this trajectory .
if @xmath547 , let us denote @xmath548 where @xmath549 are as above , @xmath550 are the index functions introduced in subsection [ s : thetas ] , and the sum is taken over all distinct billiard trajectories of ` length ' @xmath419 originating from @xmath264 and ending at @xmath551 . in view of assumption
[ a : bad - points ] , the following definition makes sense for all @xmath552 $ ] and @xmath193 .
[ def : transfer - op ] the classical transfer operators @xmath553 and the diagonal transfer operators @xmath554 in the space @xmath555 are defined for times @xmath193 by the equalities @xmath556 where @xmath557 .
the difference between @xmath558 and @xmath559 is in the contributions from recombining billiard trajectories , that is , the billiard trajectories such that @xmath560 but @xmath561 for some @xmath562 .
if the set of initial points that admit recombining billiard trajectories has measure zero then @xmath554 and @xmath563 coincide .
[ e : hemispheres ] let @xmath564 be the unit 2-dimensional sphere and @xmath47 be a great circle , splitting @xmath564 into the union of two hemispheres @xmath565 and @xmath566 .
let us provide @xmath567 with the metrics @xmath568 , where @xmath569 is the standard metric on @xmath564 and @xmath570 are positive constants , and consider the riemannian manifold @xmath571 .
in this situation , the billiard trajectories are formed by the great semicircles lying either in @xmath565 or @xmath566 , whose length is equal to @xmath572 and @xmath573 , respectively . if @xmath574 with some positive integers @xmath575 and @xmath576 then for every billiard trajectory @xmath577 there exists a distinct trajectory @xmath578 with the same end point , which is obtained from @xmath577 by replacing @xmath579 great semicircles in @xmath565 with @xmath580 great semicircles in @xmath566 or the other way round . if @xmath581 and @xmath582 are rationally independent then there are no recombining trajectories .
the following theorem reveals the link between the diagonal transfer operators and local weyl asymptotics .
[ averagingtheorem ] let @xmath583 be @xmath207s of order zero in the space of half - densities on @xmath2 such that @xmath584 . then , for all @xmath207s @xmath527 of order zero and all @xmath585 $ ] , @xmath586 if @xmath18 is the multiplication by a constant then the theorem follows from corollary [ c : local - pdo ] .
thus we can assume without loss of generality that @xmath381 .
then , by theorem [ t : parametrix ] , @xmath587 and @xmath588 can be represented as finite sums of iofs @xmath589 and @xmath590 .
it follows that @xmath591 corollaries [ c : adjoint ] and [ c : composition ] imply that @xmath592 and @xmath593 are fios associated with canonical transformation @xmath594 and @xmath467 with principal symbols @xmath595 respectively . clearly , the set of fixed points of the mapping @xmath596 consist of @xmath597 such that @xmath598 now the required result is obtained by calculating the principal symbols of @xmath408 and @xmath599 with the use of lemma [ l : split ] , applying corollary [ c : local - fios ] and summing up over all segments of the billiard trajectories .
[ c : sums ] if @xmath478 and @xmath600 then @xmath601 the first equality is an immediate consequence of .
the second is proved by applying theorem [ averagingtheorem ] to @xmath602 and comparing the obtained result with corollary [ c : local - pdo ] .
clearly , * @xmath603 and @xmath604 are positivity preserving operators .
corollary [ c : sums ] implies that * the operators @xmath603 and @xmath604 are continuous in all spaces @xmath555 with @xmath552 $ ] and their operator norms in these spaces are bounded by 1 , * @xmath603 and @xmath604 are isometries in the space @xmath605 .
note that the operators @xmath603 form a semigroup , whereas @xmath606 may not coincide with @xmath607 .
let @xmath608 be either @xmath553 or @xmath554 .
we say that @xmath608 is ergodic if for all @xmath609 @xmath610 as @xmath611 almost everywhere in @xmath9 . in view of corollary
[ c : sums ] and lebesgue s dominated convergence theorem , if @xmath608 is ergodic then @xmath612 in all spaces @xmath555 , where @xmath613 is the function identically equal to one .
[ r : ergodicity1 ] changing the order of integration , one can rewrite the condition in the following equivalent form @xmath614 [ r : ergodicity2 ] the traditional definition of ergodicity assumes that @xmath615 as @xmath616 almost everywhere .
it is easy to see that implies but , generally speaking , the converse is not true .
however , if the left hand side of does converge to a limit for all @xmath116 ( as in the von neumann ergodic theorem ) then , by , the limit coincides with @xmath617 and , consequently , the dynamics is ergodic in the classical sense . in our scenario
this happens when @xmath618 and there are no branching trajectories .
the purpose of this section is to prove the following theorem which is the main result of this paper .
[ theo : main ] suppose that assumption [ a : bad - points ] is fulfilled and that the diagonal dynamics @xmath554 is ergodic
. then quantum ergodicity ( qe ) holds , that is , for any @xmath207 @xmath619 of order zero we have @xmath620 the proof proceeds in several steps .
we can assume without loss of generality that @xmath621 simply by subtracting the constant @xmath622 from @xmath18 .
thus it is sufficient to prove that , under the assumption , @xmath623 if @xmath624 is an @xmath484-bouned operator , let us denote @xmath625 so that @xmath626 whenever the limit exists . by the cauchy
schwarz inequality , @xmath627 let us define @xmath628 and @xmath629 . here and further on integrals of operator - valued functions
are understood in the weak sense . since @xmath630 and @xmath631 for all @xmath632
, we have @xmath633 for all positive integers @xmath437 , @xmath632 and @xmath634 .
therefore , by , @xmath635 thus it is sufficient to show that @xmath636 note that , generally speaking , the operators @xmath637 and @xmath638 are not fios and do not belong to @xmath370 .
therefore we can not directly apply theorem [ t : local - fio ] or corollary [ c : local - fios ] to evaluate the upper limit . clearly , @xmath639 where @xmath640 since @xmath641 and @xmath642 is real , the integral on the right hand side of coincides with @xmath643 thus it follows that @xmath644 let @xmath645 be the operator of multiplication by a real - valued function @xmath646 such that @xmath647 , and let @xmath648 be a @xmath207 of order zero .
since @xmath649 and @xmath650 , we have @xmath651 where @xmath652 is the multiplication by @xmath653 , and @xmath654 these inequalities and imply the estimate @xmath655 now , applying the second inequality , we see that @xmath656 for all @xmath657 let us fix an arbitrary @xmath658 . in view of
, ergodicity of the diagonal dynamics implies that there exists @xmath634 such that @xmath659 let us fix such a positive @xmath660 and choose the nonnegative function @xmath646 and the @xmath207 @xmath648 of order zero such that 1 .
@xmath661 , 2 .
@xmath406 , @xmath662 and @xmath663 .
note that ( b ) can be satisfied because , by assumption [ a : bad - points ] , @xmath664 has full measure . in view of corollary
[ c : local - pdo ] , the limits @xmath665 and @xmath666 exist and are smaller than @xmath667
. therefore 1 .
the right hand side of is estimated by @xmath668 for all sufficiently large @xmath48 . by theorem [ averagingtheorem ] , the limit @xmath669 also exits and is equal to @xmath670 the lebesgue dominated convergence theorem implies that
@xmath671 substituting , integrating over @xmath672 and @xmath419 , and taking into account and ( a ) , ( b ) , we see that the absolute value of is smaller than @xmath673 . consequently , 1 .
the integral in the left hand side of is estimated by + @xmath674 for all sufficiently large @xmath48 . applying the estimates
@xmath675 and @xmath676 to , we obtain @xmath677 since @xmath678 can be chosen arbitrarily small , this implies .
in this and next sections , if @xmath679 is a vector - function of @xmath680-dimensional variable @xmath681 , we denote by @xmath682 the @xmath287-matrix function with entries @xmath683 , where @xmath437 enumerates elements of the @xmath30th row .
let @xmath684 and @xmath685 note that @xmath686 @xmath687 @xmath688 for all @xmath264 and @xmath551 in any local coordinates because the transformations @xmath340 preserves the 2-form @xmath689 and the 1-form @xmath690 .
the proof proceeds in several steps .
assume that @xmath691 .
then there exists a local coordinate system @xmath692 in a neighbourhood of @xmath693 such that @xmath694 and @xmath695
. indeed , let @xmath696 be arbitrary coordinates in a neighbourhood of @xmath693 .
one can easily show that under a change of coordinates @xmath697 the matrices @xmath698 and @xmath699 transform in the following way @xmath700 where @xmath701 , @xmath702 are symmetric matrices with entries @xmath703 ( see , for instance , ( * ? ? ?
* section 2.3 ) ) . clearly , one can choose coordinates @xmath704 in such a way that @xmath705 and @xmath706 , where @xmath707 is an arbitrary real constant . then turn into @xmath708 in view of , @xmath709 and @xmath710 map the kernels of the matrices @xmath711 and @xmath712 into the orthogonal complements of their ranges .
this implies that the matrices in the right hand sides of the equalities are non - degenerate for all sufficiently large @xmath713 .
let @xmath714 , @xmath715 and @xmath716 be as in part 1 , and let @xmath704 be an arbitrary local coordinate system such that @xmath694 and @xmath695 .
assume first that @xmath717 lies in a sufficiently small conic neighbourhood @xmath718 of the point @xmath719 such that @xmath720 for all @xmath721 and @xmath722 for all @xmath723 .
if @xmath724 then , by , @xmath725 @xmath726 and @xmath727 . therefore the phase functions @xmath728 satisfy the conditions * ( a@xmath255)**(a@xmath240 ) * of subsection [ fio : def ] for all @xmath721 , @xmath723 and @xmath704 sufficiently closed to @xmath693 .
since @xmath729 for all distributions @xmath730 , we may assume without loss of generality that @xmath731 .
then , in view of the above , the schwartz kernels @xmath732 and @xmath733 of the fio @xmath338 can be represented by oscillatory integrals @xmath734 of the form with the phase functions .
the schwartz kernel of the composition @xmath344 coincides with @xmath735 where the integrals are taken over @xmath736 , @xmath737 and @xmath738 now we are going to apply the stationary phase method to the integral with respect to the variables @xmath704 and @xmath739 , considering @xmath740 as a large parameter .
a rigorous justification of the stationary phase formula for non - convergent integrals of this type can be found , for instance , in ( * ? ? ?
* appendix c ) . the equations @xmath741 and @xmath742 are equivalent to @xmath743 respectively . since @xmath744 and @xmath695 , in a neighbourhood of @xmath745 , the second equation has a unique @xmath739-solution @xmath746 such that @xmath747
. thus the stationary point is @xmath748 .
it is unique and non - degenerate because @xmath749 by the stationary phase formula , the integral coincides modulo a smooth function with @xmath750 where @xmath751 and @xmath247 is an amplitude of class @xmath752 with the leading homogeneous term @xmath753 such that @xmath754 ( we have used the fact that the signature of the hessian is equal to zero ) . clearly , @xmath755 and @xmath756 . thus @xmath757 .
since @xmath758 for all @xmath759 , we also have @xmath760 now the second equality implies that @xmath761 . substituting @xmath235
, we obtain @xmath762 similarly , @xmath763 this equality and imply that @xmath764 and , consequently , @xmath765 differentiating the identity @xmath766 , we obtain @xmath767 from the above two equalities it follows that @xmath768 where @xmath769 , @xmath770 , and @xmath771 , @xmath772 , @xmath773 , @xmath774 are evaluated at @xmath171 . in view of and , @xmath775 consequently , @xmath776 and @xmath777 since @xmath778 , the above equality and imply that the phase function satisfies the conditions of subsection [ fio : def ] .
this shows that defines the schwartz kernel of a fio associated with the canonical transformation @xmath221 .
applying the procedure described in remark [ r : fio - def3 ] , we can remove the dependence on @xmath96 and rewrite it in the form @xmath779 where @xmath224 is a cut - off function satisfying the conditions of subsection [ fio : def ] and @xmath244 is an amplitude of class @xmath752 such that @xmath780 clearly , the leading homogeneous term @xmath310 of the amplitude @xmath226 is given by the formula @xmath781 where @xmath782 are the leading homogeneous terms of the amplitudes @xmath212 .
consider now general fios @xmath338 associated with the canonical transformations @xmath340 .
splitting @xmath338 into sums of fios with simply connected conic supports , we see that it is sufficient to prove the theorem assuming that the bundles @xmath783 are topologically trivial and the schwartz kernels of @xmath338 are given by oscillatory integrals @xmath784 of the form with phase functions @xmath785 and full symbols @xmath786 . if the conic supports of @xmath786 are sufficiently small then , choosing suitable local coordinates and transforming @xmath785 into the phase functions , we can rewrite the corresponding oscillatory integrals in the form with @xmath787 , where @xmath367 is an integer determined by the choice of branch of @xmath788 . in this case , by part 2 , the composition @xmath344 is a fio given by the oscillatory integral with the local phase function and an amplitude @xmath789 with the leading homogeneous term @xmath790 let @xmath791 be an arbitrary global phase function associated with the transformation @xmath221 . since @xmath792 is simply connected , the bundle @xmath793 is also trivial .
let us fix a continuous branch of @xmath794 . transforming the phase function @xmath225 given by into @xmath791
, we see that coincides with an oscillatory integral of the form @xmath795 where @xmath796 is another amplitude with the same leading homogeneous term @xmath797 and @xmath798 is the integer such that @xmath799 turns into @xmath800 under continuous transformation of the phase functions @xmath801 ( see remark [ r : proof2 ] ) .
thus have proved that , for @xmath338 with small conic supports , the composition @xmath344 is a fios of order @xmath345 with principal symbol @xmath802 such that @xmath348 obviously , the integer @xmath803 is uniquely defined by the choice of branches of @xmath788 and @xmath794 . therefore , using a partition of unity on @xmath215 , we see that the same result holds for all fios @xmath338 .
since the principal symbols are defined modulo a factor @xmath804 with an integer @xmath183 , this completes the proof .
let the fios corresponding to the incoming , reflected and refracted trajectories are given by the oscillatory integrals with phase functions @xmath225 , @xmath807 , @xmath808 and symbols @xmath306 , @xmath809 and @xmath810 respectively .
the first two are standard oscillatory integrals defined in section [ s : fio ] . in the case ( i ) , the third is also a standard oscillatory integral . in the case ( ii ) , it is a boundary layer oscillatory integral given by the same expression but with a complex - valued phase function satisfying the conditions of ( * ? ? ?
* section 2.6.4 ) .
since we are dealing with half densities elements @xmath811 in the domain of the operator have to satisfy @xmath812 in each local coordinate chart for each @xmath813 in the domain of the chart , where @xmath814 are the left and right boundary values of @xmath811 and @xmath815 denote the inward @xmath108-normal derivatives . substituting the sum of the integrals into the boundary conditions and equating to zero the sum of leading terms at @xmath816 and @xmath235
, we obtain the following equations , @xmath817 @xmath818 and @xmath819 where @xmath820 then condition * ( a@xmath256 ) * implies that @xmath821 similarly , in the case ( i ) , @xmath822 . in the case ( ii ) , by ( * ? ? ?
* ( 2.6.23 ) ) , we have @xmath823 . from the equalities , ( * ?
* ( 2.5.3@xmath824 ) ) and ( * ? ? ?
* ( 2.6.25 ) ) it follows that at the point @xmath816 , @xmath235 @xmath825 ( cf .
* ( 2.6.14 ) ) ) .
consequently , @xmath826 and @xmath827 where @xmath828 is a continuous branch of the square root . in view of the above equalities , the equations ,
imply that @xmath829 in the case ( i ) , and @xmath830 in the case ( ii ) , where we have used @xmath831 . solving these equations with respect to @xmath832 , @xmath833
, we obtain the required formulae for the coefficients @xmath461 , @xmath462 .
without loss of generality we shall be assuming that @xmath254 lies in a sufficiently small neighbourhood of a fixed point @xmath834 .
let @xmath835 with a sufficiently small @xmath494 .
if @xmath494 is smaller than the geodesic distance from the support of the schwartz kernel of @xmath220 to the boundary @xmath40 then , in view of theorems [ t : parametrix ] , corollary [ c : composition ] and remark [ r : negative - time ] , @xmath836 is a fio associated with the canonical transformation @xmath837 with principal symbol @xmath838 , where @xmath839 is the shift along geodesics in @xmath369 .
denote @xmath840 assume that @xmath494 and @xmath254 are small enough .
then the union @xmath841 is also small and , for @xmath835 , the schwartz kernel of @xmath836 can be represented ( modulo smoothing operators ) by an oscillatory integral @xmath842 with a phase function given by the equality @xmath843 in a local coordinate system such that @xmath844 , an amplitude @xmath226 with small conic support , and a cut - off function @xmath224 satisfying the conditions of subsection [ fio : def ] ( see appendix [ s : a ] ) . [ l : fio - local ] let @xmath845 be the set of points @xmath144 such that @xmath846 at some time @xmath847 . if the schwartz kernel of @xmath836 is given by and @xmath848 then @xmath849 as @xmath850 , where @xmath851 and @xmath310 is the leading homogeneous term of the amplitude @xmath226 . clearly , @xmath852 .
therefore @xmath853 since @xmath221 preserves the 1-form @xmath854 , we have @xmath855 and , consequently , @xmath856 . by euler
s identity for homogeneous functions , @xmath857 where @xmath858 .
now , differentiating , we see that @xmath859 for all @xmath860 provided that @xmath237 is small enough .
let @xmath861 then @xmath862 changing variables @xmath863 , where @xmath864 and @xmath865 are coordinates on the cosphere @xmath866 , we obtain @xmath867 with @xmath868 now we are going to apply the stationary phase formula with respect to @xmath419 and @xmath869 ( see ( * ? ? ?
* appendix c ) for justification of this procedure ) . in view of ,
@xmath870 on @xmath871 .
thus all stationary points of @xmath872 lying in @xmath871 are non - degenerate .
they are given by the equations @xmath873 by , the first equation either does not have any solutions , or has a unique solution @xmath874 smoothly depending on @xmath875 . here
we have used the implicit function theorem and assumed that @xmath494 is small enough so that @xmath874 is defined on @xmath871 . in the first case is a rapidly decreasing function of @xmath876 and @xmath877 for all @xmath878 , which implies that @xmath879 for all @xmath880 and @xmath881 . in the second case , by the stationary phase formula ,
@xmath882 as @xmath850 , where @xmath883 . one can easily show that the right hand side of coincides with @xmath884 where @xmath885 is the set of stationary points of the function @xmath886 . differentiating the identity @xmath887 and taking into account
, we see that @xmath888 consists of the points @xmath875 such that @xmath889 and @xmath890 the first equation implies that @xmath891 .
since @xmath892 preserves the 1-form @xmath854 , the vector - function @xmath893 in the second equation is identically equal to @xmath865
. therefore the second equation yields @xmath894 .
thus we see that @xmath895 if and only if @xmath846 . from here and it follows that @xmath896 for all @xmath895 .
recall that @xmath897 in particular , this implies that @xmath898 on a set of full measure in @xmath888 .
now , substituting into and removing lower order terms of @xmath226 , we obtain the required result .
let @xmath645 be the operator of multiplication by an arbitrary function @xmath646 such that @xmath899 .
then @xmath900 and , consequently , @xmath901 , where @xmath902 .
note that @xmath903 are linear combinations of fios lying in @xmath370 .
applying lemma [ l : fio - local ] to these fios and taking into account , we see that @xmath904 whenever @xmath905 for a sufficiently small @xmath494 .
assume , in addition , that the function @xmath497 is even and nonnegative , and @xmath906 .
then , since the functions @xmath907 are nondecreasing , standard tauberian theorems imply that @xmath908 ( see , for instance , ( * ? ? ?
* theorem 1.3 ) ) . clearly , the same estimate holds for the function @xmath909 . applying and integrating ,
we obtain @xmath910 if @xmath658 and @xmath911 then @xmath912 since @xmath913 , letting @xmath914 we see that the right hand side of coincides with @xmath915 in view of , @xmath916 on a set of full measure in @xmath523 , where @xmath225 is the phase function .
now implies that @xmath917 on a set of full measure , and follows from the weyl asymptotic formula for the counting function @xmath486 .
let @xmath728 and @xmath225 be arbitrary functions associated ( in the sense of subsection [ fio : def ] ) with the transformations and .
the condition * ( a@xmath256 ) * and imply that @xmath918 on a set of full measure in @xmath523 .
let @xmath338 and @xmath220 be arbitrary fios given by the integrals with the phase functions @xmath728 and @xmath225 , and let @xmath919 and @xmath310 be the leading homogeneous terms of corresponding amplitudes .
in view of and , we have @xmath920 on a set of full measure in @xmath523 .
in particular , this shows that the restrictions of @xmath310 and @xmath921 to this set do not depend on the choice of the phase functions .
assume now that the fios @xmath338 have sufficiently small conic supports , the phase functions @xmath728 are given by , and @xmath922 .
then , as was shown in subsection [ s : a2 ] , @xmath220 is represented by the integral with the phase function and an amplitude with leading homogeneous term @xmath923 . from here and it follows that @xmath924 on a set of full measure in @xmath523 .
the goal of this appendix is to provide an example of qe for a discontinuous riemannian metric .
we will first discuss generalities on ergodicity for markov semi - groups in order to get a simple test of ergodicity via the poincar maps associated to the singular manifold .
we will then present examples of metrics on the @xmath925sphere whose geodesic flows are ergodic .
finally , we will discuss thom - sard transversality in order to check that we have qe for a generic subset of these metrics .
we will denote by @xmath926 the unit cotangent bundle of @xmath927 and by @xmath928 the normalized liouville measure on @xmath926 .
we will denote by @xmath929 the semi - group denoted by @xmath930 in the body of the present paper ( see definition [ def : transfer - op ] ) it acts on @xmath931 and @xmath932 with norms equal to @xmath933 .
we call @xmath929 the `` geodesic flow '' even if it is not a ( deterministic ) flow on @xmath926 . the geodesic flow @xmath929 is ergodic if and only if the only functions in @xmath934 which are invariant under the semi - group @xmath935 are the functions which are constant outside a set of measure @xmath936 .
[ theo : ds ] if @xmath941 is a probability measure on a space @xmath926 and @xmath942 is a strongly measurable semi - group on @xmath943 whose norms on @xmath944 and @xmath945 are bounded by @xmath933 , then , for @xmath946 , the averages @xmath947 converge for almost all @xmath704 and in @xmath948 as @xmath949 .
the limit function @xmath950 is invariant under @xmath929 for all @xmath419 .
if @xmath929 is given by @xmath955 with @xmath956 a measurable family of probability measures on @xmath926 ( a markov kernel ) , the function @xmath957 given in theorem [ theo : ds ] is the conditional expectation of @xmath116 with respect to the @xmath958field @xmath959 of invariant sets : @xmath960 this result is proved in the book @xcite ( lemma 3.3 ) for markov chains .
let us show how to extend it to the continuous case : let us denote , for @xmath961 , by @xmath962 the function @xmath963 .
we have clearly @xmath964 hence , using the markov chain generated by @xmath965 , we deduce that @xmath957 is measurable with respect to @xmath966 , the @xmath958field of @xmath965-invariant sets . hence @xmath957 is measurable with respect to @xmath967 which is the intersection of the @xmath958fields @xmath966 .
the conclusion follows .
let us assume for simplicity that @xmath968 is smooth and cuts @xmath43 into two open disjoint parts @xmath969 .
let us denote by @xmath970 the unit ball bundles for @xmath971 in @xmath972 and let us parametrize the maximal geodesic arcs @xmath973\rightarrow x_+$ ] in @xmath565 crossing @xmath47 by a point in @xmath974 : if @xmath975 is the corresponding curve in @xmath926 and @xmath976 the final point of @xmath977 , we associate to @xmath304 the covector @xmath264 with @xmath978 and @xmath979 .
the point @xmath264 is the starting point of a maximal reflected arc @xmath980 whose associated final point corresponds in the way previously described to a point in the unit bundle @xmath981 , this point will be denoted by @xmath982 .
we get similarly a possible starting point for a refracted geodesic arc @xmath983 in @xmath566 by starting with a vector @xmath984 in @xmath985 whose restriction to @xmath986 is still @xmath987 .
this way we get a map @xmath988 defined on @xmath989 .
the poincar maps @xmath990 are not everywhere defined respectively on @xmath991 , but almost everywhere thanks to poincar recurrence theorem .
the first condition is clearly necessary , otherwise we would get at least two invariant sets of positive measure . let us assume that both conditions are satisfied and consider a set @xmath951 invariant under @xmath929 .
@xmath1005 is , up to a set of measure @xmath936 , a union of maximal geodesic arcs in @xmath1006 .
let us consider the trace @xmath246 of @xmath18 in @xmath47 .
clearly @xmath246 has to be invariant under @xmath994 and @xmath995 modulo sets of measure @xmath936 .
let @xmath1007 be the intersections of the images of @xmath246 by iterating @xmath994 and @xmath995 .
the measure of @xmath1008 vanishes .
the second condition implies that either @xmath1007 or @xmath1009 has measure @xmath936 .
the converse is easy .
let us consider two unit euclidian disks @xmath1010 and @xmath1011 and a diffeomorphism @xmath1012 so that @xmath1013 except for a finite number of values of @xmath672 .
gluing together @xmath1010 and @xmath1011 along their boundaries using @xmath1014 gives a topological manifold homeomorphic to @xmath1015 with a metric @xmath1016 which is flat outside the equator and discontinuous on the equator except at a finite number of points .
let us first describe the poincar section : as in section [ sec : poincare ] , we define @xmath1017 with the boundary @xmath1018 parametrized by the arc length @xmath672 .
then @xmath1019 with the symplectic structure @xmath1020 . using the map @xmath1014 and its extension @xmath1021 to the cotangent bundle of @xmath1011
, we get @xmath1022 - 1,+1[)$ ] or more explicitly @xmath1023 with @xmath1024 .
* @xmath1027 with @xmath1028 on which @xmath1003 acts as a rotation of angle @xmath1029 satisfying @xmath1030 * @xmath1031 with @xmath1032 on which @xmath988 acts as a rotation of angle @xmath1033 where @xmath1034 .
we have @xmath1035 where @xmath987 does not vanish .
this implies that the foliations @xmath1036 and @xmath1037 are transverse in @xmath1038 outside a finite number of segments @xmath1039 - \min ( 1,\psi(s ) ) , \min ( 1,\psi(s ) ) [ $ ] with @xmath1040 .
following the result of section [ sec : poincare ] , we have to consider a subset @xmath1044 of @xmath1045 which is invariant under @xmath994 and by @xmath995 , meaning that @xmath1046 and @xmath1047 .
we want to prove that @xmath246 or @xmath1004 has measure @xmath936 .
let @xmath1048 be the intersection of @xmath246 with the leaves @xmath1049 on which the rotation @xmath1050 is irrational
. then @xmath1051 ( i.e. @xmath1052 has measure @xmath936 ) and @xmath1053 is measurable and invariant under the rotation @xmath1054 .
hence the measure of @xmath1053 is @xmath1055 or @xmath1056 by the ergodicity of the irrational rotations of the circle . from this
we get that @xmath1057 is equivalent to a set foliated by @xmath1058 .
similarly @xmath1059 is equivalent to a set foliated by @xmath1060 .
let us now consider the set @xmath1061 with @xmath1062 . in @xmath1063 ,
both foliations are transverse .
this implies that @xmath1064 or @xmath1065 is of measure @xmath936 : using smooth coordinates @xmath1066 in @xmath1063 so that the two foliations correspond to @xmath1067 and @xmath1068 respectively , the indicator function of @xmath246 is equivalent to a function depending on @xmath96 only and to a function depending of @xmath141 only , hence is equivalent to a constant @xmath936 or @xmath933 .
if @xmath1069 is of measure @xmath936 , then @xmath1057 is of measure @xmath936 as being foliated by @xmath1070 , similarly for @xmath1071 .
the conclusion follows : all @xmath1064 are of measure @xmath936 or all @xmath1072 are of measure @xmath936 .
if we want to apply the main theorem of the paper , we have to take into account the fact that there are two ergodic components , they are equivalent by the involution @xmath1073 which on the quantum level is the complex conjugation @xmath1074 .
the semi - classical measures associated to real eigenfunctions are invariant under @xmath1075 .
we want to prove that qe holds for the manifold @xmath1076 for a generic @xmath1014 . what remains to do
is to check that @xmath1077 coincides with @xmath1078 for a generic @xmath1014 : this is a consequence of the fact that cauchy data of recombining geodesics have measure @xmath936 and will be proved in what follows to hold for a generic @xmath1014 .
we recall the let us consider a word @xmath1084 with @xmath1085 . for @xmath1086 and @xmath1087 a closed domain with a smooth boundary , so that @xmath1087 , let @xmath1088 be the manifold of all diffeomorphisms of class @xmath1089 of @xmath1090 so that @xmath1091 is defined in some open set @xmath220 containing @xmath60 .
let us denote by @xmath1092 the projection of @xmath974 onto @xmath1090 , by @xmath1093 the diagonal of @xmath1094 and , for @xmath1095 , by @xmath1096 the @xmath1097 map from @xmath1098 into @xmath1099 defined by @xmath1100 then the set of @xmath1101 s belonging to @xmath1088 so that @xmath1102 is open and dense in @xmath1103 .
the density follows from the transversality theorem as stated for example in @xcite and @xcite , page 48 ( see appendix [ app : transv ] )
. we will apply theorem [ theo : trans ] with @xmath1107 , @xmath1108 , @xmath1109 and @xmath1093 the diagonal of @xmath47 .
the transversal intersection of @xmath1110 with @xmath1093 implies that the set of @xmath704 for which @xmath1111 is a submanifold of dimension @xmath933 of @xmath981 .
let us consider the evaluation map @xmath1112 .
the differential @xmath1113 of @xmath1114 at a point @xmath1115 can be written as @xmath1116 . in order to prove transversality
it is enough to prove that @xmath1117 is surjective .
let us restrict ourselves to variations of @xmath1014 in some small neighborhood of @xmath1118 where @xmath1119 .
then we have @xmath1120 .
unique quantum ergodicity is not satisfied because there are infinitely many radial eigenfunctions corresponding to the neumann and the dirichlet problem for radial functions in the unit euclidian disk .
[ theo : trans ] let @xmath1126 and @xmath1127 and @xmath47 be @xmath1128 manifolds .
we assume that @xmath1129 is a banach manifold while @xmath1130 and @xmath1131 are finite .
the manifold @xmath43 is assumed to be compact with a smooth boundary .
consider a @xmath1128 map @xmath1132 and @xmath1133 a compact sub - manifold .
the evaluation map @xmath1134 is defined by @xmath1135 and we denote by @xmath1136 the set of the @xmath1137 s in @xmath1138 so that @xmath1139 . then if @xmath1140 , and @xmath1141 , then @xmath1142 is open and dense in @xmath1129 .
* acknowledgements . *
the authors would like to thank steve zelditch for useful discussions and comments .
d.j . and y.s .
thank the fields institute in toronto and all of the authors are grateful to the crm in montreal for the hospitality . y.cdv
thanks sbastien gouzel for illuminating discussions on ergodic theorems .
v. arnold .
_ on a characteristic class entering into conditions of quantization , _ funktsional .
i prilozhen .
* 1 * ( 1967 ) , no .
1 , 114 ( russian ) ; english translation functional anal . appl .
* 1 * ( 1967 ) , 113 .
n. dunford and j.t .
linear operators , part i : general theory .
interscience publishers , new - york _
p. grard and e. leichtnam .
_ ergodic properties of eigenfunctions for the dirichlet problem .
_ duke math . j.
* 71 * ( 1993 ) , no .
2 , 559607 . v. ivrii .
_ wave fronts of solutions of boundary - value problems for symmetric hyperbolic systems ii
. systems with characteristics of constant multiplicity . _ siberian math .
jour . * 21 * no . 4 ( 1980 ) , 527534 .
a. i. shnirelman . on the asymptotic properties of eigenfunctions in the regions of chaotic motion . in v.
lazutkin _ kam theory and semiclassical approximations to eigenfunctions .
_ ergebnisse der mathematik und ihrer grenzgebiete ( 3 ) , 24 .
springer - verlag , berlin , 1993 . | we analyze the semiclassical limit of spectral theory on manifolds whose metrics have jump - like discontinuities .
such systems are quite different from manifolds with smooth riemannian metrics because the semiclassical limit does not relate to a classical flow but rather to branching ( ray - splitting ) billiard dynamics . in order to describe this system
we introduce a dynamical system on the space of functions on phase space . to identify the quantum dynamics in the semiclassical limit we compute the principal symbols of the fourier integral operators associated to reflected and refracted geodesic rays and identify the relation between classical and quantum dynamics .
in particular we prove a quantum ergodicity theorem for discontinuous systems . in order to do this
we introduce a new notion of ergodicity for the ray - splitting dynamics . |
the concept of fame within a population has critical social and economic impact .
recently , the idea of using the number of _ hits _ returned from a search of a person s name on _ google _ as a means of quantifying that person s fame has been explored @xcite .
a seminal paper explored the fame of a unique population , that of world war i `` ace '' pilots @xcite , and found , among other things , a power - law decay in the tail of the distribution .
more recent work @xcite has applied this to a population of scientists who have published on the _ cond - mat _ e - print archive .
the tail of their fame distribution was best fit by an exponential .
on the other hand , the fame of other populations was found to follow a power - law decay .
the difference was attributed to the fact that scientists habitually use the world wide web as a professional means of communication and cite each other on the web in relation to their published work . _ google _ s goal as a service is to provide accurate search results to its users . for the purposes of determining a subject s fame ,
what is most relevant is not having accurate results listed first , as it is for most users , but to have an accurate _ count _ of those results .
unfortunately , _
google _ does not provide enough accuracy , and there are several reasons for this @xcite .
_ google _ acknowledges that the hits count given is an estimate , but does not elaborate on the accuracy of this estimation nor reveal how it is calculated .
it seems reasonable to assume that very small counts are more accurate than larger ones .
this means that the error is largest in the tail of the fame distribution , and it is this region that is of most interest . in addition , the tail of the distribution is more likely to contain results that are over - counted , further compounding the error . in @xcite , over - counting
was prevented by verifying each hit by hand , a time - consuming procedure that limited the sample size . at the time of this writing , _ google _
only returns the first 1000 hits , so it is impossible to verify the accuracy of any results beyond that number , and one must trust in _
google s _ estimation .
even manual verification is limited .
the previous searches in @xcite used a search lexicon including the boolean ` or ` operator .
we have since found out that google returns incorrect hit counts when ` or ` is used @xcite . for a simple illustration ,
a search for ` cars or automobiles ` returns 80.5 million hits ( at the time of this writing ) while searches for ` cars ` and ` automobiles ` return 94.2 million and 8.82 million hits , respectively , violating basic set theory .
thus , the previous work must be reproduced using a better lexicon . in the current work ,
all our searches avoid the problematic ` or ` operator .
see table [ table : results ] . despite these issues ,
_ google _ still provides an excellent tool for research .
it is the simplest means of getting the most information available and it commands a very large sample space . for example , the work in @xcite uses hit counts to `` teach '' the semantic meaning of words to software a central problem in artificial intelligence .
related words such as ` painter ' and ` artist ' will have many more joint occurrences than disparate words , such as ` plumber ' and ` artist ' , leading to higher hit counts .
their work confirms that _ google _ yields reasonable results when avoiding the ` or ` and using the ` and ` operator only .
_ google _ has been generous enough to open their search interface to allow tools to be created that can perform _ google _ searches automatically .
we have used this to eliminate the laborious task of entering single queries and recording the hits count .
larger populations can be searched much more quickly using an automatic tool . for the present work
, we used a script that performed the searches from a web server .
an easier way still is with the open - source _ pygoogle _ package , which integrates the _ google _ search interface with the python programming language . in both @xcite and @xcite ,
fits were performed by binning the search results with exponentially - sized bins and then fitting to the binned data using least - squares . a better technique than binning ,
when working with sparse data , is examining cumulative distributions @xcite , as we do in the present work .
also , it has been shown that there are problems with using least - squares fits to logarithmic plots @xcite .
one problem is that the log operation magnifies the error in the tail .
least - squares fitting assumes that errors for each data point are uniform and will not properly weigh the noisier tail . in this work , we use a more robust technique , that of maximum likelihood , to achieve less biased fits .
see section [ section : maxlike ] .
all distributions studied here exhibit a power - law tail , although for many the tail covers a very narrow range .
for the scientists populations , we observe a power - law tail only in the top 12% of the data .
most of the data for scientists is best fit by an exponential , just as found in the previous study @xcite .
in contrast , other populations do not fit to an exponential over any sizable range .
see section [ results ] .
a better technique to determine the parameter(s ) of a probability distribution from sampled data is that of maximum likelihood .
the results are more robust in terms of error - weighing .
this is a very common technique and is covered in many statistics and regression texts . to briefly illustrate how maximum likelihood works ,
let us derive the _ maximum likelihood estimator _ ( mle ) for @xmath0 , the parameter for an exponential probability distribution : @xmath1 the goal of maximum likelihood is to find the _ most likely _
@xmath0 given the existing data .
for this , we start with the probability of the experiment given @xmath0 , assuming independence of the data points : @xmath2 where @xmath3 is the unbinned data gathered from the experiment .
this function is the total probability of all measurements occurring in the experiment . from this , we define the _ likelihood function _ , using bayes theorem : @xmath4 this is the _ likelihood _ of @xmath0 given the experimental data . assuming @xmath5 ( the experiment has already occurred ) and @xmath6 is uniformly distributed ( all @xmath0 s are equally likely ) , then @xmath7 . to find the most likely @xmath0 , we must find the maximum of this function with respect to the parameter @xmath0 . to simplify the calculation
, we will instead maximize the log - likelihood function , @xmath8 , which is equivalent : @xmath9 this is just the inverse of the mean , exactly as expected for an exponential distribution .
we need not account for the proportionality between @xmath10 and @xmath11 because we only used the derivative of @xmath8 the other probability distribution we are concerned with is the power - law distribution : @xmath12 for the mle of @xmath13 , we reproduce the derivation given in @xcite .
the first step is to normalize eqn .
( [ powerlaw ] ) for the given data points : @xmath14 where @xmath15 is a constant of proportionality and @xmath16 is the smallest data point from the given sample .
this is then used to get the probability of the experiment : @xmath17 this is proportional to @xmath18 as before : @xmath19 again , we work with @xmath20 , which is equivalent for finding the most likely @xmath13 . then : @xmath21 the mle for @xmath13 can then be found : @xmath22 maximum likelihood derives an estimator for a distribution s parameter(s ) , regardless of whether the sampled data truly does come from such a distribution .
hence , one needs a way to test how well the estimator matches the sample . for our purposes , the kolmogorov - smirnov ( ks ) test works
quite well @xcite .
this test compares the cumulative distribution function ( cdf ) of the hypothesized probability distribution to the empirical cdf of the sampled data .
the test statistic is : @xmath23 where @xmath24 is the hypothesized cdf and s(x ) is the empirical cdf .
@xmath25 is then compared with a critical value ( for the given significance level ) which can be found in a table or generated by software .
matlab s statistics toolbox has a built - in ks - test function , ` kstest ( ) ` .
we have been able to greatly expand upon the number of searches performed compared to previous work .
in addition , due to the problems with the ` or ` operator , we have performed multiple searches of the same population using progressively inclusive lexicons . here
we describe the populations studied .
scientists : : : two populations of scientists were used in this study .
the smaller one ( of size 449 ) is the same population used in @xcite .
the larger population ( of size 1625 ) is a list of authors who have published recently on cond - mat and was harvested using arxiv s oai xml feed . aces : : : the population of 1851 aces contains the 393 german aces studied in @xcite as well as all the listed aces of other nationalities .
actors : : : the actors population contains 778 actors who were born on the second or third of each month between the years 1950 and 1955 , as collected from the archives of the internet movie database
. these selection criteria were chosen to insure a mostly uniform sample and to give all the chosen subjects roughly the same career length .
villains : : : the villains population was gathered from a user - contributed list of antagonists from fictional media . this list contains both fictitious characters and real people who have appeared in fictional works . since this list was generated by users , the characters must already enjoy a substantial level of popularity .
programmers : : : similar to the villains population , this population was collected from a user - contributed list of famous programmers ; people who have made a large contribution to computing , the internet , etc .
, such as tim berners - lee , who invented the world wide web , and bill gates , a co - founder of microsoft . as with the villains
, it seems safe to assume that this population is `` famous '' .
clarkson students : : : the students population was chosen from clarkson university s student directory .
it consists of all students ( undergraduate and graduate ) whose last name contains the letter `` e '' .
this criterion was chosen simply to make it easy to harvest a large collection of names from the online student directory .
we assume this is a `` non - famous '' population , in that the students are too young to have amassed any real fame .
runners : : : this population was used previously @xcite .
the original searches used the erroneous ` or ` operator and are here reproduced without it .
table [ table : results ] contains the power - law exponents and search lexicons for the populations studied .
many of the power - law exponents are @xmath26 , as first predicted in @xcite .
all populations display a power - law tail , regardless of whether they are `` famous '' or not .
it should be pointed out , however , that for some populations the range fitting a power - law is extremely narrow , casting doubt on this interpretation . in those cases ,
an exponential distribution may fit as well .
most of the scientists distributions fit an exponential over much of the `` non - tail '' .
see table [ table : resultssci ] .
clarkson students , another population assumed to be non - famous , does not fit to an exponential over such a range .
this is further evidence that the exponential distribution for scientists stems from their use of the world wide web as a professional means for disseminating research , rather than related to fame .
.mle power - law fits to search results .
all fits pass the ks - test ( @xmath27 ) .
[ cols=">,^,^,^ , < " , ] the power - law exponent tends to increase as the restriction due to the lexicon increases .
this is expected because a more restrictive search will make high hit counts less frequent , increasing the slope of the tail .
figure [ figure : ranks ] contains _ rank / frequency _ plots for several populations to illustrate this effect .
the plots are proportional to the empirical cdf , @xmath28 .
note that individual searches which return zero hits are not shown , changing the maximum rank between lexicons .
this is most evident in the students population : the third lexicon is very restrictive and many students garnered zero hits .
the proposed model in @xcite was shown to have a power - law exponent that approaches 2 asymptotically , from _ above _ , as the number of relevant web pages citing the population in question increases over time .
the population changes in size in table [ table : results ] are due to progressively restrictive lexicons and do not pertain to the same phenomenon . on the other hand , we are unable to account for the many instances of power - law exponents smaller than 2 observed , as any reasonable extension of the theory in @xcite yields powers @xmath29 .
a purely visual inspection of plots such as those in figure [ figure : ranks ] may lead one to conclude that a search is exponentially or power - law distributed , but this is misleading and subjective .
the eye will overweigh the number of data points in the tail , due to the logarithmic axes .
objective hypothesis tests such as the ks - test must be used .
in addition to problems with hit estimation , ` or ` , etc . , the choice of a lexicon has noticeable impact . in the rank / frequency plot for the aces population in figure [ figure : ranks ] , the second search shows a much cleaner tail , though again this region contains less than 6% of the aces .
all of these factors make it difficult to test theories . for the size of populations involved , _
google _ hits have too much `` noise '' to accurately distinguish distributions
. we are grateful to eduard vives for suggesting the analysis of cumulative plots by the method of maximum likelihood and to the universidad de granada for facilitating discussions leading to this work .
dba thanks nsf grant phy-0140094 for partial support of this work . | we study the fame distribution of scientists and other social groups as measured by the number of _ google _ hits garnered by individuals in the population .
past studies have found that the fame distribution decays either in power - law @xcite or exponential @xcite fashion , depending on whether individuals in the social group in question enjoy true fame or not . in our present study
we examine critically _ google _ counts as well as the methods of data analysis .
while the previous findings are corroborated in our present study , we find that , in most situations , the data available does not allow for sharp conclusions .
address = department of physics , clarkson university , potsdam ny 13699 - 5820 address = department of physics , clarkson university , potsdam ny 13699 - 5820 |
the cost for a superconducting linear collider can be significantly reduced by minimizing the number of microwave components , and increasing the fill factor in a machine . here
the fill factor is meant as a ratio of the active cavity length to the total cavity length ( active length plus interconnection ) .
these two conditions become partially fulfilled when the number of cells ( @xmath0 ) in a structure -fed by one fundamental mode ( fm ) coupler- increases .
unfortunately there are two limitations on the cell s number in one accelerating structure : firstly the field flatness -the sensitivity of the field pattern increases proportional to @xmath1- and secondly trapped higher order modes ( hom ) . in order to overcome these limitations on @xmath0 ,
the concept of the superstructure has been proposed for the tesla main linac @xcite . in this concept four 7-cell cavities ( sub - units )
are coupled by short beam tubes .
the whole chain can be fed by one fm coupler attached at one end beam tube .
the length of the interconnections between the cavities is chosen to be half of the wave length .
therefore the @xmath2 - 0 mode ( @xmath2 cell - to - cell phase advance and 0 cavity - to - cavity phase advance ) can be used for acceleration . in the proposed scheme hom couplers
can be attached to interconnections and to end beam tubes .
all sub - units are equipped with a tuner .
accordingly the field flatness and the hom damping can be still handled at the 7-cell sub - unit level .
the energy flow through cell - interconnections and the resulting bunch - to - bunch energy spread has been extensively studied for the superstructure with two independent codes : homdyn @xcite and mafia @xcite@xcite .
negligible spread in the energy gain , smaller than @xmath3 for the whole train of 2820 bunches , proofs that energy flow is big enough to re - fill cells in the time between two sequential bunches ; see fig .
[ spread ] .
the energy spread results from the interference of the accelerating mode with other modes from the fm passband .
the difference in energy becomes smaller at the end of the pulse due to the decay of the interfering modes .
the @xmath2 - 0 mode will be used for the acceleration of beam in the superstructure . before assembly
, each of the four 7-cell cavities will be pre - tuned for flat field profile and the chosen frequency of the @xmath2 - 0 mode .
the pre - tuning procedure is based on measurements of all modes of the fundamental mode passband .
it allows to adjust the profile with accuracy of better than 2 - 3 @xmath4 for a 9-cell tesla cavity .
this error corresponds to a frequency accuracy of the individual cells of @xmath5 30 khz .
after the cavity chain of a superstructure has been assembled and is operated in the linac at 2k , the frequency of each sub - unit can be corrected in order to equalize the mean value of the field amplitude in all sub - units ( not between cells within one sub - unit ) .
this field profile correction is possible during the linac operation , since each 7-cell structure is equipped with its own frequency tuner .
the method proposed to equalize the average accelerating field of sub - units during operation is based on perturbation theory , similar to the standard bead - pull method of l. maier and j. slater @xcite . at first , present fields of all sub - units are measured . for that , successively , the volume of each sub - unit is changed by the same amount ( stepping motor of each tuner will be moved by the same number of steps ) to measure the frequency change of the @xmath2 - 0 mode .
the change is proportional to the stored energy in the sub - unit of the superstructure . for each sub - unit
relative values can be defined and used to calculate frequency corrections needed to equalize the field .
this method has been tested on a room temperature cu model of the superstructure -see fig .
[ vergleich]- and by computer simulations , see fig .
[ tuning ] .
one should note that the method requires only one pickup probe for all 28 cells , and therefore effectively reduces the numbers of cables , feedthroughs and electronics needed for the control .
30 khz is assumed.,width=245 ]
as discussed above the field flatness in a cold superstructure can be handled at the 7-cell sub - unit level by adjusting the frequency of each sub - unit . in order to verify this ,
the field flatness in a superstructure has been calculated before and after tuning of the individual cavities .
the frequencies of the cavities have been corrected accordingly to the proposed tuning method . for the frequency of the individual cells a variation of @xmath5 30 khz
is assumed , based on the experience with the tesla 9-cell cavities .
the statistics of 10000 calculated field profiles is shown in fig .
[ statistic ] . by adjusting the frequencies of the individual cavities the field unflatness
is significantly reduced .
30 khz.,width=245 ]
the vertical normalized multibunch emittance at the interaction point of the tesla collider is desired to be @xmath6 m@xmath7rad .
simulations of the emittance growth along the tesla linac showed , that the dipole modes with dominating impedance ( r / q ) should for that be damped to the level of @xmath8 @xcite .
the interconnecting tubes of the superstructure allow to put hom couplers between the 7-cell cavities .
measurements on a cu model of the superstructure at room temperature have demonstrated , that the required damping can be achieved with five hom couplers : three attached at the interconnections and one at both ends @xcite ; see fig .
note that the sum of all listed dipole modes impedances is almost ten times smaller than the bbu limit .
a first `` proof of principle '' niobium prototype of the superstructure is under construction @xcite .
the sub - units are under fabrication and will be vertical tested similar to tesla 9-cell cavities .
the beam test for the prototype is scheduled for spring 2001 . it will allow to verify energy spread computations and rf measurements on the room temperature models .
this will include the test of the hom damping , the performance of the hom couplers at higher magnetic field and the tuning method during operation at 2k .
the presented measurements and calculations demonstrate , that in the proposed superstructure the refilling of cells , the hom damping , the field flatness and the field flatness tuning can be handled . for the final prove , that the superstructure layout can be used for acceleration
, a niobium prototype will be tested with beam at the tesla test facility linac .
this work has benefited greatly from discussions with the members of the tesla collaboration . | a new cavity - chain layout has been proposed for the main linac of the tesla linear collider @xcite .
this superstructure - layout is based upon four 7-cell superconducting standing - wave cavities , coupled by short beam pipes .
the main advantages of the superstructure are an increase in the active accelerating length in tesla and a saving in rf components , especially power couplers , as compared to the present 9-cell cavities .
the proposed scheme allows to handle the field - flatness tuning and the hom damping at sub - unit level , in contrast to standard multi - cell cavities .
the superstructure - layout is extensively studied at desy since 1999 .
computations have been performed for the rf properties of the cavity - chain , the bunch - to - bunch energy spread and multibunch dynamics .
a copper model of the superstructure has been built in order to compare with the simulations and for testing the field - profile tuning and the hom damping scheme .
a `` proof of principle '' niobium prototype of the superstructure is now under construction and will be tested with beam at the tesla test facility in 2001 . in this paper we present
latest results of these investigations . |
as a promising candidate material for nanoelectronic device , graphene has been attracted intensive attention in research in past years ( for review , see e.g. ref . ) .
it demonstrates not only peculiar electronic properties@xcite , but also very high ( as high as 5000 wm@xmath3k@xmath3 ) thermal conductivity,@xcite which is beneficial for the possible electronic and thermal device applications of graphene.@xcite a recent work has studied the structure of an interesting new allotrope of graphene by a first principle calculation.@xcite the ground state energy in this allotrope is about 0.28 ev / atom above graphene , which is 0.11 ev / atom lower than c@xmath4 . in this paper
, we calculate the ballistic phonon thermal conductance for the graphene sheet by treating the graphene as the large width limit of graphene strips , which can be described by a lattice vector @xmath5.@xcite the phonon dispersion of the graphene is obtained in the valence force field model ( vffm ) , where the out - of - plane acoustic phonon mode is a flexure mode , i.e. , it has the quadratic dispersion around @xmath6 point in the brillouin zone.@xcite our result shows that the thermal conductance has a @xmath7 dependence at low temperature , which is due to the contribution of the flexure mode.@xcite at room temperature , our result are comparable with the recent experimental measured thermal conductivity.@xcite we find that the thermal conductance in graphene depends on the direction angle @xmath0 of the thermal flux periodically with @xmath1 as the period .
the difference between maximum and minimum thermal conductance at 100 k is 1.24@xmath8wm@xmath9k@xmath3 , which is about 1@xmath10 variation .
our study shows that this directional dependence for the graphene is attributed to the directional dependence of the velocities of the phonon modes , which origins from the @xmath2 symmetry of the honeycomb structure . for the dimerite , where the @xmath2 symmetry is broken
, the thermal conductance shows more obvious anisotropy of 10@xmath10 and the value is about 40 @xmath10 smaller than that of the graphene at room temperature .
the present paper is organized as follows . in sec.ii
, we describe the graphene strip by a lattice vector .
the formulas we used in the calculation of the thermal conductance are derived in sec .
calculation results for graphene and dimerite are discussed in sec .
iv a and sec . iv b , respectively .
v is the conclusion .
[ 0.8 ] . ( a ) .
strip with @xmath11 and ( @xmath12 , @xmath13)=(12 , 1 ) ; ( b ) .
the brillouin zone for this special strip is two discrete segments ( solid ) in the reciprocal space.,title="fig:",width=264 ]
in graphene , the primitive lattice vectors are @xmath14 and @xmath15 , with @xmath16 .
@xmath17 is the c - c bond length in graphene.@xcite the corresponding reciprocal unit vectors are @xmath18 , @xmath19 . as shown in fig .
[ fig_strip ] ( a ) , a strip in the graphene sheet can be described by a lattice vector @xmath5 .
the real lattice vector @xmath20 is introduced through@xcite : @xmath21 ( @xmath22 is the greatest common divisor of @xmath23 and @xmath24 ) .
the strip is denoted by @xmath25 , where @xmath12 and @xmath13 are numbers of the periods in the directions along @xmath26 and @xmath27 , respectively . instead of @xmath14 and @xmath15 ,
we use ( @xmath26 , @xmath28 ) as the basic vectors in the following , and @xmath29 and @xmath30 are their corresponding reciprocal unit vectors : @xmath31 any wave vector in the reciprocal space can be written as : @xmath32 using the periodic boundary conditions , this strip has @xmath12 translational periods in the @xmath26 direction and @xmath33 translational periods in the @xmath27 direction .
as shown in fig .
[ fig_strip ] ( b ) , the brillouin zone for the graphene strip is @xmath33 discrete segments , which are parallel or coincide with @xmath29 .
the coordinates for the wave vectors on these lines are@xcite ( @xmath34 , @xmath35)=(@xmath36 , @xmath37 ) , with @xmath38 and @xmath39 .
the graphene sheet is actually a strip in the limit of @xmath40 and @xmath41 . in this case , the brillouin zone for the strip , i.e. , @xmath33 discrete lines , turns to the two - dimensional brillouin zone for the graphene .
the contribution of the phonon to the thermal conductance in the ballistic region is:@xcite @xmath42 where @xmath43 is the bose - einstein distribution function .
@xmath44 is transmission function . in the ballistic region , @xmath44 is simply the number of phonon branches at frequency @xmath45 . from the above expression , the thermal conductance in the graphene strip can be written as : @xmath46 where @xmath0 determines the direction of the thermal flux : @xmath47 .
@xmath48 is the wave vector in the brillouin zone of the strip , i.e. , on the @xmath33 discrete lines .
the transmission function for a phonon mode @xmath49 is assumed to be one.@xcite @xmath50 is the group velocity of mode @xmath51 in @xmath52 direction .
the value of the group velocity can be accurately calculated through the frequency and the eigen vector of this phonon mode:@xcite @xmath53 where @xmath54 is the dynamical matrix and @xmath55 is the eigen vector . only those phonon modes with @xmath56 contribute to the thermal conductance in the @xmath52 direction . in the two - dimensional graphene strip system , it is convenient to use conductance reduced by cross section : @xmath57 , where @xmath58 is the cross section .
the thickness of the strip , @xmath59 , is chosen arbitrarily to be the same as the space between two adjacent layers in the graphite .
the width for the strip is @xmath60 , where , the thermal flux in the strip is set to be in the direction perpendicular to @xmath27 , i.e. , @xmath61 . _
we address a fact that the integral parameter ( @xmath34 ) in eq .
( [ eq_conductance_strip ] ) is the quantum number along the thermal flux direction . _
the thermal conductance in @xmath52 direction of the graphene can be obtained by : @xmath62
the phonon spectrum of graphene is calculated in the vffm , which has been successfully applied to study the phonon spectrum in the single - walled carbon nanotubes@xcite and multi - layered graphene systems.@xcite in present calculation , we utilize three vibrational potential energy terms .
they are the in - plane bond stretching ( @xmath63 ) and bond bending ( @xmath64 ) , and the out - of - plane bond bending ( @xmath65 ) vibrational potential energy .
the three force constants are taken from ref .
as : @xmath66=305.0 nm@xmath3 , @xmath67=65.3 nm@xmath3 and @xmath67=14.8 nm@xmath3 . [ 1.2]@xmath68150 ) the thermal conductance of the strip can be considered as the thermal conductance value of graphene.,title="fig:",width=264 ] [ 1.2 ] v.s .
@xmath69 in extremely low temperature region .
the calculated results ( filled squares ) can be fitted by function @xmath70 ( blue line ) .
it indicates that the thermal conductance has a @xmath7 dependence in this region.,title="fig:",width=264 ] [ cols="<,^,^,^,^,^,^",options="header " , ] in fig .
[ fig_width ] , the temperature is 100 k and the direction angle for the thermal flux is @xmath71 .
it is shown that the thermal conductance for a strip decreases with increasing width . at about @xmath72=100 ,
the thermal conductance reaches a saturate value , which is actually the thermal conductance for the graphene . in the calculation ,
the width we used is about 300 , which ensures that the strip is wide enough to be considered as a graphene sheet . in fig .
[ fig_tem ] , the thermal conductance versus the temperature is displayed . in the low temperature region , the thermal conductance has a @xmath7 dependence .
this is the result of the flexure mode in the graphene sheet , which has the dispersion @xmath73 . in the very low temperature region
, this mode makes the largest contribution to the thermal conductance .
its contribution to the thermal conductance is@xcite @xmath74 , which can be seen from the figure in the low temperature region . at room temperature
@xmath75=300 k , the value for the thermal conductance is about @xmath76 wm@xmath9k@xmath3 .
this result agrees with the recent experimental value for the thermal conductance in the graphene.@xcite in the experiment , the thermal conductivity is measured to be about @xmath77 wm@xmath3k@xmath3 at room temperature .
the distance for the thermal flux to transport in the experiment is @xmath78=11.5 @xmath79 m . so the reduced thermal conductance can be deduced from this experiment as @xmath80 wm@xmath9k@xmath3 .
our theoretical result is much larger than this experimental value . because our calculation is in the ballistic region , while in the experiment
, there is scattering on defects , edges or impurities and thus the transport is partially diffusive . at the high temperature limit @xmath81
, our calculation gives the value 8.9 @xmath82 wm@xmath9k@xmath3 , which is in consistency with the previous theoretical result.@xcite as shown in fig .
[ fig_theta ] , at @xmath75=100 k , the thermal conductance varies periodically with the direction angle @xmath0 .
the calculated results can be fitted very well by the function @xmath83 .
the difference between the thermal conductance in the two directions with angle @xmath84 and @xmath85/2 is about 1.2@xmath86wm@xmath9k@xmath3 .
this difference is very stable for graphene strips with different width ( see fig .
[ fig_width_theta ] ) . at @xmath75=100 k ,
the lattice thermal conductance is about two orders larger than the electron thermal conductance.@xcite so the experimental measured thermal conductance at @xmath75=100 k is mainly due to the contribution of the phonons . as a result
, our calculated directional dependence of the lattice thermal conductance in the graphene can be carefully investigated in the experiment . in the following
, we say that two quantities @xmath87 and @xmath88 have the same ( opposite ) dependence on @xmath0 , if the signs of @xmath89 and @xmath90 are the same ( opposite ) .
[ 1.2 ] is the direction angle for the thermal flux .
the calculated results ( filled squares ) are fitted by the function @xmath91 with @xmath92 , @xmath93 and @xmath94.,title="fig:",width=264 ] [ 1.2 ] and @xmath95 directions versus width of graphene strip .
lines are guide to the eye.,title="fig:",width=264 ] [ 1.2 ] point in the brillouin zone : ( a ) . the coefficient of the out - of - plane acoustic mode with @xmath73 in the unit of @xmath96 m@xmath97s@xmath3 ; ( b)-(f ) .
velocities of the other five phonon spectrum ( from low frequency to high frequency ) , in the unit of @xmath98 .
the horizontal axises in all figures are the direction angle @xmath0 in the unit of @xmath85.,title="fig:",width=264 ] [ 1.2 ] and @xmath95 versus temperature .
this quantity has an abundant temperature dependence .
inset is the enlarged figure for the low temperature region.,title="fig:",width=264 ] to find the underlying mechanism for this directional dependence for the thermal conductance , firstly we show in fig .
[ fig_velocity ] the coefficient @xmath99 for the flexure mode and the velocities for the other five phonon modes at the @xmath6 point .
interestingly , this coefficient and velocities are also directional dependent with the period @xmath1 .
obviously , they can be fitted by function @xmath100 . in table
[ tab_velocity ] , the sign of the fitting parameter @xmath101 for this coefficient and five velocities are listed , which can be read from fig .
[ fig_velocity ] . in the third line of table [ tab_velocity ]
, we list the contribution of the six phonon modes to the thermal conductance . if the three low frequency modes are excited , the thermal conductance is in inverse proportion to their velocities.@xcite while as can be seen from eq .
( [ eq_cdt ] ) , the thermal conductance is proportional to the velocities for the three high frequency optically modes when they are excited . in each temperature region , there will be a key mode which is the most important contributor to the thermal conductance .
the direction dependence of the velocity of this key mode determines the direction dependence of the thermal conductance .
we then further study the difference between the thermal conductance in two directions with @xmath84 and @xmath102 : @xmath103 . in the fourth line of table [ tab_velocity ]
, we display the effect of different modes on @xmath104 .
it shows that @xmath104 will decrease , if the first , second and fifth phonon modes are excited sufficiently with increasing temperature .
the other three phonon modes have the opposite effect on the thermal conductance .
the dependence of @xmath105 on the temperature is shown in fig .
[ fig_anisotropic ] , where five different temperature regions are exhibited .
\(1 ) [ 0 , 4]k : in this extremely low temperature region , only the flexure mode is excited .
this mode results in @xmath106 . because the coefficient @xmath99 depends on the direction angle @xmath0 very slightly ,
the absolute value of @xmath105 is pretty small ( see inset of fig . [ fig_anisotropic ] ) .
\(2 ) [ 4 , 10]k : the second acoustic mode is excited in this temperature region . in respect that this mode has more sensitive direction dependence and favors to decrease @xmath105 , @xmath105 decreases much faster than region ( 1 ) .
\(3 ) [ 10 , 70]k : in this temperature region , the third acoustic mode begins to have an effect on the thermal conductance .
this mode s directional dependence is opposite of the previous two acoustic modes and it will increase @xmath105 .
the competition between this mode and the other two acoustic modes slow down the decrease of the value @xmath105 at temperature below @xmath75=40k .
the third acoustic mode becomes more and more important with temperature increasing , and @xmath105 begins to increase after @xmath75=40k as can be seen from the inset of fig .
[ fig_anisotropic ] .
\(4 ) [ 70 , 500]k : the third acoustic mode becomes the key mode in this temperature region . as a result ,
@xmath105 changes into a positive value and keeps increasing .
\(5 ) [ 500 , 1000]k : in this high temperature region , the optical mode will also be excited one by one in the frequency order with increasing temperature .
since there are two optical modes ( 1st and 3rd optical modes ) favors to increase @xmath105 , while only one optical mode ( 2nd optical mode ) try to decrease @xmath105 , the competition result is increasing of @xmath105 in the high temperature region .
@xmath75=100k is in region ( 4 ) , where the direction dependence of the thermal conductance is controlled by the velocity of the third mode , so the dependence of @xmath107 on @xmath0 in fig .
[ fig_theta ] is opposite to the dependence of velocity @xmath108 in fig .
[ fig_velocity ] ( c ) .
( b ) shows 49 unit cells together.,title="fig:",width=264 ] the adatom defect is used as a basic block to manufacture a new carbon allotrope of graphene , named _
dimerite_,@xcite and the relaxed configuration of this new material is investigated by a first principle calculation .
the unit cell for the dimerite is shown in fig .
[ fig_cfg ] ( a ) , where the two basic unit vectors are @xmath14 and @xmath15 , with @xmath109 , and @xmath110 .
the angle between these two vectors is 0.7@xmath85 . in each unit cell
, there is a ( 7 - 5 - 5 - 7 ) defect , which leads to two anisotropic directions : the ( 7 - 7 ) direction and the ( 5 - 5 ) direction .
( 7 - 7 ) direction is from the center of a heptagon to the opposite heptagon , and the ( 5 - 5 ) direction is from the center of a pentagon to the neighboring pentagon .
these two directions are perpendicular to each other , and corresponding to @xmath111 , 0.9@xmath85 respectively in fig .
[ fig_cfg ] ( a ) , where @xmath0 is the direction angle with respect to @xmath112 axis . from graphene to dimerite , the symmetry is reduced from @xmath2 to @xmath113 . [
1.2 ] point , there are low frequency optical modes with frequency about 200 cm@xmath3 , which do not exist in pure graphene.,title="fig:",width=264 ] the phonon dispersion for the dimerite from the above vffm is shown in fig . [ fig_phonon ] .
similar with pure graphene , there are three modes with zero frequency .
two of them are the acoustic modes in the @xmath115 plane .
the other zero - frequency mode is a flexure mode with parabolic dispersion of @xmath116 .
this flexure mode corresponds to the vibration in the @xmath117 direction .
different from the pure graphene , there are a lot of optical phonon modes with frequency around 150 cm@xmath3 in the dimerite .
[ 1.2 ] smaller than the graphene at room temperature.,title="fig:",width=264 ] [ 1.2 ] with @xmath118=1.9@xmath82 , @xmath101=1.1@xmath119 , and @xmath120= 0.1@xmath85 ( blue line).,title="fig:",width=264 ] in fig .
[ fig_tem_dimerite ] , we compare the thermal conductance of the dimerite with the graphene . at room temperature ,
the thermal conductance of the dimerite is about 40@xmath10 smaller than the graphene . as can be seen from fig .
[ fig_theta_dimerite ] , the thermal conductance in the dimerite is more anisotropic than graphene . the calculated results can be best fitted with function @xmath121 with a=1.9@xmath82 , b=1.1@xmath119 , and c= 0.1@xmath85 .
the thermal conductance has a minimum value at @xmath111 , and maximum value at @xmath122 .
as mentioned previously , the directions with @xmath111 , 0.9@xmath85 are the ( 7 - 7 ) and the ( 5 - 5 ) directions , respectively .
so , the thermal conductance in ( 5 - 5 ) direction is 12@xmath10 larger than ( 7 - 7 ) direction , which is about one order larger than that of the pristine graphene .
we expect this anisotropic effect detectable experimentally if the precision in the current experiment can be improved .
in conclusion , we have calculated the phonon thermal conductance for graphene in the ballistic region , by considering the graphene as the large width limit of graphene strips .
the calculated value for the thermal conductance at room temperature is comparable with the recent experimental results , while at high temperature region our results are consistent with the previous theoretical calculations .
we have found that the thermal conductance is directionally dependent and the reason is the directional dependence of the velocities of different phonon modes , which can be excited in the frequency order with increasing temperature . by breaking the @xmath2 symmetry in graphene
, we can see more obvious anisotropic effect of the thermal conductance as demonstrated by dimerite .
we have following two further remarks : ( 1 ) .
since the anisotropic effect of thermal conductance in graphene is small ( 1@xmath10 ) , it requires high accuracy in the calculation of phonon mode s group velocity to see this anisotropic effect .
thanks to the superiority of the vffm , we can derive an analytic expression for the dynamical matrix and calculate accurately the value of the group velocity following eq .
( [ eq_velocity ] ) .
thus we can obtain the 1@xmath10 anisotropic effect in graphene as discussed in this manuscript .
we also used the brenner empirical potential implemented in the general utility lattice program " ( gulp)@xcite to calculate the phonon dispersion and group velocity in graphene . for lack of analytic expression for the dynamical matrix , we find that the accuracy is not high enough for the group velocity and it is difficulty to see this anisotropic effect .
( 2 ) . from symmetry analysis@xcite ,
when thermal transport is in the diffusive region where the fourier s law exists , the @xmath2 symmetry of graphene constrains the thermal conductivity to be a constant value .
so the anisotropic effect in graphene can not be seen if the thermal transport is in the diffusive region , yet it can only be seen in the ballistic region as discussed in this manuscript .
but the situation changes in dimerite , where the @xmath2 symmetry is broken into @xmath113 .
the @xmath113 symmetry does not constrain the thermal conductivity to be a constant value even the fourier s law is valid .
so we can expect to see the isotropic effect in the dimerite both in the ballistic and diffusive region .
we thank lifa zhang for helpful discussions . the work is supported by a faculty research grant of r-144 - 000 - 173 - 112/101 of nus , and grant r-144 - 000 - 203 - 112 from ministry of education of republic of singapore , and grant r-144 - 000 - 222 - 646 from nus . _ note added in proof_. we have learned about a recent theoretical study on the thermal conduction in single layer graphene .
et al_.@xcite investigated the effects of umklapp , defects and edges scattering on the phonon thermal conduction in graphene .
our result for the thermal conductivity is larger than their value .
this is because our study is in the ballistic region , which gives an upper limit for their calculation results . | we investigate the phonon thermal conductance of graphene regarding the graphene sheet as the large - width limit of graphene strips in the ballistic limit .
we find that the thermal conductance depends weakly on the direction angle @xmath0 of the thermal flux periodically with period @xmath1 .
it is further shown that the nature of this directional dependence is the directional dependence of group velocities of the phonon modes in the graphene , originating from the @xmath2 symmetry in the honeycomb structure . by breaking the @xmath2 symmetry in graphene
, we see more obvious anisotropic effect in the thermal conductance as demonstrated by dimerite . |
the graduate record exam ( gre ) subject test in physics currently plays a significant role in graduate admissions for u.s .
astronomy departments . out of the 20 top - ranked astronomy graduate programs in the u.s .
( national research council rankings , 2010 ; excluding physics departments ) , 100% of them required the physics gre ( pgre ) as part of a complete application to their graduate program in 2015 . while the role that the pgre plays in assessment varies among individual departments and admissions committees , applications are often subject to formal or informal cutoffs " based on the pgre percentile score ( with a perfect score earning a percentile of 99% ) .
this practice is based on the assumption that the pgre is a useful gauge of whether applicants will be successful " , both as graduate students and as professionals .
lower - scoring applicants are either eliminated from the pool entirely or required to compensate for a low pgre score with exceptional credentials in other areas of their application .
prospective applicants with low pgre scores will sometimes choose to not even apply to particular programs - particularly those which quote an average " or typical " pgre score on their websites - out of a belief that their score will automatically render them uncompetitive in the applicant pool .
standardized tests similar to the pgre have a demonstrated record of performance being skewed by gender and race . the educational testing service , which administers the pgre as well as the general gre , has demonstrated that both gender and race impact scores on the general gre , with lower scores among women and underrepresented minorities ( see , for example , steele & aronson 1995 , miller & stassun 2014 ) .
this same demographic skew has been demonstrated for a broad array of standardized tests , including the pgre itself ( e.g. miller 2013 ) .
the predictive power of such exams for long - term outcomes is also unclear .
quantifying success " either in or beyond graduate school is a complex undertaking , and reducing this to a single metric is often considered ineffectual .
for example , while glanz ( 1996 ) shows that pgre scores weakly correlate with average coursework grades for _ physics _
graduate students at harvard , the same work also argues against heavily weighting the physics gre in graduate admissions procedures for physics programs .
glanz ( 1996 ) also demonstrates no correlation between coursework performance and other metrics of success " in graduate school , such as research productivity .
the role of the pgre in astronomy graduate admissions - or , more specifically , whether pgre score correlates with success " in the field of astronomy - has not been explicitly studied .
demographic trends in the pgre are hard to examine within small sample groups , and it is difficult to converge upon a satisfactory and broadly inclusive definition of success " .
however , we believe that even an informal examination of this specific scenario - pgre scores of successful " astronomers - can be useful in considering how the pgre should be used in astronomy graduate admissions . here
we present data from a questionnaire sent to 271 individuals who have held prize postdoctoral fellowships in astronomy between 2010 - 2015 .
the questionnaire asked recipients to self - report their scores on the pgre .
we designed a short ( 13-question ) questionnaire focused specifically on pgre performance .
the questionnaire ( figure 1 ) asked respondents to report the year that they took the pgre as well as their raw score , specific percentile scores , and/or percentile range , with percentiles split into ten equal ranges spanning from 0 - 99% .
the questionnaire also asked respondents to report their number of _ first author _ papers , both when they completed their undergraduate degrees and when they completed graduate school .
this was primarily done as a means of roughly quantifying research productivity in graduate school .
it also offered a means of examining how publications can potentially impact graduate admissions ( for example , whether respondents with low pgre scores generally had strong undergraduate publication records that would have strengthened their graduate school applications ) .
we did not collect data on respondents general gre scores or grade point averages ( gpas ) .
this was primarily done to keep the questionnaire as short as possible and maximize our potential response rate ; however , it is also worth noting that the general gre has been studied elsewhere ( see , for example , miller & stassun 2014 ) , and effectively analyzing gpas would require collecting substantial additional data on respondents undergraduate institutions , grading practices , majors , and coursework .
the questionnaire was designed to be sent to individuals who held a national prize postdoctoral fellowship ( hubble , einstein , nsf , sagan , and/or jansky ) between 2010 and 2015 . as a result of this small potential sample
we collected only minimal demographic data ( english fluency and the option to identify gender ) in order to avoid compromising the anonymity of the respondents . while respondents were asked to identify the specific fellowship or fellowships that they held along with the start year , this
was done to assess the completeness of our response rate and the data was not used in any comparative analyses .
we also asked respondents to characterize the institution that granted their bachelors degrees , but that data is not used in the analyses presented here .
all respondents were required to answer a question asking whether they consented to the use of their responses in a short astro - ph publication . for the remainder of this work
we include only the data from those who answered yes " to this question .
the questionnaire was emailed to 271 potential respondents beginning on 14 november 2015 .
all of the recipients had held a national prize postdoctoral fellowship ( hubble , einstein , nsf , sagan , and/or jansky ) between 2010 and 2015 .
responses were collected anonymously ; no email addresses or other personal identifications were logged . of the 271 questionnaire recipients , 173 responded ( response rate of 64% ) .
23 responses did not report a pgre score , leaving us with a total of 149 unique respondents with reported pgre scores .
of the 149 respondents reporting pgre scores , 75 identified as male , 53 identified as female , and 21 identified as other or chose not to identify their gender .
respondents were given the option of selecting multiple percentile ranges if they could not recall their exact scores or score ranges .
17 people ( 11% ) gave multiple percentile ranges for their scores . in 3 cases ( 2% ) , these respondents also reported their exact numerical scores and the year that the took the pgre ; this allowed us to assign them to a single percentile bin based on their reported ranges and the score conversions and percentiles published by ets .
in the remaining 14 cases ( 9% ) , the median of the reported percentile range was adopted for our analyses
. there are several important caveats worth noting in the design of this questionnaire and study .
the small sample size carries with it the inevitable impact of small number statistics .
the nature of the questionnaire is also susceptible to the standard errors of self - reporting .
finally , our distribution of the survey only to prize fellows should not in any way be interpreted as a belief that someone applying to graduate school in astronomy must eventually earn a prize fellowship in order to be considered successful " . holding
a prize fellowship is a narrow , incomplete , and non - inclusive definition of success " .
similarly , the number of first author papers published in graduate school is by no means the best or only method of assessing an individual s research productivity ; this is particularly true for people working in large collaborations or in fields such as instrumentation or computation .
at the same time , restricting our work to such narrow criteria is useful in a simplistic analysis of whether the pgre correlates with success " . holding a prize
postdoctoral fellowship is one ( though certainly not the only ) marker of success " in an astronomy career , and first author publications are one ( though not the only ) means of assessing success " in research . as a result , if the pgre is indeed an effective and unbiased means of assessing the potential for success " in graduate school , then one could reasonably expect that the pgre scores of prize postdoctoral fellows with strong graduate school publication records should be extremely high .
29% of fellows scoring below the 50th percentile .
the distribution of scores is also notably different for male and female fellows ; @xmath042% of female fellows fall below the 50th percentile.,title="fig:",scaledwidth=45.0% ] 29% of fellows scoring below the 50th percentile .
the distribution of scores is also notably different for male and female fellows ; @xmath042% of female fellows fall below the 50th percentile.,title="fig:",scaledwidth=45.0% ] figure 2 illustrates the difference between the expected " pgre scores of prize fellows and those found in our data . on the left , we show the expected appearance of a score histogram for prize fellows if the pgre was strongly correlated with success " and a 50th - percentile hard cutoff in graduate admissions was an effective means of selecting potentially successful applicants . in this scenario
we would expect scores to be distributed as an approximate half - gaussian , with most fellows scores falling into the highest bins and _ no _ prize fellows with pgre scores below the graduate admissions cutoff score .
this estimate also assumes that there is no gender disparity in pgre performance ; while the gender demographics match those of our survey respondents , the score distributions within each gender category are identical . on the right
we show our actual histogram of pgre percentile scores reported by prize fellows .
the score distribution spans the full range of percentiles , with 44 fellows ( @xmath030% ) falling below the 50th percentile .
the distribution of scores is also notably different for male and female fellows ; @xmath042% of female fellows fall below the 50th percentile .
the results show a clear disparity in gender performance on the physics gre among prize fellows .
multiple fellowships are represented in each percentile bin .
figure 3 illustrates the number ( left ) and fraction ( right ) of fellows - both in total and divided by gender - that would be eliminated from a theoretical graduate school applicant pool by applying hard thresholds in the pgre percentiles , with thresholds ranging from 10 - 60% .
table 1 offers a numerical breakdown of these values , with thresholds ranging from 10 - 90% .
cccccccccc 149.0 & 4 ( 2.7% ) & 10 ( 6.7% ) & 16 ( 10.7% ) & 32 ( 21.5% ) & 44 ( 29.5% ) & 65 ( 43.6% ) & 85 ( 57.0% ) & 104 ( 69.8% ) & 121 ( 81.2% ) + 75.0 & 1 ( 1.3% ) & 3 ( 4.0% ) & 6 ( 8.0% ) & 13 ( 17.3% ) & 15 ( 20.0% ) & 25 ( 33.3% ) & 34 ( 45.3% ) & 47 ( 62.7% ) & 59 ( 78.7% ) + 53.0 & 1
( 1.9% ) & 4 ( 7.5% ) & 6 ( 11.3% ) & 14 ( 26.4% ) & 22 ( 41.5% ) & 32 ( 60.4% ) & 42 ( 79.2% ) & 47 ( 88.7% ) & 49 ( 92.5% ) + 21.0 & 2 ( 9.5% ) & 3 ( 14.3% ) & 4 ( 19.0% ) & 5 ( 23.8% ) & 7 ( 33.3% ) & 8 ( 38.1% ) & 9 ( 42.9% ) & 10 ( 47.6% ) & 13 ( 61.9% ) + statistics and corresponding @xmath1 values are cited at the top of each panel . ]
cccccccccccc all & 147 & 18 & 47 & 71 & 131 & 186 & 267 & 353 & 426 & 490 & 635 + male & 74 & 9 & 16 & 31 & 61 & 70 & 109 & 153 & 201 & 244 & 341 + female & 53 & 2 & 21 & 28 & 55 & 96 & 133 & 165 & 183 & 189 & 199 + not specified / other & 20 & 7 & 10 & 12 & 15 & 20 & 25 & 35 & 42 & 57 & 95 + [ tab : paperselim ] in addition to fellows pgre scores , we also considered whether pgre score showed any correlation with first author publications ; a statistically robust correlation would suggest that the pgre may be indicative of research productivity in astronomy .
figure 4 compares the number of first author publications written by fellows at various stages of their career to pgre scores .
while there is marginal evidence ( @xmath2 ) of a weak ( @xmath3 ) correlation between pgre score and number of papers published as an undergraduate , there is no statistically significant ( @xmath4 ) correlation between pgre score and total first author papers during or exiting graduate school among 147 prize fellows ( 2 of the 149 respondents who reported pgre scores did not report first author paper statistics ) .
similarly , figure 5 compares histograms of first author papers written at various career stages for prize fellows whose pgre scores fell below thresholds ranging from 20 - 60% . the mean and median number of first author publications for fellows in each cutoff category
are illustrated .
the general shape of the histogram and the average number of publications are both insensitive to pgre cutoff score .
table 2 offers a numerical breakdown of these results , giving the total number of first author papers written by prize fellows with pgre scores below thresholds ranging from 10 - 90% .
based on the above results , we find no evidence that the pgre can be used as an effective predictor of success " either in or beyond graduate school .
the pgre scores of individuals who have held prize fellowships do not adhere to any minimum cutoff , and show a significant deviation from the score distribution that would be expected if pgre correlated with career performance . in addition , there is also no statistically significant correlation shown between pgre score and the number of first author publications written by prize fellows either during or exiting graduate school .
the average number of first author publications among prize fellows remains unchanged regardless of whether a pgre cutoff score is applied .
publication records also can not simply be used to counteract low pgre scores in admissions decisions ; of the 15 fellows who scored @xmath530% on the pgre , only 3 ( 20% ) had published one or more first author papers during their undergraduate careers .
it is also worth noting that this is in agreement with the overall percentage of fellows who reported publishing one or more first author papers as undergraduates ( 34 , or 23% ) .
it is true , of course , that all the prize fellows included in this study were eventually admitted to a graduate program and successfully completed their phds .
however , it is important to note that our sample includes prize fellows who received their phds overseas or from programs that do not require the pgre . while there are many factors at play in assessing a graduate school applicant pool ( including hard - to - quantify elements such as cvs , reference letters , and statements of purpose )
these data demonstrate that the use of pgre scores to either eliminate or negatively impact graduate school applications is likely to remove a significant number of potentially successful astronomers from the applicant pool .
for example , we can consider a hard " cutoff of 60% applied to a hypothetical applicant pool comprised of our poll respondents .
such a cutoff would have excluded 64 future prize fellowship recipients . in addition , while this would exclude 43.6% of the total applicant pool , the cutoff would disproportionately impact women , excluding 60.4% of female applicants .
those prospective applicants would have gone on to write 267 ( 133 ) papers during their graduate careers .
even a relatively low cutoff in pgre scores would eliminate future prize fellows ( 4 at @xmath510% , 10 at @xmath520% ) .
overall , the data suggest that any application of a pgre cutoff score in graduate admissions carries with it a substantial risk of eliminating potentially successful future astronomers from the applicant pool .
glanz , j. 1996 , science , 274 , 710 miller , c. & stassun , k. 2014 , nature , 510 , 303 miller , c. 2015 , diversity considerations for graduate admissions " , presented at the aps chairs conference ( https://www.aps.org/programs/education/conferences/chairs/2015/upload/aps-chairs-conference-diversity-miller.pdf ) steele , c. m. & aronson , j. 1995 , journal of personality and social psychology , 679 , 797 | the physics gre is currently a required element of the graduate admissions process in nearly all u.s .
astronomy programs ; however , its predictive power and utility as a means of selecting successful " applicants has never been examined .
we circulated a short questionnaire to 271 people who have held u.s .
prize postdoctoral fellowships in astrophysics between 2010 - 2015 , asking them to report their physics gre scores ( this should not in any way be interpreted as a belief that a prize fellowship is the best or only metric of success " in astronomy ) .
the response rate was 64% , and the responding sample is unbiased with respect to the overall gender distribution of prize fellows .
the responses reveal that the physics gre scores of prize fellows do not adhere to any minimum percentile score and show no statistically significant correlation with the number of first author papers published . as an example
, a physics gre percentile cutoff of 60% would have eliminated 44% of 2010 - 2015 u.s .
prize postdoctoral fellows , including 60% of the female fellows . from these data ,
we find no evidence that the physics gre can be used as an effective predictor of success " either in or beyond graduate school . |
since shor discovered a polynomial time algorithm for factorization on a quantum computer @xcite , a lot of effort has been directed towards building a working quantum computer . despite all these efforts , it is still extremely difficult to control even a few qubits .
it is thus of great interest to study exactly how few qubits are needed to factor an @xmath0-bit number .
quantum factorization consists of classical preprocessing , a quantum algorithm for order - finding and classical postprocessing @xcite ( fig .
[ circuit ] ) . we will concentrate on the quantum part of factorization and consider classical parts as being free as long as they are computable in polynomial time .
the only use of quantum computation in shor s algorithm is to find the order of @xmath4 modulo @xmath5 , where @xmath5 is an @xmath0-bit integer that we want to factor .
the order @xmath6 of @xmath4 modulo @xmath5 is the least positive integer such that @xmath7mod @xmath8 . for completeness
, we now give the full algorithm for factoring @xmath5 as given in @xcite : 1 . if @xmath5 is even , return the factor 2 .
classically determine if @xmath9 for @xmath10 and @xmath11 and if so return the factor @xmath12 ( this can be done in polynomial time ) .
3 . choose a random number @xmath4 such that @xmath13 .
using euclid s algorithm , determine if gcd(@xmath14)@xmath15 and if so , return the factor gcd(@xmath14 ) .
4 . use the order - finding quantum algorithm to find the order @xmath6 of @xmath4 modulo @xmath5 .
if @xmath6 is odd or @xmath6 is even but @xmath16mod @xmath8 , then go to step ( iii ) . otherwise , compute gcd(@xmath17 ) and gcd(@xmath18 ) . test to see if one of these is a non - trivial factor of @xmath5 , and return the factor if so .
it can be shown that with probability at least one half , @xmath6 will be even and @xmath19mod @xmath8 @xcite . the quantum part of the algorithm ( step 4 ) is known to be computable in polynomial time on a quantum computer .
using classical techniques , it is straigthforward to build the order - finding circuit ( fig .
[ circuit ] ) using a polynomial number of elementary gates and a linear number of qubits @xcite .
because the depth of the circuit is related to its running time , it is desirable to minimize this depth , and much progress has been made in that direction @xcite .
we propose to take the problem from the other side : by how much can the number of qubits be reduced for factorization in polynomial time ? answering this question would give insights on the size of a quantum computer useful for factorization .
we thus introduce a new order - finding circuit focused on reducing the number of qubits while still using only a polynomial number of elementary quantum gates .
we also somewhat try to minimize the depth of the circuit , but very little parallelization is available since we avoid using any unnecessary qubit .
the circuit for factorization that will be discussed here was inspired in part by a circuit from vedral , barenco and ekert @xcite . to reduce the number of qubits
, we use a variant of a quantum addition algorithm described by draper @xcite ( fig .
[ draperadd ] ) .
other techniques used to reduce the number of qubits are the hardwiring of classical values and the sequential computation of the fourier transform .
the quantum addition of figure [ draperadd ] takes as input @xmath0 qubits representing a number @xmath4 , and @xmath0 more qubits containing the quantum fourier transform of an other number @xmath20 , denoted by @xmath21 . after the addition
, the first register keeps the same value @xmath4 but the bottom register now contains the quantum fourier transform of @xmath22mod @xmath23 , denoted by @xmath24 .
adding together two quantum registers is , however , more than we ask for .
we are trying to find the period of the function @xmath25mod @xmath5 where @xmath4 is a classical random number smaller than @xmath5 .
since @xmath4 is classical , we only need to be able to add a classical value to a quantum register .
we can thus change the qubits representing @xmath4 in figure [ draperadd ] to classical bits .
the controlled gates are then classically controlled , and since we know what @xmath4 is beforehand , we might as well precompute the product of all gates on each single qubit and apply only one gate for every single qubit . these are one - qubit gates , which also makes them easier to implement .
since the addition takes place in the fourier space , we will call this circuit the @xmath26 gate where @xmath4 is the classical value added to the quantum register ( fig . [ phiadd ] ) . notice the thick black bar on the right , used to distinguish the gate from its unitary inverse . in order to prevent overflow
, we need @xmath27 qubits for the quantum register instead of @xmath0 , so that @xmath21 is effectively the qft of an ( @xmath27)-qubit register containing a @xmath0-bit number ( thus the most significant qubit before the qft preceding the addition is always @xmath28 ) .
if we apply the unitary inverse of the @xmath26 gate with input @xmath21 , we get either @xmath29 if @xmath30 , or @xmath31 if @xmath32 .
thus if @xmath32 , the most significant qubit of the result is always @xmath33 , whereas it is always @xmath28 if @xmath34 .
this reverse @xmath26 gate can be useful for subtraction and comparison purposes ( fig .
[ phiminus ] ) and we use a black bar on the left to distinguish it from the regular gate . the unitary inverse of a circuit is obtained by applying the unitary inverse of each elementary gate in reverse order .
now that we have a @xmath26 gate , we can use it to build a modular adder gate ( fig .
[ ccaddmod ] ) . for future use , two control qubits
are included in the circuit .
for the modular adder gate , we need to compute @xmath35 and subtract @xmath5 if @xmath36 . however , it is not so easy to implement this operation in a reversible way .
the input to the @xmath37 gate is @xmath21 with @xmath38 , and the classical number @xmath4 that we add is also smaller than @xmath5 .
we begin by applying a @xmath26 gate to the register @xmath21 .
the quantum register now contains @xmath24 with no overflow because we were careful enough to put an extra qubit in state @xmath28 along with the value @xmath20 before applying the qft .
we next run a reverse @xmath39 to get @xmath40 .
if @xmath41 , we did not have to subtract @xmath5 but now we can determine if @xmath41 by checking the most significant bit of @xmath42
. however , to access this most significant bit we need to inverse the qft on the whole register containing @xmath40 .
we can then use this qubit as the controlling qubit of a controlled - not gate acting on an ancillary qubit .
it is then possible to reapply the qft and use this ancilla as a control qubit for a @xmath39 controlled gate , so that if @xmath41 we add back the value @xmath5 that we subtracted earlier .
we now have @xmath43mod @xmath8 in the register , and we are done except for the ancilla which is now a junk bit .
we have to restore it to @xmath28 somehow , otherwise the computation will not be clean and the algorithm will not workmod @xmath5 but the period of the garbage bits can be something else . ] .
restoring the ancilla to @xmath28 is no easy task if we do not want to waste qubits .
we can still do it by using the identity : @xmath44 hence , we only have to compare @xmath45mod @xmath5 with the value @xmath4 using essentially the same trick as before .
we run an inverse @xmath26 followed by an inverse qft to get the most significant qubit of @xmath22mod @xmath46 .
this qubit is @xmath28 if @xmath22mod @xmath47 .
we apply a not gate on this qubit and use it as the controlling qubit of a controlled - not gate targeting the ancilla .
the ancilla is thus restored to @xmath28 and we can apply a not gate again on the control wire , followed by a qft and a @xmath26 gate on the quantum register . after this , we have a clean computation of @xmath22mod @xmath5 in the fourier space .
again , what we need exactly is a doubly controlled version of the @xmath37 gate . in order to reduce the complexity of the circuit
, we will doubly control only the @xmath26 gates instead of all the gates ( fig .
[ ccaddmod ] ) .
if the @xmath26 gates are not performed , it is easy to verify that the rest of the circuit implements the identity on all qubits because @xmath38 .
the next step is to use the doubly controlled @xmath37 gate to build a controlled multiplier gate that we will call @xmath48 ( fig .
[ cmul ] ) .
this gate takes three inputs , @xmath49 , and its output depends on the qubit @xmath50 . if @xmath51 , the output is @xmath52 .
if @xmath53 , then the input is unchanged and stays @xmath49 .
this gate is very straightforward to implement using doubly controlled @xmath37 gates .
we use the identity : @xmath54 thus we only need @xmath0 successive doubly controlled modular adder gates , each of them adding a different value @xmath55 with @xmath56 to get the @xmath48 gate .
we now have a controlled gate that takes @xmath57 to @xmath58 .
what we would need instead is a controlled gate that takes @xmath59 to @xmath60 .
this can however be obtained by a clever trick from reversible computing that uses two controlled multiplication gates ( fig [ ctrlu ] ) .
we first apply the @xmath48 gate to @xmath61 .
we follow with a swap between the two registers if the qubit @xmath51 ( that is effectively a controlled - swap on the registers ) .
we only need to control - swap @xmath0 qubits , not @xmath27 .
indeed , the most significant qubit of @xmath62 will always be 0 since we were careful to include one extra qubit to store the overflow in the @xmath26 gate .
we then finish with the inverse of a @xmath63 circuit .
the value @xmath64 , which is the inverse of @xmath4 modulo @xmath5 , is computable classically in polynomial time using euclid s algorithm and it always exists since gcd@xmath65 .
the fact that we apply the inverse of the circuit means that the circuit effectively takes @xmath49 to @xmath66 .
the resulting gate will be called @xmath67-@xmath68 for controlled-@xmath68 .
it does nothing if @xmath53 but if @xmath69 , then the two registers take the following values : @xmath70 since the bottom register returns to @xmath28 after the computation , we can consider this extra register as being part of the @xmath67-@xmath68 gate , thus the gate effectively takes @xmath59 to @xmath60 .
this is exactly the gate we need to run the quantum order - finding circuit ( fig [ circuit ] ) . of course
, we do nt need to apply @xmath67-@xmath68 @xmath0 times to get @xmath71-@xmath72 because we can directly run @xmath67-@xmath73 ( where @xmath74 in computed classically ) which is the same as @xmath71-@xmath72 since : @xmath75 an advantage of using the @xmath67-@xmath76 gates for shor s algorithm is the fact that we do nt really need the total @xmath77 controlling qubits .
in fact , it can be shown that only one controlling qubit is sufficient @xcite .
this is possible because the controlled-@xmath78 gates all commute and the inverse qft can be applied semi - classically . indeed , we can get all the bits of the answer sequentially as in figure [ trick ] .
each measured bit dictates which unitary transformation we have to apply after every controlled-@xmath78 step before the next measurement .
this simulates the inverse qft followed by a measurement on all qubits as in figure [ circuit ] .
we save an important number of qubits this way , and in fact we need only a total of @xmath1 qubits to factor an @xmath0-bit number as we will show in the complexity analysis section .
the implementation of the exact @xmath79 on @xmath0 qubits requires @xmath80 operations @xcite ( fig . [ qft ] ) .
however , in physical implementations , there will always be a threshold for the precision of the gates .
since many phase shifts will be almost negligible , we will in practice ignore the ones with @xmath81 greater than a certain threshold @xmath82 .
this approximate @xmath79 is in fact very close to the exact @xmath79 even with @xmath82 logarithmic in @xmath0 .
in fact , it has been shown @xcite that the error introduced by ignoring all gates with @xmath83 is proportional to @xmath84.we can thus choose @xmath85 .
the implementation of the approximate @xmath79 on @xmath0 qubits requires @xmath86 gates .
there seems to be no obvious way to reduce the depth of either the exact @xmath79 and the approximate @xmath79 on @xmath0 qubits below @xmath87 without using extra qubits @xcite .
the depth of the qft on @xmath27 qubits is thus @xmath87 with the little parallelization available without extra qubits .
the controlled - swap on one qubit is very easy to implement ( fig .
[ cswap ] ) . only two controlled - not and one toffoli are needed to perform the swap on two qubits controlled by a third .
thus , @xmath87 gates are needed to control - swap @xmath0 qubits , that is , swap @xmath0 qubits with @xmath0 others with one control qubit .
we now analyze the complexity of the given circuit for performing factorization of an @xmath0-bit number @xmath5 .
the analysis keeps track of the number of qubits , the order of the number of gates and the order of the depth of the circuit . for the depth of the circuit
, we consider that it will be possible to apply simultaneously different quantum gates that act on different qubits of the quantum computer .
however , we consider impossible to have one qubit controlling many operations in the same step
. the circuit uses only single qubit gates , up to doubly controlled conditionnal phase shifts and up to doubly controlled not gates .
these gates can be implemented using a constant number of single qubit gates and controlled - nots @xcite , so they can all be considered as elementary quantum gates .
the @xmath26 circuit ( fig . [ phiadd ] ) , where @xmath4 is a classical value , requires @xmath27 qubits and @xmath87 single qubit gates in constant depth .
the number of qubits is @xmath27 because we need an extra qubit to prevent overflows .
when a control qubit is added to the circuit , the depth becomes @xmath87 since the conditional phase shifts have to be done sequentially .
indeed , the control qubit has to control each phase shift one at a time .
the doubly controlled @xmath37 circuit ( fig . [ ccaddmod ] ) requires @xmath88 qubits .
it also requires @xmath89 gates , but has a depth of only @xmath87 regardless of @xmath82 because the qfts can be somewhat parallelized .
the @xmath48 circuit is only @xmath0 doubly controlled @xmath37 circuits .
it thus takes @xmath1 qubits , @xmath90 gates and a depth of @xmath80 to implement the @xmath48 circuit .
two of these circuits along with the controlled - swap are needed for the @xmath67-@xmath68 circuit .
the controlled - swap on @xmath0 qubits requires only @xmath87 gates and depth , so the @xmath67-@xmath68 circuit requires @xmath1 qubits , @xmath90 gates and a depth of @xmath80 again . for the whole order - finding circuit , that is , the whole quantum part of shor s algorithm , we need @xmath77 of these @xmath67-@xmath68 circuits .
the quantum resources needed are thus @xmath1 qubits , @xmath91 gates and a depth of @xmath3 .
if we decide to use the exact @xmath79 in the additions , then we would have @xmath92 . as we argued earlier ,
this would not be clever because the implementation is sure to have hardware errors anyway .
we thus should use the approximate @xmath79 with @xmath93 ) , so that the number of gates is in @xmath94 for any @xmath95 polynomial in @xmath96 .
this result of @xmath1 qubits is slightly better than previous circuits for factorization .
vedral , barenco and ekert published a circuit of @xmath97 qubits and @xmath3 elementary gates for modular exponentiation @xcite .
it is mentionned that this number can be easily reduced to @xmath98 qubits with basic optimization and further reduced to @xmath99 if unbounded toffoli gates ( @xmath0-controlled nots ) are available .
beckman , chan , devabhaktoni and preskill provided an extended analysis @xcite of modular exponentiation , with a circuit of @xmath100 qubits using elementary gates and @xmath101 if unbounded toffoli gates are available .
zalka also described a method for factorization with @xmath102 qubits using only elementary gates @xcite .
the availability of unbounded toffoli gates will of course depend on the physical implementation of the quantum computer , but it is assumed throughout our design and analysis that such gates can not be considered elementary . for that matter , if we do not restrict the type and size of the quantum gates in any way , order - finding can be achieved with @xmath27 qubits by directly using controlled multiplication gates @xcite . of the @xmath1 qubits used in the circuit provided here , one is used as an ancilla for modular addition , one is used to prevent addition overflows and @xmath0 are used as an ancillary register to get modular multiplication from successive additions .
an order - finding circuit using elementary gates and less than @xmath103 qubits is not ruled out yet , but it seems that a different method would have to be used for modular multiplication to get such a circuit .
fifteen is the smallest number on which shor s algorithm can be applied .
the circuit for factorization of @xmath104 uses eleven qubits as given here .
however , the classical computation performed to build it gives a lot of information on the order of the number @xmath4 .
indeed , for any @xmath105 , the order of @xmath4 is either two or four .
most of the multiplications in the circuit are simply the identity and can be removed , which amounts to many unused qubits .
the number 15 was factored using nmr with seven qubits in an impressive display of quantum control by vandersypen , steffen , breyta , yannoni , sherwood and chuang @xcite .
the importance of reducing the number of qubits versus reducing the depth of a quantum computation is not clear as quantum computers of useful size are not yet available .
we have to keep in mind that error correction will most probably have to be used on quantum computers , which will create an overhead in the number of qubits used @xcite .
it is however sensible to minimize the number of qubits before applying error correction if qubits are hard to come by .
putting together several tricks , we have developed a circuit for the quantum part of the factorization algorithm , that is , the order - finding algorithm , while focusing on reducing the number of qubits .
the number of qubits needed is @xmath1 and the depth is @xmath3 .
this circuit uses slightly less qubits than those previously known if restricted to elementary gates .
it is also completely general and does not rely on any properties of the number to be factored . given the values @xmath4 and @xmath5 , this circuit gives the order @xmath6 of @xmath4 modulo @xmath5 with good probability .
many runs of this algorithm may be needed to factor a number .
also , the randomly chosen value @xmath4 is hardwired in the circuit and there is a probability ( about one half ) that it will be necessary to choose a new value @xmath4 and run a new order - finding algorithm on it .
this is not a problem if the quantum computer is a physical device where the gates are interactions controlled by a classical computer such as laser pulses on trapped ions , nmr and most implementation proposals .
indeed , the circuit can easily be classically computed .
a quantum computer consisting of a physical system controlled by a classical computer is the most conceivable option at this point .
the author is very grateful to michel arsenault for many helpful comments and corrections , and would also like to thank jos manuel fernandez , christof zalka and anne broadbent .
this work was supported in part by cse of canada , and the author acknowledges support from nserc of canada .
r. cleve and j. watrous ( 2000 ) , _ fast parallel circuits for the quantum fourier transform _ , proceedings 41st annual symposium on foundations of computer science ( focs00 ) , pp .
526 - 536 . also on quant - ph/0006004 .
a. barenco , c. bennett , r. cleve , d.p .
divincenzo , n. margolus , p. shor , t. sleator , j.a .
smolin , and h. weifurter ( 1995 ) , _ elementary gates for quantum computation _ , phys .
a , 52 , pp .
3457 - 3467 .
also on quant - ph/9503016 .
vandersypen , m. steffen , g. breyta , c.s .
yannoni , m.h .
sherwood , and i.l .
chuang ( 2001 ) , _ experimental realization of shor s quantum factoring algorithm using magnetic resonance _ , nature , 414 , pp .
883 - 887 . | we try to minimize the number of qubits needed to factor an integer of @xmath0 bits using shor s algorithm on a quantum computer .
we introduce a circuit which uses @xmath1 qubits and @xmath2 elementary quantum gates in a depth of @xmath3 to implement the factorization algorithm .
the circuit is computable in polynomial time on a classical computer and is completely general as it does not rely on any property of the number to be factored . |
s.m . acknowledges support by the italian miur through grant no . prin2008xwlwf9 .
all calculations have been performed with the vienna ab - initio simulation package@xcite .
we use pseudopotentials of projector augmented wave@xcite type and the gradient corrected exchange correlation functional of perdew , burke and ernzerhof@xcite ( gga - pbe ) . a cutoff to the planewave expansion
has been converged to the value of 500 ev .
we make use of a simulation unit cell with 4 formula units ( f.u . ) of cafe@xmath3as@xmath3 . with this cell
we simulate 25% of doping level . a @xmath43 @xmath44-points mesh has been used .
rare - earth atoms pseudopotentials were constructed including f - electrons into the core .
all magnetic calculations converged to the striped antiferromagnetic phase . with this model system
, we are indeed neglecting possible disorder effects and the role of the electron dopant .
however , the model should grasp the main electronic properties of the compressed system . | we report a systematic and ab - initio electronic structure calculation of ca@xmath0m@xmath1fe@xmath2as@xmath2 with m = ca , sr , eu , la , ce , pr , nd , pm , sm , na , k , rb .
the recently reported experimentally observed structural trends in rare earths - doped cafe@xmath3as@xmath3 compounds are successfully predicted and a complete theoretical description of the pressure induced orthorhombic to collapsed tetragonal transition is given .
we demonstrate that the transition pressure is reduced by electron doping and rises linearly with the ionic size of the dopants .
we discuss the implications of our description for the realization of a superconducting phase .
the family of iron based superconductors ( sc ) continues to grow with the discovery of new systems that add to the hundreds already known@xcite .
however , in spite of this enormous amount of experimental data , an ab - initio theory describing these superconductors is still missing , and the search for new materials is guided only by some observed features that appear to be in common . even if it is not a general rule , the so - called parent compounds do not superconduct without doping@xcite or without applying high pressure@xcite .
the rationale for this behavior could be naively ascribed to the necessity to destabilize the orthorhombic ( o ) antiferromagnetic order of the parent compounds to promote a magnetic mediated pairing field@xcite .
indeed , superconductivity is realized in tetragonal ( t ) phases only after the magnetic order is suppressed , although some coexisting phases were discovered in 122 compounds@xcite . in the so - called 122 family ( mfe@xmath3as@xmath3 , m = ca , ba , sr )
superconductivity can be induced under high pressure@xcite .
the critical pressures at which superconductivity is detected varies with m@xcite and with the pressure conditions@xcite ( hydrostatic and non - hydrostatic ) . at the same time , the members of the 122 family show a pressure induced structural phase transition to a _ collapsed tetragonal _ ( ct ) phase@xcite .
the main experimental evidence of the ct phase transition is the sudden decrease of the @xmath4 lattice parameter and a subsequent increase of the in - plane @xmath5 lattice constant .
first - principles calculations were able to describe this phase transition as induced by the formation of a direct as - as bond along the @xmath4-axis of the tetragonal phase @xcite .
the ct phase is predicted to be a non - magnetic phase ( the magnetic moment being zero on the fe sites ) in agreement with experimental results@xcite which report the disappearance of the magnetic response . in ba and sr 122 compounds , where the o - t and t - ct are well separated in pressure@xcite
, it seems that the superconducting phase is realized across the o - t transition ( even if the role of non - hydrostaticity of the pressure medium is not clear ) , while it was debated in cafe@xmath3as@xmath3 where the transitions are very near in pressure@xcite and hysteresis effects are relevant .
in fact , for cafe@xmath3as@xmath3 the precise nature of the crystallographic structure of the superconducting phase is under debate , but the ct phase obtained under hydrostatic pressure conditions@xcite seems to be ruled out .
this suggests that superconductivity is realized in a tetragonal paramagnetic structure ( not collapsed ) realized only under non - hydrostatic conditions@xcite .
this phase should be characterized by the proximity of a magnetic instability absent in the ct phase .
this subject was revived with the discovery of high temperature superconductivity ( up to 45 - 49 k ) in cafe@xmath3as@xmath3 doped with trivalent rare - earth ( re ) metal atoms ( la , pr , ce , nd ) substituting the divalent ca atoms@xcite . it was demonstrated that rare earths have high solubility ( up to 27% in ref . ) in cafe@xmath3as@xmath3 mainly due to close matching between ionic - radii of re and ca , and that their incorporation suppresses the low temperature afm ordering in favor of the superconducting phase .
application of hydrostatic pressure on the doped materials is shown to increase further the critical temperature@xcite .
moreover , neutron diffraction experiments as a function of temperature revealed an interesting structural property : depending on the re dopants , the low temperature phase shows the characteristics of the ct phase@xcite . in fact ,
some of the systems ( pr , nd doped for example ) exhibit a low temperature phase in which the out - of - plane axis shrinks of about 10% its high temperature value . on the other hand , when doped with la or ce , this last structural phase transition is not observed .
nevertheless , superconductivity is found irrespective of the presence or not of the o - ct transition .
recently the scenario was further enriched with the reports of superconductivity in ca@xmath6sr@xmath7fe@xmath2as@xmath2 at high pressure@xcite and of quenched fe magnetic moment in the collapsed tetragonal phase of ca@xmath8pr@xmath9fe@xmath2as@xmath2@xcite . at the moment
many questions remain to be answered , in particular regarding the origin of the superconducting phase .
in addition , the details of the structural and electronic properties of these new phases are still unknown and it is not clear if the magnetic correlations are suppressed / enhanced upon rare - earth doping in the t phase and the ct phase@xcite . as a first step in the comprehension , we report first - principles dft@xcite prediction of structural , magnetic and electronic properties of cafe@xmath3as@xmath3 upon doping with re . in addition , we extended the study to isovalent alkaline earths and eu@xmath10 and consider the effect of hydrostatic pressure on doped systems .
we start our discussion considering the physics of the parent compound , cafe@xmath3as@xmath3 .
density functional theory successfully predicts the structural collapse of the tetragonal phase at a pressure p@xmath11 of 0.8 gpa , as calculated from the crossing point of the free energy of the two phases ( the o and ct ) .
this value is in reasonable agreement with experimental evidences of o - ct phase transition at @xmath12 gpa@xcite .
the difference is in part due to computational accuracy , to the gga functional and experimental uncertainties@xcite .
the existence of both o and ct solution at the same pressure , means that the two phases are locally stable .
this agrees with the first order nature of the phase transition and suggests possible experimental hysteresis effects , as indeed reported@xcite .
the calculated ct lattice constant ( see table [ tab : data ] and fig .
[ fig : structure ] ) are @xmath13 , @xmath14 ( near the transition pressure ) , to be compared with experimental values 3.99 and 10.62 .
the o phase has @xmath15 , @xmath16 and @xmath17 , while the experimental value@xcite is @xmath18 , @xmath19 and @xmath20 .
m@xmath1fe@xmath2as@xmath2 in the ct and o phases .
the curves follow the phase that minimizes the enthalpy as a function of p. the data points that lie off from the lines are locally stable solutions with higher enthalpy .
the structural transition is marked by a collapse of the @xmath4 lattice parameter ( see panel @xmath5 ) and an expansion in the @xmath21 directions ( @xmath22 denotes the average between the two in plane axis parameters ) as shown in panel @xmath4 .
@xmath23 is distance between as an fe layers , and @xmath24 is the interlayer distance between as atoms .
, scaledwidth=45.0% ] the structural collapse reduces the @xmath4 lattice parameter by about 8% , while the @xmath5 and @xmath25 lattice parameters have a expansion of @xmath26% ( as an average between the two orthorhombic directions ) .
the compressibility also changes abruptly .
we estimate a bulk modulus ( @xmath27 ) jump across the transition of @xmath28 gpa . the structural transition into the collapsed phase
is also accompanied with a magnetic to non - magnetic transition in which the magnetic moment at the iron site drops from @xmath29 to zero .
contrary to the low pressure tetragonal phase , the ct phase is far from any magnetic instability ( we could not converge to any magnetic solution ) , in agreement with the experimental observation of no spin - fluctuations in th ct - phase@xcite .
this is indeed the observed phenomenology of the o - ct phase transition , that is correctly captured by the dft - gga .
we switch now to the description of doped cafe@xmath3as@xmath3 , focusing our interest on those systems that experimentally show enhanced superconducting properties .
one of the main experimental evidences reported in ref .
is the observation of a clear discontinuity on the variation of the lattice constants ( @xmath5 and @xmath4 ) as a function of the temperature for 15% doped praseodymium , pure cafe@xmath3as@xmath3 and 8% doped neodymium ( but not in lanthanum doped ) . in order to predict the structural changes upon chemical doping and to understand the experimental results
, we simulated the partial chemical substitution at the ca site with three different dopants : isovalent ( sr and eu ) , aliovalent with trivalent electron - dopants re ( the experimentally realized case la , ce , pr , nd and the not yet realized pm and sm ) and hole - doping with alkali na , k , rb .
the doping level is fixed to x=25% , well within the experimentally realized doping range in cafe@xmath3as@xmath3 @xcite .
we performed structural optimization at different volumes in both non - magnetic and antiferromagnetic ( stripe ) phase .
phase stability was calculated comparing enthalpy curves as a function of the external pressure .
structural , electronic and magnetic properties are collected in tab .
[ tab : data ] .
.structural electronic and magnetic properties of m doped cafe@xmath3as@xmath3 .
@xmath30 ( gpa ) is the transition pressure of the structural collapse . @xmath31 and @xmath32 ( @xmath33 ) are the equilibrium volumes in each phase , @xmath34 ( mev / f.u . )
is the energy difference between the two phases at zero pressure . @xmath35
( @xmath36 ) represents the magnetic moment in the o phase .
[ cols="<,^,^,^,^,^",options="header " , ] the characteristic transition pressure p@xmath11 from the orthorhombic to the collapsed tetragonal phase as obtained in the pure cafe@xmath3as@xmath3 is strongly affected by the chemical substitution .
in particular , ce , pr , nd , pm and sm are stable in the collapsed phase already at ambient ( zero ) pressure , while a positive pressure is needed to induce a collapse in sr , eu , la , na , k and rb substituted systems .
this behavior is in excellent agreement with the available experimental data reported by saha and coworkers @xcite .
in fact , pr is reported to induce the structural collapse already at a concentration between 5% and 7.5% ; nd is collapsed at 8% of doping , while no collapse is reported by la doping up to 27% of doping . in particular
, we nicely predict the existence of the ct phase at 2.24 gpa in sr doped cafe@xmath3as@xmath3 as experimentally observed@xcite . on the contrary
, ce is not observed in the collapsed tetragonal phase up to a 22% of content , contrary with the theoretical predictions .
however , ce was predicted to be at the boundary of the phase transition , in fact we find the collapsed phase only very slightly more stable than the o phase by 0.7 mev / f.u . and is sufficient a tiny positive pressure to induce it ( see fig .
[ fig : structure ] ) .
the evolution of the crystal structure upon doping and pressure is reported in fig .
[ fig : structure ] .
the data from the simulations on the two phases have been connected with a spline fit jumping at the calculated transition pressure .
lattice constants , as a function of pressure , show the characteristic drop of the @xmath4-axis ( see fig .
[ fig : structure](a ) ) .
an excellent agreement between theoretical and experimental lattice parameters ( not shown ) is observed when comparing the low - temperature experimental lattice constants@xcite with the calculated ones in the non - magnetic tetragonal phase ( ct ) , confirming the validity of dft in the description of this phase . on the other hand ,
the high temperature phase is a magnetic ( probably paramagnetic tetragonal ) phase , whose description is improved , within dft - gga , including magnetic polarization of the fe sites .
the internal relaxation shows an interesting pressure dependence of the fe - as interlayer distance ( @xmath23 ) . in the o phase it decreases monotonically with the applied pressure while it becomes constant and material independent in the ct phase ( as shown in fig .
[ fig : structure](d ) ) . due to this feature in this second phase the lattice parameter @xmath4 depends only on the size of the impurity atom and can be considered a measure of its ionic radius in this class of compounds .
being originated by the formation of as - as bond , the phase transition to the collapsed phase happens when the interlayer as - as separation ( @xmath24 ) reaches the critical value of 3.0 ( about twice the as covalent radius)@xcite , independently of the dopant ( shown in fig .
[ fig : structure](b ) ) . indeed ,
this last result is confirmed by x - rays diffraction measurements of saha and coworkers which suggested@xcite that the key parameter that controls the collapse is the as - as distance .
then , if we look at the correlation between the @xmath4 lattice parameter ( in the ct phase ) and the transition pressure p@xmath11 , we find a very interesting correlation .
for all trivalent lanthanides the transition pressure scales linearly with the value of the @xmath4 lattice ( @xmath37 with the ionic size ) . the same happens for bivalent and mono - valent dopants , although with different linear coefficients , see fig .
[ fig : c_p ] .
this clearly tells that p@xmath11 is determined by the size of the impurity atom and the kind and level of doping , with electron doping favoring the ct transition ; in fact , a smaller ionic size favors the structural collapse because when as atoms get closer their interaction becomes stronger and the required pressure to induce the o - ct transition is low ( or zero ) .
the valence of the dopant , on the other hand , also affects the transition pressure .
we observe that the hole doping induced by the alkali is mainly located inside the fe - as layer , not affecting sensibly the as - as vertical bonding energy , while electron doping fills the as - as bonding states , favoring the chemical bonding . to support this picture we plot in fig .
[ fig : bands](a ) the periodic part of two states belonging to the same electronic band but taken above and below the fermi energy ( indicated in figure ) .
the interlayer bonding character is stronger above than below , therefore while electron doping contributes to the interlayer bonding , hole doping affects it only to a little extent . at this point
we switch to the second part of our analysis .
we focus on the electronic properties and their doping dependence .
the non - magnetic band structure of undoped cafe@xmath3as@xmath3 in the bct unit cell is plotted in fig .
[ fig : bands](b ) .
we observe that both o ( not shown ) and ct phases show a depletion of states at the fermi energy , with respect to this non - magnetic high - temperature phase . in the former case
this is due to a magnetic splitting , while in the latter to chemical splitting of bonding and antibonding @xmath38 state in as . considering the o phase , the electronic effects of aliovalent atoms
can not be predicted by means of a rigid band approximation because of the doping dependence of the magnetic moment ( see tab .
[ tab : data ] ) , that strongly affects the band structure near the fermi energy . on the other hand in the ct phases of all the investigated systems ( near the critical lattice constants ) a rigid shift of the fermi level reproduces
very well the dopant induced modifications of the electronic structure .
this is clearly seen comparing the dos of different compounds as shown in the lower panel of fig .
[ fig : bands ] . thus , apart from a small ( @xmath39 ev ) rigid shift of the fermi level , the non - magnetic band structure of the compressed tetragonal phase of cafe@xmath3as@xmath3 ( shown in fig .
[ fig : bands ] ) is a good approximation for that of ( superconducting ) re - doped cafe@xmath3as@xmath3 @xcite .
the structural collapse induced by the dopants profoundly changes the topology of the fermi surface with respect to the non - magnetic tetragonal phase as it does in undoped cafe@xmath3as@xmath3 .
the two - dimensional , cylindrical , hole fs s at the @xmath40 point disappear upon compression and three dimensional fs appear as a consequence of band crossing the fermi energy along the @xmath40-@xmath41 direction of the brillouin zone . thus , in the ct phase it is very difficult to recognize the characteristic topology of fs common to other pnictides superconductors , and that are thought to be the basic ingredient of the sc pairing ( namely the nearly nested cylindrical fermi surfaces ) .
in addition , the magnetic moment of fe site is predicted to be completely quenched in the ct as recently confirmed by nmr @xmath42 as spectra in pr doped cafe@xmath3as@xmath3 @xcite . a possible scenario to reconcile the theoretical description and
the experimental evidences could be the same proposed by proks and coworkers@xcite to explain the superconducting phase in pure cafe@xmath3as@xmath3 under pressure : superconductivity can exist because of the formation of a new tetragonal phase ( stabilized by chemical and/or strain / non - hydrostatic effects ) which prevents both the magnetic and the structural phase transition , then retaining the characteristic electronic properties that makes it unstable toward superconductivity .
we , thus , call for further experiments aimed to characterize the electronic , magnetic and structural properties of these new superconductors , in order to elucidate the nature of the non - collapsed tetragonal phase within the parameter - free dft - gga computational framework we explain the observed low - temperature and pressure phase transitions observed experimentally in re - doped cafe@xmath3as@xmath3 and we predict , giving a rationale , the pressure dependent structural trends .
we show that , when realized , the low temperature ct phases induced by re substitutions are indeed non - magnetic , with electronic band structures sharing the same features of the band structure of the ct phase of pure cafe@xmath3as@xmath3 .
characteristic features promoting the superconducting phase are missing .
for this reason , the superconducting instability , if it is confirmed to be present in the compressed phases of re - doped cafe@xmath3as@xmath3 seems to open a new chapter in the physics of iron - based superconductors .
however a scenario that we find more likely is that superconductivity appears in a non collapsed subphases as it was suggested by proes and coworkers for undoped cafe@xmath3as@xmath3 . |
grs 1915 + 105 is the first galactic object for which radio jets moving at superluminal speed have been observed @xcite .
radio observations have also led to the estimate of its distance ( @xmath012.5 kpc ) and the inclinations of the jets ( @xmath1 from the line of sight ) .
the source is considered to host a black hole because of its high x - ray luminosity and its similarities to gro 165540 , whose dynamical mass estimate indicate the presence of a black hole @xcite .
grs 1915 + 105 is a transient source , in the sense that it has made its appearance in the x - ray sky in 1992 @xcite , but most likely never returned into quiescence since then .
the rossi x - ray timing explorer ( rxte ) monitors regularly grs 1915 + 105 since the start of the mission . during this period ,
the source has shown a remarkable richness in variability , ranging from quasi - periodic burst - like events , deep regular dips and wild oscillations , alternated with quiescent periods @xcite .
we present here the results of the analysis of rxte data of grs 1915 + 105 and propose an interpretation for the variability in the spectral parameters .
a more detailed discussion can be found in @xcite .
rxte observed grs 1915 + 105 in many occasions . here
we present the results of the analysis of one of these observations , the one obtained on 1997 june 18 starting at 14:36 ut and ending at 15:35 ut , as it reproduces within one day most of the variability observed from this source .
the upper panel of figure 1 shows 1200 s of the 2 - 40 kev light curve .
it consists of a sequence of ` bursts ' of different duration with quiescent intervals in between .
all bursts start with a well - defined sharp peak and decay faster than they rise .
the longer bursts show oscillation ( or sub - bursts ) towards the end .
the complexity of the light curve seems to be untractable : the only obvious comment that can be made is that it consists of high flux intervals somehow alternated with low flux intervals . in order to quantify the timing structure of the oscillations
, we measured the length of the ` ups ' and ` downs ' , an unambiguous procedure given the square - wave structure of the light curve . excluding the fastest oscillations where this measurement is not easy , we obtain durations spanning from a few seconds to three minutes . to study the evolution of the spectrum of grs 1915 + 105 during these oscillations between ` up ' and ` down '
, we first produced a color - color ( c - c ) diagram of the whole observation a characteristic pattern emerged , pattern which is also observable , with small differences , in all other observations we analyzed , besides a few ( see below ) .
this strengthens the hypothesis that all the oscillations have the same nature .
an analysis of separate c - c diagrams for oscillations of different length ( see figure 2 ) shows systematic differences . as it can also be seen from the light curve , the ` ups ' are very similar to each other
, while the ` downs ' are different , in particular the longer the oscillation the deeper ( light curve ) and harder ( c - c ) is the ` down ' . both from the light curve and the c - c diagram it appears that the short ` low ' periods are similar to the last part of the long ones . since spectral analysis at the required time resolution is not possible , in order to transform the information in the c - c diagram into spectral parameters we divided it into small regions .
for all the populated regions in the diagram we accumulated 1 second time resolution energy spectra in 48 energy bands .
we measured a background spectrum from a blank sky observation , which we normalized to the highest energy channels where the contribution of the source was negligible .
we subtracted this background spectrum from each of the source energy spectra .
we used the latest detector response matrix available , and added a systematic error of 2% to account for the calibration uncertainties . for each of the regions in the c - c diagram we fitted the data with a `` standard ''
spectral model for bhcs , consisting of a disk - blackbody ( dbb ) model and a power law , both affected by interstellar absorption .
to avoid problems due to the background subtraction , and as we were only interested in the properties of the dbb component , we limited our fits to energies below 30 kev .
since both the distance to this system and the inclination of the accretion disk are known ( we assume that the jet is perpendicular to the disk ) , we can derive the inner radius of the accretion disk directly from the fits without significant additional uncertainties .
this rather complicated procedure allowed us to obtain time histories for the inner radius and the temperature of the disk ( bottom two panels of figure 1 ) . during ` ups ' the temperature is above 2 kev and the radius is stable around 20 km . during ` downs ' , the temperature drops to less than 1 kev and the radius increases .
we produced similar c - c diagrams for a number of other rxte observations of grs 1915 + 105 .
all the observations that we analyzed can be fitted in the same manner , except that of 1996 june 16th @xcite ( and some later observations in october 1997 ) .
the quasi - periodic bursts observed in many of the observations @xcite are consistent with repetitive short events like the ones described here .
this analysis brought us to consider more seriously the ` up ' and ` down ' paradigm : to be consistent with our previous work , we will call them ` outburst ' and ` quiescent ' states . a sequence of quiescence and outburst we call an ` event ' .
the main conclusion that can be derived from figure 1 is that the inner radius of the accretion disk is not constant in time .
if we interpret the small and stable radius of @xmath020 km observed during outburst as the minimum stable orbit around a black hole ( see @xcite for a more detailed discussion ) , this implies that during quiescence the central section of the disk disappears , i.e. a central hole appears in the disk .
the radius of this central hole varies between events , ranging in this observation between 30 and 90 km ( see figure 1 ) .
this can be interpreted within the model presented by us in @xcite , providing a unified picture of the variability observed in grs 1915 . in @xcite , we modeled the large amplitude changes as emptying and replenishing of the inner accretion disk caused by a viscous - thermal instability .
the small radius observed during the quiescent period was identified with the innermost stable orbit around the black hole , while the large radius during the burst phase was the radius of the emptied section of the disk .
the smaller oscillations were interpreted as failed attempts to empty the inner disk .
as it can be seen from figures 1 and 2 , from this observation we find that all variations , from major events like the ones described in @xcite to small oscillations observed at the end of a large event , can be modeled in exactly the same fashion . in this scenario ,
the `` flare state '' presented in @xcite is simply a sequence of small events following a big one , similar to the small oscillations in figure 1 .
both the spectral evolution and the duration of the event are determined by one parameter only , namely the radius of the missing inner section of the accretion disk .
it is natural to imagine that the bigger the missing inner section of the disk , the longer it will take to re - fill it . indeed ,
if we identify the re - filling time for each event with the time spent in ` quiescence ' , this is what is observed ( figure 3 ) .
following @xcite , we can naturally associate the re - filling time of an event @xmath2 to the viscous time scale of the radiation - pressure dominated part of the accretion disk .
this can be expressed as @xmath3 , where @xmath4 . using the expressions for the scale - height @xmath5 and sound speed @xmath6 found in @xcite
, we obtain : @xmath7 where @xmath8 , @xmath9 is the radius in units of @xmath10 cm , @xmath11 the central object mass in solar masses , and @xmath12 is the accretion rate in units of @xmath13 g / s .
notice that even the largest radii derived here are well within the radiation - pressure dominated part of the disk ( see equation 2 in @xcite ) .
the line in figure 3 represents the best fit to the data with a relation of the form @xmath14 .
the fit is excellent , with the exception of the point corresponding to the longest event .
the qualitative agreement with the theoretical expectation is striking , although by substituting the appropriate values for the mass and accretion rate we find that our best fit predicts rather small values of @xmath15 ( 0.004 and 0.05 for the schwarzschild and extreme kerr cases respectively ) .
this indicates a small viscosity in the disk , although we stress that @xmath16 is only a time scale , so that additional corrections might be necessary in order to allow a precise quantitative comparison .
let us follow an event and describe it in terms of the model .
the accretion rate @xmath17 provided by the secondary is constant ( within one observation ) . at the start of a quiescent period
, the disk has a central hole , whose radius is r@xmath18 . outside the hole ,
@xmath17 lies on a stable branch in the @xmath19 curve ( see @xcite ) . at all radii inside the hole ,
@xmath17 lies on the unstable branch : the disk lies on the lower ( stable ) branch .
this means that the hole does not have to be empty , but rather filled with gas whose radiation is too soft to be detected .
slowly the hole in the disk is re - filled by a steady accretion rate @xmath17 from outside .
each annulus of the disk will move along the lower branch of its s - curve in the @xmath19 plane trying to reach @xmath17 .
the local density at each radius increases as the annulus moves towards the unstable point at a speed determined by the local viscous time scale . during this period
, no changes are observed in the radius of the hole , since all the matter inside does not radiate in the pca band . in the xte data the disk appears to have a roughly constant radius .
the observed accretion rate during this phase is @xmath17 since it is determined from the spectrum of the radiation coming from stable regions . at the end of this phase
, one of the annuli will reach the unstable point and switch to the high-@xmath20 state , where the accretion rate is larger than @xmath17 , causing a chain - reaction that will `` switch on '' the inner disk .
the observed accretion rate is now higher than the external value @xmath17 since it is determined from the spectrum of radiation originating from the unstable part of the disk , through which a higher @xmath20 is flowing .
a smaller , hot radius is now observed . at the end of the outburst ,
the inner disk runs out of fuel and switches off , either jumping back to the @xmath21 state or possibly emptying completely .
a new hole in the disk is formed and a new cycle starts .
notice that in this scenario the more `` normal '' state for the source is the one at high count rates , where the disk extends all the way to the innermost stable orbit : in this state the energy spectrum is similar to that of conventional black - hole candidates ( see @xcite ) .
not only the start and end of a major burst , but also all the small amplitude oscillations within a burst show the same timing signature of decaying faster than they rise .
this is in agreement with what was already noticed in @xcite : the rise time is determined by the speed at which a heating wave moves through the central disk , while the faster decay time is due to the rapid fall of matter into the black hole ( or ejection into a relativistic jet ) .
it has been found that when the hardness ratio hr2 exceeds 0.1 , the power density spectrum is similar to the one observed in black - hole transient systems during the very high state ( see @xcite ) . in these occasions , a strong 1 - 6 hz qpo peak was found , positively correlated with the count rate @xcite .
the limit [email protected] is an indication that the source was in a quiescent state .
our spectral results show that the fast timing features ( both qpo and band - limited noise , see @xcite ) can not originate from the innermost regions of the optically thick accretion disk , since those are missing during the quiescent phases .
the fact that the qpo frequency increases with count rate is in qualitative agreement with the model , since a higher count rate indicates a smaller inner disk radius , and therefore shorter time scales
. however , the qpo seems to be associated to the power law spectral component and not from the disk component , making this phenomenology very difficult to understand .
the radii for the disappearing region of the disk found here are considerably smaller than that reported in @xcite from an observation where the length of the events was substantially longer .
this is entirely due to the improved knowledge of the spectral response of the pca .
interestingly , the length of the quiescent period and the maximum inner radius observed in @xcite are in agreement with the curve in figure 3 when the time is properly re - normalized to take into account the difference in accretion rate ( see equation 1 ) .
our model reduces the complication of the spectral evolution of grs 1915 + 105 to one parameter : the radius of the missing section of the accretion disk .
the structure of the variability is however yet to be explained .
the question to be answered is : what determines the length of the next outburst ?
the model outlined above does not provide an answer , but allows us to reformulate the question in more physical terms .
what determines how large the next missing section of the disk will be ? in some observations the events are very regular , in some others they seem to be random , and in some others no events are observed at all .
the latter might be part of extremely long quiescent intervals . in the framework of this model ,
it is clear that a regular quasi - periodic structure of events arises naturally if the radii of the holes are all similar . in the observation reported here a striking one - to - one relation between quiescent and burst time
is observed @xcite , a relation which applies to the burst and i quiescent states of the observation presented in @xcite but is obviously not satisfied in other observations ( see e.g. @xcite ) nor during the `` flare '' state of @xcite . moreover , as already mentioned , a few observations among the ones we analyzed does not fit this pattern and requires a different interpretation ( 1996 june 16th is a template ) .
both the timing structure and the spectral evolution in these observation is radically different from what presented here .
nevertheless , our model provides a satisfactory interpretation of the cause of the changes in the x - ray emission .
grs 1915 + 105 is a remarkable x - ray source . despite it uniqueness ( and because of it ) , we have the chance to learn something fundamental about accretion disks around black holes .
some of its characteristics ( the x - ray variability being the major one ) are indeed peculiar , but others ( the band - limited noise and qpo in power spectra , the thermal disk and the hard power law components in the energy spectra ) are remarkably similar to many ` normal ' black hole candidates .
whatever the different states in these sources are , so far we have been observing them on long time scales ( years for persistent sources like cyg x1 and gx 3394 , months for black hole transients ) . here
we can observe more spectral and timing variations within one hour of observation .
in addition , coordinated observations in different bands are beginning to provide us with the first links between what happens in the accretion disk ( observed in x - rays ) and the ejection of relativistic jets ( observed in ir and radio ) @xcite
. 9 mirabel i.f . ,
& rodrguez l.f . , 1994 ,
nature , 371 , 46 .
bailyn c.d . , et al . , 1995 , nature , 378 , 157 .
castro - tirado a.j .
, brandt s. , & lund s. , 1992 , iauc 5590 .
greiner j. , morgan e. , & remillard r.a .
, 1996 , apj , 473 , l107 .
morgan e. , remillard r.a . , & greiner j. , 1997 , apj , 482 , 993 .
chen x. , swank j.h . , & taam r.e . , 1997 ,
apj , 477 , l41 .
belloni t. , et al . , 1997a , apj , 479 , l145 .
belloni t. , et al .
, 1997b , apj , 488 , l109 .
taam r.e , chen x. , & swank j.h . , 1997 ,
apj , 485 , l83 .
frank j. , king a. , & raine d. , 1992 , `` accretion power in astrophysics '' , cambridge univ .
press , cambridge .
tanaka y. , & lewin w.h.g .
, 1995 , in `` x - ray binaries '' , eds .
lewin w.h.g . ,
van paradijs j. , & van den heuvel e.p.j .
, cambridge univ . press , cambridge . van der klis m. , 1995 , in `` x - ray binaries '' , eds .
lewin w.h.g . ,
van paradijs j. , & van den heuvel e.p.j .
, cambridge univ .
press , cambridge .
pooley g.g .
, & fender r.p . , 1997 , mnras , 292 , 925 .
eikenberry s.s . , et al . , 1998 , apj , in press .
mirabel i.f . , et al . , 1998 ,
a&a , 330 , l9 . | grs 1915 + 105 , one of the two known galactic microquasars , shows an extremely complex variability in the x - ray band , comparable to no other x - ray source in the sky . making use of rxte / pca data
, we have analyzed the x - ray spectral distribution throughout the variability .
we find that all variations can be attributed to the rapid appearance and disappearance of the inner region of an optically - thick accretion disk .
since the time scale for each event is related to the maximum radius of the disappearing region , the difference in time structure is due to the time distribution of such radii .
the observed relation between the extent of the missing inner region of the disk and the duration of an event is in remarkable agreement with the expected radius dependence of the viscous time scale of the radiation - dominated region of an accretion disk . |
dynamical relaxation processes near massive black holes ( mbh ) in galactic centers affect the rates of strong stellar interactions with the mbh , such as tidal disruption , tidal dissipation , or gravitational wave ( gw ) emission ( e.g. * ? ? ?
these relaxation processes may also be reflected by the dynamical properties of the different stellar populations there @xcite , as observed in the galactic center @xcite .
of particular importance , in anticipation of the planned laser interferometer space antenna ( lisa ) gw detector , is to understand the role of relaxation in regulating the rate of gw emission events from compact remnants undergoing quasi - periodic extreme mass ratio inspiral ( emri ) into mbhs . two - body relaxation , or non - coherent relaxation ( nr ) , is inherent to any discrete large - n system , due to the cumulative effect of uncorrelated two - body encounters .
these cause the orbital energy @xmath1 and the angular momentum @xmath2 to change in a random - walk fashion ( @xmath3 ) on the typically long nr timescale @xmath4 .
in contrast , when the gravitational potential has approximate symmetries that restrict orbital evolution ( e.g. fixed ellipses in a keplerian potential ; fixed orbital planes in a spherical potential ) , the perturbations on a test star are no longer random , but correlated , leading to coherent ( @xmath5 ) torquing of @xmath2 on short timescales , while the symmetries hold . over longer times , this results in resonant relaxation ( rr ) ( @xcite ; [ ss : rr ] ) , a rapid random walk of @xmath2 on the typically short rr timescale @xmath6 .
rr in a near - keplerian potential can change both the direction and magnitude of @xmath7 ( `` scalar rr '' ) , thereby driving stars to near - radial orbits that interact strongly with the mbh .
rr in a near - spherical potential can only change the direction of @xmath7 ( `` vector rr '' ) .
rr is particularly relevant in the potential near a mbh , where compact emri candidates originate .
@xcite show that rr dominates emri source dynamics . depending on its still poorly - determined efficiency , rr can either increase the emri rate over that predicted assuming nr only , or if too efficient , it can strongly suppress the emri rate by throwing the compact remnants into infall ( plunge ) orbits ( cf fig .
[ f : emri ] below ) that emit a single , non - periodic and hard to detect gw burst .
a prime motivation for the systematic numerical investigation of rr efficiency presented here , are the still open questions about the implications of rr for emri rates and orbital properties .
this paper is organized as follows . in
[ s : theory ] we briefly review the theory of nr and rr relaxation and derive a new relation between scalar and vector rr . in
[ s : acf ] we describe our method of analyzing and quantifying the effects of rr in @xmath0-body simulations , which are described in [ s : simulations ] .
we present our results in [ s : results ] and discuss and summarize them in [ s : conclusions ] .
the nr time for @xmath1-relaxation , @xmath8 , corresponds to the time it takes non - coherent 2-body interactions to change the stellar orbital energy by order of itself , @xmath9 ( by stellar dynamical definition convention , @xmath10 for a bound orbit ) .
similarly , the nr time for @xmath2-relaxation , @xmath11 , corresponds to the time it takes the stellar orbital angular momentum to change by order of the circular angular momentum @xmath12 , where near a mbh of mass @xmath13 , @xmath14 .
the @xmath1-relaxation timescale can be estimated by considering the rate @xmath15 of gravitational collisions in a system of size @xmath16 at a relative velocity @xmath17 , between a test star and @xmath0 field stars of mass @xmath18 and space density @xmath19 , at the minimal impact parameter where the small angle deflection assumption still holds , @xmath20 .
the collision rate is then @xmath21 .
taking into account also collisions at larger impact parameters increases the rate by the coulomb logarithm factor @xmath22 .
therefore , @xmath23 . near the mbh @xmath24 , and so @xmath25 , where @xmath26 is the mass ratio .
when the stars move under the influence of the central mbh ( @xmath27 ) , the relaxation time can be expressed as @xmath28 , where @xmath29 is the keplerian period . following the notation of @xcite ( rt96 ) , the nr changes in @xmath1 , @xmath2 and @xmath7 over the dimensionless time - lag @xmath30 are@xmath31 where @xmath32 and @xmath33 are dimensionless constants , whose exact values are system - dependent and
can not be estimated with accuracy without detailed calculations or simulations .
the corresponding nr timescales are related to these coefficients by @xmath34 , @xmath35 and @xmath36 . when the potential has symmetries that restrict the orbital evolution , for example to fixed ellipses in the potential of a point mass , or to planar annuli in a spherical potential , the perturbations on a test star are no longer random , but correlated .
this leads to a coherent changes in @xmath7 on times @xmath37 , @xmath38 , by the residual torque @xmath39 exerted by the @xmath0 randomly oriented , orbit - averaged mass distributions of the surrounding stars ( mass wires for elliptical orbits in a kepler potential , mass annuli for rosette - like orbits in a spherical potential ) .
the coherence time @xmath40 is set by deviations from the true symmetry , which lead to a gradual orbital drift and to the randomization of @xmath41 .
for example , the enclosed stellar mass leads to non - keplerian retrograde precession ; general relativity leads to prograde precession .
ultimately , the coherent torques themselves randomize the orbits ( alternatively , this can be viewed as the result of potential fluctuations due to the finite number of stars ) . the effective coherence time is set by the shortest de - coherence ( quenching ) process in the system . the accumulated change over @xmath40 , @xmath42 , then becomes the basic step - size , or mean free path in @xmath7-space , for the long - term ( @xmath43 ) non - coherent ( @xmath3 ) relaxation of @xmath7 .
since this step - size is large , rr can be much faster than nr .
the rr timescale @xmath44 is then defined by @xmath45 .
note that the relaxation of _
@xmath1 _ is not affected by rr because the potential of the system is stationary on the coherence timescale , and so @xmath1 changes incoherently on all time scales .
the torques exerted by elliptical mass wires in a kepler potential can change both the direction and magnitude of * j*. in contrast , the torques exerted by planar annuli can only change the direction of * j*. here we consider only newtonian dynamics . the coherence timescale for scalar rr
is determined by the time it takes for the orbital apsis to precess by angle @xmath46 due to the potential of the enclosed stellar mass ( `` mass precession''),@xmath47 where @xmath48 is an @xmath49 factor reflecting the approximations in this estimate .
the coherence timescale for vector rr is determined by the time it takes for the coherent torques to change @xmath50 ( alternatively , this is the timescale @xmath51 to accumulate @xmath49 fluctuations in the stellar potential @xmath52 relative to the total gravitational potential @xmath53 as the stars rotate by @xmath46 on their orbits ) , @xmath54p\,,\ ] ] where @xmath55 is an @xmath49 factor , @xmath56 and where the approximate equality is for the keplerian limit @xmath57 .
following rt96 , the rr changes in @xmath2 and @xmath7 during the coherent phase @xmath58 can be expressed as @xmath59 where the @xmath49 dimensionless coefficients @xmath60 and @xmath61 depend on the parameters of the system and reflect the uncertainties introduced by the various approximations and simplification of this analysis .
accurate determination of their values requires detailed calculations or simulations .
the scalar rr change @xmath62 on time - lags @xmath63 is then @xmath64 and the scalar rr timescale is @xmath65p\,.\ ] ] the rr efficiency factor @xmath66 defined by @xcite expresses how much shorter @xmath67 is relative to the value estimated by rt96 .
scalar rr is faster than nr by a factor @xmath68 .
similarly , the vector rr change @xmath69 on time - lags @xmath70 in the keplerian limit is @xmath71 and the vector rr timescale is @xmath72p\,.\ ] ] rt96 performed a limited set of near - keplerian simulations to check their predictions .
they analyzed the results both star by star and in the average and observed the coherent growth of @xmath73 and @xmath74 relative to the simulation s initial values .
although the evolution of the of these quantities for any single star was very noisy and the proportionality factors had a very large scatter , rr was clearly observed , as predicted .
the population averages of the scalar and vector coefficients @xmath75 and @xmath76 , are not independent quantities .
rather , @xmath77 , and @xmath78 ( on timescales @xmath79 ) , where the constant @xmath80 depends on the averaging procedure . here
we average @xmath81 and @xmath82 by the rms over the stellar population ( [ s : acf ] ) .
we focus on the limit where @xmath83 , and assume that the change is isotropic on average , @xmath84 , ( _ @xmath85 _ ) , as is indicated by our simulations ( [ s : simulations ] ) .
the rms of vector rr is @xmath86 . defining the @xmath87-axis along @xmath7 , then @xmath88 in the small change limit , and
its rms is @xmath89 .
therefore ( see eqs .
[ e : etasl ] , [ e : etavl ] , [ e : betas ] , [ e : betav ] ) @xmath90 for all @xmath91 and @xmath92 for @xmath93 , as is indeed seen in the simulations is used to define @xmath76 , then @xmath94 . ]
[ f : q1e6 ] ) .
the constant ratio between the population averages of the vector and scalar coefficients is in part a geometrical effect ( 1d vs 3d changes ) and in part a reflection of the symmetries of the perturbations ( isotropic @xmath95 ) .
this strict proportionality is not expected to hold when @xmath96 , for example when only very eccentric stars ( @xmath97 ) are included in the sample , or for very long time - lags .
given the high computational cost of the @xmath0-body simulations , and the very large variance in the evolution of individual orbits , it is essential to extract the rr signal as efficiently and robustly as possible from the simulated data .
after some experimentation , we adopted the auto - correlation analysis for detecting and measuring rr in @xmath0-body simulation snapshots .
our procedure is as follows .
\(1 ) the stellar phase space coordinates are transformed to the rest - frame of the mbh , which is almost identical with the center of mass in the near - keplerian system close to the mbh .
\(2 ) the energy @xmath98 , angular momentum @xmath99 , circular angular momentum @xmath100 and keplerian period @xmath101 are calculated for the @xmath102th star @xmath103 at discrete times @xmath104 in the simulation ( `` snapshots '' ) .
\(3 ) to make full use of the data , we assign a normalized time - lag @xmath105 for each pair of times @xmath106 . for each of the _ @xmath0 _
stars , we calculate the normalized energy and angular momentum differences at all lags , @xmath107 , @xmath108 and @xmath109 .
\(4 ) the differences from all stars are binned into discrete @xmath91-bins according to their associated @xmath110 .
the bin rms values and their standard deviations are calculated and plotted against the bin s average time - lag @xmath111 , thereby creating the auto - correlation curve . by using all possible time - lags recorded in the simulation ,
this approach makes maximal use of the entire data set and averages over the strongly fluctuating individual relaxation curves ( cf the rt96 procedure , [ ss : rr ] ) .
however , this method is not entirely free of bin - to - bin bias .
since the number of orbital periods completed by a star with a mean period @xmath112 over the simulation time @xmath113 is @xmath114 , long - period stars will not contribute to a high-@xmath91 bin , @xmath115 if @xmath116 .
conversely , short - period stars will not contribute to a low-@xmath91 bin , @xmath115 , if @xmath117 , where @xmath118 is set by the minimal time difference between consecutive snapshots .
in extreme cases some stars may not contribute to the relaxation curve at any @xmath115 .
since long and short period stars could well have systematically different responses to rr , this introduces bias to the relaxation curve .
this bias can be minimized , at the cost of losing some information , by using only the middle range of the @xmath91 -bins , or at a computational cost , by longer simulations with a higher snapshot rate .
phase drift in a 2-body system over a time of @xmath119 for a @xmath120 mass ratio and an extreme eccentricity of @xmath121 .
the line is the initial phase curve ( @xmath122 as function of @xmath123 ) ; the dots are the simulated data .
the phase drift of @xmath124 corresponds to @xmath125 . ] the measured ( points ) and fitted ( lines ) auto - correlation curves for @xmath126 , @xmath127 and @xmath74 in a @xmath128 , @xmath129 simulation with @xmath130 particles .
the points are the results of the @xmath0-body simulation , the thick lines are the predicted theoretical curves ; the thin straight lines show the asymptotic ( @xmath131 ) linear behavior .
the best fit parameters and related quantities are also listed . ]
[ f : q1e6 ] our @xmath0-body code uses a 5th order runge - kutta integrator with individual time - steps and pair - wise k - s regularization @xcite , without gravity softening .
the time steps were chosen to conserve total energy at the level of @xmath132 , well below the nr energy changes expected in the simulations .
the @xmath0-body code must maintain phase coherence to a high enough accuracy to reliably simulate rr .
it is thus important to verify that the numerical phase drift in the simulation is much smaller than that due to physical processes , such as mass precession .
experience shows that most of the phase drift @xmath133 accumulates near periapse , where the acceleration is largest , and is therefore more pronounced in eccentric orbits .
we estimated the phase drift by integrating a highly eccentric ( @xmath134 ) 2-body system over many orbits , and plotted the relative distance between the 2 masses as function of the orbital phase ( fig .
[ f : phase ] ) , as it evolved during the simulation .
the phase spread relative to the theoretical phase curve , @xmath133 , was measured near @xmath135 .
we define the phase de - coherence timescale as @xmath136 .
we estimate conservatively that our near - keplerian simulation typically have @xmath137 ( this is for highly eccentric orbits .
we do not detect any phase drift up to @xmath138 in orbits with moderate eccentricities ) . since @xmath139 for the models simulated here ( cf fig .
[ f : evol ] ) , it can be safely neglected .
our simulations consisted of 200 particles ( including the mbh as a free particle ) .
the initial orbital semi - major axes were randomly drawn from a @xmath140 distribution for @xmath141 , for @xmath142 in the range @xmath143 , with eccentricities drawn from a thermal @xmath144 distribution , with random phases , orbital orientations and isotropic velocities .
this distribution corresponds to an @xmath145 number density distribution with an outer cutoff at radius @xmath146 from the mbh ( in dimensionless units where @xmath147 , @xmath148 ) .
these stellar cusps span a wide range of possible physical scenarios ( e.g. * ? ? ? * ) , and in particular those of lisa targets , which are expected to be relaxed galactic nuclei @xcite .
a typical simulation lasted a @xmath149 system orbital times and resulted in @xmath149 snapshots of the system configuration . in order to decrease the statistical errors we ran @xmath150@xmath151 simulations with different initial conditions for each of the ( @xmath152,@xmath153 ) models we studied . in fig .
( [ f : q1e6 ] ) we plot the measured auto - correlation curves in a typical simulation for @xmath154 : @xmath155 , @xmath156 and @xmath157 .
the curves reflect the joint effects of nr and rr . the coefficients @xmath75 and @xmath76 are measured by fitting the auto - correlation curves in the coherent phase to the functions @xmath158 where the two terms in the square root express the contributions of nr and rr , respectively .
note that on short timescales , @xmath159 , the rr auto - correlation curve rises as @xmath160 due to the effect of nr .
it then rises as @xmath91 in the rr - dominated coherent phase at times @xmath161 , before turning over again to a @xmath160 rise in the accelerated random - walk phase at @xmath162 ( not shown in fig .
[ f : q1e6 ] ) .
the nr and rr parameters were derived from the best 2-parameter fits of eqs .
( [ e : alphal ] , [ e : djs ] , [ e : djv ] ) to the data
. to control the star to star variance , we limited our analysis to time - lag bins that sampled at least 0.75 of the stars in the simulation ( [ s : acf ] ) .
the excellent fit of the data points to the predicted auto - correlation curves seen in fig .
( [ f : q1e6 ] ) indicates that rr is present and measurable .
although the auto - correlation analysis stabilizes against star to star scatter in a single simulation , we still find a large simulation to simulation scatter in the derived values of the coefficients .
we therefore constructed a large grid of near - keplerian models ( within the computational time limitations ) , where rr should be clearly detected .
@xmath163 .[t : coeff]measured nr and rr coefficients @xmath164 . [ cols="^,^,^,^,^,^,^,^ " , ] as function of mass ratio @xmath153 and for stellar density profiles with logarithmic slopes of @xmath165 .
a @xmath166 curve is shown to guide the eye . ]
[ f : alphal ] ) , for the measured nr scalar angular momentum coefficient @xmath167 ( bottom points ) and the vector angular momentum coefficients @xmath168 ( top points ) . ]
[ f : etasl ] ) , for the measured rr scalar angular momentum coefficient @xmath60 ( bottom points ) and vector angular momentum coefficient @xmath61 ( top points ) .
rr does not depend on a coulomb factor . ]
[ f : betas ] we summarize our results in table ( [ t : coeff ] ) and in figs .
( [ f : alphal ] , [ f : etasl ] , [ f : betas ] ) . the coefficients @xmath169 and @xmath170 do not directly express the intrinsic properties of nr , since they should vary as @xmath171 over the @xmath153-range of our models .
this small fractional difference and the relatively low statistics of our simulation suite may explain why we do not detect a clear @xmath153-dependence in these quantities .
since we do not see a clear @xmath152 or @xmath153 dependence , we adopt as the best fit estimates of the values of @xmath172 , @xmath75 and @xmath76 their grand average over all the simulations ( table [ t : coeff ] ) .
the measured ratio @xmath173 is consistent with the predicted ratio of @xmath174 ( [ sss : sv ] ) .
rt96 derived from their simulations smaller mean values for @xmath60 ( @xmath175 , as compared to @xmath176 here ) and for @xmath169 and @xmath167 ( @xmath177 and @xmath178 as compared to @xmath179 and @xmath180 here ) .
at least part of the difference in @xmath169 and @xmath167 can be traced to their use of a softening length @xmath181 in the calculation of the gravitational force .
this decreases the effective value of the coulomb factor , since @xmath182 and so @xmath183 , about twice as small as in our simulations .
indeed , rt96 noted that decreasing the softening length to @xmath184 led to an increased value of @xmath185 .
the change in the emri and plunge rates relative to that predicted assuming only nr , as function of the rr efficiency @xmath186 ( adapted from @xcite , fig .
the rt96 estimate of @xmath60 ( @xmath187 ) predicts an increase of @xmath188 in the emri rate due to rr , close to the maximum .
our ( eka08 ) new measured efficient rr ( @xmath189 ) predicts a higher plunge rate and thus a smaller increase of @xmath190 in the emri rate . ]
it is difficult to trace the specific reason for the discrepancy between our best estimate value for @xmath60 and that derived by rt96 , given the many differences in both the simulations and the methods of analysis .
our statistics are better due to the larger number of simulations and more efficient use of the data , and the analysis here is more rigorous .
we therefore briefly consider the implications of this revised value of @xmath60 for emri rates .
@xcite parametrized the rr efficiency by a factor @xmath191^{2}$ ] , and derived the @xmath186 dependence of the branching ratio of the emri and infall ( plunge ) rates ( fig .
[ f : emri ] ) .
the rt96 value @xmath187 happens to lie very close to the maximum of the rr - accelerated emri rate .
we find here @xmath192 , which corresponds to a factor @xmath193 increase in the emri rate compared to that estimated for nr only , but is a factor @xmath194 smaller than implied by the rt96 value , because the higher rr efficiency leads to a higher plunge rate at the expense of the inspiral rate .
we characterized and measured the mean efficiency coefficients of nr ( @xmath195 ) and rr ( @xmath196 ) in newtonian @xmath0-body simulations of isotropic , thermal , near - keplerian stellar cusps around a mbh .
we derived a simple analytical form for the rms rr auto - correlation curves of @xmath7 and @xmath2 , and showed that the two are simply proportional to each other .
we then measured these coefficients in a large suite of small scale @xmath0-body simulations with different stellar density distributions and mbh / star mass ratios .
we found no statistically significant trends in the values of these coefficients as function of the system properties .
this may require better statistics .
our measured rr efficiency suggests that rr increases the emri rate by a factor of @xmath193 above what is predicted for nr only .
this estimate of rr efficiency is consistent with that suggested by the analysis of the dynamical properties of the different stellar populations in the galactic center @xcite .
however , this conclusion is still preliminary , since several important open issues remain , which should be addressed by larger scale simulations . these include ( 1 ) the dependence of the rr coefficients on the orbit of the test star , for example its eccentricity @xcite .
we find that @xmath197 is not enough for reliable statistics on sub - samples within given _ e _ or
_ j_-bins .
the eccentricity dependence of rr is particularly relevant for the supply rate of stars to the mbh from @xmath198 orbits ( the loss - cone refilling problem ) .
( 2 ) the effects of a stellar mass spectrum .
this will likely affect rr by decreasing the rr timescale , and changing the stellar density distribution through strong mass segregation ( @xcite ; alexander & hopman 2008 , in prep . ) .
( 3 ) the robustness of rr against perturbations from the larger non - keplerian stellar system in which the inner near - keplerian region of interest is embedded . ( 4 ) the role of post - newtonian effects in rr , such as general relativistic precession and gw emission .
these are expected to play a key role in enabling inspiral by quenching rr just as the compact remnant enters the emri phase , and in regulating the gw inspiral rate @xcite . | resonant relaxation ( rr ) is a rapid relaxation process that operates in the nearly - keplerian potential near a massive black hole ( mbh ) .
rr dominates the dynamics of compact remnants that inspiral into a mbh and emit gravitational waves ( extreme mass ratio inspiral events , emris ) .
rr can either increase the emri rate , or strongly suppress it , depending on its still poorly - determined efficiency .
we use small - scale newtonian @xmath0-body simulations to measure the rr efficiency and to explore its possible dependence on the stellar number density profile around the mbh , and the mass - ratio between the mbh and a star ( a single - mass stellar population is assumed ) .
we develop an efficient and robust procedure for detecting and measuring rr in @xmath0-body simulations .
we present a suite of simulations with a range of stellar density profiles and mass - ratios , and measure the mean rr efficiency in the near - keplerian limit .
we do not find a statistically significant dependence on the density profile or the mass - ratio .
our numerical determination of the rr efficiency in the newtonian , single - mass population approximations , suggests that rr will likely _ enhance _ the emri rate by a factor of a few over the rates predicted assuming only slow stochastic two - body relaxation . |
0 the study of harmonic functions on complete riemannian manifolds , i.e. , the solutions of the equation @xmath1 where @xmath2 is the laplace - beltrami operator , lies at the interface of analysis , geometry and stochastics .
indeed , there is a deep interplay between geometry , harmonic function theory , and the long - term behaviour of brownian motion .
negative curvature amplifies the tendency of brownian motion to move away from its starting point and , if topologically possible , to wander out to infinity .
on the other hand , non - trivial asymptotic properties of brownian paths for large time correspond with non - trivial bounded harmonic functions on the manifold .
there is plenty of open questions concerning the richness of certain spaces of harmonic functions on riemannian manifolds . even in the case of
simply connected negatively curved riemannian manifolds basic questions are still open .
for instance , there is not much known about the following problem posed by wu in 1983 .
[ conj : wu ] if @xmath0 is a simply - connected complete riemannian manifold with sectional curvature @xmath3 , do there exist @xmath4 bounded harmonic functions ( @xmath5 ) which give global coordinates on @xmath0 ? it should be remarked that , under the given assumptions , it is even not known in general whether there exist non - trivial bounded harmonic functions at all .
this question is the content of the famous greene - wu conjecture which asserts existence of non - constant bounded harmonic functions under slightly more precise curvature assumptions .
[ conj : greene - wu ] let @xmath0 be a simply - connected complete riemannian manifold of non - positive sectional curvature and @xmath6 such that @xmath7 for some @xmath8 compact , @xmath9 and @xmath10 .
then @xmath0 carries non - constant bounded harmonic functions . concerning conjecture [ conj : greene - wu ]
substantial progress has been made since the pioneering work of anderson @xcite , sullivan @xcite , and anderson - schoen @xcite .
nevertheless , the role of lower curvature bounds is far from being understood . from a probabilistic point of view ,
conjecture [ conj : greene - wu ] concerns the eventual behaviour of brownian motion on these manifolds as time goes to infinity . indeed , for any riemannian manifold , we have the following probabilistic characterization . for a riemannian manifold @xmath11 the following two conditions are equivalent : 1
. there exist non - constant bounded harmonic functions on @xmath0 .
bm has non - trivial exit sets , i.e. , if @xmath12 is a brownian motion on @xmath0 then there exist open sets @xmath13 in the @xmath14-point compactification @xmath15 of @xmath0 such that @xmath16 more precisely , brownian motion @xmath12 on @xmath0 may be realized on the space @xmath17 of continuous paths with values in the @xmath14-point compactification @xmath15 of @xmath0 , equipped with the standard filtration @xmath18 generated by the coordinate projections @xmath19 up to time @xmath20 .
let @xmath21 be the lifetime of @xmath12 and let @xmath22 denote the shift - invariant @xmath23-field on @xmath24 . then there is a canonical isomorphism between the space @xmath25 of bounded harmonic functions on @xmath0 and the set @xmath26 of bounded @xmath22-measurable random variables up to equivalence , given as follows : @xmath27 ( bounded shift - invariant random variables are considered as equivalent , if they agree @xmath28-a.e .
, for each @xmath29 . )
note that the isomorphism is well defined by the martingale convergence theorem , and that the inverse map to is given by taking expectations : @xmath30.\ ] ] in particular , @xmath31 is a bounded harmonic function on @xmath0 , and _ non - constant _ if and only if @xmath13 is a _ non - trivial _ exit set .
now let @xmath11 be a _ cartan - hadamard manifold _ , i.e. , a simply connected complete riemannian manifold of non - positive sectional curvature .
( all manifolds are supposed to be connected ) . in terms of the exponential map @xmath32 at a fixed base point @xmath6
, we identify @xmath33 . via pullback of the metric on @xmath0
, we get an isometric isomorphism @xmath34 , and in particular , @xmath350,\infty[}\times s^{n-1}$ ] . in terms of such global polar coordinates , brownian motions @xmath12 on @xmath0
may be decomposed into their radial and angular part , @xmath36 where @xmath37 and where @xmath38 takes values in @xmath39 . for a cartan - hadamard manifold @xmath0 of dimension
@xmath4 there is a natural geometric boundary , the _ sphere at infinity _
@xmath40 , such that @xmath41 equipped with the _ cone topology _ is homeomorphic to the unit ball @xmath42 with boundary @xmath43 , cf .
@xcite , @xcite . in terms of polar coordinates on @xmath0 , a sequence @xmath44 of points in @xmath0 converges to a point of @xmath40 if and only if @xmath45 and @xmath46 .
given a continuous function @xmath47 the _ dirichlet problem at infinity _ is to find a harmonic function @xmath48 which extends continuously to @xmath40 and there coincides with the given function @xmath49 , i.e.,@xmath50 the dirichlet problem at infinity is called _ solvable _ if this is possible for every such function @xmath49 . in this case a rich class of non - trivial bounded harmonic functions on @xmath0 can be constructed via solutions of the dirichlet problem at infinity . in 1983
, anderson @xcite proved that the dirichlet problem at infinity is indeed uniquely solvable for cartan - hadamard manifolds of pinched negative curvature , i.e. for complete simply connected riemannian manifolds @xmath0 whose sectional curvatures satisfy @xmath51 where @xmath52 are arbitrary constants .
the proof uses _ barrier functions _ and perron s classical method to construct harmonic functions . along the same ideas
choi @xcite showed in 1984 that in rotational symmetric case of a _ model _ @xmath11 the dirichlet problem at infinity is solvable if the radial curvature is bounded from above by @xmath53 .
hereby a riemannian manifold @xmath11 is called _ model _ if it possesses a pole @xmath54 and every linear isometry @xmath55 can be realized as the differential of an isometry @xmath56 with @xmath57 . choi
@xcite furthermore provides a criterion , the _ convex conic neighbourhood condition _ , which is sufficient for solvability of the dirichlet problem at infinity .
a cartan - hadamard manifold @xmath0 satisfies the _ convex conic neighbourhood condition at @xmath58 _ if for any @xmath59 , @xmath60 , there exist subsets @xmath61 and @xmath62 containing @xmath63 and @xmath64 respectively , such that @xmath61 and @xmath65 are disjoint open sets of @xmath66 in terms of the cone topology and @xmath67 is convex with @xmath68-boundary .
if this condition is satisfied for all @xmath69 , @xmath0 is said to satisfy the _ convex conic neighbourhood condition_. it is shown in choi @xcite that the dirichlet problem at infinity is solvable for a cartan - hadamard manifold @xmath0 with sectional curvature bounded from above by @xmath70 for some @xmath9 , if @xmath0 satisfies the convex conic neighbourhood condition . in probabilistic terms ,
if the dirichlet problem at infinity for @xmath0 is solvable and almost surely @xmath71 exists in @xmath40 , where @xmath72 is a brownian motion on @xmath0 with lifetime @xmath73 , the unique solution @xmath48 to the dirichlet problem at infinity with boundary function @xmath49 is given as @xmath74.\ ] ] here @xmath75 is a brownian motion starting at @xmath29 .
conversely , supposing that for brownian motion @xmath75 on @xmath0 almost surely @xmath76 exists in @xmath40 for each @xmath29 , one may consider the _ harmonic measure _
@xmath77 on @xmath40 , where for a borel set @xmath78 @xmath79 then , for any borel set @xmath78 , the assignment @xmath80 defines a bounded harmonic function @xmath81 on @xmath0 . by the maximum principle @xmath81 is either identically equal to @xmath82 or @xmath14 or takes its values in @xmath830,1[$ ] .
furthermore , all harmonic measures @xmath77 on @xmath40 are equivalent .
thus , by showing that the harmonic measure class on @xmath40 is non - trivial , we can construct non - trivial bounded harmonic functions on @xmath0 . to show that , for a given continuous boundary function @xmath47 , the harmonic function @xmath84 extends continuously to
the boundary @xmath40 and takes there the prescribed boundary values @xmath49 , we have to show that , whenever a sequence of points @xmath85 in @xmath0 converges to @xmath86 then the measures @xmath87 converge weakly to the dirac measure at @xmath88 .
in particular , for a continuous boundary function @xmath89 , the unique solution to the dirichlet problem at infinity is given by formula .
the first results in this direction have been obtained by prat @xcite between 1971 and 1975 .
he proved that on a cartan - hadamard manifold with sectional curvature bounded from above by a negative constant @xmath90 , @xmath91 , brownian motion is transient , i.e. , almost surely all paths of the brownian motion exit from @xmath0 at the sphere at infinity @xcite .
if in addition the sectional curvatures are bounded from below by a constant @xmath92 , @xmath93 , he showed that the angular part @xmath38 of the brownian motion almost surely converges as @xmath94 . in 1976 , kifer @xcite presented a stochastic proof that on cartan - hadamard manifolds with sectional curvature pinched between two strictly negative constants and satisfying a certain additional technical condition , the dirichlet problem at infinity can be uniquely solved .
the proof there was given in explicit terms for the two dimensional case .
the case of a cartan - hadamard manifold @xmath11 with pinched negative curvature without additional conditions and arbitrary dimension was finally treated in kifer @xcite . independently of anderson , in 1983 , sullivan @xcite gave a stochastic proof of the fact that on a cartan - hadamard manifold with pinched negative curvature the dirichlet problem at infinity is uniquely solvable .
the crucial point has been to prove that the harmonic measure class is non - trivial in this case .
the harmonic measure class on @xmath95 is positive on each non - void open set .
in fact , if @xmath85 in @xmath0 converges to @xmath88 in @xmath40 , then the poisson hitting measures @xmath87 tend weakly to the dirac mass at @xmath88 .
in the special case of a riemannian surface @xmath0 of negative curvature bounded from above by a negative constant , kendall @xcite gave a simple stochastic proof that the dirichlet problem at infinity is uniquely solvable .
he thereby used the fact that every geodesic on the riemannian surface `` joining '' two different points on the sphere at infinity divides the surface into two disjoint half - parts .
starting from a point @xmath63 on @xmath0 , with non - trivial probability brownian motion will eventually stay in one of the two half - parts up to its lifetime .
as this is valid for every geodesic and every starting point @xmath63 , the non - triviality of the harmonic measure class on @xmath40 follows . concerning the case of cartan - hadamard manifolds of arbitrary dimension several results
have been published how the pinched curvature assumption can be relaxed such that still the dirichlet problem at infinity for @xmath0 is solvable , e.g. @xcite and @xcite .
let @xmath11 be a cartan - hadamard manifold .
the dirichlet problem at infinity for @xmath0 is solvable if one of the following conditions is satisfied :
1 . there exists a positive constant @xmath96 and a positive and nonincreasing function @xmath97 with @xmath98 such that @xmath99 2 .
there exist positive constants @xmath100 , @xmath101 and @xmath102 such that @xmath103 for all @xmath29 with @xmath104 .
it was an open problem for some time whether the existence of a strictly negative upper bound for the sectional curvature could already be a sufficient condition for the solvability of the dirichlet problem at infinity as it is true in dimension 2 . in 1994 however , ancona @xcite constructed a riemannian manifold with sectional curvatures bounded above by a negative constant such that the dirichlet problem at infinity for @xmath0 is not solvable .
for this manifold ancona discussed the asymptotic behaviour of brownian motion .
in particular , he showed that brownian motion almost surely exits from @xmath0 at a single point @xmath105 on the sphere at infinity .
ancona did not deal with the question whether @xmath0 carries non - trivial bounded harmonic functions .
borbly @xcite gave another example of a cartan - hadamard manifold with curvature bounded above by a strictly negative constant , on which the dirichlet problem at infinity is not solvable .
borbly does not discuss brownian motion on this manifold , but he shows using analytic methods that his manifold supports non - trivial bounded harmonic functions .
this paper aims to give a detailed analysis of manifolds of this type and to answer several questions .
it turns out that the manifolds of ancona and borbly share quite similar properties , at least from the probabilistic point of view . on both manifolds the angular behaviour of brownian motion degenerates to a single point , as brownian motion drifts to infinity .
in particular , the dirichlet problem at infinity is not solvable .
nevertheless both manifolds possess a wealth of non - trivial bounded harmonic functions , which come however from completely different reasons than in the pinched curvature case .
since borbly s manifold is technical easier to handle , we restrict our discussion to this case .
it should however be noted that all essential features can also be found in ancona s example .
unlike borbly who used methods of partial differential equations to prove that his manifold provides an example of a non - liouville manifold for that the dirichlet problem at infinity is not solvable , we are interested in a complete stochastic description of the considered manifold . in this paper we give a full description of the poisson boundary of borbly s manifold by characterizing all shift - invariant events for brownian motion .
the manifold is of dimension 3 , and the @xmath23-algebra of invariant events is generated by two random variables .
it turns out that , in order to see the full poisson boundary , the attracting point at infinity to which all brownian paths converge needs to be `` unfolded '' into the @xmath106-dimensional space @xmath107 .
the asymptotic behaviour of brownian motion on this manifold is in sharp contrast to the case of a cartan - hadamard manifold of pinched negative curvature .
recall that in the pinched curvature case the angular part @xmath108 of @xmath12 carries all relevant information and the limiting angle @xmath109 generates the shift - invariant @xmath23-field of @xmath12 .
all non - trivial information to distinguish brownian paths for large times is given in terms of the angular projection of @xmath12 onto @xmath40 , where only the limiting angle matters .
as a consequence , any bounded harmonic function comes from a solution of the dirichlet problem at infinity , which means on the other hand that any bounded harmonic function on @xmath0 has a continuous continuation to the boundary at infinity .
more precisely , denoting by @xmath110 the banach space of bounded harmonic functions on a cartan - hadamard manifold @xmath0 , we have the following result of anderson .
let @xmath11 be a cartan - hadamard manifold of dimension @xmath111 , whose sectional curvatures satisfy @xmath112 for all @xmath29 . then the linear mapping @xmath113 is a norm - nonincreasing isomorphism onto @xmath110 . in the situation of borbly s manifold brownian motion almost surely exits from the manifold at a single point of the sphere at infinity , independent of its starting point @xmath63 at time @xmath82 . in particular , all harmonic measures @xmath77 are trivial , and harmonic functions of the form @xmath114 are necessarily constant . on the other hand , we are going to show that this manifold supports a huge variety of non - trivial bounded harmonic functions , necessarily without continuation to the sphere at infinity .
it turns out that the richness of harmonic functions is related to two different kinds of non - trivial exit sets for the brownian motion : \1 .
non - trivial shift - invariant information is given in terms of the direction along which brownian motion approaches the point at the sphere at infinity . even if there is no contribution from the limiting angle itself , angular sectors about the limit point allow to distinguish brownian paths for large times .
brownian motion converges to a single point , in appropriate coordinates the fluctuative components ( martingale parts ) of brownian motion for large times are relatively small compared to its drift components . roughly speaking
, this means that for large times brownian motion follows the integral curves of the deterministic dynamical system which is given by neglecting the martingale components in the describing stochastic differential equations .
we show that this idea can be made precise by constructing a deterministic vector field on the manifold @xmath0 such that the corresponding flow induces a foliation of @xmath0 and such that brownian motion exits @xmath0 asymptotically along the leaves of this foliation . to determine the poisson boundary the main problem is then to show that there are no other non - trivial exit sets .
this turns out to be the most difficult point and will be done by purely probabilistic arguments using time reversal of brownian motion . to this end ,
the time - reversed brownian motion starting on the exit boundary and running backwards in time is investigated .
it is interesting to note that on the constructed manifold , despite of diverging curvature , the harmonic measure has a density ( poisson kernel ) with respect to the lebesgue measure . recall that in the pinched curvature case the harmonic measure may well be singular with respect to the surface measure on the sphere at infinity ( see @xcite , @xcite , @xcite , @xcite for results in this direction ) .
typically it is the fluctuation of the geometry at infinity which prevents harmonic measure from being absolutely continuous .
pinched curvature alone does in general not allow to control the angular derivative of the riemannian metric , when written in polar coordinates .
the paper is organized as follows . in section [ section2 ]
we give the construction of our manifold @xmath0 which up to minor technical modifications is the same as in borbly @xcite : we define @xmath0 as the warped product @xmath115 where @xmath116 is a unit - speed geodesic in the hyperbolic space @xmath117 of constant sectional curvature @xmath118 and @xmath119 is one component of @xmath120 .
the riemannian metric @xmath121 on @xmath0 is the warped product metric of the hyperbolic metric on @xmath119 coupled with the ( induced ) euclidean metric on @xmath122 via the function @xmath123 , @xmath124 by identifying points @xmath125 and @xmath126 with @xmath127 and @xmath128 and choosing the metric `` near '' @xmath116 equal to the hyperbolic metric of the three dimensional hyperbolic space @xmath129 the manifold @xmath0 becomes complete , simply connected and rotationally symmetric with respect to the axis @xmath116 .
we specify conditions the function @xmath130 has to satisfy in order to provide a complete riemannian metric on @xmath0 for which the sectional curvatures are bounded from above by a negative constant , and such that the dirichlet problem at infinity is not solvable . as the construction of the function @xmath130
is described in detail in @xcite we mainly sketch the ideas and refer to borbly for detailed proofs .
we explain which properties of @xmath130 influence the asymptotic behaviour of the brownian paths .
the probabilistic consideration of the manifold @xmath0 starts in section [ section3 ] .
we specify the defining stochastic differential equations for the brownian motion @xmath12 on @xmath0 , where we use the component processes @xmath131 , @xmath132 and @xmath133 of @xmath12 with respect to the global coordinate system @xmath134 for @xmath0 .
the non - solvability of the dirichlet problem at infinity is then an immediate consequence of the asymptotic behaviour of the brownian motion ( see theorem [ bmconverges ] ) .
it is obvious that the asymptotic behaviour of the brownian motion on @xmath0 is the same as in the case of the manifold of ancona . in particular , it turns out that the component @xmath131 of the brownian motion @xmath12 almost surely tends to infinity as @xmath94 .
it is a remarkable fact that in contrast to the manifold of ancona , where the brownian motion almost surely has infinite lifetime , we can show that on @xmath0 the lifetime @xmath73 of the brownian motion is almost surely finite . in section [ nontrivialshiftinvariant ]
we start with the construction of non - trivial shift - invariant events . to this end
, we consider a time change of the brownian motion such that the drift of the component process @xmath131 of @xmath12 equals @xmath20 , i.e. such that the time changed component @xmath135 behaves then comparable to the deterministic curve @xmath136 . we further show that for a certain function @xmath137 , the process @xmath138 almost surely converges in @xmath139 when @xmath140 .
it turns out that the non - trivial shift - invariant random variable @xmath141 can be interpreted as one - dimensional angle which indicates from which direction the projection of the brownian path onto the sphere at infinity attains the point @xmath142 .
the shift - invariant random variable @xmath143 indicates along which surface of rotation @xmath144 inside of @xmath0 the brownian paths eventually exit the manifold @xmath0 .
thereby @xmath145 is the trajectory starting in @xmath146 of the vector field @xmath147 section [ section5 ] finally gives a description of the poisson boundary .
the main work in this part is to verify that there are no other invariant events than the specified ones .
this is achieved by means of arguments relying on the time - inversed process . as a consequence a complete description of
all non - trivial bounded harmonic functions on the manifold @xmath0 is obtained .
concerning `` boundary harnack inequality '' the class of non - constant bounded harmonic functions shares an amazing feature : on any neighbourhood in @xmath0 of the distinguished point at the horizon at infinity , a non - trivial bounded harmonic function on @xmath0 attains each value lying strictly between the global minimum and the global maximum of the function .
finally it is worth noting that our manifold also provides an example where the @xmath23-field of terminal events is strictly bigger than the @xmath23-field of shift - invariant events .
0 let @xmath148 be a fixed unit speed geodesic in the hyperbolic half plane @xmath149 equipped with the hyperbolic metric @xmath150 of constant curvature @xmath118 . for our purposes one can assume without loss of generality @xmath151 to be the positive @xmath64-axis
let @xmath119 denote one component of @xmath152 and define a riemannian manifold @xmath0 as the warped product : @xmath153 with riemannian metric @xmath154 where @xmath155 is a positive @xmath156-function to be determined later . by identifying points @xmath125 and @xmath126 with @xmath157 and @xmath158
, we make @xmath0 a simply connected space .
we introduce a system of coordinates @xmath159 on @xmath0 , where for a point @xmath54 the coordinate @xmath160 is the hyperbolic distance between @xmath161 and the geodesic @xmath116 , i.e. the hyperbolic length of the perpendicular on @xmath116 through @xmath161 .
the coordinate @xmath162 is the parameter on the geodesic @xmath163{-\infty},\infty[}\,\}$ ] , i.e. the length of the geodesic segment on @xmath116 joining @xmath164 and the orthogonal projection @xmath165 of @xmath161 onto @xmath116 .
furthermore , @xmath166 is the parameter on @xmath122 when using the parameterization @xmath167 .
we sometimes take @xmath168 , in particular when considering components of the brownian motion , thinking of @xmath139 as the universal covering of @xmath122 . in the coordinates @xmath159 the riemannian metric on @xmath169 takes the form @xmath170 where @xmath171 .
, width=377,height=207 ] let @xmath172 for @xmath173 ( the complete definition is given below ) , then the above metric smoothly extends to the whole manifold @xmath0 , where @xmath0 is now rotationally symmetric with respect to the axis @xmath116 and for @xmath173 isometric to the three dimensional hyperbolic space @xmath129 with constant sectional curvature @xmath118 , cf .
@xcite . from
that it is clear that the riemannian manifold @xmath11 is complete . from now on
we fix the basis @xmath174 for the tangent space @xmath175 in @xmath54 .
herein the christoffel symbols of the levi - civita connection can be computed as follows
the indices refer to the corresponding tangent vectors of the basis : @xmath176 all others equal @xmath82 .
herein @xmath177 denotes the partial derivative of the function @xmath130 with respect to the variable r , @xmath178 the partial derivative with respect to @xmath162 , etc . + for the computation of the sectional curvatures @xmath179 write @xmath180 and @xmath181 for tangent vectors @xmath182 in terms of the basis @xmath183 .
then one gets @xmath184 as well as @xmath185 we conclude that the manifold @xmath0 has strictly negative sectional curvature , i.e. @xmath186 for some @xmath91 , all @xmath182 and all @xmath54 , if and only if the following inequalities hold : @xmath187 this is explained by the fact that the quadratic form @xmath188 is non - positive for all @xmath189 if and only if @xmath190 as it is obvious that , for @xmath191 fixed , the curves @xmath192 form a foliation of @xmath0 by geodesic rays , we can describe the sphere at infinity @xmath40 as the union of the `` endpoints '' ( i.e. equivalence classes ) @xmath193 of all geodesic rays
@xmath194 foliating @xmath0 together with the equivalence classes @xmath195 and @xmath196 determined by the geodesic axis @xmath116 of @xmath0 , which sums up to : @xmath197 , width=302,height=207 ] this also explains why it suffices to show that the @xmath162-component @xmath198 of the brownian motion @xmath199 converges to @xmath200 as @xmath20 approaches the lifetime @xmath73 of the brownian motion , if we want to show that @xmath199 converges for @xmath201 to the single point @xmath195 on the sphere @xmath40 at infinity .
we give a brief description of the properties which the warped product function @xmath202 in has to satisfy to provide an example of a riemannian manifold where the dirichlet problem at infinity is not solvable whereas there exist non - trivial bounded harmonic functions .
we give the idea how to construct such a function @xmath130 and refer to borbly @xcite for further details .
[ propofg ] there is a @xmath156-function @xmath202 , @xmath203 satisfying the following properties : 1 .
[ prop1 ] @xmath204 and @xmath205 .
[ prop2 ] @xmath206 for @xmath207 , and for @xmath208 it holds that : @xmath209 3 . [ prop3 ] denoting @xmath210 one has @xmath211 and for all @xmath212 : @xmath213 4 .
[ prop4 ] the function @xmath214 satisfies : @xmath215 5 .
[ prop5 ] we further need that @xmath216 and that there exists @xmath2170,1/4[}$ ] such that @xmath218 6 .
[ prop6 ] the function @xmath214 finally has to fulfill : @xmath219 for every @xmath220 and every @xmath221 .
the following remark summerizes briefly the significance of these conditions . 1 .
as mentioned in the definition of the riemannian metric on @xmath0 , property ( [ prop2 ] ) assures that the riemannian metric smoothly extends to the geodesic @xmath116 and that curvature condition ( [ eq:12 ] ) is satisfied for a suitable @xmath222 .
the conditions in ( [ prop4 ] ) stated for @xmath214 assure validity of the curvature condition ( [ eq:14 ] ) at least for @xmath223 .
3 . the integral condition forces brownian motion @xmath224 on @xmath0 to converge to the single point @xmath225 as @xmath94 which as an immediate consequence implies non - solvability of the dirichlet problem at infinity for @xmath0 .
the given condition assures that the drift term in the stochastic differential equation for the component @xmath132 of the brownian motion @xmath12 compared with the drift term appearing in the defining equation for the component @xmath131 grows `` fast enough '' to ensure @xmath198 going to @xmath226 as @xmath20 approaches the lifetime of the brownian motion .
+ in @xcite , lemma 2.1 , borbly uses this condition to show that the convex hull of any neighbourhood of @xmath142 is the whole manifold @xmath0 .
this is a natural first step in the construction of a manifold for which the dirichlet problem at infinity is not solvable ( cf .
the introduction and @xcite , @xcite ) .
4 . property ( [ prop6 ] ) is needed for the brownian motion to almost surely enter a region where @xmath198 has nonzero and positive drift ( see the construction of @xmath227 and @xmath228 below ) ; otherwise @xmath198 might converge in @xmath139 with a positive probability .
the idea to construct a function @xmath130 with the wanted properties is as follows .
as @xmath130 is given as solution of the partial differential equation @xmath229 one has to find an appropriate function @xmath214 and the required initial conditions for @xmath130 to obtain the desired function .
we will see later that the construction of the metric on @xmath0 is similar to that given in ancona @xcite , as borbly also defines the function @xmath161 `` stripe - wise '' to control the requirements for @xmath161 , @xmath130 respectively , on each region of the form @xmath230 \times { \mathbb{r}}$ ] .
yet he is mainly concerned with the definition of the `` drift ratio '' @xmath161 , what makes it harder to track the behaviour of the metric function @xmath130 and to possibly modify his construction for other situations , whereas ancona gives a more or less direct way to construct the coupling function in the warped product . as a consequence this allows to extend his example to higher dimensions and to adapt it to other situations as well .
we start with a brief description how to construct the function @xmath231 as given in @xcite : @xmath232 is defined inductively on intervals @xmath233 $ ] , where @xmath234 and @xmath235 sufficiently large , see below .
we let @xmath236 on @xmath237 $ ] and define it on @xmath238 $ ] as a slowly increasing function satisfying conditions ( [ prop4 ] ) and ( [ prop5 ] ) . for @xmath239
$ ] let @xmath240 be constant , where @xmath241 is chosen big enough such that @xmath242 and @xmath243
. on the interval @xmath244 we choose the function @xmath232 to be decreasing with @xmath245 , whereas @xmath246 is still increasing and strictly convex ; see @xcite , lemma 2.3 and lemma 2.4 . on the interval
@xmath247 $ ] for @xmath248 big enough as given below ( and in general on intervals of the form @xmath249 $ ] ) one extends @xmath246 via a solution of the differential equation @xmath250 note that carefully smoothing the function @xmath232 on the interval @xmath251 $ ] ( and on @xmath252 $ ] respectively ) to become @xmath156 does not interfere with the properties ( [ prop4 ] ) and ( [ prop5 ] ) of @xmath232 and can be done such that @xmath253 is still valid .
lemma 2.4 in @xcite shows that in fact @xmath254 decreases on @xmath249 $ ] and again by @xcite , lemma 2.3 , for given @xmath255 one can choose the upper interval bound @xmath256 such that @xmath257 which guarantees property ( [ prop6 ] ) for @xmath232 . on the interval
@xmath258 $ ] with @xmath259 sufficiently large as given below and in general on intervals of the form @xmath260 $ ] , let @xmath261 for some well chosen constant @xmath262 .
this differs from the construction of @xcite p. 228
where @xmath232 is chosen to be constant in these intervals , but it does not change the properties of the manifold . as above , smoothing on intervals @xmath263 $ ] preserves the conditions ( [ prop4 ] ) , ( [ prop5 ] ) and @xmath264 for @xmath265 small enough and independent of @xmath4 , but depending on the choice of @xmath232 on @xmath266 .
if one chooses for given @xmath248 , @xmath256 respectively , the interval bounds @xmath259 , @xmath267 respectively , large enough we can achieve that @xmath268 which finally assures property ( [ prop3 ] ) for @xmath232 .
following borbly @xcite , one defines @xmath214 via @xmath269 as @xmath270 by using a `` cut - off function '' @xmath271 .
herein @xmath271 is given as @xmath272 , where @xmath273 is smooth and increasing with @xmath274 for @xmath275 , @xmath276 for @xmath277 and @xmath278 , @xmath279 .
the function @xmath227 satisfies @xmath280 , @xmath281 on the interval @xmath237 $ ] and @xmath282 on the interval @xmath283 , with the same @xmath284 as in .
then the two pieces are connected smoothly such that for @xmath285 @xmath286 the function @xmath227 is nondecreasing such that @xmath287 ( see in lemma [ propofg ] ) , and one can check that @xmath214 satisfies the required properties listed in lemma [ propofg ] ( see @xcite , p.228ff ) . for an appropriate initial condition to solve the partial differential equation ( [ pdeforg ] ) , we set @xmath288 on @xmath289 $ ] .
on @xmath290 let @xmath291 be the solution of the differential equation @xmath292 smoothing @xmath293 yields a @xmath156-function @xmath294 such that @xmath295 on @xmath296}\cup{[\frac1{10}+\delta,\infty[}$ ] for an appropriate @xmath297 .
in particular , there exist @xmath298 such that @xmath299 the function @xmath294 serves as initial condition to solve the partial differential equation @xmath300 , width=377 ] more precisely , it is proven in @xcite that all trajectories of the vector field @xmath301 intersect the domain @xmath302 letting @xmath303 be the integral curve of @xmath304 satisfying @xmath305 , we define @xmath306 by @xmath307 then ( see sect . [ section5 ] for details ) @xmath308
let @xmath309 be a filtered probability space satisfying the usual conditions and @xmath12 a brownian motion on @xmath0 considered as a diffusion process with generator @xmath310 taking values in the alexandroff compactification @xmath311 of @xmath0 .
further let @xmath73 denote the lifetime of @xmath12 , i.e. @xmath312 for @xmath313 , if @xmath314 .
as we have fixed the coordinate system @xmath315 for our manifold @xmath0 , we consider the brownian motion @xmath12 in the chosen coordinates as well and denote by @xmath316 , @xmath317 and @xmath318 the component processes of @xmath72 . the generator @xmath310 of @xmath12 is then written in terms of the basis @xmath319 of @xmath320 as : @xmath321
therefore we can write down a system of stochastic differential equations for the components @xmath322 and @xmath133 of our given brownian motion : @xmath323 with a three - dimensional euclidean brownian motion @xmath324 .
as already mentioned , we are going to read the component @xmath133 of the brownian motion with values in the universal covering @xmath139 of @xmath122 .
[ lifetimeofb ] inspecting the defining stochastic differential equations for the components @xmath131 , @xmath132 and @xmath133 of the brownian motion @xmath12 on @xmath0 , it is interesting to note that the behaviour of the component @xmath325 does not influence the behaviour of the components @xmath326 and @xmath198 .
the first two equations can be solved independently of the third one ; their solution then defines the third component of the brownian motion .
hence it is clear that the lifetime @xmath73 of @xmath199 does not depend on the component @xmath325 in particular does not depend on the starting point @xmath327 of @xmath325 .
we are going to use this fact later to prove existence of non - trivial bounded harmonic functions on @xmath0 . as we we are going to see in section [ geominterpretation ] the `` drift ratio '' @xmath328 influences the interplay of the components @xmath198 and @xmath326 of the brownian motion and therefore determines the behaviour of the brownian paths .
for this reason it is more convenient to work with a time - changed version @xmath329 of our brownian motion , where the drift of the component @xmath330 is just @xmath20 and the drift of @xmath331 is essentially given by @xmath161 .
this can be realized with a time change @xmath332 defined as follows : let @xmath333 and @xmath334 for @xmath335 .
the components @xmath336 of the time - changed brownian motion @xmath337 are then given for @xmath338 by the following system of stochastic differential equations : @xmath339 we are now going to state the main theorem of this chapter which shows that from the stochastic point of view the riemannian manifold @xmath340 constructed by borbly @xcite has essentially the same properties as the manifold of ancona in @xcite .
we further give a stochastic construction of non - trivial bounded harmonic functions on @xmath0 , which is more transparent than the existence proof of borbly relying on perron s principle .
[ bmconverges ] 1 .
[ i1 ] for the brownian motion @xmath12 on the riemannian manifold @xmath340 constructed above the following statement almost surely holds : @xmath341 independently of the starting point @xmath342 .
in particular the dirichlet problem at infinity for @xmath0 is not solvable .
[ i2 ] the component @xmath325 of brownian motion @xmath199 almost surely converges to a random variable @xmath343 which possesses a positive density on @xmath122 .
[ i3 ] the lifetime @xmath73 of the process @xmath12 is a.s .
finite . from eqs .
, and we immediately see that the derivatives of the brackets of @xmath135 and @xmath344 are bounded , and that the drift of @xmath345 takes its values in @xmath346 $ ] and is therefore bounded . as a consequence
the process @xmath347 has infinite lifetime .
moreover for all @xmath348 , we have almost surely @xmath349 eventually . from this and
the fact that @xmath350 for @xmath351 we deduce that the martingale parts of @xmath135 and @xmath344 together with the process @xmath352 converge as @xmath20 tends to infinity .
next we see from @xmath353 that for @xmath20 sufficiently large the drift of @xmath344 is larger that @xmath354 .
consequently @xmath344 tends to infinity as @xmath355 . to prove it is sufficient to establish @xmath356 and
this is a consequence of the fact that @xmath357 converges to infinity .
the proof of is a direct consequence of the convergence of @xmath358 to a random variable which conditioned to @xmath135 and @xmath344 is gaussian ( in @xmath139 ) and non - degenerate .
we are left to prove .
changing back the time of the process @xmath359 , we get @xmath360 but , for large @xmath160 and @xmath361 , we have @xmath362 as a.s .
@xmath363 eventually and as @xmath357 converges to infinity as @xmath20 tends to infinity , we get the result .
0 as explained in the introduction there is a one - to - one correspondence between the @xmath23-field @xmath364 of shift - invariant events for @xmath12 up to equivalence and the set of bounded harmonic functions on @xmath0 .
we are going to use this fact and give a probabilistic proof for the existence of non - trivial shift - invariant random variables , which in turn yields non - trivial bounded harmonic functions on @xmath0 .
in addition , we get a stochastic representation of the constructed harmonic functions as `` solutions of a modified dirichlet problem at infinity '' .
in contrast to the usual dirichlet problem at infinity however , the boundary function does not live on the geometric horizon @xmath365 , and the harmonic functions are not representable in terms of the limiting angle of brownian motion . in the discussed example it turns out that the shift - invariant random variable @xmath343 is non - trivial and can be interpreted as `` @xmath14-dimensional angle '' on the sphere @xmath40 at infinity .
the variable @xmath343 gives the direction on the horizon , wherefrom the brownian motion @xmath12 converges to the limiting point @xmath142 .
despite the fact that the limit @xmath366 itself is trivial , and hence the limiting angle @xmath367 as well , brownian paths for large times can be distinguished by their projection onto @xmath40 .
taking into account that in the pinched curvature case the angular part carries all shift - variant information , one might conjecture that the random variable @xmath343 already generates the shift - invariant @xmath23-field @xmath364 . in turn
this would imply a stochastic representation for bounded harmonic functions @xmath97 on @xmath0 as @xmath368 = { \mathbb{e}}^x\left[\lim_{t\to\zeta}f(\operatorname{pr}_3(x_t))\right]\ ] ] with @xmath369 measurable
. however , borbly @xcite describes a way to construct a family of harmonic functions @xmath370 which are rotationally invariant , i.e. independent of @xmath371 , and therefore can not be of the above form .
as he uses `` perron s principle '' for the construction , these harmonic functions do not come with an explicit representation . put in the probabilistic framework
, we learn from this that there must be a way to obtain non - trivial shift - invariant events also in terms of the components @xmath198 and @xmath326 of the brownian motion .
indeed , as will be seen in theorem [ zisnotconstant ] , there exists a non - trivial shift - invariant random variable of the form @xmath372 herein @xmath344 and @xmath135 are time - changed versions of @xmath132 and @xmath131 and @xmath373 is a function already constructed by borbly , whose properties are listed in lemma [ propsofq ] below .
this finally leads to additional ( to that depending on the component @xmath371 ) harmonic functions via the stochastic representation @xmath374,\ ] ] where @xmath375 is a bounded measurable function .
[ propsofq ] there is a @xmath156-function @xmath373 with the following properties : 1 .
[ propq1 ] @xmath376 2 .
[ propq2 ] for @xmath377 the function @xmath137 satisfies the inequalities @xmath378 3 .
[ propq3 ] there is a @xmath379 such that @xmath380 see section [ constructionofq ] below and @xcite .
[ zisnotconstant ] consider @xmath381 as before
. the limit variable @xmath382 exists .
moreover the law of the random variable @xmath383 has full support @xmath107 and is absolutely continuous with respect to the lebesgue measure .
let @xmath384 and @xmath385 then a straight - forward calculation shows that @xmath386 when @xmath387 is sufficiently large , this simplifies as @xmath388 for @xmath20 sufficiently large , we know that @xmath389 ( for any small @xmath390 ) and that @xmath357 is positive ; furthermore @xmath391 is larger than @xmath392 for @xmath160 large and @xmath361 .
this together with the fact that the functions @xmath161 , @xmath393 @xmath394 , @xmath137 and @xmath395 are bounded allows to conclude that @xmath396 is converging . to prove that the law of @xmath397 has full support on @xmath139
, it is sufficient to establish convergence of the finite variation part of @xmath396 , and that its diffusion coefficient is eventually bounded below by a continuous positive deterministic function and above by @xmath398 for some @xmath348 .
indeed , with these properties it is easy to prove that , taking any open non empty interval @xmath399 of @xmath139 , there is a time @xmath20 such that after time @xmath20 , the process @xmath396 will hit the center of @xmath399 and then stay in @xmath399 with positive probability .
the convergence of the drift has already been established , together with the upper bound of the diffusion coefficient . to obtain the lower bound , we use eq . of lemma [ l1 ] below , along with the fact that eventually @xmath400 and @xmath401 , to obtain @xmath402 for some @xmath403 , where @xmath49 is defined at the end of sect .
[ section2 ] .
this immediately yields a lower bound for the diffusion coefficient @xmath404 of the form @xmath405 ( for some @xmath9 ) which is a positive continuous function .
the assertion on the support of the law of @xmath406 is then a direct consequence of the fact that conditioned to @xmath407 , the random variable @xmath408 is a non - degenerate gaussian variable .
we are left to prove that the law of @xmath409 is absolutely continuous . using the conditioning argument for @xmath408 ,
it is sufficient to prove that the law of @xmath410 is absolutely continuous . to this end
we use the estimates in sect .
[ section5 ] .
it is proven there that there exists an increasing sequence of subsets @xmath411 of @xmath412 satisfying @xmath413 such that on @xmath411 , for each @xmath414 , the diffusion process @xmath415 has a continuous extension to @xmath416 $ ] and that this extension has bounded coefficients with bounded derivatives up to order @xmath106 .
consequently , following @xcite theorem 4.6.5 , there exists a @xmath417 flow @xmath418 which almost surely realizes a diffeomorphism from @xmath419{-\tan u_1},\infty[}\times { \mathbb{r}}$ ] to @xmath419{-\tan u_2},\infty[}\times { \mathbb{r}}$ ] if @xmath420 and from @xmath419{-\tan u_1},\infty[}\times { \mathbb{r}}$ ] to its image in @xmath421 if @xmath422 ( for the first assertion one uses the fact that brownian motion has full support ) . as a consequence
, the density @xmath423 of @xmath424 ( which exists and is positive on @xmath419{-\tan \pi/4,\infty}[}\times { \mathbb{r}}$ ] ) is transported by the flow to @xmath425 where @xmath426 is the jacobian matrix of @xmath427 .
integrating with respect to @xmath428 proves existence of a positive density @xmath429 for @xmath430 : @xmath431.\ ] ] finally projecting onto the second coordinate gives the density of @xmath397 , which proves that its law is absolutely continuous .
as the explicit construction of @xmath137 is already done in @xcite we just give a short sketch ( following borbly ) how to get a function @xmath137 with the required properties : let @xmath432 and @xmath433 such that @xmath434 and @xmath435 for @xmath436 .
let further @xmath437 such that @xmath438 for @xmath439 and @xmath440 . for @xmath441 the function @xmath137 is defined as @xmath442 for @xmath439 choose a strictly increasing @xmath156-extension of @xmath443 such that @xmath444 for @xmath160 large enough and @xmath445 as @xmath446 with @xmath447 ( due to the construction of @xmath232 , see sect .
[ section1 ] ) there is an @xmath377 with @xmath448 .
let @xmath449 for @xmath450 set @xmath451 the desired function @xmath137 is then a smoothened version of this function .
borbly shows in @xcite , p.232 , that the function @xmath137 obtained this way meets indeed the additionally required properties @xmath452 we conclude this chapter with some explanations how the behaviour of the brownian paths on @xmath0 should be interpreted geometrically . as seen in sect .
[ section1 ] and sect .
[ nontrivialshiftinvariant ] , the random variables @xmath453 serve as non - trivial shift - invariant random variables for @xmath12 and hence yield non - trivial shift - invariant events for @xmath12 . to give a geometric interpretation of the shift - invariant variable @xmath454 we have to investigate again the stochastic differential equations for @xmath357 and @xmath387 : @xmath455 we have seen in sect .
[ nontrivialshiftinvariant ] that the local martingale parts @xmath456 of @xmath387 and @xmath357 converge almost surely as @xmath457 .
this suggests that the component @xmath387 , when observed at times @xmath20 near @xmath226 ( or when starting @xmath12 near @xmath142 ) , should behave like the solution @xmath458 of the deterministic differential equation @xmath459 from the stochastic differential equation above we get @xmath460 where the local martingale @xmath461 converges as @xmath462 , and @xmath387 is expected to behave like @xmath463 , when the starting point @xmath464 of @xmath12 is close to @xmath142 . one might therefore expect @xmath357 to behave ( for @xmath20 near to @xmath226 ) like the solution @xmath465
, starting in @xmath466 , of the deterministic differential equation @xmath467 it remains to make rigorous the meaning of `` should behave like '' . considering the solutions @xmath468 of the deterministic differential equations above , we note that @xmath469 given as @xmath470 with @xmath471 is the trajectory of the `` drift '' vector field @xmath472 starting in @xmath473 . as we are going to see below ( see remark [ limfortrajectories ] ) , the `` endpoint '' @xmath474 of all the trajectories is just @xmath225 .
furthermore , for every point @xmath475 , there is exactly one trajectory @xmath476 of @xmath477 with @xmath478 , in other words , the union @xmath479 defines a foliation of @xmath119 . recall that @xmath119 is one component of @xmath480 and @xmath481 . defining a coordinate transformation @xmath482 where @xmath466 is the starting point of the unique trajectory @xmath476 with @xmath483
, we obtain coordinates for @xmath484 where the trajectories @xmath476 of @xmath477 are horizontal lines . applying the coordinate transformation @xmath485 to the components @xmath387 and @xmath357 of the brownian motion , @xmath486 we are able to compare the behaviour of the components @xmath387 and @xmath357 with the trajectories @xmath476 of @xmath477 , in other words , with the deterministic solutions @xmath487 and @xmath465 .
the component @xmath488 obviously equals @xmath387 . yet , knowing that for @xmath355 the new component @xmath489 possesses a non - trivial limit , would mean that the brownian paths ( their projection onto @xmath490 , to be precise ) finally approach the point @xmath491 along a ( limiting ) trajectory @xmath476 , where @xmath492 .
this would contribute another piece of non - trivial information to the asymptotic behaviour of brownian motion , namely along which trajectory ( or more precisely , along which surface of rotation @xmath493 ) a brownian path finally exits the manifold @xmath0 .
it remains to verify that the above defined component @xmath494 indeed has a non - trivial limit as @xmath462 .
as already seen , @xmath494 is the starting point of the deterministic curve @xmath465 , satisfying the differential equation ( [ deterministiccurve ] ) with @xmath495 .
the solution @xmath465 is of the form @xmath496 where @xmath497 .
in particular , @xmath465 explicitly depends on @xmath498 for @xmath499 .
that is the reason why , when applying it s formula to @xmath500 , there appear first order derivatives of the flow @xmath501 with respect to the variable @xmath162 . estimating these terms does not seem to be trivial , nor to provide good estimates in order to establish convergence of @xmath494 as @xmath355 .
a possibility to circumvent this problem is to find a vector field @xmath502 on @xmath503 of the form @xmath504 whose trajectories also foliate @xmath119 and are not `` far off '' the trajectories @xmath476 of @xmath477 in particular the trajectories of @xmath502 have to exit @xmath0 through the point @xmath225 as well . as seen in sect .
[ nontrivialshiftinvariant ] , we have @xmath505 furthermore , for @xmath450 the function @xmath506 is defined as @xmath507 , in particular @xmath506 does not differ much from the function @xmath214 which equals @xmath508 for @xmath160 and @xmath162 large .
hence @xmath506 is a good approximation for @xmath509 for @xmath160 large , and is independent of the variable @xmath162 .
we therefore consider the vector field @xmath510 starting in @xmath511 the trajectories @xmath145 of @xmath502 have the form @xmath512 as we are going to see below , we also have @xmath513 , see remark [ limfortrajectories ] , and the union @xmath514 forms a foliation of @xmath119 . for @xmath515 there is exactly one trajectory @xmath145 of @xmath502 with @xmath516 .
its starting point @xmath466 can be computed as @xmath517 .
we can therefore define a coordinate transformation , width=453 ] @xmath518 as seen in figure [ figure4 ] , the trajectories @xmath145 of @xmath502 are horizontal lines in the new coordinate system . in the changed coordinate system the components
@xmath387 and @xmath357 of @xmath12 then look like @xmath519 as we have proven in sect .
[ nontrivialshiftinvariant ] , @xmath520\equiv \lim_{t\to\widetilde \zeta } \phi_s(\widetilde r_t,\widetilde s_t)\ ] ] exists and is a non - trivial shift - invariant random variable .
therefore the non - triviality of @xmath521 allows to distinguish brownian paths when examining along which of the trajectories @xmath145 of @xmath502 , or more precisely along which surface of rotation @xmath522 , the path eventually exits the manifold .
this gives the geometric significance of the limit variable @xmath523.\ ] ] it finally remains to complete the section with the proof that the trajectories of the vector field @xmath502 , as well as the trajectories of the vector field @xmath477 , exit the manifold @xmath0 in the point @xmath142 .
this is done in the final remark .
[ limfortrajectories ] we have : @xmath524 it is enough to show that the `` @xmath162-component '' of each trajectory @xmath145 , @xmath476 resp .
, converges to @xmath226 with @xmath462 .
the s - component of @xmath145 is @xmath525 the s - component of @xmath476 @xmath526 since for @xmath527 we have @xmath528 it follows immediately that @xmath529 because @xmath530 and @xmath531 due to lemma [ propofg ] , property .
for the second term we notice that @xmath532 with @xmath533 is nondecreasing as the integrand is positive .
moreover we have seen above and in the foregoing sections that @xmath534 as @xmath535 it suffices to show that @xmath536 .
this is true as @xmath537 for all @xmath538 and therefore for @xmath160 large enough we have @xmath539 .
the claimed result then follows exactly as above .
0 in this section we prove that any shift - invariant event for brownian motion in @xmath0 is measurable with respect to the random variable @xmath540 constructed in sect .
[ nontrivialshiftinvariant ] . as a consequence
this allows a complete characterization of the poisson boundary of @xmath0 .
we perform the change of variable @xmath541 with @xmath542 .
further let @xmath384 and @xmath543 in the new coordinates , the three components @xmath544 of time - changed brownian motion satisfy @xmath545 in case when @xmath546 exceeds the positive constant @xmath547 defined in lemma [ propsofq ] , the equations simplify to @xmath548 in the subsequent estimates we always assume @xmath549 and @xmath361 .
since @xmath232 is nonincreasing and @xmath246 is increasing and strictly convex , we have @xmath550 which yields @xmath551 and consequently @xmath552 , @xmath553 and @xmath554 . we know that @xmath555 .
recall that for @xmath556 , @xmath220 , the curve @xmath303 is the integral curve to the vector field @xmath557 satisfying @xmath305 .
recall further that @xmath306 is defined by @xmath558 see eq . , and that by definition of the metric @xmath130 , we have @xmath559 for some @xmath560 , see eq . . letting @xmath561 the function @xmath562 satisfies @xmath563 which yields @xmath564 there exists two constants @xmath565 such that on the set @xmath566 , @xmath567 and @xmath568 to establish , let @xmath573 and denote by @xmath574 the derivative of @xmath575 with respect to the @xmath576-th variable , @xmath577 .
we then have @xmath578 and @xmath579 differentiating yields @xmath580 since @xmath581 and @xmath582 ( see @xcite p. 229 ) .
consequently it is sufficient to prove that @xmath583 to this end , we differentiate with respect to @xmath160 and obtain @xmath584 solving the last equation with initial condition @xmath585 ( which comes from @xmath305 ) gives @xmath586 where for the last equality we used .
we thus get @xmath587 since for @xmath588 $ ] , @xmath589 and @xmath590 for @xmath572 ( see @xcite p. 229 and
[ section2 ] ) .
hence we have @xmath591 where we used the fact that @xmath592 when @xmath593 , and @xmath594 ( @xcite ( 2.15 ) and @xmath595 ) .
to establish we first note that @xmath604 and @xmath605 hence , we have @xmath606 using @xmath607 , along with the boundedness of the integral in the exponential , the bound for @xmath608 , the boundedness of @xmath596 , @xmath609 and @xmath610 ( @xcite ( 2.15 ) ) , and the fact that @xmath611 we get the wanted bound . since @xmath615 for @xmath361 , it is sufficient to establish for @xmath616
. let @xmath100 be such that @xmath617 .
for @xmath613 , define @xmath618 as @xmath619 since @xmath620 , we have @xmath621 . hence it is sufficient to establish with @xmath618 in place of @xmath622 .
writing @xmath623 we obtain @xmath624 but when @xmath625 $ ] , we have @xmath626 hence @xmath627 and this implies @xmath628 thus , since @xmath629 , we obtain @xmath630 differentiating with respect to @xmath160 yields @xmath631 which combined with @xmath632 gives @xmath633 from the convexity of @xmath246 it follows @xmath634 by eq .
, taking into account that @xmath246 is nondecreasing , we derive the inequality @xmath635 .
choosing @xmath6360,1[}$ ] such that @xmath637 , this implies @xmath638 for @xmath160 sufficiently large .
we get @xmath639 and with @xmath640 integrating from @xmath100 sufficiently large and noting that @xmath641 , we finally arrive at the desired result .
we are now in a position to estimate @xmath642 and @xmath643 .
since @xmath644 and its derivatives are bounded , we can estimate @xmath645 by @xmath646 . recall that @xmath647 which gives @xmath648 and @xmath649 the last equation , along with and , and @xmath650 for some @xmath651 and all @xmath160 sufficiently large ( recall that @xmath652 ) , gives @xmath653 and estimating @xmath654 by @xmath655 , we obtain @xmath656 similarly to @xmath553 , we estimate @xmath657 with @xmath658 . since @xmath659 , we have @xmath660 recall that @xmath661 . the first term on the right can be bounded in absolute value by @xmath662 which is smaller than @xmath663 since @xmath664 by means of lemma [ l1 ] and lemma [ l2 ] , the second term on the right can be bounded in absolute value by @xmath665 consequently we obtain @xmath666 the following lemma is a consequence of lemma [ l3 ] , and eqs . , .
we recall that @xmath670 for some @xmath651 .
let us briefly explain how bounds for higher order derivatives can be established .
differentiating expression with respect to @xmath160 yields , for @xmath361 and @xmath160 sufficiently large , @xmath671 ( we remark that differentiation of each term in amounts to multiplication by an expression smaller than @xmath672 in absolute value ) .
this easily yields , for @xmath361 and @xmath160 sufficiently large , @xmath673 and @xmath674 now exploiting the fact that @xmath675 , a straightforward calculation shows ( for @xmath361 and @xmath160 sufficiently large ) : @xmath676 all the estimates are either immediate or easy consequences of the ones for @xmath657 and @xmath681 . on the other hand ,
since the proof is the same for these two functions , we only establish the estimate for @xmath681 .
note that the final @xmath682 should lie in @xmath830,\f{\b-1}{2(\b+1)}[$ ] where @xmath612 is given by lemma [ l3 ] , due to the estimate @xmath683 . from lemma [ l3 ] and @xmath687
, we have @xmath688 if @xmath689 , we are done
. otherwise we have @xmath690 which yields @xmath691 in this case , since @xmath692 and @xmath232 is nonincreasing , @xmath693 we finally get @xmath694/(\b-1)}(r)\right);\ ] ] since @xmath6840,\f{\b-1}{\b+1}[}$ ] the right hand side is clearly bounded .
let @xmath696 and define @xmath697 , @xmath698
. furthermore denote @xmath699 from we get @xmath700 for some three - dimensional brownian motion @xmath701 . letting @xmath702 and @xmath703 , we can write @xmath704 where @xmath705 , @xmath706 ( the last line equals @xmath707 for @xmath708 large ) , @xmath709 adopting proposition [ p1 ] and defining @xmath710 and @xmath711 , we obtain easily a @xmath712 extension of @xmath713 and @xmath714 on @xmath715\times { \mathbb{r}}^3,\ x^1>-\tan u\}$ ] with vanishing derivatives at @xmath716 .
let @xmath8 be a compact subset of @xmath717 .
further let @xmath718 , @xmath719 be two @xmath712 maps defined on @xmath720\times { \mathbb{r}}^3 $ ] which are bounded together with their derivatives bounded up to order @xmath106 , and which coincide on @xmath8 with @xmath713 and @xmath714 .
we fix @xmath7210,\pi/2[}$ ] and let @xmath722 be defined by @xmath723 for the diffusion @xmath724}$ ] conditions ( h1 ) , ( h2 ) of @xcite are satisfied , and we may conclude with corollaire 2.4 of @xcite that the time - reversed process @xmath725}\ ] ] satisfies @xmath726 where @xmath727 is a brownian motion independent of @xmath728 and where the coefficients are given by @xmath729 here we have @xmath730 and @xmath731 is the density of @xmath732 .
[ p2 ] for every @xmath733 with @xmath285 , up to negligible sets , the pre - image of @xmath734 under the path map @xmath485 to the brownian motion @xmath735 , starting from @xmath63 , is contained in the @xmath23-algebra @xmath736 generated by @xmath737 .
fix @xmath7210,\pi/2[}$ ] . since @xmath738 , the @xmath23-algebra @xmath739
is clearly included in @xmath740 .
on the other hand , we have @xmath741\a,\pi/2[}\s(\breve x_v,\ v\ge u)\\ & = \bigcap_{u\in]\a,\pi/2[}\s(\hat x_v,\ v\ge u)\quad \hbox{up to negligible sets}\\ & = \bigcap_{u\in]0,\pi/2-\a[}\s(\check x_v,\ v\le u).\end{aligned}\ ] ] for @xmath742 , consider the compact subset @xmath743 of @xmath717 , @xmath744\times { \mathbb{r}}^3,\ x^1\ge -\tan u+\f1n,\ |x^i|\le n,\ i=1,2,3\right\}\ ] ] and let @xmath745\big\}.\ ] ] since @xmath746 has a.s .
continuous paths with positive first component , we have for any @xmath742 @xmath747 now define @xmath748 and @xmath749 . for @xmath7500,\pi/2-\a[}}\s(\check x_v,\ v\le u),\ ] ] we find from eq .
@xmath751 denoting by @xmath752 ( resp .
@xmath753 ) the path map corresponding to @xmath754 ( resp .
@xmath755 ) and by @xmath756 the @xmath713-field of terminal events for continuous paths @xmath757\to d$ ] , there exists @xmath758 such that @xmath759 then on @xmath411 , we have @xmath760 , and hence @xmath761 where @xmath762\a,\pi/2]}\s(\hat x_v^n,\ v\ge u).\ ] ] on the other hand , since @xmath763 is the strong solution of a stochastic differential equation driven by a brownian motion @xmath764 independent of @xmath765 ( see @xcite corollaire 2.4 ) , we have @xmath766\a,\pi/2[}\s(\hat x_v^n,\ v\ge u)&=\bigcap_{u\in]0,\pi/2-\a[}\s(\check x_v^n,\ v\le u)\\ & = \bigcap_{u\in]0,\pi/2[}\left(\s(\check x_{0}^n)\vee \s(\check w_v^n,\ v\le u)\right)\\ & = \s(\check x_{0}^n)\vee \bigcap_{u\in]0,\pi/2[}\s(\check w_v^n,\ v\le u)\quad \hbox{(up to negligible sets)}\\ & = \s(\check x_{0}^n ) = \s(\hat x_{\pi/2}^n).\end{aligned}\ ] ] as a consequence , there exists a borel subset @xmath767 of @xmath768 such that @xmath769 this yields , along with , @xmath770 the inclusion @xmath771 together with yields for all @xmath742 , @xmath772 this is true for all @xmath742 , so letting @xmath773 , we get @xmath774 now let us establish the other inclusion .
if @xmath775 then choose @xmath4 such that @xmath776 for all @xmath777 $ ] . then choose @xmath778 so that @xmath779 .
we clearly have @xmath780 , so that @xmath781 by .
finally we proved that up to a negligible set , @xmath782 this shows that @xmath783 up to negligible sets .
let @xmath788 and @xmath791 the sphere of radius @xmath14 centered at @xmath792 . define a time - changed brownian motion @xmath793 ( time - changed in the sense of a solution to ) on a product probability space as follows :
@xmath794 is a time - changed brownian motion started at @xmath792 , and for any @xmath795 , @xmath796 is a time - changed brownian motion started at @xmath63 . letting @xmath797 , we define @xmath798 for every @xmath29 , let @xmath803 be a borel subset of @xmath804 such that @xmath805 where @xmath806 is a time - changed brownian motion on @xmath0 started at @xmath63 , with coordinates @xmath807 . since @xmath808 , we have @xmath809 and hence @xmath810 but by lemma [ p2 ] , @xmath811-a.s . , @xmath812 and for all @xmath795 , @xmath813-a.s . , @xmath814 on the other hand ,
it is easy to see that @xmath815 let @xmath816 be the law of @xmath817 .
a consequence of the equalities above is that for @xmath816-almost all @xmath63 , we have @xmath818 a.s . , @xmath819 on the other hand , we know that the law of @xmath820 given @xmath821 and @xmath822 has a positive density on @xmath830,\infty[$ ] and depends only on @xmath823 .
hence equality of the second and the last term in implies that for @xmath816-almost all @xmath63 , @xmath824 a.s .
, if @xmath825 then @xmath826 contains almost all @xmath827 such that @xmath828 let @xmath829 we proved that for @xmath816-almost all @xmath63 , @xmath830 a.s .
, @xmath831 since @xmath832 we get almost surely , @xmath833 to establish the other inclusion , let us define , for @xmath834 , @xmath835 we know that on @xmath836 , for @xmath837-almost all @xmath63 , @xmath838 a.s . , @xmath839 recall that @xmath840 , and that @xmath841 by the positiveness of the density of the conditional law of @xmath820 , for @xmath816-almost all @xmath63 and @xmath842 such that @xmath843 we find @xmath844 this yields @xmath845 which implies by that @xmath846 taking into account that above probability takes its values in @xmath847 , it must be equal to @xmath14 . consequently for @xmath816-almost
all @xmath63 , @xmath848-almost surely , @xmath849 which rewrites as @xmath850 we get , @xmath811-almost surely , @xmath851 with and we finally obtain : @xmath852 this achieves the proof .
the map @xmath856\big)\end{aligned}\ ] ] is one to one .
more precisely , the inverse map is given as follows .
for @xmath857 , letting @xmath858 be the density of @xmath859 with respect to the lebesgue measure on @xmath107 , for all @xmath860 there exists a unique @xmath861 such that @xmath862 moreover , for all @xmath29 , the kernel @xmath858 is almost everywhere strictly positive .
m. cranston and c. mueller , _ a review of recent and older results on the absolute continuity of harmonic measure _
, geometry of random motion ( ithaca , n.y . , 1987 ) ,
, vol . 73 , amer . math .
soc . , providence , ri , 1988 , pp .
919 .
kifer , _ brownian motion and positive harmonic functions on complete manifolds of nonpositive curvature _ , from local times to global geometry , control and physics ( coventry , 1984/85 ) , pitman res .
notes math .
150 , longman sci .
harlow , 1986 , pp . | the liouville property of a complete riemannian manifold @xmath0 ( i.e. , the question whether there exist non - trivial bounded harmonic functions on @xmath0 ) attracted a lot of attention . for cartan - hadamard manifolds
the role of lower curvature bounds is still an open problem .
we discuss examples of cartan - hadamard manifolds of unbounded curvature where the limiting angle of brownian motion degenerates to a single point on the sphere at infinity , but where nevertheless the space of bounded harmonic functions is as rich as in the non - degenerate case .
to see the full boundary the point at infinity has to be blown up in a non - trivial way .
such examples indicate that the situation concerning the famous conjecture of greene and wu about existence of non - trivial bounded harmonic functions on cartan - hadamard manifolds is much more complicated than one might have expected . |
it was realized a few years ago ( anosova & orlov 1991 ) that the multiple system ads 6175 , which contains three spectroscopic binaries ( castor a , castor b and yy gem ) , could be just the most prominent members of a moving group .
this moving group would include about 18 stars , having spectral types between a1 v and m6 ve , including one of the prototypes of the @xmath1 pic stars .
the @xmath1 pic type stars show ir excesses , which are associated the the presence of circumstellar dusty disks .
the prototypes for this class of stars are @xmath1 pic , vega and fomalhaut . as a result of their proximity and brightness
, these three stars have been studied in great detail , and estimates of the mass , radial distribution and structure , dust - grain properties of their disks have been made ( e.g. , backman & paresce 1993 ) .
it is generally believed that these circumstellar dusty disks are either the direct descendents of t tauri disks or the secondary products of the planet formation process .
knowledge of the ages of @xmath1 pic stars is therefore one of the keys to understand the formation and evolution of their disks .
however , there is only a handful of estimates of the ages of these systems , and no member of this class has yet been detected with certainty in an open cluster ( although see backman , angelova & stauffer 1998 ) .
all three prototypes are a stars , and estimates from their post - zams evolution give approximate ages of 100 , 200 , and 400 myr for @xmath1 pic , fomalhaut and vega , respectively ( backman and paresce 1993 ) . recently , barrado y navascus et al .
( 1997a ) estimated an age of 200@xmath2100 my .
for fomalhaut , based on a number of properties ( x - ray emission , rotation , lithium abundance , isochrones ) of its physical companion gl879 .
another relevant star belonging to this class , hr4796a , has been studied in a similar way by stauffer et al .
these authors derived an age of 8@xmath22 myr . lately , holland et al .
( 1998 ) have imaged the thermal emission from the disks of the three prototypes at submillimeter wavelengths .
they concluded that any earth like planet must have already formed .
moreover , the central holes found by holland et al .
( 1998 ) in the orbit of fomalhaut and by jayawardhana et al .
( 1998 ) and koerner et al . (
1998 ) in the case of hr4796a , support the idea of the early formation of planets . in this paper
, we examine the proper motions of several nearby stars . based on hipparcos data ,
we show that , indeed , castor shares its galactic motion with other stars , including fomalhaut and vega and several late spectral type stars .
this fact allows us to derive the age of the group .
a few years ago , barrado y navascus et al . ( 1997a ) determined the age of one of the prototypes of the @xmath1 pic class , fomalhaut , based on the properties of a physical late spectral type companion , gl879 .
they noted that anosova & orlov ( 1991 ) had suggested that both stars could be part of a moving group which would include castor a@xmath3b and yy gem .
however , uncertainties at that time were too large to derive any firm conclusion .
the wealth of data obtained by hipparcos allows a more precise determination of distances , positions , and proper motions .
using these new values , we have recomputed the galactic velocities for the anosova & orlov ( 1991 ) sample , to try to verify whether there is a moving group or not .
in addition to the 18 possible members studied by anosova & orlov ( 1991 ) , we have included several stars from two different sources .
the first source provides the kinematics of several types of cool dwarfs in the solar neighborhood ( soderblom 1990 ) .
the second source is a large study which searched for kinematic groups in the solar neighborhood ( agekyan & orlov 1984 ) .
these latter authors computed the galactic velocities for more than 1,000 stars selected from the gliese catalog , and suggested the existence of several moving groups .
from these two sources , we have selected a total of 26 stars , and we have re - analyzed their kinematic properties .
table 1 lists our sample of stars .
the first column contains the common name for well known stars , whereas the other three columns provide the numbers in the gliese , hd and hipparcos / tycho catalogs , respectively .
a color - magnitude diagram of the whole sample is shown in figure 1 , where the absolute visual magnitude is plotted versus the ( b v ) color index . a schmidt - kaler ( 1982 ) zero age main sequence ( zams )
is also included as a solid line .
the group includes eight a stars with different levels of ir emission , as well as five df , two dg , three dk and seven dm .
note that we have not performed a systematic search for possible members ( either using the whole gliese catalog or taking advantage of the hipparcos database ) and , therefore , our sample is not complete and may be biased .
table 2 contains additional information about the sample .
we list the apparent and absolute visual magnitudes , ( b v ) and ( v i ) cousin colors and , when available , the equivalent width of h@xmath0 ew(h@xmath0) , x - ray luminosity and lithium abundance .
based on the positions , parallaxes , and proper motions provided by the hipparcos database , and the radial velocities ( simbad database and references therein ) , we have computed the kinematic properties of our sample of stars .
table 3 shows the numbers in the gliese catalog , the positions , parallaxes , proper motions , radial velocities , and the galactic velocities ( u , v and w ) , with their associated errors for all these quantities . galactic velocities and errors were computed following the description provided by johnson & soderblom ( 1987 ) and follow the right - handed coordinate system ( positive toward the galactic center , galactic rotation and north galactic pole ) .
since the time span for the hipparcos mission was short , the proper motions based on hipparcos data might be not accurate .
for this reason , we have also computed the galactic velocities using the proper motion values from the ppm ( roeser & bastian 1988 ; bastian & roeser 1993 ; roeser & bastian 1994 ) , and the hipparcos positions and parallaxes . except in very few cases ( e.g. gl351 , gl521.2 , castor itself )
, the discrepancies do not affect the conclusions .
however , we note that the differences between the hipparcos and the ppm proper motions differ by more than the quoted uncertainties for all stars in the sample .
narayanan & gould ( 1998 ) have shown that hipparcos proper motions are consistent with its parallaxes ( at least in the direction toward the hyades ) , an indication that there are not systematics errors due to the short baseline .
the derived v component of the galactic velocities also differ by more than the nominal uncertainties in 30% of the sample .
this fact indicates that some stars catalogued as members of the moving group could be spurious members .
we have made a special effort to use the most accurate radial velocities .
primarily , we selected the values from the web catalog ( duflot et al .
1995 ) , a compilation of three different works ( wilson 1963 ; evans 1978 ; batten et al .
we also provide other values quoted in the literature . only in three cases ( gl426 , gl564.1 and gl842.2 ) , there are disagreements between these values . for gl426 and gl842.2 , the kinematics and other information seem to indicate that they are not associated with the moving group . the case of gl564.1
is discussed below .
figures 2a and 2b allow us to identify possible members based on their kinematic properties ( v against u and w , respectively , from hipparcos data ) .
soderblom & mayor ( 1993 ) pointed out that the components of a stellar kinematic group must have the same motion in the direction of the galactic rotation ( the v component ) .
visual inspection indicates that the velocities are clustered around ( -10,-8,-10 ) .
the dispersions of these velocities are rather small , in particular for the the 7 a spectral type stars ( note that castor is itself composed of two pairs of spectroscopic binaries , but we have counted them as one star ) .
however , there are several stars which have velocities in disagreement with these average values . in particular , the galactic velocities ( specifically , using the v component as a criterion ) seem to indicate that gl426ab , gl466 and gl696 are probable non - members of the moving group , and that gl803 and gl842.2 are possible non - members ( indicated with the labels `` n '' and `` n ? '' in the second column of table 4 ) .
the table also indicates if the stars are probable or possible members ( `` y '' and `` y ? '' , respectively ) .
gl564.1 ( @xmath0@xmath4 lib ) deserves special attention .
it is physically associated to gl563.4 ( poveda et al .
the web value of the radial velocities of both stars leads to a a similar value of the v component of the galactic velocity , but poveda et al .
( 1994 ) provide a radial velocity in large disagreement with the web value .
we have preferred to keep this last value .
on the other hand , the u component of the velocity of gl563.4 is quite different from the average of the group .
the only relevant difference between both stars is the radial velocity and this could be due to the orbital motion as well as to uncertainties in the measured values .
therefore , we do not rule out the possibility that it indeed belongs to the moving group .
figure 3c displays the velocities computed with the ppm data .
the agreement in the kinematics of these stars also suggest a common origin . however , when computed from the ppm values , castor has a v component of the velocity quite different from the average of the moving group , with a difference much larger than the quoted uncertainties
. therefore , it would be possible that these system would not be a member of the moving group ( and the association should be called vega moving group ) .
in fact , if a very restricted definition would be used , this figure indicates that there could be 2 different groups ; one including vega , fomalhaut and @xmath5 phe and another with 14 lep and @xmath6 lib .
alderamin could be in either of them , due to its large error bars .
however , the data based on hipparcos and the evidences provided below seem to indicate that , indeed , all of them share a common origin . further information on the membership can be obtained from color - magnitude diagrams .
figure 3a displays the absolute visual magnitude against the ( b v ) color , whereas figure 3b has ( v i)@xmath7 in the x - axis .
the initial sample of late spectral type stars are shown as circles .
filled symbols indicate final members and open symbols those stars rejected as physically associated ( see section 3.4 ) .
both figures include the dantona and mazzitelli ( 1994 ) isochrones for the ages 3 , 10 , 35 and 70 myr ( top to bottom , dashed lines ) .
the solid line is the zams .
details concerning the convertion between theoretical values ( t@xmath8 and luminosity ) and the observational plane can be found in stauffer et al .
( 1995 ) and barrado y navascus et al .
( 1997a ) .
we have selected those stars located below the 35 myr isochrone and above the zams as possible members and indicated this in the third and fourth columns of table 4 .
figure 4 compares the equivalent width of h@xmath0 for our sample of stars ( filled circles ) with the hyades ( open circles ) .
this graph can be used to support rejection of several candidates or to accept them as members : gl696 presents a minimum value in its h@xmath0 absorption ( it is a very inactive star ) , so it seems to be older than the hyades and a non - member of the moving group .
gl896a , gl896b and gl803 are very active stars .
their activity is similar to the most active hyades stars , and indication that they could be younger .
however , barrado y navascus et al .
1998 ) have shown that some dm praesepe stars have larger activity than equivalent hyades stars ( both clusters are coeval ) .
therefore , these stars could be as old as 600 - 800 myr .
the spectroscopic binary yy gem is a by dra system , whose rotational period is synchronized with its orbital one , phenomenon which produces an enhanced stellar activity , not related with the age of the system .
its location on the plot can not provide any information about its membership to the moving group .
similar conclusions can be reached if the coronal activities are compared .
we have combined the information provided by the kinematics and the cmd to select a list of members .
we have also taken into account the lithium abundance and the h@xmath0 equivalent width when that information is available .
we find that there is a good agreement between these criteria ( if the kinematics are different from the average value , then the location in the cmd usually desagree with membership ) .
column six in table 4 provides our final classification for the membership .
the hotter stars are classified as members based only on the kinematic .
however , they are relatively young , and the probability of having several stars of a spectral type , leaving the main sequence , with the same spatial velocity , and not having been born at the same time , is quite low .
we have a total of 16 probable members and another star which could be part of the group ( @xmath0@xmath9 lib ) . a significant fraction of the stars studied here fulfill the soderblom & mayor ( 1993 ) definition of a stellar kinematic group .
the most restrictive requirement is that of having identical v component of the galactic velocity . in our case
, there are 8 stars with v within 1 sigma of the average of the final members .
this group includes castor , vega and fomalhaut .
although we do not rule out the possibility of having a spurious kinematic group , we believe that the data presented here provide strong indications for a common origin of , at least , the majority of the sample .
however , the membership of individual stars can not be known without doubt .
in fact , whitmire et al .
( 1992 ) studied the kinematics of several a stars in the solar vicinity , including several stars in common with our study , namely , gl20 , gl217.1 , gl278 , gl564.1 , gl721 , gl826 and gl881 .
although they discussed the possibility of a common origin for several of these stars , they concluded that the small velocity dispersion is due to an observational selection .
their model implies an interaction with an interstellar cloud , a phenomenon which could enhance the circumstellar disks . however ,
not all a stars in our sample display ir excesses , a subproduct of the disks . in any case , some caveats about particular membership should be kept in mind .
palou@xmath10 & piskunov ( 1985 ) have shown that there is a relation between the kinematic properties and the age of b and a stars . based on their table 3
, it is possible to constrain the age of the hot members of the castor moving group .
they have average values of the velocities of @xmath11u@[email protected] , @xmath11v@[email protected] and @xmath11w@[email protected] km / s , leading to ages of 100400 , 200 - 600 and 50 - 100 myr , respectively . these ages are indicative , since the velocity dispersion components in the palou@xmath10 & piskunov ( 1985 ) study are quite large and , in the last case , the w component of the velocity of the castor group is slightly out the range of velocities examined by palou@xmath10 & piskunov ( 1985 )
. a combined age would be 100 - 400 myr , in good agreement with the values derived for the late spectral type stars ( see below ) .
figures 3a and 3b can provide an estimate of the age of the association .
the final set of members lies below the 35 myr isochrone and above the 70 myr .
both the proximity to the zams and the unknown metallicity of the group do not allow an accurate determination of the age .
however , it seems that these stars are clearly above the zams .
then , they would be premain sequence ( pms ) objects , still approaching at the ms .
we believe that due to the errors in the conversion from the theoretical plane to observational plane , together with the errors associated with the observational data ( very small in most cases ) , the two color - magnitude diagrams put only a lower limit to the age , of 35 myr . on the other hand , the hotter real members of the moving group should be close to the end of their lives in the main sequence ( in particular , castor a and b , and @xmath0@xmath4 lib ) .
this particular evolutionary status allows for the possibility of isochrone fit .
again , uncertainties in the individual actual distances negligible , see table 3 , magnitudes and colors are too large to provide accurate ages . backman & paresce ( 1993 ) , by fitting isochrones , derived ages close to 400 and 200 myr for vega and fomalhaut , respectively , with uncertainties of 30% . for our data
set , we use meynet et al .
( 1993 ) set of evolutionary tracks .
we find that fomalhaut is too close to the ms to fit an isochrone .
the location of vega , hotter than fomalhaut , in the cmd is completely compatible with the turn - off position of the open cluster ngc 6494 , and provides an age close to 300 myr . in any case
, it seems that at least some stars , including the castor system , yy gem , fomalhaut and gl 879 , are physically associated and , for this reason , they have the same age .
the common origin and age of these stars would have important implications .
in particular , yy gem , an eclipsing binary , is one of the two m dwarfs with accurate measurements of radii and temperatures .
chabrier & baraffe ( 1995 ) , using new evolutionary models , fit isochrones to both components of the system ( either in the radii - mass or the t@xmath8-mass planes ) , and concluded that the system has to be on the late pre - main sequence contraction phase .
they estimated an age of @xmath13100 myr .
we note that if it is really slightly a pms system , then it should not be used as a calibrator of theoretical models ( unless one can accurately correct for its pms nature or accurately know its age ) .
the lithium abundance of late spectral type stars can be used as a membership criterion , as well as a statistical method to estimate the age of an association of stars , since the lithium abundance depends on both age and mass ( e.g. , balachandran 1994 ) .
figures 5a and 5b depict the abundance against the effective temperature for those stars with published values ( crosses ) . in the first case ( fig .
5a ) , we have included data corresponding to hyades stars ( filled circles ) and pleiades stars ( open circles ) .
these two clusters have standard ages of 600 - 800 and 70 - 80 myr , respectively ( see stauffer et al .
1998 for a new determination of the age of the pleiades , which yields 125@xmath25 myr , based on the position of the lithium boundary for very low mass stars and brown dwarfs ) .
the second figure contains data from the m34 cluster ( open circles ) and the uma moving group ( filled circles ) , which are 200 and 300 myr old , respectively .
figure 5b also includes a 300 myr lithium depletion isochrone from chaboyer ( 1993 ) .
the sources of the lithium data for all four associations can be found in barrado y navascus et al .
( 1997a ) . the comparison between the lithium abundances of cluster stars of different ages and gl879 , a physical companion of fomalhaut , was used by barrado
y navascus et al .
( 1997a ) to establish that these stars are younger than 300 myr , probably in the range 100 - 300 myr .
figure 5a and 5b indicates that the lithium abundance of @xmath14 lib is compatible with this age ( see our caveats about the membership based on the kinematic , section 3.5 ) .
gl848 seems too old to be part of the of the moving group , whereas gl896a is too young .
the case of yy gem is more complex .
it is well known that it is a very close sb2 eclipsing binary .
its rapid rotation ( v sini=40 kms@xmath15 ) causes significant for the important spectral line blending .
moreover , since the system is composed of 2 similar stars ( m=0.62 and 0.57 m@xmath16 , bopp 1974 ) , lines arising from both components appear in the spectrum . for the particular observation used to derive the lithium abundance ( phase=0.21 , barrado y navascus 1996 ) , the lii6708.8 doublets of each component were well detached .
the final abundance is log n(li)[email protected] , in the customary scale where log n(h)=12 .
however , close binaries in clusters ( barrado y navascus & stauffer 1996 in the case of hyades , and ryan & deliyannis 1995 in the case of m67 ) and chromospherically active binaries , such as yy gem ( barrado y navascus et al .
1997b ) inhibit partially the lithium depletion , probably due to mixing related to rotation . therefore
, this binary is not a good candidate to be used to estimate the age of the castor moving group , although it is very well established that there is a physical connection between these stars
. all these comparisons , which are compatible with the well determined age of gl879 and fomalhaut , allows us to conclude that , indeed , the castor moving group has an age of 200@xmath2100 myr .
note that at this point , we have assumed that the stellar kinematic group is real and that these stars are members of it .
however , as mentioned before , the group , as found in open clusters , could contain spurious members , and even the possibility of having a sample with the same kinematic properties and compatible cmd and a different origin can not be rejected .
if these stars do have a common origin , as it seems , the same age derived for them is a very relevant fact , since the group has several a stars with different infrared excesses ( or none at all ) , indicating that , at least at this age , the evolution of the protoplanetary disks does not depend on their age .
we have used data provided by hipparcos to select several stars as possible members of the castor moving group .
additional information , such as the position in color - magnitude diagrams , isochrones , lithium abundances , stellar activity , was used to add new constraints to the membership and to estimate the age of the association , which is 200@xmath2100 myr .
the moving group contains seven a spectral type stars with different ir excesses .
this fact could indicate that , at this stage , the evolution of these excesses , when present , does not depend on age , and that , therefore , the time scale of the formation of planetary does not seem to be universal .
the excesses would be produced by collisions and sublimation of large bodies and not from protoplanetary disks .
figure 3.- color - magnitude diagrams for the late spectral type candidates .
the isochrones ( 3 , 10 , 35 , 70 and zams ) are those from dantona & mazzitelli ( 1994 ) .
final members are indicated as filled circles , whereas rejected members appear as open circles . * a * mv against ( b v ) . * b * mv against ( v i)@xmath7 .
figure 5.- lithium abundances against effective temperatures .
crosses represent the proposed members of the moving group . * a * pleiades and hyades data ( open and filled circles ) . *
b * m34 and uma group data ( open and filled circles ) . | we have recomputed the kinematic properties of several of dozens nearby stars , to try to verify if castor ( @xmath0 gem ) indeed has a cohort of stars sharing the same space motion and age .
we used kinematics , the location of the stars in color - magnitude diagrams , their lithium abundances , and their activity , to establish that the moving group seems to be real and to reject several stars which were believed to be part of this group .
of the initial 26 stars , we show that probably only 16 stars are physically associated .
the moving group contains several a spectral type stars . among them ,
vega and fomalhaut , two of the prototypes of the @xmath1 pic type stars .
if these stars are coeval , their different levels of ir emission suggest that the time scale for the formation of planets is not universal . due to the age of the group
, these ir excesses would appear as a consequence of collisions and sublimation of larger bodies and they would not arise from protoplanetary structures .
since this association includes several late spectral type stars , we used their properties to estimate their age and , therefore , the age of the group and that of vega and fomalhaut .
our estimate for that age is 200@xmath2100 myr . |
unconventional superconductors are a subject of intensive experimental and theoretical studies during past decade @xcite-@xcite .
coexistence of superconductivity and magnetism results in a number of unusual phenomena , which have both fundamental and application interests .
unconventional superconductors , such as @xmath1 @xcite , @xmath2 @xcite and @xmath3 @xcite , which are characterized by the state with broken time - reversal symmetry ( trsb ) , possess the magnetic structure related to non - zero orbital magnetic moment due to spin - triplet @xmath4-wave state of multi - component sc order parameter .
the properties of such superconductors are not trivial .
in particular , the macroscopic magnetization is not well - defined quality , and the equations of motion can not be expressed in terms of local magnetization @xcite .
starting from the pioneering work @xcite , @xcite on investigation of non - uniform states of superconducting order parameter ( see , also review @xcite ) , the planar defects such as domain walls ( dw ) and surfaces have been attracted recently a lot of interest @xcite , @xcite , @xcite , @xcite due to new experiment possibilities , e. g. scanning superconducting quantum interference device ( squid ) and hall probe technique , on investigation of magnetic field distribution resulting from spontaneously generated superconducting currents which can serve as a prove of trsb origin of superconducting order parameter in unconventional chiral superconductors .
although , the magnetostatic fields are screened by superconducting current , metastable domain walls ( dws ) , as topologically stable planar defects , may exist even in the meissner state @xcite .
thus , the domain structure of scfm can not be ignored .
the anomalies in the local magnetization loop for sr@xmath5ruo@xmath6 near @xmath7 are considered as a strong indication of the presence of chiral domains @xcite .
the strong coordinate dependence of magnetization of such a @xmath8d magnetic defect creates the intrinsic magnetic field which interacts with superconducting ( sc ) current . for unconventional superconductors
the discontinuity of magnetic induction at the dw can be interpreted as an effective magnetization , and the contribution of dw current to the energy density can be transformed to an effective zeeman term providing the possibility of excitation of orbital magnetization waves by incident electromagnetic field @xcite . in present publication
we solve the maxwell equations for distribution of magnetic field inside the sample and above it due to the presence of planar defect , such as a dw , show rapid decrease of the magnetic field in the vicinity of sample surface , discuss the long - range origin of this field resulting in existence of the tails at the distance @xmath9 ( london penetration depth ) and compare our results with recent experimental data on measurements of the magnetic field in sr@xmath5ruo@xmath6 by squid @xcite and hall probe technique @xcite .
it is shown that for infinitely thin dw the stray fields above the sample depend only the magnetic field jump at the dw , but not on the details of dw structure .
let us consider the @xmath10 dw of bloch type in a semi - infinite sample of superconducting ferromagnet occupying @xmath11 .
we assume , that the surface of a crystal is parallel to the @xmath12 plane at @xmath13 , and the domain wall , being parallel to the @xmath14 plane , separates two domains with magnetization @xmath15 along the @xmath16 or @xmath17 direction provided that the quality parameter @xmath18 , e.g. the field of magnetic crystallographic anisotropy of the easy - axis ( @xmath19-axis ) type ( @xmath20 ) is high enough in order to stabilize the system against the flip of the domain magnetization into the @xmath21-@xmath22 plane .
we restrict ourselves to the simplest case when the london penetration length @xmath23 exceeds the dw thickness @xmath24 .
this means that at the spatial scales of the order of @xmath24 , the domain wall structure is governed by the exchange and magnetic crystallographic energies and is not affected by the kinetic energy associated with superconducting currents .
thus , we assumed that the next hierarchy of characteristic lengths for scfm is fulfilled : @xmath25 , where @xmath26 is the size of the domain .
we consider the steady - state properties of dw deep in the bulk .
-component @xmath27 and @xmath22-component @xmath28 of magnetic induction for bloch dw at constant value of the domain wall width @xmath24 and variable london penetration depth @xmath23 .
the indexes 1 , 2 and 3 correspond to @xmath29 , @xmath30 and @xmath31 respectively .
the dashed line shows the @xmath19-component @xmath27 and @xmath22-component @xmath28 of magnetization which are the limiting values for corresponding components of magnetic induction at @xmath32 , e.g. in neglecting of screening by superconducting current.,scaledwidth=40.0% ] the free energy of the superconducting ferromagnet can be written as follows @xcite : @xmath33+\frac{1}{8\pi\lambda^{2}}\right .
\nonumber\\ & & \times\left.\left(\frac{\phi_{0}}{2\pi}\mathbf{\nabla}\phi-\mathbf{a}\right)^{2}+\frac{b^{2}}{8\pi}-\mathbf{b}\cdot\mathbf{m } \right\},\quad%\end{aligned}\ ] ] where @xmath34 is the phase of the superconducting order parameter , @xmath35 is the magnetic flux quantum , @xmath36 is the magnetization , and @xmath37 is the magnetization component perpendicular to the easy axis @xmath38 , which is the normal to the sample surface .
the vector potential @xmath39 determines the magnetic induction @xmath40 .
the first and second terms in eq .
( [ eq : free energy ] ) describe the magnetic crystallographic and exchange energies correspondingly , the third term is kinetic energy of superconducting current , and the last two terms relate to magnetostatic energy .
the length @xmath24 characterizes the stiffness of the spin system and is also the dw width .
the magnetization can be expressed in polar coordinates , @xmath41 with @xmath42 the saturation magnetization and @xmath43 the lattice constant .
-component @xmath27 and @xmath22-component @xmath28 of magnetic field for bloch dw at constant value of the domain wall width @xmath24 and variable london penetration depth @xmath23 .
the indexes 1 , 2 and 3 correspond to @xmath29 , @xmath30 and @xmath31 consequently.,scaledwidth=40.0% ] -components of magnetic induction ( @xmath43 ) and magnetic field ( @xmath44 ) for bloch dw at constant value of the london penetration depth @xmath23 and variable dw width @xmath24 .
the indexes 1 , 2 and 3 correspond to @xmath29 , @xmath30 and @xmath31 consequently .
the dash lines correspond the limit values for magnetic induction ( @xmath43 ) and magnetic field ( @xmath44 ) in case of infinitely thin dw @xcite.,scaledwidth=40.0% ] the ground state of the scfm is defined by : @xmath46 where @xmath47 .
minimization of the free energy density eq .
( [ eq : free energy ] ) with respect to azimuthal , @xmath34 , polar , @xmath48 , angles and vector potential @xmath39 leads to the system of non - linear differential equations describing the ground state of scfm with dw of bloch type . in case of @xmath49 the solution of these equations are as follows : @xmath50 where we used the reduced values for the components of magnetic induction , @xmath51 and @xmath52 .
the condition @xmath53 allows us to neglect the influence of meissner current on the structure of the dw , which is defined entirely by the magnetic anisotropy and exchange energy . the screening action of the current results in decreasing of magnetic induction at the distance of @xmath54 from the center of the domain wall @xmath55 .
the graphical representation of the solutions ( [ eq : ground state ] ) are shown in fig .
[ induction ] for different values of london penetration depth @xmath23 at constant dw width @xmath24 .
the distribution of magnetic field @xmath56 , created by meissner current , are shown in fig .
[ magnetic field ] .
it follows from fig .
[ induction ] and fig .
[ magnetic field ] that the screening effect of superconducting current degreases with increasing of london penetration depth @xmath23 . at constant @xmath23 in limit case when the dw width @xmath57 , the jump of the tangential component of magnetization @xmath58 at the plane of the geometric domain boundary @xmath59 defines the current sheet responsible for the jump of tangential component of magnetic induction @xmath60 @xcite .
the distributions of tangential components of magnetic induction @xmath60 and magnetic field @xmath61 of meissner current at different values of the dw width @xmath24 and constant london penetration depth @xmath23 are shown in fig .
[ limit ] .
the results of our calculation based on eq .
( [ eq : ground state ] ) show the decrease of maximum of magnetic induction splitting with increase of dw width @xmath24 ( see , fig .
[ limit]@xmath43 ) which confirms the effect of smoothing due to finite dw width @xcite .
it follows from maxwell equation @xmath62 that there exists a two - component screening current @xmath63 , which is the result of 2-dimensional structure of the dw eq .
( [ eq : ground state ] ) .
to calculate the distribution of magnetic field near the surface of scfm we neglect the domain wall @xmath24 , assuming that @xmath64 .
this assumption does not affect the results qualitatively , but essentially simplifies the problem . in the end
we shall discuss the effects of finite dw width @xmath65 .
the difference between a normal fm and a scfm is important at distances larger than @xmath24 : while in normal fms the magnetic field @xmath67 vanishes and the magnetic induction @xmath68 is constant inside domains , in scfms the magnetic induction @xmath69 is confined in the meisssner layers of width @xmath23 @xcite : @xmath70 where the upper and the lower signs correspond to @xmath71 and @xmath72 respectively , and @xmath73 is the magnetic induction splitting . thus the meissner currents @xmath74 screen out the main bulk of domains from the magnetic induction .
this screened magnetic induction eq .
( [ eq : magnetic induction ] ) influences at the distribution of magnetic field above the sample surface and , in principle , can be detected by hall probes . to find the distribution of magnetic field above the surface of the sample due to magnetic induction eq .
( [ eq : magnetic induction ] ) the standard procedure of solving maxwell equation in magnetic media ( @xmath75 ) and vacuum ( @xmath76 ) with appropriate boundary conditions at the surface located at @xmath13 , e. g. the continuity of normal component of magnetic induction and tangential component of magnetic field , is used ( see , e. g. @xcite ) .
the results of calculation can be represented in explicit form : @xmath77 for the components of the reduced magnetic field @xmath78 in vacuum , and @xmath79}{\widetilde{k}^{2}(k+\widetilde{k})}\sin{(kx)}\end{aligned}\ ] ] for the components of the reduced magnetic induction @xmath80 in the sample . in eqs .
( [ eq : vacuum]),([eq : sample ] ) we used the notation @xmath81 .
the expression for @xmath82 ( last equation in ( [ eq : vacuum ] ) ) in slightly different form was calculated in @xcite .
-component of magnetic field in vacuum above the surface of scfm ( blue line ) and in the sample ( red line ) for one bloch dw lying in @xmath14 plane .
the presence of the identical tails in the distribution at large distances are due to long - range origin of dipole - dipole interaction in accordance to eq .
( [ eq : distribution]).,scaledwidth=40.0% ] the components of magnetic field in vacuum eqs .
( [ eq : vacuum ] ) can be calculated by introducing the complex potential @xmath83 which is analytical function of complex variable @xmath84 .
thus , the complex magnetic field @xmath85 is derived from @xmath86 by @xmath87 .
the complex potential @xmath86 can be expressed through special functions @xmath88,\ ] ] where @xmath89 and @xmath90 are zero - order struve and neumann functions correspondingly ( see , e. g. @xcite ) . for @xmath91 , the difference @xmath92 , therefore , the main contribution to the potential @xmath93 is due to the first term in eq .
( [ eq : complex potential ] ) which allows to calculate the asymptotic distribution of the component of magnetic field above the sample ( @xmath94 ) @xmath95 in particular , it follows from eq .
( [ eq : distribution ] ) , that at @xmath96 the value of magnetic field decreases as inverse power law , e. g. @xmath97 , due to long - range magnetostatic interaction . the results of numerical calculations of normal to the sample surface component of magnetic field in vacuum above the surface of the sample @xmath82 eq .
( [ eq : vacuum ] ) and in the sample @xmath98 eq .
( [ eq : sample ] ) are represented in fig .
[ normal component ] which shows that at the sample surface @xmath990 , magnetic induction splitting equals half of it bulk value @xmath73 and are characterizes by the rapid decrease with the distance above the dw .
thus , at the distance of @xmath990.1@xmath23 above the dw the magnetic field decays till the third of the bulk value @xmath73 . at @xmath100 , the magnetic field splitting reaches about 0.05 of it bulk value and decays with @xmath21 as inverse power low eq .
( [ eq : distribution ] ) due to the long - range origin of dipole - dipole interaction . in a limit of infinitely thin dw , @xmath66 ,
the stray fields in vacuum above the superconducting ferromagnet depend only on the jump of magnetic induction at dw plane , but not on the details of dw structure . in this case , the tangential to the sample surface component of magnetic field is non - analytic function of @xmath19 and is characterized by the discontinuous jump at the dw position , @xmath13 . in next subsection
we take into account the effects of small ( @xmath64 ) , but finite dw width @xmath65 . to consider the effects related to the finite dw width , @xmath101 , we assume the linear distribution of magnetic induction in the dw deep in the bulk , e.g. @xmath102 at @xmath103 , which is not affected by superconducting currents , inasmuch as @xmath64 .
this assumption significantly simplifies the calculation and allows to express the results in analytic form .
it can be shown that for finite dw width the kernels in eqs .
( [ eq : vacuum ] ) and ( [ eq : sample ] ) are modified by the multiplier @xmath104 .
it results in slight suppression of @xmath19-component of magnetic field and restores the analyticity of @xmath21-component of magnetic field .
the discontinuous jump of tangential component of magnetic field at the position of dw is replaced by the value @xmath105 , where @xmath106 is euler - mascheroni constant .
recent experiments on imaging of magnetic field distribution above the surface of the sample of sr@xmath5ruo@xmath6 by scanning squid and hall probe microscopy has revealed no evidence for existence of dws in this unconventional superconductor @xcite .
this negative result can be understood in the framework of the developed theory .
if the dws exist , the maximum value of magnetic induction due to permanent current flowing in the plane of the dw does not exceed the lower critical field , e. g. @xmath10730 g above which the domain structure is unstable due to formation of abrikosov vortices .
thus , at the distance of @xmath108190 nm , the normal component of magnetic field due to the presence of dw is of order of 3 g which definitely improbable by the used methods .
i. felner , u. asaf , y. levi , and o. milo , phys .
b * 55 * , 3374 ( 1997 ) .
m. r. eskildsen , k. harada , p. l. gammel , a. b. abrahamsen , n. h. andersen , g. ernst , a. p. ramirez , d. j. bishop , k. mortensen , d. g. naugle , k. d. d. rathnayaka , and p. c. canfield , nature ( london ) * 393 * , 242 ( 1998 ) .
g. m. luke , y. fudamoto , k. m. kojima , m. i. larkin , j. merrin , b. nachumi , y. j. uemura , y. maeno , z. q. mao , y. mori , h. nakamura and m. sigrist , nature ( london ) * 394 * , 558 ( 1998 ) . c. pfleiderer , m. uhlarz , s. m. hayden , r. vollmer , hv .
l@xmath109hneysen , n. r. bernhoeft , and g. g. lonzarich , nature ( london ) * 412 * , 58 ( 2001 ) .
s. s. saxena , p. agarwal , k. ahilan , f. m. grosche , r. k. w. haselwimmer , m. j. steiner , e. pugh , i. r. walker , s. r. julian , p. monthoux , g. g. lonzarich , a. huxley , i. sheikin , d. braithwaite , and j. flouquet , nature ( london ) * 406 * , 587 ( 2000 ) .
e. b. sonin and i. felner , phys .
b * 57 * , r14000 ( 1998 ) . v. braude and e. b. sonin , phys
b * 74 * , 064501 ( 2006 ) .
e. b. sonin , phys .
b * 66 * , 100504 ( 2002 ) . v. braude and e. b. sonin , phys . rev .
b * 93 * , 117001 ( 2003 ) .
g. e. volovik and l. p. gorkov , sov .
jetp * 61 * , 843 ( 1985 ) .
m. sigrist , t. m. rice , and k. ueda , phys .
lett . * 63 * , 1727 ( 1989 ) .
m. sigrist and k. ueda , rev .
* 63 * , 239 ( 1991 ) .
g. bjrnsson , y. maeno , m. e. huber , and k. a. moller , phys .
b * 72 * , 012504 ( 2005 ) . p. g. kealey , t. m. riseman1 , e. m. forgan , l. m. galvin , a. p. mackenzie , s. l. lee , d. m. paul , r. cubitt , d. f. agterberg , r. heeb , z. q. mao , and y. maeno , phys .
lett . * 84 * 6094 ( 2000 ) .
t. tamegai , k. yamazaki , m. tokunaga , z. mao , and y. maeno , physica c volumes * 388 - 389 * 499 ( 2003 ) .
v. o. dolocan , c. veauvy , f. servant , p. lejay , k. hasselbach , y. liu , and d. mailly , phys .
95 * , 097004 ( 2005 ) .
d. vollhardt and p. w@xmath110lfle , _ the superfluid phases of hellium 3 _ ( taylor and francis , london , 1990 ) . s. chikazumi , _ physics of ferromagnetism _ ( clarendon press , oxford , england 1997 ) . j. r. kirtley , c. kallin , c. w. hicks , e .- a .
kim , y. liu , k. a. moler , y. maeno , and k. d. nelson , phys .
b * 76 * , 014526 ( 2007 )
. h. bluhm , arxiv:0705.4118v1 [ cond-mat.sup-con ] . m. abramowitz and i. a. stegun , _ handbook of mathematical functions with formulas , graphs , and mathematical tables _ ( dover , new york , 1972 ) | steady - state properties of 180@xmath0 bloch domain wall ( dw ) in superconducting ferromagnet ( scfm ) are studied .
the distribution of magnetic field above and below the surface of the scfm due to the permanent magnetization supercurrent flowing in the dw plane is calculated by solving maxwell equations supplemented by london equation .
it is shown that part of the magnetic flux of the two neighboring domains closures in the nearest vicinity of the surface of the sample giving rise to declination of the line of the force from being parallel to the dw plane . as a result , the value of the normal component of magnetic field at the surface of the sample reaches only half of the value of the bulk magnetic flux . at the distances of the order of value of london penetration depth the magnetic field decreases as inverse power law due to the long - range character of dipole - dipole interaction .
the last two circumstances are important for comparison the calculated magnetic field with the data obtained by the methods on measurement of normal component of magnetic field , e. g. hall probe technique , aimed to confirm the existence of magnetic order parameter in unconventional superconductor . |
the realization of bose - einstein condensates ( bec ) of alkali atom vapors has attracted much interest into new aspects of photon - matter interaction arising from the quantum nature of the atomic sample .
recently attention has been paid to the photoionization of a bec by monochromatic laser light @xcite .
the products of the photoionization process ( electrons and ions ) obey to fermi - dirac statistics . due to the quantum nature of the initial atomic target and the narrow spectral width of the laser ionization source ,
the occupation number of the electron / ion final states may become close to unity , especially for a laser excitation close to threshold .
if this regime is reached the ionization rate could be slowed down by the pauli blockade , with the occupation of ionized states determined by the balance between the laser ionization of the condensate and the rate of escape of the ionization products from the condensed system . moreover
, the electrostatic interaction between the ions and neutral atoms in the condensate is expected to originate localized deformations in the condensate density distribution .
either an increase or a decrease in the neutral atom spatial density , depending on the sign of the atom - ion phase shift , could be originated in small regions surrounding the ions @xcite .
+ the experimental study of these processes requires the production of a large ionic concentration within a condensate on time shorter than the timescale on which the ions escape from the system , of the order of hundred ns for typical condensate samples , without simultaneously destroying the condensate . instead , in the two - photon ionization experiments reported in refs .
@xcite the photoionization laser produced a significant transfer of atoms from the condensate into the thermal cloud .
those experimental investigations demonstrated that the rubidium two - photon ionization was not performed in efficient way , _
i.e. _ without producing a large depletion of the condensate atoms with their transformation into thermal atoms .
indeed , in typical two - photon ionization experiments the ionization is increased only at the expenses of a large production of thermal atoms .
+ the present work reports a theoretical investigation of the efficiency realized in the ionization of atomic rubidium .
this work addresses a target which is different from those associated with the typical investigations of multiphoton processes .
our aim is not to increase the multiphoton ionization probability per se .
in contrast , the production of ions is compared to optical pumping rate , produced by the spontaneous emission , transferring condensate atoms into the thermal cloud . the ionization
_ efficiency _ will be determined by the decrease in the loss towards atomic states initially not occupied , i.e. a decrease of optical pumping into sink thermal states , and by the increase in the number of ions .
+ our condensate ionization scheme takes place in a three - level cascade scheme , the final one being the atomic continuum .
the condensate losses are produced by spontaneous emission from the intermediate excited state .
an efficient ionization is realized for an atomic transfer into the continuum with a reduced real excitation of the intermediate level , whence a reduced role of spontaneous emission processes from that intermediate level .
the task of reducing the transfer towards sink levels is usually reached applying the stirap technique for the coherent transfer between the initial and final levels of a three level scheme , as analyzed in @xcite , and for transfers to and from a continuum as in @xcite .
an efficient stirap transfer , which relies on the coherent coupling of the initial and the final state , can be realized only using very short laser pulses , i. e. shorter than the characteristic decoherence time of the system . in the case of photoionization
, the use of laser pulses shorter than the typical electron - ion decoherence times would prevent from creating the narrow kinetic energy distribution for the electrons and ions required for the pauli blockade .
instead we will analyze here the two - photon ionization produced by narrow band cw radiation .
+ for long interaction times the photoionization and optical pumping processes can be described through a simple rate equation approach . at shorter interaction times , comparable to or smaller than the spontaneous emission lifetime from the first excited state ,
the determination of the photoionization efficiency requires the solution of density matrix equations .
our solution of the density matrix equations searching for a regime of efficient ionization has strong connections with the investigations of refs .
@xcite for the lineshapes of atomic excitation and ionization in a cascade three - level system under two - color excitation as a function of the time delay in the application of the two lasers .
+ section ii describes the characteristics of our three - level scheme .
section iii analyzes the case of simultaneous application of the excitation and photoionization lasers for long interaction times on the basis of the rate equation solution . in section iv
we analyze the ionization efficiency for different temporal shapes of the excitation and ionization laser pulses using the density matrix approach .
the lengths of the laser pulses and their time delay represent the experimental parameters controlling the ionization efficiency .
section v analyzes ionization based on a rapid adiabatic transfer . state .
the thermal sink state @xmath0 is occupied by the spontaneous emission decay from the @xmath1 state.,title="fig : " ] +
we investigate the rubidium two - photon ionization scheme from the @xmath2 ground state to the continuum with intermediate resonant state the @xmath3 state at 2.94 ev above the @xmath2 ground state as in the experiment of ref .
the two - photon ionization was produced either by two photons a 421 nm or by a two color process , one photon at 421 nm and a second one at 1002 nm .
the photoionization cross sections for absorption of one 421 nm or one 1002 nm photon from the @xmath4 state have a small difference , not playing an important role for the present investigation of the photoionization efficiency .
instead a larger intensity was available at the 1002 nm wavelength , and the contribution to ionization of the two color process was much larger than the contribution of the one - color photoionization by two 421 nm photons . the level scheme of fig . 1 schematizes the two - photon ionization process from the ground state @xmath5 of the condensate to the condensate ionic state @xmath6 with a near resonant excitation of the intermediate level @xmath1 .
the population of the intermediate state @xmath1 may decay back into the ground state through emission of spontaneous emission photons . for a sample composed by a bose - einstein condensate with atoms having a momentum @xmath7
, the spontaneous emission produces ground state atoms with an atomic momentum @xmath8 different from zero .
these atoms in the @xmath9 ground state constitute a thermal component separate from the condensate , to be described through a sink state @xmath0 whose atomic properties are equivalent to those of the @xmath5 state .
the @xmath0 state is filled by the spontaneous emission with rate @xmath10 from the @xmath1 state .
the interaction of the @xmath0 state with the photoionization radiation is the same one of the @xmath5 state .
however the photoionization from the @xmath0 state produces ions having an initial momentum @xmath9 and weakly interacting with the condensate .
therefore that photoionization process is not relevant for our target of electrons / ions with a small kinetic energy@xcite and it will not be considered into our analysis . + a laser with angular frequency @xmath11 , rabi frequency @xmath12 and detuning @xmath13 acts on the @xmath14 and intensity @xmath15 acts on the @xmath16 transition .
the intensity of the first laser is @xmath17 .
the ionization from the excited state @xmath1 is described by the @xmath18 ionization cross section for the laser with intensity @xmath19 . for rubidium
the atomic parameters are @xmath20 s@xmath21 , @xmath22 m@xmath23 for the 421 nm radiation and @xmath24 m@xmath23 for the 1002 nm radiation@xcite . + the evolution of the atomic system interacting with the laser radiation is given by the optical bloch equations for the atomic density matrix@xcite on the hilbert space spanned by the four atomic states introduced above @xmath25 with the additional condition @xmath26 , where @xmath27 denotes the interaction representation of @xmath28 .
we have not included into the density matrix equations the excitation and ionization of the sink state @xmath0 by the two lasers , because those processes do not modify the number of atoms lost or ionized from the condensate phase .
the numerical results reported in the following are valid not only for the simple three - level scheme of fig . 1 , but also for the real rubidium level scheme , with the excited state occupation spontaneously decaying to other intermediate states before reaching the final sink state@xcite .
both the ions and the losses into the sink states deriving from the condensate component are equivalent in the two cases .
the efficiency of the ionization will be measured through the ratio @xmath29 between the atomic population in the sink state and the atomic population in the ionic state , after the end of the laser interaction @xmath30 where the time @xmath31 is longer than the laser pulse length @xmath32 in order to include also the spontaneous emission losses towards the sink state at the end of the laser - atom interaction .
notice that in this analysis a large ( small ) value of @xmath29 corresponds to a low ( high ) ionization efficiency .
+ for any temporal shape of the applied laser pulses , a key parameter for the photoionization efficiency is the laser detuning @xmath33 from the @xmath34 transition . for decreasing detunings the two - photoionization process becomes more efficient owing to the large excitation of the intermediate @xmath1 state ; at the same time , however , also the spontaneous emission becomes more efficient . as a consequence , for decreasing detunings the loss rate towards the @xmath0 state usually increases faster than the ionization rate , resulting in the decrease of efficiency for the condensate atoms photoionization .
+ ) ( solid lines ) and the numerical solution of eqs .
( [ rogeq]-[density ] ) ( dashed lines ) for the laser pulse shape shown as dotted line in ( a ) .
in ( a ) fraction @xmath35 , in ( b ) sum of ionized fraction @xmath36 and of sink - state fraction @xmath37 .
laser parameters @xmath38 wcm@xmath23 , corresponding to @xmath40 mhz , @xmath41 wcm@xmath23.,title="fig : " ] +
during the simultaneous application of the two lasers , the atomic population decays from the excited state @xmath1 and is lost from the system with total rate equal to @xmath42 .
therefore , at every time @xmath43 the total population in the @xmath5 and in the @xmath1 states decays with a time dependent loss rate equal to @xmath44 . in the case
in which @xmath45 the ionized fraction of population and the fraction lost by spontaneous emission can be estimated by treating the losses as a perturbation of the steady state solution for the eqs.([rogeq]-[density ] ) restricted to the @xmath46 $ ] two - level system .
this perturbative solution is @xmath47\ ] ] supposing the two lasers applied for a time @xmath32 the ionic and sink state occupations may be written as @xmath48 where we have introduced the timescale @xmath49 therefore all the population is lost from the three - level system if the duration @xmath32 of the pulses is much longer than @xmath50 . in this limiting case , with no atom remaining in the @xmath5 state , the efficiency of the process is given by @xmath51 with the @xmath52 parameter defined by @xmath53 for instance , at laser intensity @xmath54w cm@xmath23 corresponding to a rabi frequency @xmath55 mhz and laser intensity @xmath56 w cm@xmath23 , the @xmath52 parameter is equal to 53.8 .
as demonstrated in refs @xcite , under the interaction with coherent pulsed radiation the atomic excitation presents different properties than in measurements performed under the cw excitation regime . in the regime of pulsed excitation ,
the time evolution of populations and coherences becomes very important . for calculating the ionization efficiency
it is necessary to perform the numerical solution of the density matrix equations ( [ rogeq]-[density ] ) .
in fact , it should be taken into account that the actual occupation @xmath35 of the excited state is characterized by rabi oscillations with effective rabi frequency @xmath57 . because the ionization depends on the occupation of that state , the atomic ionized fraction increases in time with oscillations at the same frequency
[ longpulse - fig ] compares the exact solution of the density matrix equations and the perturbation solution given by eqs.([longpulse - eq ] ) .
the average ionized and the lost fraction are well described by the perturbative solution .
however , due to the oscillatory evolution of the excited state population , the numerical solution of eqs .
( [ rogeq]-[density ] ) is required in order to obtain the correct value of the ionization efficiency parameter @xmath29 .
+ whenever @xmath58 , while no more ionization occurs after the duration @xmath32 of the pulse and the total ionized fraction remains equal to @xmath59 , a finite fraction of population @xmath60 remains in the excited state at the end of the atom - laser interaction and decays by spontaneous emission into the sink state .
therefore for a pulse of duration @xmath32 the ionization efficiency ratio of eq .
[ efficiency - eq ] is given by @xmath61 in general @xmath62 and the maximum efficiency of the ionization process corresponds to the minimum value @xmath63 . and in ( b ) , continuous line , parameter @xmath29 of the efficiency ratio versus the blue laser detuning @xmath33 calculated through the numerical solution of the density matrix equations . the dashed line reports the limiting value given by the @xmath52 parameter of eq .
[ fraction ] .
laser parameters @xmath38 wcm@xmath23 , corresponding to @xmath64 mhz , @xmath41 wcm@xmath23 , interaction time @xmath65s.,title="fig : " ] + the oscillating behavior of the ionization process appears also measuring the ionization as function of the detuning @xmath33 . such a behavior is shown by the results of fig .
[ efficiency ] where the ionized fraction @xmath36 and the efficiency ratio @xmath29 are plotted as a function of the laser detuning @xmath33 from the @xmath1 excited state , at fixed interaction time @xmath32 .
owing to the choice @xmath66 , @xmath67 , so that @xmath29 has a minimum ( and the efficiency has a maximum ) at zero detuning . at larger values of the detunings
@xmath36 decreases with oscillations , and @xmath68 assumes values dependent on the effective frequency of the rabi oscillations .
the role of the rabi oscillations is more clear on the plot of fig .
[ efficiency](b ) for the efficiency ratio @xmath29 . + for fixed pulse duration the oscillating time dependence of the excited state population with detuning - dependent effective rabi frequency @xmath69 produces an oscillating dependence of the ionized and of the lost fractions , and thus also of the ionization efficiency , as a function of the detuning @xmath70 .
in particular , the large oscillation of the ionization efficiency is determined by the oscillating amount of population remaining in the excited state @xmath1 at the end of the laser pulse and decaying totally by spontaneous emission into the sink state @xmath0 . as a consequece
, @xmath29 assumes oscillating values which are always larger than the @xmath52 value of eq .
( [ fraction ] ) , and can be equal to that value only for isolated detunings where the effective rabi frequency @xmath71 produces a 2@xmath72 pulse for the applied duration of the atom - laser interaction . + and in c ) parameter @xmath29 describing the efficiency ratio versus laser detuning @xmath70 , in mhz .
@xmath73 mhz , @xmath74 w/@xmath75 , @xmath76 .
results for different time lengths of the ir pulse , 30 , 60 , 125 and 250 ns respectively , with the blue pulse length scaled as shown in a ) and discussed in the text.,title="fig : " ] + + the occupation of the excited state at the end of the laser pulse can be annulled in the regime defined as coherent population return@xcite . for the coherent excitation of two - level atomic systems subjected to light pulses of duration @xmath32 and detuning @xmath33 from resonance , the dynamics of the atomic population is adiabatic if the detuning is significantly larger than the fourier width of the pulse , _
i.e. _ , if @xmath77 . in this case
the dressed state occupied at the end of the excitation is the same one as before the excitation . for a smooth pulse at constant detuning , that state coincides with the atomic ground state , so that even if the excited state is populated during the application of the pulse , all the population is adiabatically brought back to the ground state at the end of the process .
therefore no population fraction remains in the excited state , @xmath78 .
this regime of coherent population return does not produces any loss into the sink state at the laser pulse end .
results showing the advantages of this regime in the two - photon ionization are reported in fig .
[ pulsed ] .
+ we considered several durations of pulses all of them with the shape shown in fig . [ pulsed](a ) . for each duration of the ir square pulse ,
the blue intensity had a gaussian profile such that the intensity decreased to @xmath79 of the peak value in correspondence of the ir pulse start and end . in fig .
[ pulsed ] ( b ) and ( c ) the ionized fraction @xmath36 and the parameter @xmath29 of the the ratio between the atomic fraction in the sink state and the ionized fraction are shown , respectively , as a function of the laser detuning @xmath33 , for several durations of the laser pulse and at a given blue rabi frequency . in all cases ,
the effects of adiabatic excitation are visible , because for each pulse duration there is a minimum detuning necessary to obtain the optimal ratio @xmath29 .
the value of the minimum detuning is only slightly sensitive to the rabi frequency of the transition , because , as pointed out in @xcite , no power broadening occurs in a scheme of coherent population return .
the limiting value of @xmath29 reached at large detunings corresponds to the minimum value @xmath52 calculated at the laser intensities of the numerical analysis .
another excitation scheme allowing an efficient and robust transfer of the population from the ground to the excited state is the rapid adiabatic passage ( rap ) technique@xcite .
this technique takes advantage of the population adiabatic evolution in order to transfer almost the whole population from the ground state to the excited state , blocking completely the population return .
rap is produced by a time dependence of the laser detuning , and for an efficient transfer the laser detuning changes from a very large and negative value to a very large and positive value , or vice versa , during the application of the light .
the laser frequency experiences a chirp , and the chirp duration should be shorter than the excited state lifetime . under these circumstances ,
the lowest energy adiabatic state of the system coincides with the ground state before the application of the light and with the excited state at the end of the process .
when this scheme is applied to the lower atomic transition of the rubidium two - photon ionization , it allows to transfer the ground state population to the @xmath80 state , from where a pulse of ir light , with the time sequence shown in fig .
[ rap](left - hand side ) , produces the ionization shown in fig .
[ rap](right - hand side ) . because the ir pulse ionizes the excited state atoms and also modifies the energy of the excited state , for any temporal shape of the two laser pulses the largest two - photon ionization is obtained for a particular delay of the ir pulse with respect to the blue laser pulse . for the pulse shapes of fig .
[ rap ] the maximum ionization was obtained when the center of the ir pulse was delayed by 0.02 @xmath81s with respect to the center of the blue pulse .
a linear chirp of the blue detuning was examined , but the chirp exact shape had a minor influence on the ionization efficiency , provided that the detuning difference between initial and final value was larger than @xmath82100 mhz . suppressing the oscillatory behavior of the excited population , the rap led to an excited state occupation and an ionized fraction larger than those obtained by pulses of the same shape but at constant blue detuning . for the blue laser scanning of fig .
[ rap ] the ionization fraction was larger by a factor 2 than produced at constant blue detuning ( 0.0057 against 0.0027 ) .
the rap technique is thus suitable for producing the maximum possible fraction of ionized population within a laser pulse sequence of short duration , during which the ratio between the atomic population in the sink state and the atomic population in the ionic state corresponds to the optimal value @xmath52 . due to the suppression of the coherent population return ,
on the other hand , a large fraction of population typically remains in the excited state at the end of the laser pulse and decays into the sink state thereafter . as a consequence , in rap at long interaction times the efficiency parameter @xmath29 , calculated according to eq .
[ efficiency - eq ] , reaches values larger than those associated to the adiabatic regime . for the parameters considered above the efficiency ratio @xmath29 was 165 against the minimum value @xmath83 .
, in mhz , at @xmath84 mhz and @xmath85 w / cm@xmath23 . in right plot time evolution of the excited state occupation ( solid line ) and of ionized fraction excited state ( dashed line).,title="fig : " ] +
we have examined a problem to be classified as a quantum control one : reach a specific target , the ionization of the condensate , minimizing the losses into a different output channel , the production of thermal atoms .
we have explored different two - photon ionization schemes , producing different quantum controls for the atomic excited state .
however the quantum control reached for the two - photon ionization is very poor . in effect the ionization process transferring atoms from the excited state to the ionization state is governed by rate equations , whence a classical response instead of a quantum one .
our final result is that the maximum ionization efficiency to be reached through the quantum control on the first excitation , depends on the probabilities of two processes , the spontaneous decay rate and the ionization rate .
these rate are not modified within the quantum control of our analysis .
+ while the qualitative behavior of the ionized fraction depends only on the properties of the laser pulse exciting the first step of the ionization sequence , the absolute value of the ionized fraction depends only on the intensity of the ir light coupling the excited state to the continuum .
an analogous statement applies also to the efficiency ratio between the thermal fraction and the ionized fraction .
the criteria we have derived for the optimization of the two - photon ionization have a general validity , whereas the absolute numbers we have discussed correspond to the intensities available in the analyzed experiment .
the fraction of ionized atoms in these conditions is much smaller than the number of photons scattered by each atom .
the application of the optimization criteria we have derived , however , together with ir intensities larger than those we have considered , will allow to reach experimental conditions leading to an efficient large photoionization of bose - einstein condensates .
this research was supported by the infm - italy through the photonmatter project , by the miur - italy through a prin project , and by the european commission through the cold quantum - gases network , contract hprn - ct-2000 - 00125 .
thanks n. vitanov for useful discussions .
m. anderlini , d. ciampini , e. courtade , f. fuso , o. morsch , j.h .
mller and e. arimondo , in `` laser spectroscopy '' , proceedings of the xvi international conference , eds .
p. hannaford , a. sidorov , h. bachor and k. baldwin , ( world scientific , 2004 ) p.353 - 361 . | the efficient photoionization of a bose - einstein condensate requires the creation of ions with the smallest possible transfer of atoms from the condensate into the thermal phase .
the spontaneous decay from excited states into atomic states with momentum different from the initial one should be reduced .
we investigate theoretically the two - photon ionization of a rubidium condensate using near resonant excitation to the 6p state and second photon at 421 nm or 1002 nm into the continuum .
different ionization schemes with coherent control of the first excitation and reduction of the spontaneous decay are presented . |
the last decade has seen a crystallisation of ideas regarding the nature of the dressed - gluon and -quark propagators in qcd . in landau gauge the dressed gluon two - point function
is widely held to be described by a momentum - dependent mass function , @xmath0 .
its magnitude is large at infrared momenta : @xmath1 .
however , it vanishes with increasing spacelike momenta : @xmath2 , thereby maintaining full accord with perturbative qcd .
background and context for these observations may be found , e.g. , in refs.@xcite , and citations therein and thereto .
similarly , the dressed quark two - point function is described by two momentum - dependent functions : a wave - function renormalisation , @xmath3 , and mass function , @xmath4 , both of which are strongly modified from their perturbative forms for @xmath5 . in fact , from the confluence of results on @xmath4 obtained with dyson - schwinger equations ( dses ) in qcd and numerical simulations of the lattice - regularised theory , evident , e.g. , in refs.@xcite , a widespread appreciation has emerged of the reality and impact of dynamical chiral symmetry breaking ( dcsb ) in the strong interaction .
since the two - point functions of elementary excitations are strongly modified in the infrared , one must accept that the same is generally true for three - point functions ; i.e. , the vertices .
this was actually realised early on , with studies of the fermion gauge - boson vertex in abelian gauge theories @xcite that have inspired numerous ensuing analyses .
the importance of this dressing to the reliable computation of hadron physics observables was exposed in refs.@xcite , insights from which have subsequently been exploited effectively ; e.g. , refs.@xcite .
many studies of hadron physics observables have employed an _
ansatz _ for the fermion
photon vertex .
the best are informed by analyses that emphasise the constraints of quantum field theory , amongst which are that the vertex should @xcite : be free of kinematic singularities ; ensure gauge covariance and invariance ; and assist in providing for the multiplicative renormalisability of solutions to the dses within which it appears .
_ anstze _ that are largely consistent with these constraints have also been used to represent the dressed quark - gluon vertex . in this connection ,
perhaps , it is clearest that such considerations are not sufficient to fully determine the vertex . as an example , the ball - chiu vertex @xcite augmented by the curtis - pennington extension @xcite is unable to explain the mass splitting between the @xmath6- and @xmath7-mesons , parity partners in the spectrum .
the minimum required to understand this is inclusion of a dressed - quark anomalous chromomagnetic moment @xcite , the presence and strength of which are driven by dcsb @xcite and confirmed by numerical simulations of lattice - qcd @xcite .
we note that extending lattice analyses to the entire kinematic domain of spacelike momenta relevant to the numerous uses of the fermion gauge - boson vertex is numerically challenging @xcite .
we suspect that absent an appreciation within the lattice - qcd community of the physical importance of this problem , much time will elapse before new results are available .
this magnifies the importance of studies in the continuum .
it is in this context that we are motivated to readdress the task of constructing an efficacious and workable vertex _
ansatz_. owing to sensible considerations regarding tractability , the most recent detailed studies @xcite have deliberately overlooked the role of dcsb in building the fermion gauge - boson vertex .
herein , informed by recent developments in hadron physics phenomenology @xcite , notably the dynamical generation of an anomalous magnetic moment for perturbatively massless fermions , we develop a practical _ ansatz _ for the fermion - photon vertex that produces a gauge - independent critical coupling for dcsb in qed and shows some promise as a tool for hadron physics phenomenology .
section [ sec : gapqed ] provides some background on the coupling of a dressed - fermion to a photon .
our _ ansatz _ is developed in sec.[sec : ansatz ] and employed to find a critical coupling for dcsb in sec.[sec : alphac ] .
section [ sec : sandw ] illustrates a phenomenological utility of the model and explains some obvious weaknesses , and sec.[sec : epilogue ] is an epilogue .
much of the progress toward understanding dcsb and the fermion gauge - boson vertex has followed from studies of the gap equation , which in qed can be written @xmath8 where : we employ a poincar invariant regularisation of the integral , with @xmath9 the regularisation mass - scale , which typically doubles as the renormalisation point in dse studies of qed ; @xmath10 is the fermion wave - function renormalisation ( @xmath11 in qed ) ; @xmath12 is the fine - structure constant ; and the dressed - photon propagator is @xmath13\frac{1}{q^2[1+\pi(q^2 ) ] } + \xi \frac{q_\mu q_\nu}{q^4}\,,\ ] ] with @xmath14 the covariant - gauge parameter . whilst our vertex _
ansatz _ is always consistent with one - loop qed perturbation theory , aspects of the infrared behaviour we elucidate are particular to the quenched theory ; viz .
, @xmath15 .
we use a euclidean metric : @xmath16 ; @xmath17 ; @xmath18 $ ] ; @xmath19 ; and @xmath20 spacelike @xmath21 @xmath22 .
the solution of eq . has the form @xmath23 in order to study dcsb , one must define a chiral limit and explore the behaviour of @xmath4 as the value of the fine - structure constant is varied .
this is straightforward if the fine - structure constant is less than some critical value , denoted by @xmath24 , for then @xmath25 defines the chiral limit and @xmath26 is the only solution for the mass function .
it might , however , be viewed as problematic for @xmath27 because four - fermion operators become relevant in strong - coupling qed and must be included in order to obtain a well - defined continuum limit @xcite .
this complication is not of concern to us because one can obtain the critical coupling by approaching this value from below and at strong coupling one can view the cutoff version of the theory as an illustrative model .
owing to the ward - takahashi identity : @xmath28 ( or the slavnov - taylor identity in non - abelian theories ) , eleven independent tensor structures are required to fully express a fermion gauge - boson vertex .
furthermore , @xmath29 can always be decomposed into two pieces : @xmath30 with @xmath31 and @xcite @xmath32 , \label{bcvtx}\\ & : = & \sum_{i=1}^3 \lambda_i(k^2,p^2 ) \ , i l^i_\mu(k , p ) \ , , \label{bcvtx2}\end{aligned}\ ] ] for an abelian theory , where @xmath33 , @xmath34\ , , \\
\delta_{\phi}(k^2,p^2 ) & = & \frac{\phi(k^2)-\phi(p^2)}{k^2-p^2}\,,\end{aligned}\ ] ] @xmath35 , @xmath36 .
we remark that some hints for a practical extension of eq .
to qcd can be found in ref.@xcite and it is conceivable that transverse symmetry transformations might assist in placing constraints on @xmath37 @xcite .
eight independent tensors are required in order to specify the transverse vertex : @xmath38 the following decomposition was introduced in ref.@xcite [ tseuclidean ] @xmath39,\,q = k - p\ , , \\
t^2_{\mu } ( k , p ) & = & -i t_{1\mu } ( \gamma \cdot k + \gamma \cdot p)\,,\\ \label{t3mu } t^3_{\mu } ( k , p ) & = & q^2 \gamma_\mu - q_\mu \gamma \cdot q = : q^2 \gamma_\mu^{\rm t}\ , , \\ t^4_{\mu } ( k , p ) & = & it_{1\mu } p_\nu k_\rho \sigma_{\nu\rho } \,,\\ t^5_{\mu } ( k , p ) & = & \sigma_{\mu\nu } q_\nu \,,\\
t^6_{\mu } ( k , p ) & = & -\gamma_\mu ( k^2- p^2 ) + ( k+p)_\mu \gamma \cdot q \ , , \\ \nonumber t^7_{\mu } ( k , p ) & = & \frac{i}{2 } ( k^2 - p^2 ) [ \gamma_\mu ( \gamma \cdot k + \gamma \cdot p ) -(k+p)_\mu]\\ & & + ( k+p)_\mu p_\nu k_\rho \ \sigma _ { \nu\rho } \,,\\
t^8_{\mu } ( k , p ) & = & k_\mu \gamma \cdot p - p_\mu \gamma \cdot k -i \gamma_\mu p_\nu k_\rho \sigma _ { \nu\rho } \,,\end{aligned}\ ] ] and has since been used widely . as we shall see
, however , it has a couple of pitfalls .
a model for the vertex consists in a choice for the scalar - valued functions @xmath40 . following but expanding upon ref.@xcite
, we choose [ ansatz ] @xmath41 } \ , , \\
\tau_5(k^2,p^2 ) & = & a_5\ , \delta_b(k^2,p^2 ) \,,\\
\tau_6(k^2,p^2 ) & = & \frac{a_6(k^2+p^2 ) \ , \delta_a(k^2,p^2)}{[(k^2-p^2)^2+(m^2(k^2)+m^2(p^2))^2]}\,,\rule{2em}{0ex } \\
\tau_7(k^2,p^2 ) & = & \frac{a_7\ , \delta_b(k^2,p^2)}{(k^2+p^2 ) }
\,,\\ \tau_8(k^2,p^2 ) & = & a_8 \delta_a(k^2,p^2)\,,\end{aligned}\ ] ] where @xmath42 are momentum - independent constants .
this construction draws from a direct comparison with the structural dependence of the ball - chiu vertex on the functions that constitute the fermion propagator ; and the momentum - dependence of each term guarantees that from our _ ansatz _ one recovers a vertex which possesses the appropriate leading - order perturbative behaviour for @xmath43 . the coefficients @xmath44 are not independent .
as we now illustrate , they are interconnected by numerous constraints from perturbative qed and gauge covariance . at one - loop in an arbitrary covariant gauge
, the fermion - photon vertex obeys : @xmath45\,.\end{aligned}\ ] ] in the context of eqs .
, this demands @xcite @xmath46 in addition , given the anticipated asymptotic behaviour of the scalar functions in the dressed - fermion propagator , then for @xmath43 the other terms in eqs .
decay as follows , up to @xmath47$]-factors : @xmath48 i.e. , damping in agreement with one - loop perturbation theory @xcite . in building an _ ansatz
_ , it is natural to insist on such correspondences with perturbation theory . some comments on eq .
are necessary .
consider the choice @xmath49 , which corresponds precisely to the _ ansatz _ of ref.@xcite .
taken as a statement about the vertex on a carefully defined domain of asymptotically - large spacelike momenta , this assumption is internally consistent . however , whilst there are mitigating considerations , implemented as a constraint on the vertex for the entire domain of @xmath50 , the assumption @xmath51 is generally mistaken .
the quantity @xmath52 is associated with @xmath53 in eq . and hence contributes as follows to the complete vertex : @xmath54 since @xmath55 is the leading tensor structure associated with a vector meson bound - state and given the existence of the @xmath6-meson , no realistic solution of the inhomogeneous bethe - salpeter equation for the fermion - photon vertex can produce a coefficient of @xmath55 that is identically zero . notwithstanding this , the choice @xmath49 is not worse than using @xmath56 alone . in the wigner phase , multiplicative renormalisability of the fermion propagator requires that @xmath57 , where @xmath58 is an anomalous dimension @xcite .
it is multiplicative renormalisability that ensures the absence of overlapping divergences in the tower of dses . in quenched - qed
, one finds @xcite @xmath59 in connection with our vertex _
, power law behaviour of @xmath3 is guaranteed so long as @xmath60 this condition ensures the absence of @xmath61-divergences and exposes the constraint @xmath62 = 0 \ , .
\label{adnan1}\end{aligned}\ ] ] the choice @xmath63 , discussed in connection with eq . , produces that anomalous dimension associated solely with gauge covariance @xcite .
this is the leading - order result .
it is known to be incomplete but corrections are higher - order in @xmath12 and depend on the truncation , as evident via eq .. they become crucial in strongly coupled theories @xcite . as one approaches the bifurcation point associated with the onset of dcsb from below ,
the behaviour of @xmath3 begins to influence the dressed - fermion mass function , which itself possesses power - law behaviour in the ultraviolet @xcite . in this instance , elimination of @xmath64-divergences requires @xmath65 the tensor @xmath66 is that matrix structure associated directly with the pauli form factor of an on - shell fermion .
one should therefore expect a realistic vertex _ ansatz _ to have @xmath67 . as a consequence ,
at least one of @xmath68 must be nonzero .
pursuing this point , consistency with the one - loop result for the anomalous magnetic moment of an on - shell fermion with mass @xmath69 @xcite entails @xmath70 where the functions are evaluated at @xmath71 , @xmath72 . in the context of our model , the content of eq .
is readily illustrated : in landau gauge ( @xmath73 ) @xmath74 at one - loop order and eq .
becomes the constraint @xmath75 using the one - loop expression for @xmath76 , neglecting @xmath77 terms for simplicity , then the constraint entails @xmath78 .
( inclusion of the @xmath79-term reduces this value by 11% , which is a negligible effect for our purposes . )
observe now that the tensors @xmath80 can be reexpressed : @xmath81 following numerous earlier studies , we have composed a vertex _ ansatz _ based on these tensors and required that their coefficient functions be free of kinematic singularities ; viz . , regular at @xmath72 .
, their net contribution to the vertex is @xmath82 the asymptotic domain @xmath43 corresponds to large values of @xmath83 , upon which the contribution is therefore @xmath84 plainly , the kinematic dependence on @xmath85 plays no role asymptotically and the standard analysis proceeds without reference to it . on the infrared domain , however , the kinematic dependence on @xmath85 is crucial . with the choices described above , the contribution from @xmath86 to the fermion s anomalous magnetic moment vanishes .
this is a pitfall of the basis in eq .. with hindsight , one could equally have chosen to use the tensor basis in ref.@xcite , which corresponds to @xmath87 in this case eqs .
, suggest alternative natural choices for the coefficient functions : @xmath88 , @xmath89 ; eq .
represents the vertex contribution for all values of @xmath85 ; and the fermion s anomalous magnetic moment receives contributions from both these terms .
equation is then modified as follows : @xmath90 a comparison with leading - order perturbation theory would now inform constraints on additional structures in the vertex .
for example ,
eq . and ref.@xcite together entail that @xmath91 should assume functional forms that vanish at leading - order in landau gauge . at this point
a comparison with ref.@xcite is useful .
the function @xmath92 here corresponds to @xmath93 therein , which does not vanish at @xmath72 .
similarly , @xmath94 in ref.@xcite , which correspond to @xmath95 herein , are nonzero and significant at @xmath72 .
this suggests that a tensor basis which avoids multiplicative factors of @xmath85 is better suited to building vertex _ anstze _
intended for use on the entire domain of @xmath50 that is sampled in both nonperturbative solutions of truncated dses and applications in hadron physics .
having detailed an _ ansatz _ and a number of constraints , we are now in a position to solve the gap equation
; i.e. , the coupled equations for @xmath3 , @xmath4 . these equations simplify in the neighbourhood of @xmath24 ; viz .
, the coupling whereat a @xmath96 solution bifurcates away from the @xmath97 solution , which alone is possible in perturbation theory @xcite .
the behaviour of the solutions near the bifurcation point may be investigated by performing functional differentiation of the gap equations with respect to @xmath4 and evaluating the results at @xmath98 .
practically , this amounts to analysing linearised forms of the original gap equations ; i.e. , the equations obtained by eliminating all terms of quadratic or higher order in @xmath4 . since the only mass - scale at the bifurcation point is the regularisation parameter , then in the vicinity of @xmath24 one
may uniformly approximate the mass function as @xmath99 , @xmath100 with @xmath101 the mass anomalous dimension .
combined with our vertex _
ansatz _ , power - law behaviour for both @xmath102
, @xmath103 enables one to evaluate the angular and radial integrals in the gap equation .
this produced eq . and here yields the following equation in the quenched truncation : @xmath104 - \left ( \frac{a_2}{3 } -2a_6 \right ) \left [ \frac{1}{s-\nu } + \frac{1}{s-\nu -1 } -\frac{1}{s}+\frac{1}{1-s } \right]\\ \nonumber & & - \frac{2a_2}{3 } \bigg[\frac{1}{s } -\frac{1}{s-\nu } - \overline{\psi}\left ( \frac{s}{2 } \right)+\overline{\psi } ( s ) + \overline{\psi}\left ( \frac{\nu -s}{2 } \right ) \nonumber - \overline{\psi}(\nu - s ) \bigg ] \\ \nonumber & & - \frac{a_4}{12 } \left [ \frac{1}{1 + s } + \frac{1}{2 - s } + \frac{1}{\nu - 1 } - \frac{1}{\nu + 2 } \right ] + \left ( \frac{a_7}{6 } - \frac{a_1}{3 } \right ) \bigg [ \frac{1}{2 } \bigg ( \frac{3}{s } + \frac{1}{1 - s } + \frac{3}{\nu } - \frac{1}{\nu + 1 } - \pi \cot\pi s \\
\nonumber & & - \pi \cot\pi \nu + \pi \cot\frac{\pi s}{2 } + \pi \cot\frac{\pi \nu}{2 } \bigg ) - \psi\left(-\frac{s}{2}\right ) + \psi\left(\frac{\nu}{2}\right ) + \overline{\psi}(s ) - \frac{1}{2 } \overline{\psi}\left(\frac{s}{2}\right ) - \overline{\psi}(\nu ) + \frac{1}{2 } \overline{\psi}\left ( \frac{\nu}{2 }
\right ) \bigg ] \\ & & + a_7 \bigg [ \frac{1}{4 } \left ( \frac{5}{s } + \frac{3}{1 - s } + \frac{5}{\nu } - \frac{3}{\nu + 1 } - \pi \cot\pi s - \pi \cot \pi \nu ] \right ) + \psi(-s ) - \psi\left(-\frac{s}{2}\right ) - \psi(\nu ) + \psi\left ( \frac{\nu}{2 } \right ) \bigg ] \bigg\ } \,,\end{aligned}\ ] ] where : @xmath105 is the euler function ; @xmath106 , its logarithmic derivative ( digamma function ) ; and @xmath107 .
the absence of power - law divergences is guaranteed so long as @xmath108 , @xmath109 .
in addition , eq . means that @xmath110 is not an independent parameter .
our initial goal is to locate the bifurcation point , @xmath24 .
consider eqs .
one finds that for given values of @xmath111 and vertex _ ansatz _ parameters , eq . produces a solution for @xmath58 . using these parameters in eq .
, one finds no solutions for @xmath112 if @xmath113 and two solutions if @xmath114 : @xmath115 has the appearance of a flattened catenary ; and the bifurcation point is found when the locations of this function s roots and its interior extremum coincide .
this condition may be imposed by requiring @xmath116 in addition to eqs . , .
we now demand a little more ; namely , that our _ ansatz _ produce a landau - gauge value of @xmath24 which agrees with that produced by the ball - chiu _ ansatz _ @xcite and minimises @xmath117 on @xmath118 $ ] .
( we choose the ball - chiu result to define @xmath119 because this _ ansatz _ is the minimal vertex consistent with the ward and ward - takahashi identities . ) for this purpose @xmath120 ; i.e. , a function that describes a smooth surface in six dimensions , and our demand is straightforwardly mapped into a multidimensional extremisation problem .
namely , find that set @xmath121 which produces @xmath122\}$ ] .
we obtain a solution with @xmath123 where @xmath124 was fixed following eq . , and @xmath125
are determined from the other _ ansatz _ parameters via eqs .
n.b . with @xmath126 , we comply with an observation made after eq .. our result is the solid curve in fig.[fig : alphac ] , which illustrates that it is straightforward to obtain a critical coupling that is almost insensitive to @xmath14 if all amplitudes are retained in the vertex .
notably , this _ ansatz _ depends neither on @xmath14 nor the angle defined by @xmath127 . moreover , if any one of the parameters @xmath44 in eq .
is allowed to depend on @xmath14 , then @xmath128 is guaranteed .
it is important to observe that @xmath129 in landau gauge .
this result emphasises that four - fermion operators become relevant to qed for @xmath114 because the operator @xmath130 has dimension @xmath131 . as a consequence
, one can not sensibly compute a fermion condensate unless this operator is included .
that , however , introduces a new parameter , the operator s coupling strength , which can not readily be constrained . in these circumstances ,
no attempt to enforce @xmath14-independence of the condensate can produce additional meaningful constraints on the vertex .
it is worth digressing here and reiterating that landau gauge occupies a special place @xcite .
for example , it is a fixed point of the renormalisation group ; and that gauge for which the one - loop contribution to @xmath132 vanishes in any number of dimensions in any renormalisable gauge theory ( see eq . and ref.@xcite ) .
it follows from the latter that in landau gauge any sensitivity to model - dependent differences between _ anstze _ for the fermion - photon vertex are least noticeable .
one may therefore argue that all discussion of the gauge - parameter dependence of a vertex _ ansatz _ is moot , since the vertex in anything other than landau gauge should simply be defined as the landau - khalatnikov - fradkin ( lkf ) transform @xcite of the landau gauge form .
the sensible implementation of this procedure guarantees gauge covariance and hence obviates any question about the gauge dependence of gauge invariant quantities . notwithstanding these observations ,
the lkf transform of a general vertex _
ansatz _ is practically difficult to obtain @xcite , and , as we have illustrated herein , requiring gauge - parameter independence of physical quantities computed with a given vertex can inform and constrain the construction of an _ ansatz_.
we now address aspects of an additional issue ; namely , whether the _ ansatz _ expressed in eqs. , is adequate for use in nonperturbative studies of truncated dses or to describe the dressed - quark photon coupling in hadron physics phenomenology .
both applications sample domains of @xmath50 that stretch far outside those considered in asymptotic analyses .
we limit ourselves to the context provided by the anomalous magnetic moment of a dressed fermion , in part because above we have highlighted dcsb and this emergent phenomenon produces an anomalous magnetic moment for a perturbatively massless fermion : it is impossible for a truly massless fermion to possess a measurable anomalous magnetic moment @xcite .
we follow ref.@xcite in constructing an anomalous magnetic moment distribution . at each value of spacelike-@xmath133 ,
define spinors to satisfy @xmath134 where @xmath135 ; i.e. , the mass - function evaluated at that value of @xmath133 . in this case @xmath136u(p,\mathpzc{m } ) \ , , \rule{1em}{0ex}\end{aligned}\ ] ] owing to a gordon identity , and the fermion s anomalous magnetic moment is @xmath137 . with the _ ansatz _ specified by eqs. , , @xmath138 where @xmath139 , @xmath140 .
the tensor denominated @xmath141 herein is associated with @xmath142 in ref.@xcite . )
the distribution is plotted in fig.[fig : dist ] , where the euclidean constituent fermion mass @xmath143 .
it is immediately apparent that dcsb produces a distribution that is large on the nonperturbative infrared domain : @xmath144 , but vanishes with the strength of this dynamical effect .
more striking in the present context , however is the degree of similarity between the dashed and dot - dashed curves .
this signals that our vertex _
ansatz _ passes a nontrivial test .
namely , whilst expressed solely in terms of the functions in the dressed - fermion propagator , it nevertheless produces a magnetic moment distribution in fair agreement with the most realistic symmetry - preserving solution of the inhomogeneous vector bethe - salpeter equation that is currently available .
it is therefore worth employing this ansatz more widely ; e.g. , in the computation of hadron electromagnetic form factors @xcite . notwithstanding this
, there are some caveats that should be borne in mind . as observed in sec.[subsec : oneloop ] , no _ ansatz _ for the transverse part of the fermion - photon vertex is completely satisfactory if it does not express what might be called leakage into the spacelike region of spectral strength from both the two - pion continuum and the region of the @xmath6-meson pole .
if one concentrates on the spacelike domain , then such effects are maximal near @xmath72 and may be characterised by observing that they alter charge radii by @xmath145% @xcite . related , and more important perhaps , are the limitations of the tensor basis in eqs . that we discussed in sec.[subsec : amm ] .
namely , analyses concerned with domains of ultraviolet momenta are not necessarily a fair guide to the @xmath85-dependence of the vertex in the infrared .
explicit computations @xcite present conflicts with expectations fed by _ anstze
_ built upon eqs .
, . finally , it is worth investigating how our _ ansatz _ fares in the dse for the photon vacuum polarisation , which is known to be sensitive to features and kinematic domains that are not strongly constrained by the gap equation @xcite .
such analyses would likely lead to a refinement of our model .
motivated by the fact that knowledge of the dressed - fermion gauge - boson vertex is critical to any continuum study of a gauge field theory , we have detailed a workable model for the dressed - fermion - photon vertex , @xmath29 .
it is expressed solely in terms of functions which appear in the dressed - fermion propagator , is independent of the angle defined by @xmath127 , and does not explicitly depend on the covariant - gauge parameter .
ansatz _ is nevertheless consistent with constraints that have long been held important , namely : it is free of kinematic singularities ; ensures gauge covariance and invariance in the application tested ; and assists in providing for the multiplicative renormalisability of solutions to the dses within which it appears .
significantly , the _ ansatz _ contains nontrivial factors associated with those tensors that are even in the number of dirac matrices ; i.e. , whose appearance is expressly driven by dynamical chiral symmetry breaking in a perturbatively massless theory .
this novel feature enables a direct and positive comparison with the best available symmetry - preserving solutions of the inhomogeneous bethe - salpeter equation for the vector vertex .
the positive outcome indicates that our model might provide a much needed tool for use in poincar - covariant symmetry - preserving studies of hadron electromagnetic form factors .
furthermore , given the general nature of our constraints and the simplicity of our construction , there is some room to hope that a straightforward extension of our approach might yield an _ ansatz _ adequate to the task of representing the dressed - quark - gluon vertex .
we are grateful for useful input from c. chen , , p.c .
tandy and d.j .
this work was supported by the _ programa de cooperacin bilateral mxico - estados unidos _ ( conacyt 2009 - 2011 ) with counterpart funding from the u.s
. national science foundation , under grant no .
nsf - phy-0903991 ; conacyt project 46614-f ; coordinacin de la investigacin cientfica ( cic ) project no . 4.10 ; and the u.s .
department of energy , office of nuclear physics , contract no .
de - ac02 - 06ch11357 . | we present a workable model for the fermion - photon vertex , which is expressed solely in terms of functions that appear in the fermion propagator and independent of the angle between the relative momenta , and does not explicitly depend on the covariant - gauge parameter .
it nevertheless produces a critical coupling for dynamical chiral symmetry breaking that is practically independent of the covariant - gauge parameter and an anomalous magnetic moment distribution for the dressed fermion that agrees in important respects with realistic numerical solutions of the inhomogeneous vector bethe - salpeter equation . |
the x - ray binary gx 339@xmath04 comprises a compact primary which is a strong black hole candidate ( bhc ) , with mass function of [email protected] ( hynes 2003 ) and a secondary which is likely to be an evolved low mass star ( shahbaz , fender & charles 2001 ; chaty 2002 ) .
the system has an orbital period of 1.75 days ( hynes 2003 ) and is located at a distance of at least 4 kpc ( zdziarski 1998 ; see also shahbaz , fender & charles 2001 and maccarone 2003 ) , with an orbital inclination of less than 60 to the line of sight , as inferred from the lack of eclipses ( cowley 2002 ) .
gx 339@xmath04 has been a key source in our understanding of the relation between accretion and the production of relativistic jets .
it was the first bhc to reveal a positive correlation between radio and x - ray fluxes in the low / hard x - ray state ( hannikainen 1998 , later quantified by corbel 2003 ) , and to demonstrate an association between the ` quenching ' of core radio emission and the transition to a high / soft x - ray state ( fender 1999 ; see e.g. done 2001 , for a review on x - ray states and fender 2004 for a comparison of radio and x - ray behaviour in galactic bhcs ) .
after spending almost three years in ` quiescence ' , gx 339@xmath04 re - brightened in x - rays at the end of 2002 march ( smith 2002 ) and changed rapidly to a soft outburst state , undergoing a dramatic state change in 2002 may .
this transition was associated with a bright radio flare ( fender 2002 ) , reaching four to five times the brightest radio level ever observed from the source ( see corbel 2000 for the long - term behaviour of gx 3394 ) . by analogy with other systems ( xte j1550564 , corbel 2001 ) , this flare was likely to be the signature of a powerful ejection event .
repeated radio observations of gx 339@xmath04 have indeed confirmed this association : the 2002 radio flaring has led to the formation of a large - scale relativistic radio jet , whose morphology and dynamics will be presented in the course of this letter .
the australia telescope compact array ( atca ) performed eight continuum observations of gx 339@xmath04 at roughly regular intervals between 2002 april and august , simultaneously at 4800 mhz ( 6.3 cm ) and 8640 mhz ( 3.5 cm ) .
three further observations were performed in 2003 january , march and may at four frequencies : 1384 mhz ( 21.7 cm ) , 2368 mhz ( 12.7 cm ) , 4800 and 8640 mhz .
the atca synthesis telescope is an east - west array of six 22 m antennas with baselines ranging from 31 m to 6 km ; it uses orthogonal polarized feeds and records full stokes parameters ( i , q , u , v ) . the target was systematically offset by about 10 arcsec from the array phase centre , in order to avoid possible artefacts due to system errors such as dc - offsets . in each observation ,
pks 1934638 was used for absolute flux and bandpass calibration , while either pmn 16034904 , pmn 16505044 or pmn 17265529 was the phase calibrator for antenna gains and phases .
the data reduction process and image analysis have been carried out with the multichannel image reconstruction , image analysis and display ( miriad ) software package ( sault , teuben & wright 1995 ; sault & killeen 1998 ) .
dates of atca observations are indicated in fig .
[ fig1 ] , superimposed on the _ rossi x - ray timing explorer_/all sky monitor ( rxte / asm ) 212 kev light curve of the system . smith ( 2002 ) reported an increase in the x - ray flux from gx 339@xmath04 , which had been in quiescence for almost three years ( kong 2000 ) , on 2002 march 26 ( mjd 52360 ) .
the source reached a peak flux ( 212 kev ) of @xmath50.8 crab on 2002 may 15 ( mjd 52410 ) , followed by a decrease to @xmath50.4 crab , and a secondary rise to @xmath50.9 crab around mjd 52500 , after which gx 339@xmath04 has started a slow return to ` quiescence ' ( reached by the time of writing ) .
the hardness ratio hr2 suggests that a rapid transition from a low / hard to a softer x - ray state took place a few days before the ( first ) outburst peak , although colour and timing analysis of the rxte proportional counter array data indicates a smooth transition from a low / hard to a very high state in the rising phase , followed by a high / soft state ( or possibly an ` oscillating ' very - high state ) after the peak ( belloni 2002 ; nespoli 2003 ) .
atca observations performed between 2002 april and june have detected the brightest radio flare ever observed by the system , which reached a peak flux density of about 55 mjy on may 14 ( mjd 52408 ) , almost contemporaneously with the ( first ) x - ray peak . [
cols="<,^,^,^,^,^,^,^,^ " , ] persistent large - scale radio jets observed in 1e 1740.72942 ( mirabel 1992 ) and grs 1758258 ( rodrguez , mirabel & mart 1992 ) have been found to extend up to 13 pc ; in the case of grs 1915 + 105 instead , relativistic ejecta were tracked up a projected distance of 0.08 pc ( mirabel & rodrguez 1999 ) , while in the large - scale x - ray jet powered by xte j1550564 ( corbel 2002 ) , the eastern jet has been detected to a projected physical separation of 0.75 pc covered in 4 years .
the large - scale radio jet of gx 339@xmath04 displays something , with a projected extension of 0.23 pc ( at 4 kpc ) covered in about 250 days ( if associated with the may 2002 radio flare ) . scheduled _
observation will discover if the jet is active in x - rays as well , possibly confirming the similarity with xte j1550564 ( corbel 2002 ; kaaret 2003 ; tomsick 2003 ) , where the x - ray jet is still capable of accelerating particles up to tev energies four years after the main ejection event .
we thank dick hunstead and duncan campbell - wilson for kindly providing us with preliminary results of the most observations .
the australia telescope is funded by the commonwealth of australia for operation as a national facility managed by csiro .
rxte / asm results are provided by the asm / xte team at mit .
belloni t. , nespoli e. , homan j. , van der klis m. , lewin w. h. g. , miller j. m. , mendz m. , 2002 , new views on microquasars , the fourth microquasars workshop , ed .
durouchoux , y. fuchs , and j. rodrguez , center for space physics : kolkata ( india ) , 75 | we report on the formation and evolution of a large - scale , synchrotron - emitting jet from the black hole candidate and x - ray binary system gx 339@xmath04 . in 2002
may , the source moved from a low / hard to a very high x - ray state , contemporaneously exhibiting a very bright optically thin radio flare .
further observations with the australia telescope compact array have tracked the formation of a collimated structure extending to about 12 arcsec , with apparent velocity greater than 0.9@xmath1 .
the luminosity of the outflow seems to be rapidly decreasing ; these observations confirm that transient large - scale jets are likely to be common events triggered by x - ray state transitions in black hole x - ray binaries .
2@xmath2 1s@xmath3 2@xmath2 1s@xmath3 accretion , accretion discs binaries : general ism : jets and outflows radio continuum : stars x - rays : stars individual : gx 339@xmath04 |
the purpose of this paper is two - fold : to show how standard schrdinger equation analysis familiar in a simple physics situation may be used to understand an important problem in biology , and to study , via such an analysis , the effect of spatial variation of resources on the ecologically relevant problem of extinction of populations of biological species .
extinction in biology is intimately related , as we shall show below , to abrupt transitions in physical systems , and is important whether the focus is on the dynamics of bacteria in a petri dish @xcite or of the movement of larger animals such as rodents involved in the spread of epidemics @xcite .
consider the logistic equation @xmath0 which may describe the time dependence of the density @xmath1 of a spatially uniform population subject to a growth rate @xmath2 and a resource competition term controlled by the parameter @xmath3 .
the steady state density undergoes a transcritical bifurcation at @xmath4 as @xmath2 is varied from negative to positive values .
the physical meaning of the bifurcation is straightforward . whereas for positive @xmath2 the stable steady state value of @xmath1 is @xmath5 , for negative @xmath2 the density vanishes in the steady state because the growth rate is then really a death rate .
introduction of space as a mere parameter @xmath6 brings in no new behavior .
a new aspect does enter , however , if spatial _ derivatives _ are introduced into eq .
( [ logistic ] ) through , for instance , diffusion terms which represent the motion of the members of the species from one spatial location to another .
it is this situation , represented by a fisher equation @xcitewith spatially varying growth rate @xmath7 and diffusion constant @xmath8 , @xmath9 that is the subject of the present paper .
we study the 1-d case for simplicity . as expected , the steady state @xmath10 has a tendency to follow the spatial variation of the growth rate @xmath7 but is modified by the effects of diffusion .
for instance , if @xmath7 has the value @xmath2 in a segment of space , and the value @xmath11 outside it , as shown first by skellam @xcite , @xmath10 is given by the square of the elliptic @xmath12 function @xcite .
thus @xmath10 , which is maximum at the center of the favorable region and drops off gradually towards the edges , is not proportional to @xmath7 .
the latter has the shape of a step function
. equation ( [ fisher ] ) , which has been quite successful in the description of ecological dynamics @xcite , predicts interesting behavior in addition to the mere difference in shapes of @xmath7 and @xmath10 .
the maximum value @xmath13 of the density within the favorable region , i.e. , the spatial segment in which @xmath7 is positive , naturally depends on the length @xmath14 of the segment , and decreases if @xmath14 is decreased .
what is striking is that there is a minimum value of @xmath14 below which @xmath13 , and therefore the density throughout , vanishes identically .
this minimum ( critical ) value can be obtained , except for a proportional constant equal to @xmath15 , by equating the growth time @xmath16 to the diffusion time @xmath17 : @xmath18 the underlying physical idea is that the critical length of the favorable segment corresponds to the situation that the random walker ( bacterium or animal ) traverses the length of the segment diffusively in the time necessary for growth , and falls prey to the harsh conditions outside the segment .
our interest in the present paper is to ask how this well - known result @xcite is modified by the simultaneous introduction of two items of realism into the growth @xmath7 : a gradual , rather than step - like , spatial variation of the growth rate at the border of the segment , and a non - infinite value for destruction outside the segment .
the latter , by itself , has been analyzed earlier @xcite but the former has not , to the best of our knowledge .
we provide an analysis of the combined situation below on the basis of wave function calculations from the time - independent schrdinger equation describing a quantum particle in a uniform gravitational or electrostatic field .
the analysis we present is applicable to at least two specific biological systems that have been studied recently : a bacterial population in a petri dish , and rodents responsible for hantavirus infection spread . in several experiments @xcite on the former system
, the bacterial population is bathed in the presence of deadly ultraviolet radiation , except in a localized region covered with a mask that protects the bacteria .
the growth rate @xmath7 is positive in the protected region and negative outside . in observations of the second system , there are patches of resources which make possible population growth and sustainance of rodents such as mice in parts of the landscape but not beyond those patches .
earlier analyses have treated the patches or the masks to correspond to _ sharply _ changing resources .
the real experimental situation is naturally different .
the ultraviolet light that falls on the mask seeps under the edges of the mask . as a consequence ,
the change in the growth rate is gradual rather than sudden .
similarly , the available water and food in the rodent patches change gradually in the landscape .
transitions have been reported in experiments on the bacterial system @xcite following quantitative predictions made earlier @xcite .
transitions also seem to be present in rodent experiments although the observations are more difficult to unravel . the observational relevance of our present investigation should be , thus , clear in both cases .
the paper is set out as follows . a simple model that incorporates a gradual variation of resources in space
is presented in section 2 and analytical formulae are derived in terms of airy functions .
limits are studied in section 3 and concluding remarks appear in section 4 .
the model we analyze is as given in eq .
( [ fisher ] ) with the growth rate @xmath7 as in fig.1 .
it equals the constant @xmath2 for a spatial segment of length @xmath14 around the origin @xmath19 , and drops gradually ( linearly ) to the constant negative value @xmath20 outside the segment , the spatial variation being linear ( constant slope ) . during the linear drop
, the rate remains positive on either side of the segment for an additional distance @xmath21 as shown . using the suffixes @xmath22 , @xmath23 and @xmath24 for the respective regions i , ii and iii in fig.1
, we have @xmath25 we do not attempt to solve either this nonlinear partial differential equation or the corresponding nonlinear ordinary differential equation for the steady state @xmath10 obtained by setting the time derivative equal to zero .
rather , we argue that , because our interest lies only in the extinction phenomenon , i.e. , the vanishing of @xmath10 , we can eliminate the bilinear terms of the density , and match solutions of the ensuing _ linear _ equation across the interfaces of the three regions i , ii and iii ( see fig .
1 ) . this argument appears to have been given first by ludwig et al .
@xcite in his investigations of the spatial patterning of the budworm .
it has been used recently by the present authors to propose double - mask experiments for the bacterial problem @xcite .
the crux of the argument is that , at the transition , the densities vanish and therefore terms of order higher than the first may be safely neglected even while the linear terms are compared to extract the extinction condition . the critical value of @xmath14 signifying extinction which result for _ any positive _ @xmath3 can therefore be calculated precisely by considering the linear equation obtained by putting @xmath3 equal to zero .
the problem we are presented with here is simply of finding of the wavefunction for a time - independent schrdinger equation of a quantum particle in a potential which is linear in @xmath6 as in the case of a uniform gravitational or electrostatic field . defining the dimensionless position @xmath26 and the quantities @xmath27 and @xmath28 ,
all obtained by dividing @xmath6 , @xmath29 , and @xmath21 respectively by the diffusion length @xmath30 , we focus on the linear steady state counterpart of eq .
( [ pde ] ) , with the notation @xmath31 : @xmath32 the growth rate plays the role of the potential in a 1-dimensional quantum mechanical problem in a constant force field .
the first of these equations is solved trivially in terms of trigonometric functions , the third in terms of exponentials , and the second in terms of airy functions .
showing its value as the constant @xmath2 in a segment of length @xmath14 joined through a linear decrease to the value @xmath20 on both sides .
the width of the transition region is @xmath21 on either side.,width=302 ] specifically , using the fact that the system has symmetry about @xmath33 , the solution in the three regions is given by @xmath34\nonumber\\ & & + b_2bi\left[-r^{2/3}\left(1-\frac{x - l}{r}\right)\right ] , \nonumber\\ u_3(x)&=&a_3e^{-\xi x } , \label{solns } \end{aligned}\ ] ] where @xmath35 and @xmath36 are airy functions and the @xmath37 s and @xmath38 are constants to be determined from the boundary conditions . matching the logarithmic derivative of the solutions across the interfaces and introducing @xmath39 for notational convenience , we arrive at the following condition for extinction .
@xmath40 } { ai(-\eta^2)bi^{\prime}(\xi^2\eta^2 ) -ai^{\prime}(\xi^2\eta^2)bi(-\eta^2)+\xi\eta\left[ai(-\eta^2)bi(\xi^2\eta^2)-ai(\xi^2\eta^2)bi(-\eta^2)\right]}\right ) .
\label{central}\ ] ] primes denote derivatives .
the normalized extinction length @xmath41 is the arctangent of the right hand side if the latter is positive , and is zero otherwise
. equation ( [ central ] ) is the central result of this paper .
the crucial parameters in the expression are @xmath42 which measures the transition region relative to the diffusion length and @xmath43 which measures the death rate outside , relative to the growth rate inside , the segment . as is clear from the expression , the second parameter
can also be chosen naturally as the product @xmath44 .
if the transition region vanishes ( @xmath45 ) , we take the limit @xmath46 in eq . ( [ central ] ) and recover the ludwig formula @xcite @xmath47 in the general case @xmath48 , the extinction value normalized to this ludwig value is given by plotting the analytic expression in eq .
( [ central ] ) .
2 shows , in the form of the solid line in the main figure , this dependence of @xmath49 on @xmath28 , the normalized extent of the transition region , for a specific value of @xmath43 , viz . , 3 .
the solid circles are obtained via the numerical solution of the full nonlinear equation ( [ pde ] ) , for a nonzero @xmath50 in appropriate units .
note that eq .
( [ pde ] ) can not be solved analytically , and that the complete agreement between the numerical results from the nonlinear equation ( whatever the value of @xmath3 ) and the analytic result from the linear extinction analysis ( that has no @xmath3 ) vindicates our theoretical procedure . )
based on the linear theory ( solid line ) and the numerical solution of the full nonlinear equation ( circles ) .
plotted in the main figure is the extinction length normalized to the ludwig value , @xmath49 , versus the extent of the transition region @xmath28 for @xmath51 .
the inset shows steady state density profiles corresponding to three different values of @xmath14 larger than the critical , for @xmath51 and @xmath52 .
distance @xmath26 is expressed in units of diffusion length @xmath30 while the steady state density @xmath53 is expressed in units of @xmath5 .
the density decreases as the patch length decreases and vanishes at the critical value , at which @xmath54 . ]
various limiting cases may be studied from our eq .
( [ central ] ) through ascending and asymptotic expansions of the airy functions .
an example of the former is @xcite @xmath55 where , @xmath56 and @xmath57 , @xmath58 .
one of the goals of the present paper is the generalization of the skellam result ( [ dirichlet ] ) in the presence of a transition region . for this , we let @xmath59 in ( [ central ] ) and obtain @xmath60 .
\end{aligned}\ ] ] ) in the presence of a transition region of extent @xmath21 .
we plot @xmath61 as a function of @xmath62 .
the solid line is the exact result ( eq .
[ skelgen ] ) and the dashed line the approximation retaining the lowest power of @xmath63 .
note that @xmath64 for @xmath65 . ]
we divide eq .
( [ skelgen ] ) by the skellam result ( [ dirichlet ] ) , and plot @xmath61 as a function of the transition region extent ( i.e. , @xmath62 ) in fig .
the exact expression is the solid line .
it is approximated reasonably well for small @xmath28 by the ascending series approximation @xmath66 denoted by a dashed line , with @xmath67 .
important to note is that the exact @xmath14 becomes zero at a finite value of the transition region extent ( at @xmath68 ) .
this follows from the existence of a zero of @xmath69 at finite @xmath28 ; the physical significance is commented on below .
( in units of @xmath30 ) on the transition region extent in ( a ) and against the ration of the death to the growth rate in ( b ) , plotted from the airy function formula eq .
( [ central ] ) .
note the threshold values of @xmath70 and @xmath43 below which extinction does _ not _ occur.,width=366 ] the extinction length @xmath14 given by the exact expression ( [ central ] ) is plotted in fig .
4 against ( a ) the reciprocal of the transition region extent , @xmath70 , for different values of @xmath43 as shown , and ( b ) the square root of the ratio of the death rate to the growth rate , @xmath43 for different values of @xmath28 . the extinction length becomes zero for large enough @xmath28 in ( a ) and small enough @xmath43 in ( b ) .
this means that there is no extinction for @xmath71 and @xmath72 where @xmath73 and @xmath74 are the threshold values .
these threshold values are plotted in fig .
5 directly from the graphs .
we have verified that the curves coincide precisely with those obtained by putting the numerator of eq .
( [ central ] ) equal to zero .
and @xmath73 obtained graphically .
they are found to agree with the prediction of eq .
( [ central ] ) obtained by putting its numerator equal to zero.,width=366 ] this vanishing of the extinction length @xmath14 , signifying that the extinction phenomenon disappears for large enough transition region or small enough death rate , stems from the definition of @xmath14 as the top segment of the @xmath75 trapezoid in fig .
1 . thus , fig . 6 shows with this segment of zero extent ( @xmath64 ) that the favorable patch is still wide enough to make possible for the walkers to survive in the protected patch because of the transition region .
, there being no extinction because the transition region itself provides a large enough favorable patch for survival.,width=302 ]
our focus in this paper has been to study the combined effect of a gradual change in resources at the borders of a patch , and of finite death rates outside the patch , on the extinction transition of bacteria in a petri dish , rodents in a landscape , or of similar random walkers that can move over an inhomogeneous spatial region .
although the system is essentially nonlinear , and the extinction transition is a consequence of the nonlinearity ( coupled to the inhomogeneity ) , we have been successful in obtaining exact solutions through an analytic procedure .
the procedure draws on an analogy to the solution of a time - independent schrdinger equation and results in an expression for the critical length of the patch in terms of airy functions and their derivatives .
our central result is eq .
( [ central ] ) .
one of the crucial parameters is @xmath42 which is @xmath63 , where @xmath28 is the length of the transition region in units of the diffusion length @xmath30 .
the appearance of the latter is natural since the underlying process is the movement of the random walker ( the diffusing entity ) from within the favorable patch to the unfavorable space outside within the growth time .
the emergence of the third power is a consequence of the airy function which arises from the linear dependence of the gradual decrease of the resources .
the other quantity of importance is the ratio of the death rate outside the patch to the growth rate inside , specifically @xmath31 .
this second parameter can also be taken to be the combination @xmath76 .
we have recovered the skellam result eq .
( [ dirichlet ] ) and given its generalization in eq .
( [ skelgen ] ) .
we have also recovered the ludwig result and provided its generalization as well .
a. l. lin , b. mann , g. torres , b. lincoln , j. kas , and h. l. swinney , biophys .
j. * 87 * , 75 ( 2004 ) ; see also a. l. lin in _ modern challenges in statistical mechanics : patterns , growth , and the interplay of nonlinearity and complexity , _ eds . v. m. kenkre and k. lindenberg , aip , new york ( 2003 )
. n. perry , j. r. soc .
interface * 2 * , 379 ( 2005 ) .
k. a. dahmen , d. r. nelson , n. m. shnerb , j. math .
biol . * 41 * , 1 ( 2000 ) .
n. m. shnerb , phys .
e * 63 * , 011906 ( 2000 ) . v. m. kenkre and m. n. kuperman , phys .
e * 67 * , 051921 ( 2003 ) .
v.m . kenkre and niraj kumar , _ nonlinearity in bacterial population dynamics : proposal for experiments for the observation of abrupt transitions in patches _ , proc .
, accepted for publication
. t. l.yates , j. n. mills , c. a. parmenter , t. g. ksiazek , r. r. parmenter , j. r. vande castle , c. h. calisher , s. t. nichol , k. d. abbott , j. c. young , m. l. morrison , b. j. beaty , j. l. dunnum , r. j. baker , j. salazar - bravo , and c. j. peters .
bioscience * 52 * , 989 - 998 ( 2002 ) .
g. abramson and v. m. kenkre , phys .
e * 66 * , 011912 ( 2002 ) . v. m. kenkre , l. giuggioli , g. abramson , and g. camelo - neto , europhysics j. b * 55 * , 461 ( 2007 ) . c. escudero , j. buceta , f. j. de la rubia , k. lindenberg , phys .
e * 69 * , 021908 ( 2004 ) .
j. d. murray , _ mathematical biology _ , 2nd edn , new york , springer(1993 ) .
m. kot , _ elements of mathematical ecology _ ( cambridge university press , cambridge , u. k. , 2003 ) .
j. g. skellam , biometrika * 38 * , 196 ( 1951 ) .
d. ludwig , d. g. aronson , h. f. weinberger , j. math . biol . * 8 * , 217 ( 1979 ) .
r. a. fisher , ann .
eugen , london * 7 * , 355 - 369(1937 ) .
m. abramowitz and i. a. stegun , handbook of mathematical functions ( dover , n. y. , 1972 ) .
o. vallee and m. soares , _ airy functions and applications to physics _
( imperial college press , london , u. k.,2004 ) | we study bifurcations in a spatially extended nonlinear system representing population dynamics with the help of analytic calculations based on the time - independent schrdinger equation for a quantum particle subjected to a uniform gravitational field . despite the linear character of the schrdinger equation
, the result we obtain helps in the understanding of the onset of abrupt transitions leading to extinction of biological populations .
the result is expressed in terms of airy functions and sheds light on the behavior of bacteria in a petri dish as well as of large animals such as rodents moving over a landscape . |
parkinson has studied socio - dynamical systems with special focus on bureaucratic and administrative bodies . in his famous book @xcite
he comprises 10 essays which give detailed descriptions of microscopic mechanisms leading to macroscopic phenomena like exponential growth of administrations , impact on the size of executive councils on efficient decision - making , and the relationship between the age of retirement and a company s vital promotion scheme .
these essays are not quantitative , however , the description of these microscopic mechanisms is so clear and brilliant , that they can be easily cast into dynamical models .
we pick three of these essays which are especially well suited for this purpose . in the first essay ( chapter 4 in @xcite ) , parkinson conjectured on historical evidence that there is a characteristic group size beyond which the ability of this group for efficient decision - making , i.e. the finding of consensus , is considerably diminished .
the empirical findings based on government size , found in the 1950s @xcite , are still valid today to a large extend @xcite . by proposing an opinion formation model for small groups
we provide an understanding for this hypothesis through the existence of finite size effects which lead to this characteristic size ( coefficient of inefficiency ) of decision - making bodies . in the second essay ( chapter 1 in @xcite )
, we formalize the so - called parkinson s law. we propose dynamical equations for his model of bureaucratic growth and solve them explicitly .
as the main result we derive the phase diagram of this model for the macroscopic evolution of an administrative body .
one phase corresponds to an exponentially growing body , the other to a shrinking one .
the third essay ( chapter 10 in @xcite ) , studies the interrelation between the age of retirement and promotion schemes with the motif force to ensure the presence of career opportunities and prevent the lack thereof , a phenomenon also known as prince charles syndrome. there we show the existence of a unique maximum for the efficiency of a bureaucratic body at the optimum life - time of a worker in the system .
given a promotion scheme and the hierarchy in the system , the optimal age of retirement can be computed .
this paper is organized as follows : in section ii we review work on the relationship between the size of a decision - making body and its efficiency @xcite . in section iii
we propose and solve the dynamical equations for a model of an administrative body which evolves according to parkinson s law .
the question when an official within such a body should retire to ensure a maximum of overall efficiency of the administrative body is left for the final section iv .
we interpret our results and conclude in section v.
parkinson discovered what he called the life cycle of a cabinet .
he studied the membership of britain s highest executive council , the cabinet , spanning 700 years , from 1257 to 1955 .
he found that an initially small council steadily increased in size , until growing beyond a size of 20 , shortly after which it was superseded by a new council with the same fate .
the british cabinet went five times through this life cycle within the aforementioned timespan .
he also lists the cabinet sizes of his time revealing that cabinets beyond a size of 21 are only found in communistic countries .
he concludes that it is critical for cabinets to have memberships below this characteristic size ( 21 ) for efficient decision - making within this council ( the coefficient of inefficiency ) .
his explanation for this phenomenon was that as group size increases the influence of individual members decreases ( not only because there are simply more of them , but also because the group is more likely to dissociate into separate subgroups ) .
the less influence a member bears , the more easily new members are admitted to the council which in turn decreases their influence further . in the following
we ask if similar observations can be made with today s data on cabinets .
we then propose a model for opinion formation in small groups whose dependence on the group - size nicely resembles the transition conjectured by parkinson .
versus ( a ) the human development indicator and ( b ) and political stability . the @xmath0values indicate the high significance of negative correlation between them .
we computed the average indicator value for each @xmath1 and binned them with intervals of two ; errors were calculated by gaussian error propagation .
the green line is a linear least squares fit which intersects the median of the respective distributions at around 20.,title="fig:",height=219 ] versus ( a ) the human development indicator and ( b ) and political stability .
the @xmath0values indicate the high significance of negative correlation between them .
we computed the average indicator value for each @xmath1 and binned them with intervals of two ; errors were calculated by gaussian error propagation .
the green line is a linear least squares fit which intersects the median of the respective distributions at around 20.,title="fig:",height=219 ] we compiled a database of cabinet sizes @xmath1 , for 197 countries and self - governing territories with data drawn from the cia @xcite .
the results reported in @xcite showed that the majority of cabinet sizes falls in the range between 13 and 20 . to compare cabinet size with its efficiency we made use of indicators provided by the united nations development programme @xcite and the indicator political stability ( ps ) provided by the world bank @xcite .
we processed the data by computing the average of the indicator values for each @xmath1 and tested against the hypothesis that there exists no correlation between them , see fig .
[ indplots ] .
we find negative correlations between the cabinet size and the human development indicator ( hdi ) , see fig .
[ indplots](a ) , and between cabinet size and a series of governance indicators from @xcite revealing that larger cabinets coincide with a more unstable political climate , fig .
[ indplots](b ) .
a linear fit to the data reveals that the value of @xmath1 separating above - average ranking countries from below - average ranking ones is found around 20 for _ each _ indicator . for more details
see @xcite . ) .
each individual is a cell with binary opinion ( black or white ) . for each value of @xmath1
we show 4 independent update - runs , all starting with the same initial configurations and the same network . in each run
we show the first four time steps 0,1,2,3 , where time 0 corresponds to the initial configuration .
line 1 is obtained by the iterative application of the opinion formation protocol on the initial configuration in a random sequence of updates . the subsequent lines ( 2,3 )
are obtained analogously .
the actual sequence is seen to play an important role for the final state at time 3 .
we show two different trajectories leading to consensus ( everything white or black at time 3 ) or dissensus ( mixed colors , i.e. opposing factions ) for each size @xmath1.,height=188 ] decision - making in small groups can be cast into a dynamical opinion formation model @xcite .
each member of the committee is represented as a node in a network and holds a binary opinion , say 0/1 .
each node has a connectivity @xmath2 which is the number of undirected links to other nodes , representing a social influence ( interactions , such as discussions ) two agents exert upon each other @xcite .
each node shares @xmath2 undirected links with other nodes .
thus for @xmath3 the graph is not fully connected and nodes appear which are not directly linked .
we assume the total network to be connected , i.e. there exist no disjoint subnetworks . according to parkinson groups such as cabinets
are typically highly clustered .
therefore a sensitive choice for the fixed network architecture of our model is a small - world network , which has already been shown to be of paramount importance in modelling social interactions @xcite .
each node is connected with its @xmath2 nearest neighbors , then with probability @xmath4 each link is randomly rewired . as a dynamical rule governing the interactions between connected agents we chose a majority rule @xcite with a predefined threshold
@xmath5 $ ] , see @xcite .
when the update is carried out random sequentially , i.e. within one iteration each node is updated one after the other in a random order , this yields exactly the opinion formation model studied in @xcite . here
a node adopts the state of the majority of its @xmath2 neighbors only if this majority lies above @xmath6 , otherwise the node keeps its previous internal state . to anticipate the evolution of a system specified by the above characteristics , consider fig .
[ traj ] , where four different runs for three choices of cabinet size @xmath1 are shown .
we used parameter settings @xmath7 .
the system is initiated with a random configuration of internal states ( lowest rows , labelled 0 ) , each cell represents one agent in the network with a color corresponding to his binary opinion .
the above rows are derived after updating each node once in a random sequential order .
it becomes apparent that this model favors the forming of clusters , i.e. neighboring nodes tend to share the same opinion coalitions emerge .
note that for each of the four runs we start with the same initial conditions , thus the difference of the final states is solely due to the random sequential update .
further we see that the final states ( rows number 3 ) can be characterized whether they consist of agents with the same opinion ( consensus ) or whether there are at least two factions of opposite opinions ( dissensus ) . for four different settings of the model parameters @xmath8 . setting @xmath9 @xmath10
is shown by the blue squares .
for @xmath11 we obtain always consensus here , except the @xmath12 case .
we then find two regimes approximated by a linear fit and separated by a shaded area , where we conjecture the coefficient of inefficiency to lie .
decreasing the connectivity of the group`s network , parameter setting @xmath13 , red left triangles , shifts the position of the shaded transition area and increases the tendency toward dissensus .
adjusting the threshold such that we recover the pure majority rule , @xmath14 , green right triangles , has no impact on @xmath10 due to our choice of @xmath2 , except that we do not find the charles i ' scenario in this case .
lowering @xmath4 and therefore increasing the spatial correlations in the network hardens the finding of consensus too , as can be seen from the settings @xmath15 ( magenta diamonds).,height=226 ] the question whether efficient decision - making in the group is possible or not translates into the question how likely the group evolves into final state of consensus or dissensus .
we define dissensus @xmath10 as the order parameter of the model : let @xmath16 be the initial population of nodes with internal state 0 and @xmath17 the final population in this state .
we then have @xmath18 with @xmath19 being the heaviside step function and @xmath20 the average over all possible initial conditions .
@xmath16 is drawn with uniform probability from @xmath21 .
accordingly , the opinions are randomly assigned to the individual nodes .
@xmath10 gives the expectation value of a final state without consensus and measures the group s proneness to end up in dispute .
we show @xmath10 for four different parameter settings in fig .
[ dn ] . for each curve
we find three distinct regimes . in the region
where the network is fully connected we always find consensus ( except the notable @xmath12 case , to which we will return later ) . then we find two regions of approximately linear growth . between these , the shaded areas in the plot , the slope considerably decreases .
we further see that for lower values of @xmath2 , @xmath10 increases ( because of the emergence of disconnected clusters ) and for less shortcuts , i.e. lower values of @xmath4 and therefore higher spatial correlations @xmath10 also increases .
both mechanisms , lower @xmath2 and @xmath4 , favor the emergence of clusters . the decrease of the slope after @xmath1 has passed the sizes of the shaded areas can be naturally understood as a finite size effect present for small groups .
this effect is constituted by the considerably increased likelihood to form coalitions in this model , is in one - to - one correspondence to the diminishing influence of the assembly s members .
since the gap between two values for @xmath10 can be understood as the resistance to the cabinet s equivalent expansion , the dynamics proposed by parkinson and the existence of a characteristic size for its functioning , the coefficient of inefficiency , are both reproduced by the behavior of @xmath10 in fig .
[ dn ] .
parkinson has noted that there were no cabinets found with eight members .
this is still the case today .
our findings show that there is good reason to avoid this number , at least for certain parameter values , see fig .
historically there exists a famous exception : eight was the number preferred by king charles i for his committee of state . and look what happened to him!.
the famous sentence work expands as to fill the resources available for its completion was a guiding principle for parkinson when he demonstrated that the growth of the administrative staff of the british navy stood in no relation to the work that actually had to be administered .
for example , in the peacetime periods between 1935 and 1954 ( i.e. excluding the even stronger growing staff during world war ii ) the staff of the colonial office steadily increased by an average percentage of 6% a year , from 372 to 1661 , while the colonial territories in the same timespan shrunk dramatically .
the same figure of 6% turns out if one traces the growth of the dockyard staff of the british navy between 1914 and 1928 . during this timespan
the admiralty officials increased by around 80% , while the ships in commission decreased from 62 to 20 .
parkinson suggested to explain this growth by the principle that an official tries to maximize subordinates , not rivals .
this is nowadays soetimes referred to as parkinson s law .
that is , if an official gets promoted ( happening with probability @xmath22 ) this is in real terms carried out by allocating @xmath23 subordinates to him .
we now present a model to understand parkinson s intuition for administrative growth .
we derive the growth rate @xmath24 explicitly , depending on the model parameters @xmath23 and @xmath22 .
further we introduce a drop - out rate @xmath25 , the probability for an official to quit his job , as well as the age of retirement @xmath26 , which gives the typical administrative life - span .
consider @xmath25 and @xmath26 to be fixed and measurable .
the question to determine @xmath26 is left to the next section .
let us denote the number of promoted staff that has served for @xmath27 years at time @xmath28 , @xmath29 and the number of subordinate staff @xmath30 , analogously .
let the age of retirement be @xmath26 , the time between entry and end of service .
we then have for the number of total promoted / subordinate staff @xmath31 .
further , let @xmath32 be the probability to get promoted after @xmath27 years of service , @xmath23 the average number of subordinates one gets appointed if promoted , and @xmath25 the drop - out rate .
we then arrive at the update equations @xmath33 assume that no official enters at a promoted level , @xmath34 . by introducing @xmath22 as the average probability to get promoted
, @xmath35 , eq . [ d3 ] takes the form @xmath36 combining this with eq .
[ d2 ] in the definition of @xmath37 we obtain the update equation for the total number of subordinate staff , @xmath38 \nonumber \\ & = & \left(1-\gamma+\left(r-1\right)p\right ) \left(1-q_-\right ) n^-\left ( t \right ) \quad , \label{n_min}\end{aligned}\ ] ] where we define @xmath39 and assume that the growth of the system is stationary . using eq .
[ d1 ] in the definition for @xmath40 we get the update equation for the total number of promoted staff as , @xmath41+p \left[n^-\left(t\right)-n_{\tau_r}^-\left(t\right)\right ] \nonumber \\ & = & \left(1-\gamma\right ) \left(1-q_+\right ) n^+\left(t\right ) + p \left(1-q_-\right ) n^-\left(t\right ) \quad , \label{n_plu}\end{aligned}\ ] ] where @xmath42 .
we define @xmath43 , @xmath44 and @xmath45 to rewrite eqs .
[ n_min ] , [ n_plu ] in matrix form @xmath46 to obtain @xmath47 let us rewrite eq .
[ n_min ] as @xmath48 . from eq .
[ d2 ] we find that @xmath49 .
here we can plug in eq .
[ d3 ] , iterate @xmath50 forward for @xmath51 times and divide by an overall @xmath37 to obtain a self - consistent equation for @xmath47 , @xmath52^{\tau_r+1 } } \quad .
\label{qmin}\ ] ] for @xmath53 we start with eq .
[ d1 ] and iterate it @xmath26 times backward . making use of eqs .
[ d2 ] and [ d3 ] and using @xmath34 we get @xmath54 applying eq .
[ d2 ] for @xmath27 times we obtain a sum which can be written as an incomplete geometric series leading to @xmath55 combining eq . [ d3 ] and eq .
[ n_min ] we get the helpful relation @xmath56 plug this in eq .
[ qpl02 ] and divide both sides by @xmath57 to finally obtain @xmath58}{\left[1-\gamma+\left(r-1\right)p\right]^{\tau_r+1 } \left(1-q_-\right)^{\tau_r } } \quad , \label{qpl}\ ] ] where @xmath59 is the ratio of promoted to subordinate staff @xmath60 .
let @xmath24 be the growth rate of the entire staff @xmath61 \propto \mathrm e^{\lambda t}$ ] .
parkinson s law then takes the form of the eigenvalue problem of a triangular matrix @xmath62 we find the eigenvalues immediately in the diagonal .
this system allows one meaningful solution for @xmath24 and @xmath59 , namely @xmath63 and @xmath64 . for the other solution , @xmath65
, @xmath59 diverges . with @xmath66 .
the solid line ( given by @xmath67 ) separates a regime of a exponential body ( white area ) from shrinking growth ( grey area).,height=226 ] we are finally in the position to answer the century , if not millennium old question how a system of bureaucrats can be organized such that it does _ not _ grow exponentially . the phase diagram in fig .
[ lambda ] shows the two phases , one of exponential growth , one of shrinking , for administrative bodies evolving according to parkinson s law .
the line separating these two phases ( given by @xmath67 ) shows how the parameters @xmath23 ( number of subordinates ) and @xmath22 ( average promotion probability ) should be adjusted to keep the size of the bureaucracy constant ( drop - out rate was set to @xmath66 ) .
to fully understand parkinsonian growth of administrations we now determine the age of retirement @xmath26 .
parkinson suggests that the optimal value for @xmath26 has to be found on the basis of an administrative body s promotion scheme .
the age of retirement thus does not depend on the actual age of the official whose retirement we are considering , but on his possible successors which should be given the possibility to advance in their careers .
if this is not ensured a mechanism nowadays called prince charles syndrome sets in , i.e. the official gets frustrated by a lack of promotion opportunities , and individual efficiency decreases .
.stages of efficiency of an official .
he enters at stage 1 . after the respective number of years in service ( number in brackets next to parkinson s coding ) he advances to the next stage .
if he does not reach a promoted position by reaching stage 6 , he will enter the career described by the right column .
we also show the values of the efficiency for official @xmath68 , @xmath69 , we assign to each stage . + [
cols=">,<,^,<,^ " , ] [ eff ] we assume the hierarchical structure of the administrative body to be built according to the motif force that officials try to maximize subordinates , not rivals , i.e. each official seeks to achieve a number @xmath23 of subordinates .
one can picture this structure as a pyramid consisting of horizontal layers of officials holding an equally high office . on the hierarchical level above
there are then @xmath70 times less , on the level below @xmath23 times more officials .
let us label the levels in the pyramid by an index @xmath71 , then there are @xmath72 officials at level @xmath71 .
if we fix the number of levels to @xmath73 we have @xmath74 and @xmath75 .
we assign to each official @xmath68 an efficiency @xmath69 which depends on the years he has served and his current position .
we use the career stages given as in tab .
[ eff ] , and assign an @xmath69 to each stage , increasing linearly .
these career stages are used as a tribute as in @xcite , but one could equally well work with other efficiency curves .
the only important feature is the existence of a bifurcation point reflecting the age where the prince charles syndrome sets in , here this is stage 6 . at stage 1
we have @xmath76 .
for each subsequent stage ( left column ) it increases by @xmath77 .
if after reaching stage 6 the official @xmath68 reached a promoted position his efficiency proliferates . if at stage 6 we still find him in a subordinate position , his efficiency will decline by @xmath77 in each of the following stages ( right column ) .
we introduce the parameter @xmath78 which is the lowest level which is counted as promoted .
the subordinate staff is then present on the levels @xmath79 .
the number of subordinate levels is called @xmath80 . .
for three values of @xmath81 ( @xmath82 , green slash - dotted , @xmath83 , red slashed , @xmath84 , blue solid line ) we show results for systems of sizes @xmath85 ( curves from bottom to top ) . with increasing system size
the efficiency becomes size - independent .
there is a unique maximum which position depends on @xmath81 for @xmath26.,height=226 ] we numerically study this system in the following way .
it is initialized with randomly distributed ages ( and therefore efficiencies ) and @xmath86 .
one time step in the simulation corresponds to one year . at each iteration
each official ages by one year and his efficiency is updated according to tab .
if he reaches the age @xmath26 he retires and his position becomes vacant .
one of his two subordinates is chosen randomly to replace him , leaving a vacant position one level below , triggering a cascade of promotions . when the lowest hierarchical level @xmath73 is reached the vacant position
is taken by a new official who enters the system at stage 1 .
this process is iterated until the average efficiency in the system
@xmath87 , the mean of @xmath69 over all officials @xmath88 approaches a stationary value .
simulation results are shown in fig .
we vary @xmath26 and compute the average efficiency @xmath87 for systems of sizes @xmath89 and @xmath901 ( green slash - dotted lines ) , 2 ( red slashed lines ) , 3 ( blue solid lines ) .
for increasing @xmath73 the system approaches a stationary state where the efficiency only depends on @xmath81 .
accordingly we find three bundles according to @xmath91 with curves for different values of @xmath73 discernible .
curves of same color and style correspond to increasing @xmath73 from bottom to top . for each bundle
we find a unique maximum indicating the optimal choice for @xmath26 . for values of @xmath26 below this
maximum officials retire too early to reach higher efficiencies , above this point the prince charles syndrome becomes a danger .
we further see that the flatter the hierarchy , i.e. higher @xmath81 , the lower the optimal age of retirement .
recall that we defined the ratio between promoted and subordinate staff as @xmath92 .
by re - labelling the summation index we can write this as @xmath93 which for large @xmath73 and @xmath94 converges to @xmath95 for large system sizes the ratio @xmath59 indeed approaches a stationary value .
so @xmath81 determines @xmath59 and the best choice for @xmath26 , which in turn can be used to compute the growth rate @xmath24 from the previous model .
we quantified three famous , descriptive essays of c.n .
parkinson on bureaucratic inefficiency in a dynamical socio - physical framework . in the first model we showed how
the use of recent opinion formation models for small groups could be used to understand parkinson s observation which we showed is still valid in modern data that decision making bodies such as cabinets or boards become highly inefficient once their size exceeds a critical coefficient of inefficiency. we showed how this characteristic size arises due to finite - size effects in the process of forming of coalitions in small systems .
parkinson s law states that the growth of bureaucratic or administrative bodies usually goes hand in hand with a drastic decrease of its overall efficiency . in our second model
we pictured a bureaucratic body as a system of a flow of workers through an administrative body .
officials enter , become promoted to various internal levels within the system over time , and leave the system after having served for a certain time , they retire . within the proposed model we showed how to compute the phase diagram under which conditions bureaucratic growth can be confined .
we thereby link parkinson s microscopic interaction rules to macroscopic properties observable in administration and bureaucracy .
it is possible to adjust the microscopic model parameters which are altogether observable and measurable , the individual promotion probability and number of subordinates , to control the growth rate of the bureaucratic body . in our last model
we assign individual efficiency curves to workers throughout their life in administration , and compute the optimum time to send them into retirement , in order to ensure a maximum of efficiency within the body , parkinson s pension point. we implement parkinson s observation that an individual s efficiency declines if he resides in a subordinate position for too long , i.e. the individual efficiency curves over time are characterized by the existence of bifurcation points . this effect necessitates the need of promotion opportunities , which can be ensured by two mechanisms here . on
the one hand one can lower the age of retirement allowing young workers to advance in their career faster , or on the other hand one can increase the levels of internal hierarchy . | we formulate three famous , descriptive essays of c.n .
parkinson on bureaucratic inefficiency in a quantifiable and dynamical socio - physical framework . in the first model
we show how the use of recent opinion formation models for small groups can be used to understand parkinson s observation that decision making bodies such as cabinets or boards become highly inefficient once their size exceeds a critical coefficient of inefficiency , typically around 20 .
a second observation of parkinson which is sometimes referred to as parkinson s law is that the growth of bureaucratic or administrative bodies usually goes hand in hand with a drastic decrease of its overall efficiency . in our second model
we view a bureaucratic body as a system of a flow of workers , which enter , become promoted to various internal levels within the system over time , and leave the system after having served for a certain time .
promotion usually is associated with an increase of subordinates . within the proposed model it becomes possible to work out the phase diagram under which conditions bureaucratic growth can be confined . in our last model
we assign individual efficiency curves to workers throughout their life in administration , and compute the optimum time to send them to old age pension , in order to ensure a maximum of efficiency within the body in parkinson s words we compute the pension point . |
over the course of the last decade , work by a number of groups has led to the development of a widely accepted picture for the formation of the first stars , the so - called population iii ( or pop .
iii ) stars . in this picture , the very first stars ( population iii.1 in the nomenclature of @xcite and @xcite ) form within small dark matter halos that have total masses @xmath0 , virial temperatures of around a thousand k , and that are assembled at redshifts @xmath130 or above @xcite .
gas falling into one of these small dark matter halos is shock - heated to a temperature close to the virial temperature of the halo , and thereafter cools via h@xmath2 rotational and vibrational line emission .
the h@xmath2 that enables the gas to cool is primarily formed via the gas - phase reactions @xmath3 where the required free electrons are those that remain in the gas after cosmological recombination at @xmath4 .
the low abundance of these electrons , plus the fact that the gas starts to recombine further once it collapses into the dark matter halo , limits the amount of h@xmath2 that can form in this manner ; the typical fractional abundances of h@xmath2 that result are a few times @xmath5 .
the limited h@xmath2 abundance , the low critical density of the h@xmath2 molecule , and the large energy separation of its lowest accessible rotational levels combine to significantly limit the extent to which the gas can cool .
the minimum temperature reached depends upon the dynamical history of the collapse , but is typically around @xmath6 .
this minimum temperature is reached at a density of around @xmath7 , comparable to the critical density of h@xmath2 ( the density at which its level populations reach their local thermodynamical equilibrium ( lte ) values ) , and at higher densities , the gas begins to reheat .
gas falling into the dark matter halo therefore tends to fragment once it reaches this temperature and density , and the resulting fragments have masses of the order of the local jeans mass , @xmath8 . following this period of fragmentation , the gas in the fragments undergoes a further period of gravitational collapse , eventually reaching a density of @xmath9@xmath10 , at which point three - body h@xmath2 formation becomes effective .
this rapidly converts almost all of the available atomic hydrogen into h@xmath2 , but the significant energy release that accompanies this process heats the gas , which typically attains a temperature of 10002000 k during this phase of the collapse
. continued collapse next leads to the gas becoming optically thick in the main h@xmath2 cooling lines ( at @xmath11 ; @xcite , @xcite ) , the onset of collision - induced emission cooling ( at @xmath12 ; @xcite ) , and finally to the gas becoming optically thick in the continuum as well as in the h@xmath2 lines ( @xmath13 ; @xcite ) .
dissociation cooling allows for further collapse , but once all of the h@xmath2 is gone , the collapse becomes fully adiabatic .
detailed simulations of the collapse , using adaptive mesh refinement ( amr ; see e.g. @xcite ) or smoothed particle hydrodynamics ( sph ; see e.g. @xcite ) show little or no evidence for fragmentation between @xmath14 and the onset of this adiabatic evolution , and it is natural to associate the latter with the formation of a primordial protostar .
the amount of gas incorporated into the protostar at this point is small , approximately @xmath15 , but it is surrounded by a massive , dense envelope and hence begins to accrete rapidly . if the envelope does not fragment , and if accretion onto the protostar is unhindered by radiative feedback , then the final mass of the star can be very large .
for example , @xcite estimate that its mass could lie anywhere in the range @xmath16@xmath17 , depending on the environment and dynamical history of the gas .
the degree to which feedback will suppress accretion remains uncertain , but the most effective potential feedback mechanisms , lyman-@xmath18 scattering and photoionization @xcite become important only once the mass of the star exceeds 20 m@xmath19 .
the expectation is therefore that all population iii.1 stars were massive , with masses typically of the order of @xmath20 or more , thereby also explaining why none of these stars appear to have survived until the present day .
if a population iii.1 star of this mass does form , then it will rapidly dissociate h@xmath2 throughout the halo , thereby suppressing further star formation @xcite .
this fact , together with the lack of fragmentation seen in the simulations , is often taken to imply that only a single population iii.1 star will form in each dark matter halo .
this widely accepted picture for population iii star formation also provides for a second mode of pop .
iii star formation .
this occurs in metal - free gas which has been ionized by radiation from a previous generation of population iii stars . following the death of these stars , the gas recombines , and the elevated fractional ionization in the recombining gas allows more h@xmath2 to form .
the enhanced h@xmath2 fraction enables the gas to cool to a lower temperature , which in turn increases the effectiveness of cooling by the singly - deuterated hydrogen molecule , hd . in the standard pop .
iii.1 scenario , hd cooling is of limited importance , but in this second scenario , termed population iii.2 star formation by @xcite , it becomes dominant , cooling the gas down to the temperature of the cosmic microwave background @xcite .
the lower gas temperature , plus the higher critical density of hd ( @xmath21 , compared with @xmath22 for h@xmath2 ) , means that fragmentation occurs somewhat later , and produces significantly lower mass fragments , with characteristic masses @xmath23 .
subsequently , the evolution of these fragments is believed to proceed in the same fashion as described above , resulting ultimately in the formation of a primordial protostar of similar mass . however , since this protostar is embedded in a less massive envelope , with a lower accretion rate , the final stellar mass is thought to be an order of magnitude or so smaller , @xmath24 .
nevertheless , this still corresponds to what we would call , by galactic standards , a massive star , and moreover one which is more than capable of dissociating h@xmath2 throughout a large volume of the halo , thereby suppressing further star formation . in the past few years
, it has become possible to partially test this picture using numerical simulations that begin with the proper cosmological initial conditions and follow the collapse of the gas all the way to protostellar densities .
these simulations confirm that fragmentation during the initial collapse phase is ineffective : in general , only a single protostar forms , although in roughly 20% of cases , the gas fragments into two clumps , forming a wide binary ( @xcite ; m. turk , private communication ) .
however , most of these studies were prevented by technical limitations from following the evolution of the infalling gas after the formation of the first protostar .
the technical problem involves the hydrodynamical timestep : as the collapse is followed down to protostellar densities , this becomes extremely short ( @xmath25 ) , making it computationally infeasible to follow the evolution of the surrounding dense gas over any significant time period . in studies of contemporary star formation
, a similar problem is avoided by the use of sink particles @xcite . when gravitationally bound regions of gas collapse below the scale on which they can be spatially resolved , they are replaced in the simulation by a sink particle of the same mass .
this can accrete gas from its surroundings , but otherwise interacts only via gravity .
recent simulations of population iii star formation that have used sink particles to allow the evolution of the gas to be followed beyond the formation of the first protostar consistently find evidence for more extensive fragmentation than is assumed in the conventional picture of population iii star formation ( @xcite ; clark et al .
, in prep . ) .
if fragmentation is as common as these results suggest , and most population iii stars form in multiple systems rather than as single stars , then this has profound implications for the final masses of the stars , their production of ionizing photons and metals , the rate of high redshift gamma ray bursts , and many other issues .
it is therefore important to better understand the physical basis of fragmentation in these systems . to do this
, we would ideally like to have a large sample of simulations , as basing our arguments on only one or two realizations of population iii star formation leaves one open to the possibility that these realizations may not be typical .
unfortunately , simulations starting from cosmological initial conditions and following collapse all the way to protostellar densities are computationally costly , making it difficult to explore a large region of parameter space . in these simulations , it is also difficult to be certain about which aspects of the included physics are the most important for driving fragmentation .
simulations that start from simpler initial conditions and that focus on exploring the importance of a single free parameter therefore play an important role , which is complementary to that of fully cosmological simulations .
a good example is the recent work by machida and collaborators @xcite which examined the influence of the initial rotational energy of the gas , and found that zero metallicity clouds with sufficient initial rotational energy could fragment into tight binaries .
another example , and one which explores a much lower density regime , is the study by @xcite , who showed that the fragmentation behavior depends sensitively on the adopted density profile of the primordial halo . in this paper , we perform a similar study into the effects of the initial turbulent energy of the gas .
our motivation comes from the fact that recent high - resolution cosmological simulations have shown that the self - gravitating regions in which pop .
iii star formation occurs contain significant turbulent motions ( see e.g. * ? ? ?
* ) . from studies of present - day star formation
, we know that such turbulent motions can lead to the fragmentation of initially marginally unstable gas clouds , and so it is instructive to investigate how such conditions could affect the ` standard ' picture of primordial star formation . in a pop .
iii analogue of the studies on present - day star formation , we will look at how subsonic turbulence affects the collapse of marginally super - critical bonnor - ebert spheres , in an attempt to quantify the effects of turbulently induced fragmentation on the mass function of pop .
iii stars .
central to this study will be the use of the sink particles to allow us to capture the evolution of the gas cloud beyond the collapse of the first region .
the paper is laid out in the following manner .
our modifications to the cosmological sph code gadget 2 @xcite are outlined in [ sec : gadget2 ] ( with further details of the chemical networks and heating and cooling described in the appendix ) .
the initial conditions for this numerical experiment are described in
[ sec : ics ] , including the definitions of pop .
iii.1 and iii.2 that are used in our study .
the details of the fragmentation process seen in the simulations are discussed in [ sec : frag ] and the long - term chemical and thermodynamical evolution of the infalling envelope are described in [ sec : evol ] .
we discuss the implications of this study for the initial mass function ( imf ) of primordial stars in [ sec : chat ] , and summarize the main points of this paper in [ sec : sum ] .
we model the evolution of the gas in our simulations using a modified version of the gadget 2 smoothed particle hydrodynamics code @xcite .
we have modified the publicly available code in several respects .
first , we have added a sink particle implementation , based on the prescription in @xcite , to allow us to follow the evolution of the gas beyond the point at which the first protostar forms .
our particular implementation is derived from the one first described in @xcite .
we briefly describe the ideas behind the algorithm here .
the actual numerical values used for the parameters discussed here are given later , in
[ sec : ics ] .
sink particles are not really added to the code in the sense that a new particle is introduced , but instead a normal sph particle is turned into a sink particle once certain criteria have been met .
the particle undergoes a series of tests once it has reached a threshold density .
the first is to see whether the candidate sph particle is sufficiently far away from any other sinks , measured in terms of the sink particle s accretion radius , @xmath26 .
we adopt a conservative value of @xmath27 .
next , we check to see whether the smoothing length of the particle is less than the ` accretion radius ' of the sink particle that it will become .
this ensures that when the sink particle forms it can instantly accrete at least @xmath28 neighboring particles ( we adopt 50 neighbors in these simulations ) .
the third test is to make sure that the candidate sink particle and its neighbors are on the same integration time - step .
once these three preliminary criteria are met , the dynamical state of the possible sink particle and its neighbors are assessed to ensure that the particles are indeed undergoing gravitational collapse and are not about to re - expand from their dense state .
this takes the form of a further four tests .
firstly , we require that @xmath29 where @xmath18 is the ratio of the thermal energy to the magnitude of the gravitational energy of the particles .
secondly , we ensure that @xmath30 where @xmath31 is the ratio of rotational energy to the magnitude of the gravitational energy .
the third condition is that the total energy of the particles must be negative ( which actually renders the above checks redundant , but can help to improve the computational efficiency of the code ) .
finally , the forth test is that the divergence of the accelerations must be less than zero .
this final check ensures that the group of particles is not in the process of being tidally disrupted or bouncing .
if all of these tests are passed , the particle becomes a sink and the mass and linear momentum of the neighbors are added to the sink particle .
as the simulation progresses , the sink particles are then allowed to accrete other gas particles that fall within the accretion radius .
as in the bate et al .
( 1995 ) prescription , several tests must be passed before any sph particles can be accreted by the sink .
first , it must obviously be bound and moving toward the candidate sink .
second , in the case where there is more than one sink present , it must be more bound to the candidate sink than any other sink in the simulation .
finally , the sph and sink particles need to be on the same integration time - step ( to ensure temporal momentum conservation ) .
once these conditions are passed , the mass and linear momentum of the sph particle are added to the sink particle .
note that accretion onto existing sink particles is done _ before _ any new candidate sinks are considered .
in addition to the sink particles , we have also implemented an external pressure term ( e.g. * ? ? ?
* ) , that enables us to model a constant pressure boundary , as opposed to the vacuum or periodic boundary conditions that are the only choices available in the standard version of gadget .
we modify the standard gas pressure contribution to the gadget 2 momentum equation , @xmath32 , \ ] ] by replacing @xmath33 and @xmath34 with @xmath35 and @xmath36 respectively , where @xmath37 is the external pressure , and all quantities have the usual meaning , consistent with those used by @xcite .
the pair - wise nature of the force summation over the sph neighbors ensures that @xmath37 cancels for particles that are surrounded by other particles . at the edge , where the term does not disappear
, it mimics the pressure contribution from a surrounding medium .
the values of @xmath37 used in this study are @xmath38 for the pop iii.1 clouds , and @xmath39 for the pop iii.2 clouds , where @xmath40 is the boltzmann constant .
these confining pressures are similar to the internal pressure of the clouds in each case ( see the temperatures and densities in [ sec : ics ] below ) .
finally , we have added a treatment of primordial gas chemistry and cooling to the code . to model the thermal evolution of the gas
, we use an operator - split formalism , which treats the effects of radiative and chemical heating and cooling separately from compressional heating .
the influence of radiative and chemical heating and cooling on the thermal energy of each particle can thus be formulated as an ordinary differential equation ( ode ) , which can then be solved simultaneously with the chemical rate equations ( also odes ) that describe the chemical evolution of the gas .
as @xcite have previously discussed , the strong coupling between the chemical and thermal evolution of high density primordial gas that results from the importance of three - body h@xmath2 formation heating and h@xmath2 collisional dissociation cooling renders it vital to treat cooling and chemistry simultaneously ; treatments that do not do so will produce numerically stable results only with great difficulty .
full details of our chemical model and cooling function are given in the appendix .
the top panel shows the mass functions from those pop .
iii.1 simulations in which fragmentation occurs . in all cases the mass function is plotted at the point at which the total mass of gas converted to sink particles is 100 . note
that as accretion is ongoing , and the system is still young ( @xmath41 yr ) , these will often not be the final masses of the sinks .
the mass functions in the individual simulations differ substantially , although the combined mass function , shown in the bottom panel , exhibits a broad and flat distribution between masses of 0.4 and 20 .,height=240 ] the top panel shows the mass functions from those pop . iii.1 simulations in which fragmentation occurs . in all cases
the mass function is plotted at the point at which the total mass of gas converted to sink particles is 100 . note
that as accretion is ongoing , and the system is still young ( @xmath41 yr ) , these will often not be the final masses of the sinks .
the mass functions in the individual simulations differ substantially , although the combined mass function , shown in the bottom panel , exhibits a broad and flat distribution between masses of 0.4 and 20 .,height=240 ] temperature as a function of number density for the pop .
iii.1 ( dark blue ) and pop . iii.2 ( light blue ) @xmath42 simulations . in both cases
, the curves denote the state of the cloud at the point just before the formation of the sink particle . ,
height=288 ] accretion rates as a function of enclosed gas mass in the pop .
iii.1 ( upper lines ; blue ) and pop . iii.2 ( lower lines ; magenta ) simulations , estimated as described in section [ pop31 ] .
note that the sharp decline in the accretion rates for enclosed masses close to the initial cloud mass is an artifact of our problem setup ; we would not expect to see this in a realistic pop .
iii halo.,height=288 ] time taken for the first sink particle to form in each simulation . for reference
, the horizontal dashed line denotes the free - fall time of the clouds at their initial density . ]
( top ) mass functions for our two 1000 pop .
iii.2 clouds . as with the pop .
iii.1 data in fig .
[ fig : sinkmf ] , we show the mass function after 100 of gas has been accreted by the sink particles . due to the reduced fragmentation in these calculations ,
the sink particles are on average more massive than their pop .
iii.1 counterparts since they have less competition for the available mass .
( bottom ) the combined mass function for the two simulations.,height=240 ] ( top ) mass functions for our two 1000 pop .
iii.2 clouds . as with the pop .
iii.1 data in fig .
[ fig : sinkmf ] , we show the mass function after 100 of gas has been accreted by the sink particles . due to the reduced fragmentation in these calculations , the sink particles are on average more massive than their pop .
iii.1 counterparts since they have less competition for the available mass .
( bottom ) the combined mass function for the two simulations.,height=240 ]
since the aim of this study is to investigate the fragmentation properties of primordial gas , we use initial conditions that allow us to perform a controlled numerical experiment .
the clouds start as unstable bonnor - ebert ( be ) spheres into which we inject a subsonic turbulent velocity field .
the be sphere is made by allowing a gravitationally stable cloud of gas to evolve under its own self - gravity until the system has settled into a stable , centrally condensed configuration . during this initial settling phase , the gas temperature is held constant and the chemical evolution is not followed . on the scales of interest in this study ,
gravitational forces from the dark matter are negligible compared to the self - gravity of the gas , and so for simplicity , we do not include dark matter in our models . by re - scaling the mass and temperature of the cloud ,
we are then able to choose initial conditions that are gravitationally unstable , and which are similar to the initial conditions for the pop .
iii.1 and pop .
iii.2 star formation channels that occur in cosmological simulations .
all of the simulations presented in this paper start with a maximum central cloud number density of @xmath43 @xmath44 .
the pop .
iii.1 simulations start with an initial temperature of 300k and contain 1000 of gas . for the pop .
iii.2 simulations , we adopt an initial temperature of 75 k and examine two different initial masses .
the first set of pop .
iii.2 simulations contain 150 of gas , chosen such that they have the same ratio of thermal to gravitational energy as the pop .
iii.1 simulations , roughly 0.15 . as such , these pop .
iii.2 simulations contain the same initial number of jeans masses of gas as the pop .
iii.1 simulations . for a uniform sphere
, the number of jeans masses is given by ( @xmath45 , so these clouds have roughly 3 jeans masses in the initial configuration .
all else being equal , if these clouds were to evolve isothermally from this point on , they would have the same propensity to fragment as the pop .
iii.1 clouds , since this is directly related to the initial ratio of gravitational to thermal energy .
note that we have chosen a slightly subvirial configuration for our initial setup , to ensure that the clouds are still able to collapse when the turbulent motions are included .
in addition to these simulations , we also performed a second set of pop .
iii.2 simulations that start with the same gas mass as in the pop .
iii.1 case ( 1000 m@xmath19 ) . in this case , the pop .
iii.2 simulations are initially more jeans unstable than their pop . iii.1 counterparts , and therefore might be expected to fragment significantly more . in our pop .
iii.1 simulations , we set the initial fractional abundances of h@xmath2 , h@xmath46 , hd and d@xmath46 to @xmath47 , @xmath48 , @xmath49 and @xmath50 , respectively .
our values for @xmath51 , @xmath52 and @xmath53 are typical of the values found at these densities in cosmological simulations of pop .
iii.1 star formation ( see e.g. * ? ? ?
* ) , and account for the fact that the hd / h@xmath2 ratio is elevated over the cosmological d / h ratio of @xmath54 @xcite owing to the effects of chemical fractionation @xcite . in the case of d@xmath46 , fractionation is unimportant at our starting temperature , and so we simply set @xmath55 . in our pop .
iii.2 simulations , we adopt the same initial h@xmath46 and d@xmath46 abundances , but set @xmath56 and @xmath57 , following @xcite . in both the pop .
iii.1 and pop .
iii.2 simulations , we assumed that all of the helium remained neutral , and set the initial abundances of all of our other tracked species to zero .
within the be spheres , we impose a turbulent velocity field that has a power spectrum of @xmath58 .
we assume that the turbulence considered here has its origin in gravitationally driven flows that arise as the gas and dark matter virialize in mini - halos ( wise & abel 2007 ; greif et al .
2008 ; klessen & hennebelle 2010 ) .
as the gas is compressible in nature the turbulent velocity field will have a power spectrum that is somewhat steeper than the standard kolmogorov ( 1941 ) description for incompressible flows .
however , this deviation is small and we note that the the ability of a cloud to fragment is only weakly dependent on the power spectrum of the turbulence @xcite . the three - dimensional root - mean - squared velocity in the turbulent field which we will refer to as @xmath59 is then scaled to some fraction of the sound speed in the initial conditions . for the simulations presented here
we use four different rms velocities : 0.1 , 0.2 , 0.4 and 0.8 @xmath60 . for an isothermal sound speed and an adiabatic index of @xmath61 ,
the corresponding ratios of the turbulent to thermal energy are given by @xmath62 , yielding 0.0033 , 0.0133 , 0.0533 and 0.2133 respectively , for our chosen values of @xmath60 . in order to focus on the effects of the turbulence
, we do not include any ordered rotation of the initial gas cloud .
note , however , that this does not imply that the initial angular momentum of the cloud is zero , since the imposed turbulent velocity field gives the cloud a small amount of angular momentum .
note that we only consider subsonic turbulence in this study since our clouds have only a few jeans masses , and supersonic turbulence would unbind them . to study the effects of supersonic turbulence , one would have to look at clouds that are initially more jeans unstable than those we study here .
the clouds in our study are all modeled using 2,000,000 sph particles .
although this means that the mass resolution is higher in the 150 pop .
iii.2 clouds than in the other simulations , the jeans mass at the point where sink particles form is well resolved in every case ( see below ) . for this study
, it is more important that the turbulence in all simulations is evolved with the same resolution , hence our choice of a constant particle number throughout .
sink particles are created once the number density of the gas reaches @xmath63 @xmath44 , at which point the gas has a temperature of around 1200k .
the corresponding jeans mass at this density and temperature is 0.08 .
our mass resolution in the pop .
iii.1 and the 1000 pop .
iii.2 clouds is @xmath64 , where @xmath65 is the number of neighbors employed for force evaluations ( in our case 50 ) , and @xmath66 is the mass of an sph particle . the 150 pop .
iii.2 simulations have a mass resolution of 0.008 . once the candidate particle has passed the criteria described in bate et al .
( 1995 ) , it is replaced by a sink particle that can accrete gas particles that fall within its accretion radius @xmath26 , which we fix at 20 au . note that this radius is significantly larger than the jeans radius at the point that the gas reaches the density threshold for sink creation , which is around 6 au . as such , all the fragmentation that we capture in this study is well resolved and not lying close to the limits of our resolution ; if any ` artificial ' fragmentation were to occur , the sink particle would immediately swallow the offending region and replace it with a single accreting point .
we also prevent sink particles from forming within @xmath67 of one another .
this prevents the formation of sinks out of gas that in reality would by accreted by a neighboring sink particle before it could go into direct collapse by itself .
lastly , gravitational interactions between sinks , and between the sinks and the gas , are softened in the standard way in gadget 2 , using a fixed softening parameter of 5 au for the sinks , and a variable softening parameter for the gas particles that is equivalent to their smoothing length .
we note that the sink algorithm employed here does not allow sink particles to merge , in contrast to the ` sticky ' sinks introduced in bromm et al .
( 2002 ) . as such
, our simulations may be biased towards low masses .
small , secondary sinks could be driven towards more massive ones via dynamical friction and coalesce with them .
this effect , which acts to reduce the resulting number of fragments , has been seen in amr simulations of massive star forming regions in the present - day universe ( krumholz et al .
2007 , 2009 ) .
however , given the size of our sink particles , it is unclear whether sink - merging is really applicable : the protostellar radius is at most expected to be 0.5 au ( and then only for a short time , e.g. @xcite ) , which is significantly smaller than both our adopted accretion radius and the gravitational softening . in fact , in the simulations from present - day star formation in which extremely small sink particles are used along with little , or no softening merging events are found to be fairly rare ( e.g. @xcite ) .
they do however occur , and this should be borne in mind when interpreting the results of this study .
also , given the size of our sink particles , we are unable to resolve any of the disks that would invariably form around the young protostars and it has been suggested that such structures would be unstable to fragmentation on these scales ( e.g. @xcite ) . finally , our calculations do not include any model for the feedback processes that accompany the formation of a young star ( e.g. @xcite ) .
given these uncertainties , the sink particles are unable to say exactly what the shape of the final imf will be , but rather measure how the gas can fragment at a given scale ( our resolution ) , and how these fragments are likely to evolve , assuming feedback processes play only a minor role over the timescales investigated in this study .
these caveats should be borne in mind when interpreting the results in the following sections .
in our comparison of the different simulations performed in this study , we will compare the properties of the clouds after roughly 10 percent of their mass has been accreted ( although in some cases we will mention in passing what happens as the simulations are advanced further ) . in terms of looking at the ability of the clouds to fragment , comparing the different simulations at this point in their evolution
ensures that they have turned the same faction of their initial number of jeans masses into the sink particles .
the exception in our analysis is the comparison between the 1000 pop .
iii.1 and iii.2 simulations , since in this case the pop .
iii.2 clouds are initially more jeans unstable .
we now go on to describe the fragmentation in the pop .
iii.1 and iii.2 channels in some detail .
the panels in fig .
[ fig : image3.1 ] show the column density distribution in the pop .
iii.1 simulations after 10 percent of their mass has been accreted onto the sink particles .
the images show the inner 1300 au of the cloud , centred on the first sink particle to form in each simulation .
we can see that the clouds with @xmath68 fragment into small clusters of sink particles ; the runs with @xmath69 , @xmath70 and @xmath71 form 5 , 31 and 15 sink particles respectively .
these calculations clearly demonstrate that turbulent subsonic motions are able to promote fragmentation in primordial gas clouds . only in the case with @xmath72 does the cloud form just a single sink particle . in this case
an extended disk builds up around the star , since the seed turbulence gives rise to some low level of rotation in the collapsing core , but otherwise the gas contains little structure .
interestingly , there is no clear trend linking the number of fragments that form and the initial turbulent energy : the 0.4 run fragments more than the 0.8 run , despite containing only one quarter of the initial turbulent energy .
the effects at play here are somewhat complex .
first , the turbulent velocity field contained in the collapsing region will differ with each value of @xmath73 , since the different strengths of flow will push the gas around to different degrees .
in addition , the nature of the turbulence that survives in the collapsing core will also affect the fragmentation . as we see from the images in fig .
[ fig : image3.1 ] , the cloud with @xmath74
the most successful in terms of fragmentation forms a large disk - like structure . as we will discuss in [ sec : evol ] , this configuration appears to aid the fragmentation of the infalling envelope .
therefore , much of the cloud s ability to fragment depends on the level of rotation that happens to become locked - up in the collapsing region .
further , and to a lesser degree , the extra delay in the collapse caused by the increased turbulent support also gives the cloud more time to wash out anisotropies in the gas ( see fig .
[ fig : tsf ] for the collapse times ) .
thus the ability of the cloud to fragment is a competition between these conflicting processes .
for the randomly generated velocity field that we used in this study , the ability to fragment is better for @xmath75 , than @xmath76 , but we note that this may not always be the case . we stress that to make a quantitative statistical statement about the number of fragments that form as a function of the turbulent mach number , we would need to run a series of different realizations of the turbulent velocity field in each case .
such a comparison lies outside the scope of our current study .
the mass functions of the sink particles from the simulations that undergo fragmentation are shown in fig .
[ fig : sinkmf ] . for clarity
, we have omitted the single 100 sink particle that forms in the 0.1 cloud . for the 0.2 and 0.8 clouds
, we see that the sink masses cluster around some central value roughly 12 and 4 respectively while the 0.4 cloud has a mass function that is skewed to lower masses , with a peak at around 1 , a sharp fall - off below this , and a broad distribution towards higher masses , extending up to around 13 .
we emphasize that the mass functions presented here ( as well as in fig .
10 ) do not yet represent the final imfs , as they correspond to an intermediate time in the overall accretion process , where only 10% of the cloud has been accreted .
they also might be affected by the numerical details of our sink technique , in particular the absence of any sink - sink mergers , as discussed in section 3 .
the reason for the spread of masses becomes apparent when we look at the accretion properties of the sink particles , which are shown in fig .
[ fig : maccp1 ] . focusing on the 0.4 and 0.8 runs , and looking at the evolution of the individual sinks in more detail
, we see that some appear to accrete rapidly and then suddenly stop .
this behaviour is typical of what is seen in simulations of bound , fragmenting cores in the context of present - day star formation @xcite , and is a result of velocity kicks from dynamical three ( or more ) body interactions .
the mass accretion rate in such a system @xcite is given by @xmath77 where @xmath78 is the mass of the sink particle , @xmath79 is the gas density , and @xmath80 is the velocity of the sink particle relative to the gas .
the ability of a sink to accrete more mass from the available reservoir is significantly reduced once its velocity increases .
the effect is then exacerbated by the fact that an increase in velocity results in the sink particle moving to a more distant orbit ( or even being kicked out of the system entirely ) and hence into a position where the gas density is lower . since this sink
particle is now in a position where further accretion is difficult , its siblings are able to accrete its ` share ' of the mass reservoir , with the majority going to those few sinks that sit right in the middle of the cloud s potential well . in general , as the sinks accrete from the background gas , they tend to move towards the centre , due to mass - loading .
further , their increased mass makes them more likely to survive dynamical encounters with their less massive siblings . as such , the ` rich get richer ' , with a few sinks ending up significantly more massive than the rest .
the process is typically termed ` competitive accretion ' , @xcite and normally leads to the type of distribution of masses seen in our 0.4run ( fig .
[ fig : sinkmf ] ) .
looking at the mass evolution of the individual sink particles , we also see that the formation of new sink particles occurs in bursts .
the turbulence generates structure in the gas which is enhanced by the gravitational collapse as the gaseous envelope falls in towards the central system of sink particles .
the bursts in sink formation reflect the moments when these structures detach from the flow and become self - gravitating in their own right .
figure [ fig : maccp1 ] also shows the accretion rate of the cluster as a whole , and how that compares to an estimate made from the radial infall profile . to construct this estimate ,
we first compute the mass infall rate as a function of the radial distance @xmath81 from the densest sph particle : @xmath82 where @xmath83 is the gas density in a spherical shell with radius @xmath81 and width d@xmath81 , and @xmath84 is the radial velocity of the gas in this shell and all quantities are volume - averaged .
given the enclosed mass as a function of radial distance , @xmath85 , it is straightforward to convert from @xmath86 to an infall rate as a function of enclosed mass , @xmath87 , which we identify with the total mass accretion rate of the system of sink particles . figure [ fig : maccp1 ] demonstrates that this estimate provides a fairly close match to the actual accretion rate onto the sink particle population , except at very early times , which is a numerical artifact : the sink particles instantly accrete all gas within their accretion radius when they form . at later times , the main difference between the estimated accretion rate and the true accretion rate is that the latter is significantly noisier , owing to the bursts of sink formation and the clumpiness of the infalling gas .
finally , we note that the accretion rates are also fairly insensitive to the level of turbulence in the cloud , suggesting that random turbulent motions with the magnitudes considered here should not lead to a variation in the overall accretion rates of primordial stars or star clusters from minihalo to minihalo .
other sources of support against gravity , such as rotation and magnetic fields ( assuming the latter can be efficiently generated , as in @xcite and @xcite ) , are likely to play a greater role in regulating the accretion rate .
indeed , the important role played by rotation in regulating the infall and accretion of gas can be appreciated if we compare the accretion rates measured in our simulations with those estimated or measured in previous studies of pop .
iii star formation starting from more realistic cosmological initial conditions ( e.g. * ? ? ?
* ; * ? ? ?
the absence of initial rotational support in our simulations leads to higher infall velocities , and hence to an accretion rate that is a factor of a few larger than these previous values .
the first clouds we will examine in the pop .
iii.2 case are those with an initial energy balance similar to those studied in the pop .
iii.1 case : that is , clouds that have only a few jeans masses initially .
since the pop .
iii.2 channel is cooler at number densities around @xmath43 @xmath44 , with a typical temperature of around 75k , the same initial number of jeans masses requires that the clouds have a lower mass of 150 .
the column density distribution in these clouds after 10 percent of the gas has been accreted by the sink particles is shown in the column density images in fig .
[ fig : image3.2 ] .
we see from the images that these clouds undergo significantly less fragmentation than their pop .
iii.1 counterparts , forming at most 3 sink particles after 10 percent of the cloud mass has been accreted .
while these calculations demonstrate that this primordial star formation channel is susceptible to fragmentation if turbulence is present in the collapsing gas , it appears to be significantly more stable than the pop .
iii.1 channel that gives rise to the first stars in the universe .
the main reason why this mode of star formation is less susceptible to fragmentation has to do with the thermal evolution of the gas as it collapses . in fig .
[ fig : rhotevol ] we show the temperature of the gas as a function of its number density , for the pop .
iii.1 and pop .
iii.2 cases , taken from the simulations with @xmath88 .
the elevated @xmath89 and hd fractions in the pop .
iii.2 simulations do not provide enough cooling to keep the gas close to the cmb temperature at these densities , owing to the increasing inefficiency of the hd as a coolant as it nears the critical density at which the populations of its rotational level reach their local thermodynamic equilibrium ( lte ) values .
the gas therefore heats up significantly as it collapses , increasing its temperature from 75 k at @xmath90 to roughly 700 k at @xmath91 , corresponding to evolution that is almost adiabatic .
this sharp increase in temperature temporarily increases the local jeans mass at the center of the cloud , and significantly slows the collapse , giving the turbulence in the cloud time to decay . in the absence of additional physical processes
able to replenish the turbulence , there is nothing to sustain the density inhomogeneities in the cloud that act as the seeds for later fragmentation ( since they are not yet self - gravitating ) , and so the end result is a collapse with a much lower level of fragmentation than in most of the pop .
iii.1 simulations .
a similar effect has previously been noted by @xcite in their study of fragmentation in very metal poor gas clouds .
they find that for metallicities of around @xmath92 , heat input due to three - body h@xmath2 formation at @xmath93 leads to a sharp jump in the gas temperature , which delays the collapse , reduces the elongation of the collapsing core , and suppresses any fragmentation .
furthermore , @xcite also briefly addressed this issue in their study of pop .
iii.2 star formation , and showed that their simulated pop .
iii.2 prestellar core would be stable against gravitational deformation at similar densities , owing to its hard effective equation of state .
although the evolution of the temperature with density is significantly different between the pop . iii.1 and pop .
iii.2 channels , we see that the accretion rates are similar when we consider the inner 0.01 to 10 solar masses of the collapsing envelope ( fig . [
fig : maccp2 ] ) , consistent with the results from @xcite .
however , if we consider the evolution of the accretion rates over the whole cloud , we see that at later times the two channels depart significantly from one another , with accretion occurring at a significantly slower rate in the pop .
iii.2 case ( fig .
[ fig : mdotcloud ] ) . also , at very early times , the accretion rates in the pop .
iii.2 clouds are more sensitive to the level of turbulence than the pop . iii.1 clouds , and the turbulence has a somewhat stronger effect in delaying the onset of star formation in these calculations ( fig .
[ fig : tsf ] ) . given that the pop .
iii.2 clouds seem to be much more stable against fragmentation , it is worthwhile investigating whether they can be made to fragment when the gas is initially more jeans unstable .
recent simulations of the formation of the first galaxies show that regions where pop .
iii.2 star formation occurs are fed by cold , supersonic turbulent streams of gas ( e.g. * ? ? ?
* ; * ? ? ?
as such , the initial condition for the pop .
iii.2 channel in this picture may have significantly more than one jeans mass , due to the rapid assembly of the self - gravitating core . in this section
we consider pop .
iii.2 clouds that contain 1000 , but otherwise have the same properties as the 150 clouds ( i.e. , same initial temperature , density , and initial chemical composition ) . since these clouds are colder than the pop .
iii.1 clouds of the same mass and density , they are initially more jeans unstable , having around 26 jeans masses at the start of the simulation . the results of two such calculations are shown in fig . [
fig : image3.2big ] , in which the initial levels of turbulence have been set to 0.4 and 0.8 @xmath60 .
again we show the evolution at the point where 10 percent of the cloud s mass has been converted into ( or accreted onto ) sink particles .
although these clouds are now initially more unstable than the pop .
iii.1 clouds , they form significantly fewer sink particles , only 8 in the 0.4 cloud and 7 in the 0.8 cloud ( see fig . [
fig : sinkmf32big ] ) .
the fact that the mass functions of the sink particles appear to have two peaks is due to the cloud undergoing two distinct bursts of sink formation .
the interesting result here is that these clouds exhibit _ less _ fragmentation than their pop .
iii.1 counterparts , despite containing _ more _ jeans masses in the initial configuration .
the main reason for this is the gas in the pop .
iii.2 clouds follows an extremely ` stiff ' effective equation of state ( essentially adiabatic evolution ) , as can be seen from the temperature - density relationship shown in fig .
[ fig : rhotevol ] .
such a rapid increase in the temperature during the initial collapse makes generating structure in the gas difficult , helping to remove any anisotropies introduced by the turbulent flows .
an additional effect is that for equations of state with effective adiabatic index of @xmath94 ( as is the case with these clouds ) , the jeans mass increases with increasing density . as such
, the collapse halts until sufficient mass has been assembled , and the new jeans mass has been reached . since the cooling time is significantly longer than the free - fall time , the gas has a chance to remove structure that could potentially assist the fragmentation at higher densities , where the effective @xmath95 is more similar to that found in the pop .
iii.1 clouds . the combination of these effects results in a gas which is much more stable to fragmentation during collapse than in the pop .
iii.1 simulations .
more insight into the stability of the clouds against fragmentation can be gained from fig . [
fig : evol ] , where we show the temperature and @xmath89 fraction as a function of density , for the 1000 pop .
iii.1 and iii.2 clouds in which the initial level of turbulence was 0.4 @xmath60 .
the different colors correspond to different points in the evolution of the clouds , with red , green and blue corresponding to the formation of the first sink particle , 50 accreted , and 100 accreted , respectively .
the temperature - density plots show the same behavior as reported in @xcite .
the particle evolution appears to diverge at a density of around @xmath96@xmath44 , with one group of sph particles heating up to a maximum temperature of around 7000k and the other staying around 1500k and cooling significantly at higher densities . as discussed in @xcite , the hot part of the diagram corresponds to gas that falls in at a later stage in the evolution of the cloud . at that point ,
the enclosed gas mass is larger than at earlier times , and so the free - fall velocity is correspondingly larger .
the gas therefore shocks more strongly than at early times , causing it to become hot enough to collisionally dissociate its h@xmath2 rapidly . with
the h@xmath2 gone , there is nothing to cool the gas until its temperature reaches @xmath97 . at this point ,
lyman-@xmath18 cooling becomes effective , allowing it to resist further heating .
although we see the same trends in both the pop .
iii.1 and pop .
iii.2 cases , in the latter the amount of gas departing from the ` standard ' temperature - density evolution is significantly reduced , since the elevated @xmath89 fractions allow the post - shock gas to cool more effectively , limiting the temperature rise and allowing more of the gas to retain its h@xmath2 . from the accompanying graphs we see that the @xmath89 is rapidly dissociated as the temperature rises , but given the relatively low amounts of @xmath89 present in the gas at these densities , the cooling provided by the dissociation is clearly unable to offset the compression .
however we see that at higher densities the @xmath89 suddenly reforms , coinciding with the region in the temperature - density diagrams where the gas is cold .
this prompts the question of whether the cold gas is the result of an isochoric cooling instability , brought on by rapid @xmath89 formation ? does the turbulence trigger an instability that naturally exists in pure primordial gas , but that is less effective in gas that has been influenced by the presence of previous star formation ? by harnessing the lagrangian nature of sph , the plots in fig . [ fig : ptrace ] shed some light on this issue .
they show the temporal evolution of the density , temperature and @xmath89 fraction for several sph particles that eventually become sink particles , and as such trace the conditions in the gas in the run - up to gravitational fragmentation and collapse . again
the simulations are those shown in fig .
[ fig : evol ] .
the quantities are calculated by averaging over each particle s 50 nearest neighbors at each instant in time . in the case of the pop .
iii.1 cloud , these sinks are the last to form in the simulation , while in the pop .
iii.2 cloud , the five lines represent all of the sink formation that occurs between the formation of the first and last sinks .
the figure shows a number of interesting features .
first , we see that none of those sph particles destined to become sinks undergo the rapid rise in temperature and accompanying loss of @xmath89 that is shown in fig .
[ fig : evol ] .
in contrast , their temperatures remain close to the temperatures found within the first collapsing core ( see fig .
[ fig : rhotevol ] ) .
this demonstrates that the cool particles seen in fig .
[ fig : evol ] do not come from regions that undergo shock heating and subsequent loss of @xmath89 .
instead , we see that they come from regions of gas that first undergo a relatively quiescent collapse , before being involved in several expansions and contractions .
the fact that this first stage of the density evolution is fairly slow needs to be stressed : the free - fall times at densities of @xmath98 and @xmath99 are approximately 1600 yr and 500 yr , respectively . as such
, these particles are not experiencing as much compression as those that end up losing their @xmath89 .
in fact we see that they actually have a higher than average @xmath89 fraction , when we compare them with the lower right - hand plot in fig . [
fig : evol ] , a property that helps them remain fairly cool as they collapse .
the reason why they experience less compression is that they are collapsing in a rotating structure that has been formed during the collapse of the turbulent gas .
the evolution of these particles and their immediate surroundings alters abruptly once they enter the dynamically complicated swirling regions that we can see in the column density figures .
first , they collide with other material , which results in a sharp increase in their temperature and density .
this increase in the density in turn increases the rate of @xmath89 formation , which very rapidly turns them fully molecular . as they re - expand , the adiabatic cooling and the now significantly enhanced @xmath89 line - cooling act together to reduce the temperature . in some cases
this happens several times , but for all of these particles , the end result is the same and they find themselves in a clump of gas that is now jeans unstable and fully molecular : the perfect conditions for forming a new protostar .
interestingly , we see similar behaviour leading up to the formation of the sink particles in both the pop .
iii.1 and iii.2 clouds . the relative lack of fragmentation in the pop .
iii.2 case simply results from the lack of structure in the collapsing envelope , rather than any thermal properties of the gas at these high densities .
the calculations presented in this study suggest that the first stars in the universe may form in small dense clusters , provided that the turbulent initial conditions we adopt are close to those found in minihalos .
this suggests that the population iii imf covered a broad range in masses , possibly exhibiting a scale - free , power - law extension similar to the present - day case .
previous arguments in favor of a peaked imf , in the shape of a narrow gaussian or even a delta function , would then need to be revisited ( e.g. , bromm & larson 2004 ) . although a similar prediction can be drawn from the study of clark et al .
( 2008 , hereafter cgk08 ) , there are some subtle differences .
our present study employs a fully self - consistent treatment of the thermodynamics , rather than the piece - wise polytropic equation of state ( eos ) approach used by cgk08 , which was taken as a fit to the detailed one - zone calculations of @xcite . although both start with similar ( pop .
iii.1 ) initial conditions , the fragmentation seen in the current simulations has a different origin . in the self - consistent treatment employed here ,
the fragmentation is driven by the complicated thermodynamics of high density clumpy gas as it enters the disk - like regions that surround the first protostar .
in contrast , the fragmentation in cgk08 was evident at much lower densities , with the turbulent flows forming structures that were enhanced during the collapse , and rotation providing a ` window of opportunity ' for those structures to become gravitationally unstable in their own right , rather than being simply accreted onto the central protostellar core .
note however that the simulations in cgk08 also contained systematic , solid - body rotation in the initial conditions , which we do not study in this paper .
this may help some of the structure to survive . in general ,
the calculations presented in this paper suggest that the structure formed early on in the collapse is less likely to survive when the full thermodynamical behavior of the gas is taken into account .
this is evident in the fact that the current calculations yield significantly fewer fragments than the pop iii simulations in cgk08 25 sink particles for 19 of accreted gas in cgk08 , compared to 31 sink particles for 100 of accreted gas in the current study .
one feature that should be stressed ( and which was originally pointed out in @xcite ) is that the long - term thermodynamic evolution of the envelope differs significantly from the one - zone models .
this brings into question the practice of using a piece - wise barotropic eos as a proxy for the full thermodynamics in simulations that intend to study the evolution of the gas beyond the collapse of the first protostellar core . assuming that the turbulent initial conditions used in this paper are indeed representative of the gas in minihalos , it is worth considering the implications of the fragmentation that we see .
the mass spectrum of the fragments in our calculations ranges from a few 0.1 to a few 10 .
we would not expect stellar feedback to change this result significantly , since previous studies ( e.g. * ? ? ?
* ) have shown that feedback becomes effective at limiting accretion onto pop .
iii protostars only for protostellar masses greater than about 20 .
however , such feedback effects are expected to become important , and in particular to suppress further fragmentation ( e.g. , krumholz et al . 2009 ) during the further evolution of the cloud .
the fragments will then likely grow in mass , and possibly even merge with each other .
it is very difficult to extrapolate to the final situation where accretion and merging stops .
but our main result that the population iii imf was likely broad seems robust . as we have discussed above , it is difficult to reach any definitive conclusions regarding the final population iii masses and the resulting imf
. it appears at least possible , however , that in rare cases truly metal - free stars with masses less than @xmath100 could have formed , and would still be present in the milky way today , thus providing a unique opportunity to directly probe the physical conditions at the end of the dark ages with galactic observations . current models of hierarchical galaxy formation predict that the first and most metal - poor halos to merge will become part of the bulge component of the resulting spiral galaxy ( e.g. * ? ? ?
it is therefore highly interesting to survey the bulge of the milky way for extremely metal - poor and metal - free stars .
however , this is also very challenging as the bulge is far away , has high stellar density , and contains a wide range of stellar populations of all ages and metallicities .
the odds of finding a few truly metal - free stars amongst millions of other stars are very low , or even zero due to pollution from the interstellar medium ( frebel et al .
2009 ) . with the same caveat as above , it is interesting to consider the implications of a pop .
iii imf with a characteristic mass significantly smaller than the canonical @xmath20 .
for instance , the flux of ionizing photons produced by a population of pop .
iii stars has a significant dependence on the form of the pop .
iii imf , as does the metal - enrichment pattern produced by a collection of pop .
iii supernovae @xcite .
indeed , the abundance patterns observed to date in extremely metal - deficient stars in the galactic halo ( see e.g. the review by * ? ? ?
* ) are far more consistent with an imf that produces primarily core - collapse supernovae , with progenitor masses of 10@xmath101 , rather than with an imf that produces only very massive pair - instability supernovae , or pisne @xcite .
on the other hand , there may be subtle selection effects at work that bias current surveys against finding pisn - enriched stars ( karlsson et al .
the basic argument here is that pisne have such high metal yields that abundances in stars that form out of this material are already quite high ( see greif et al .
2010 ) , and would therefore be missed in searches that target the lowest metallicities .
the interpretation of the large carbon enhancements seen in the population of carbon - enhanced metal - poor stars as the result of the enrichment of these stars by winds from binary companions that have passed through the agb phase also implies the existence of a large number of intermediate - mass pop .
iii stars , with masses @xmath102@xmath103 @xcite .
again , there are alternative models to explain the carbon enhancement in metal - poor stars in terms of nucleosynthesis in faint supernovae ( e.g. , iwamoto et al . 2005 ) , in line with a higher characteristic pop .
iii mass .
it is important to note that one factor that may limit the impact of turbulent fragmentation on the primordial imf is if most pop .
iii stars form in conditions resembling our pop .
iii.2 clouds .
it is relatively straightforward to show that most pop .
iii stars will form in halos that have been affected in some fashion by a previous episode of pop .
iii star formation @xcite .
thus , using the @xcite terminology , most pop .
iii stars will be pop . iii.2 stars .
however , it is less obvious how many of these stars will have formed out of gas that has been cooled to temperatures @xmath104 by hd .
the effectiveness of hd cooling in these systems depends on the balance between the enhanced formation of h@xmath2 and hd , owing to the enhanced initial fractional ionization in this gas , and the destruction of h@xmath2 and hd due to lyman - werner band absorption of ultraviolet photons from an extragalactic background ( haiman , abel & rees 2000 ; johnson , greif & bromm 2008 ) , or from local sources @xcite .
the relative importance of these effects has not been explored in great detail , and in any case , the outcome is likely to be sensitive to uncertainties in the microphysics that have only recently been resolved @xcite . to sum up
, our calculations suggest that turbulent fragmentation may play an important role in the formation of pop .
iii stars , and may strongly influence the form of the population iii imf .
however , the approximate nature of our calculation specifically , our simplified choice of initial conditions means that we should regard this at present as no more than a plausible hypothesis . to better establish the role of turbulent fragmentation in real pop .
iii minihalos , we will need to be able to simulate a representative sample of both pop .
iii.1 and pop .
iii.2 minihalos with sufficiently high resolution , such that they can resolve the turbulent flows in the gas at the moment when it becomes self - gravitating .
this is a challenging prospect , lying far beyond of the scope of this preliminary study , but our results suggest that it would prove extremely worthwhile .
we have explored the effects of subsonic turbulence on the gravitational collapse of primordial gas clouds . the study employed sink particles to model the run - away collapse of protostellar cores , which allowed us to follow the evolution of the collapsing clouds beyond the formation of the first protostar .
the calculations also used a full time - dependent chemical network that accounts for the thermodynamic behaviour of the gas .
the current calculations contain neither magnetic fields nor feedback from the protostars .
our main findings can be summarized as follows : * turbulent primordial gas is unstable to fragmentation when one considers the evolution _ beyond _ the formation of the first protostellar core .
* gas starting from conditions appropriate to pop .
iii.1 collapse , rather than pop .
iii.2 star formation , is more susceptible to fragmentation . as suggested by @xcite , the thermal evolution of pop .
iii.2 gas as it collapses helps to suppress further gravitational instability over and above the main collapse mode .
however some fragmentation in the pop .
iii.2 case is seen in this study , caused by the inhomogeneities introduced by the turbulence . * in the cases where fragmentation is efficient ( in particular the pop iii.1 cloud with turbulent rms velocity @xmath74 ) , the masses of the fragments extend over a large range , which results in a distribution of stellar masses exhibiting a power - law extension towards high mass , as seen in the present - day imf @xcite
. however it should be noted that the exact form of the sink particle mass function may depend on our assumed initial conditions ( see appendix [ icdiscuss ] ) and the simplifications employed in the sink particle implementation itself . * due to the relative lack of fragmentation in the pop .
iii.2 clouds compared to the pop .
iii.1 clouds , the stars formed in the pop .
iii.2 calculations are of higher mass on average ( for the same accreted mass ) than their pop
. iii.1 counterparts , even though the mass accretion rate of the star cluster as a whole is higher in the pop .
iii.1 clouds .
* fragmentation tends to occur in gas which has temperatures of around 200
400 k at densities above @xmath105 , significantly lower than the 1000 1500 k associated with the collapse of the first core .
such cold temperatures are a result of the expansion that occurs as gas enters the rotating , disk - like regions around the central core , coupled with relatively high fractions of @xmath89 , which can provide efficient line - cooling .
these rotating structures are themselves a consequence of the angular momentum that is present in the turbulence . in summary , we propose that if even small levels of turbulence , with velocity dispersions of order 20% of the sound speed or more , are present in the baryonic component of dark matter minihalos , primordial stars are likely to be born in small stellar groups , rather than in isolation , and to have a wide range of stellar masses .
further , the very first stars ( pop .
iii.1 ) may have lower masses on average than the second generation of stars ( pop .
iii.2 ) , contrary to what has previously been assumed .
the authors would like to thank tom abel and thomas greif for stimulating discussions that helped shape this paper .
the work presented in this paper was assisted by the european commission fp6 marie curie rtn constellation ( mrtn - ct-2006 - 035890 ) .
p.c.c . acknowledges support by the _ deutsche forschungsgemeinschaft _ ( dfg ) under grant kl 1358/5 .
r.s.k . acknowledges financial support from the _ landesstiftung baden - wrrtemberg _ via their program international collaboration ii ( grant p - ls - spii/18 ) and from the german _ bundesministerium fr bildung und forschung _ via the astronet project star format ( grant 05a09vha ) .
r.s.k . furthermore acknowledges subsidies from the dfg under grants no .
kl1358/1 , kl1358/4 , kl1358/5 , kl1358/10 , and kl1358/11 , as well as from a frontier grant of heidelberg university sponsored by the german excellence initiative . r.s.k . also thanks the kavli institute for particle astrophysics and cosmology at stanford university and the department of astronomy and astrophysics at the university of california at santa cruz for their warm hospitality during a sabbatical stay in spring 2010 .
v.b . acknowledges support from nsf grant ast-0708795 and nasa atfp grant nnx08al43 g .
part of the simulations were carried out at the texas advanced computing center ( tacc ) , under teragrid allocation tg - ast090003 .
abel , t. , bryan , g. l. , & norman , m. l. 2002 , science , 295 , 93 aldrovandi , s. m. v. & pequignot , d. 1973 , a&a , 25 , 137 attwood , r. e. , goodwin , s. p. , stamatellos , d. , & whitworth , a. p. 2009 , a&a , 495 , 201 barlow , s. g. 1984 , phd thesis , univ .
colorado bate , m. r. 2009 , mnras , 392 , 590 bate , m. r. , bonnell , i. a. , & price , n. m. 1995 , mnras , 277 , 362 beers , t. c. , & christlieb , n. 2005 , ara&a , 43 , 531 benz , w. , 1990 , in ` proceedings of the nato advanced research workshop on the numerical modelling of nonlinear stellar pulsations problems and prospects ' , ed .
j. r. buchler , ( dordrecht : kluwer ) , 269 bonnell , i. a. , bate , m. r. , clarke , c. j. , & pringle , j. e. 1997 , mnras , 285 , 201 bonnell , i. a. , bate , m. r. , clarke , c. j. , & pringle , j. e. 2001a , mnras , 323 , 785 bonnell , i. a. , clarke , c. j. , bate , m. r. , & pringle , j. e. 2001b , mnras , 324 , 573 bonnell , i. a. , vine , s. g. , & bate , m. r. 2004 , mnras , 349 , 735 bonnell , i. a. , & bate , m. r. 2006 , mnras , 370 , 488 bromm , v. , coppi , p. s. , & larson , r. b. 2002 , apj , 564 , 23 bromm , v. , & larson , r. b. 2004 , ara&a , 42 , 79 bromm , v. , & loeb , a. 2004 , new astron . 9 , 353 bromm , v. , yoshida , n. , hernquist , l. & mckee , c. f. 2009 , nature , 459 , 49 chabrier , g. 2003 , pasp , 115 , 763 clark , p. c. , glover , s. c. o. , & klessen , r. s. 2008 , apj , 672 , 757 dalgarno , a. , & lepp , s. 1987 , in astrochemistry , ed .
m. s. vardya & s. p. tarafdar ( dordrecht : reidel ) , 109 delgado - donate , e. j. , clarke , c. j. , & bate , m. r. 2004 , mnras , 347 , 759 dove , j. e. , rusk , a. c. m. , cribb , p. h. , & martin , p. g. 1987 , apj , 318 , 379 ferland , g. j. , peterson , b. m. , horne , k. , welsh , w. f. , & nahar , s. n. 1992 , apj , 387 , 95 flower , d. r. , & harris , g. j. 2007 , mnras , 377 , 705 frebel , a. , johnson , j. l. , & bromm , v. 2009 , mnras , 392 , l50 frommhold , l. 1993 , collision - induced absorption in gases , ( cambridge university press : cambridge ) galli , d. , & palla , f. 1998 , a&a , 335 , 403 gerlich , d. 1982 , in ` symposium on atomic and surface physics ' , eds .
lindinger , w. , howorka , f. , mrk , t. d. , ( dordrecht : kluwer ) , p. 304
glover , s. 2005 , space sci .
, 117 , 445 glover , s. c. o. 2008 , in first stars iii , eds . b. oshea , a. heger & t. abel , ( new york : aip ) , 25 ; arxiv:0708.3086 glover , s. c. o. , & abel , t. 2008 , mnras , 388 , 1627 glover , s. c. o. , & brand , p.
w. j. l. 2001 , mnras , 321 , 385 glover , s. c. o. , & jappsen , a .- k .
2007 , apj , 666 , 1 glover , s. c. o. , savin , d. w. , & jappsen , a .- k .
2006 , apj , 640 , 553 goodwin , s. p. , whitworth , a. p. , & ward - thompson , d. 2004a , a&a , 414 , 633 goodwin , s. p. , whitworth , a. p. , & ward - thompson , d. 2004b , a&a , 423 , 169 greif , t. h. , johnson , j. l. , klessen , r. s. , & bromm , v. 2008 , mnras , 387 , 1021 greif , t. h. , glover , s. c. o. , bromm , v. , & klessen , r. s. 2010 , apj , 716 , 510 haiman , z. , abel , t. , & rees , m. j. 2000 , apj , 534 , 11 hosokawa , t. , & omukai , k. 2009 , apj , 691 , 823 hummer , d. g. , & storey , p.j .
1998 , mnras , 297 , 1073 huq , m. s. , doverspike , l. d. , champion , r. l. , & esaulov , v. a. 1982 , j. phys .
b , 15 , 951 iwamoto , n. , umeda , h. , tominaga , n. , nomoto , k. , & maeda , k. 2005 , science , 309 , 451 janev , r. k. , langer , w. d. , evans , k. , & post , d. e. 1987 , elementary processes in hydrogen - helium plasmas , springer jappsen , a .- k . , klessen , r. s. , larson , r. b. , li , y. , & mac low , m .-
2005 , a&a , 435 , 611 jappsen , a .-
low , m .- m . m. , glover , s. c. o. , klessen , r. s. , & kitsionas , s. 2009 , apj , 694 , 1161 jappsen , a .-
klessen , r. s. , glover , s. c. o. , & mac low , m .-
2009 , apj , 696 , 1065 joggerst , c. c. , almgren , a. , bell , j. , heger , a. , whalen , d. j. , & woosley , s. e. 2010 , apj , 709 , 11 johnson , j. l. , & bromm , v. 2006 , mnras , 366 , 247 johnson , j. l. , greif , t. h. , & bromm , v. 2008 , mnras , 388 , 26 karlsson , t. , johnson , j. l. , & bromm , v. 2008 , apj , 679 , 6 karpas , z. , anicich , v. , & huntress , w. t. 1979 , j. chem .
phys , 70 , 2877 kimura , m. , lane , n. f. , dalgarno , a. , & dixson , r. g. 1993 , apj , 405 , 801 klessen , r. s. , & burkert , a. 2000 , apjs , 128 , 287 klessen , r. s. , & burkert , a. 2001 , apj , 549 , 386 klessen , r. s. 2001 , apj , 550 , l77 klessen , r. s. , & hennebelle , p. 2009 , arxiv:0912.0288 kreckel , h. , bruhns , h. , ek , m. , glover , s. c. o. , miller , k. a. , urbain , x. , & savin , d. w. , 2010 , science , 329 , 69 kroupa , p. 2002 , science , 295 , 82 krumholz , m. r. , klein , r. i. , & mckee , c. f. 2007 , apj , 656 , 959 krumholz , m. r. , klein , r. i. , mckee , c. f. , offner , s. s. r. , & cunningham , a. j. 2009 , science , 323 , 754 larson , r. b. 1981 , mnras , 194 , 809 machida , m. n. 2008 , apj , 682 , l1 machida , m. n. , omukai , k. , matsumoto , t. , & inutsuka , s .- i .
2008 , apj , 677 , 813 machida , m. n. , omukai , k. , matsumoto , t. , & inutsuka , s .-
2009 , mnras , 399 , 1255 , m .- m . , & klessen , r. s. 2004 , rev .
phys . , 76 , 125 martin , p. g. , schwarz , d. h. , & mandy , m. e. 1996 , apj , 461 , 265 martin , p. g. , keogh w. j. , & mandy , m. e. 1998 , apj , 499 , 793 mckee , c. f. , & tan , j. c. 2008 , apj , 681 , 771 mielke , s. l. , peterson , k. a. , schwenke , d. w. , garrett , b. c. , truhlar , d. g. , michael , j. v. , su , m .- c . , & sutherland , j. w. 2003 , phys .
lett . , 91 , 063201 molaro , p. 2008 , in ` pathways through and eclectic universe ' , asp conf .
390 , eds .
j. h. knapen , t. j. mahoney & a. vazdekis , ( san francisco : asp ) , 472 nagakura , t. , & omukai , k. 2005 , mnras , 364 , 1378 neufeld , d. a. , & kaufman , m. j. 1993 , apj , 418 , 263 omukai , k. , & nishi , r. 1999 , apj , 518 , 64 omukai , k. , tsuribe , t. , schneider , r. , & ferrara , a. 2005 , apj , 626 , 627 oshea , b. w. , & norman , m. l. 2007 , apj , 654 , 66 oshea , b. w. , mckee , c. f. , heger , a. , & abel , t. 2008 , first stars iii , 990 , 13 palla , f. , salpeter , e. e. , & stahler , s. w. 1983 , apj , 271 , 632 peart , b. , & hayton , d. a. 1994 , j. phys .
b , 27 , 2551 poulaert , g. , brouillard , f. , claeys , w. , mcgowan , j. w. , & van wassenhove , g. 1978 , j. phys .
b , 11 , l671 ramaker , d. e. , & peek , j. m. 1976 , phys .
rev . a , 13 , 58 ripamonti , e. , & abel , t. 2004 , mnras , 348 , 1019 savin , d. w. 2002 , apj , 566 , 599 savin , d. w. , krstic , p. s. , haiman , z. , & stancil , p. c. 2004 , apj , 606 , l167 ; erratum apj , 607 , l147 schleicher , d. r. g. , banerjee , r. , sur , s. , arshakian , t. g. , klessen , r. s. , beck , r. , & spaans , m. 2010 , mnras , submitted ; arxiv:1003.1135 schmeja , s. , & klessen , 2004 , a&a , 419 , 405 schneider , i. f. , dulieu , o. , giusti - suzor , a. , & roueff , e. 1994 , apj , 424 , 983 ; erratum apj , 486 , 580 schulz , g. j. , & asundi , r. k. 1967 , phys .
, 158 , 25 shapiro , p. r. , & kang , h. 1987 , apj , 318 , 32 shavitt , i. 1959 , j. chem .
phys . , 31 , 1359 springel , v. 2005 , mnras , 364 , 1105 stacy , a. , greif , t. h. , & bromm , v. 2010 , mnras , 403 , 45 tan , j. c. , & blackman , e. g. 2004 , apj , 603 , 401 tan , j. c. , & mckee , c. f. 2008 , in first stars iii , eds .
b. oshea , a. heger & t. abel , ( new york : aip ) , 47 trenti , m. , & stiavelli , m. 2009 , apj , 694 , 879 trevisan , c. s. , & tennyson , j. 2002a , plasma phys . controlled fusion , 44 , 1263 tsuribe , t. , & omukai , k. 2008 , apj , 676 , l45 tumlinson , j. , venkatesan , a. , & shull , j. m. 2004 , apj , 612 , 602 tumlinson , j. 2007a , apj , 664 , l63 tumlinson , j. 2007b , apj , 665 , 1361 tumlinson , j. 2010 , apj , 708 , 1398 turk , m. j. , abel , t. , & oshea , b. w. 2009 , science , 325 , 601 turk , m. j. , clark , p. c. , glover , s. c. o. , greif , t. h. , abel , t. , klessen , r. s. , & bromm , v. 2010 , apj , _ accepted _ , arxiv:1011.1908 walkauskas , l. p. , & kaufman , f. 1975 , symp .
combust .
proc . , 15 , 691 wise , j. h. , & abel , t. 2007 , apj , 665 , 899 wise , j. h. , & abel , t. 2008 , apj , 685 , 40 wishart , a. w. 1979 , mnras , 187 , 59p wolniewicz , l. , simbotin , i. , & dalgarno , a. 1998 , apjs , 115 , 293 yoshida , n. , omukai , k. , hernquist , l. , & abel , t. 2006 , apj , 652 , 6 yoshida , n. , omukai , k. & hernquist , l. 2007 , apj , 667 , l117 yoshida , n. , omukai , k. , & hernquist , l. 2008 , science , 321 , 669 zygelman , b. , dalgarno , a. , kimura , m. , & lane , n. f. 1989 , phys . rev .
a , 40 , 2340
to model the chemical evolution of the metal - free gas , we use the network for primordial hydrogen , helium and deuterium chemistry detailed in table [ tab : chem ] .
our treatment of the hydrogen and helium chemistry largely follows @xcite , but our treatment of the deuterium chemistry is significantly simplified .
this simplification arises from our neglect of @xmath106 , @xmath107 and @xmath108 , none of which play a significant role in controlling the hd abundance in the physical conditions relevant to our present study . for the most part
, our choice of rate coefficients also follows @xcite .
( reaction number 7 in their table ) should use natural logarithms , and not base 10 logarithms , as listed .
however , the listed fitting coefficients are correct .
second , the temperature dependence of the reaction @xmath109 ( their reaction 24 ) should be @xmath110 , and not @xmath111 . ]
we adopt case b rate coefficients for the recombination of h@xmath46 and he@xmath112 , and treat he@xmath46 recombination as decribed in section 2.1.4 of @xcite .
we adopt rate coefficients from @xcite for the associative detachment of h@xmath113 ions by atomic hydrogen ( reaction 2 ) , and for the mutual neutralization of h@xmath113 by h@xmath46 ( reaction 5 ) .
we also note that the uncertainties in these reaction rate coefficients discussed in @xcite are unlikely to be significant in the high density , low ionization gas modelled in our present simulations . for the rate coefficient of reaction 30 ,
the three - body formation of h@xmath2 with atomic hydrogen as the third body @xmath114 we adopt the rate coefficient proposed by @xcite .
the value of the rate coefficient for this reaction is highly uncertain @xcite , but our chosen value lies intermediate between the fastest and slowest rates to be found in the literature ( @xcite and @xcite , respectively ) , and is based on the best currently available data for the inverse reaction ( i.e. collisional dissociation of h@xmath2 by h , reaction 9 ) .
the influence of the uncertainty in this rate coefficient has been studied in detail elsewhere @xcite . to fix the rate coefficient for
the analogous reaction @xmath115 we follow @xcite and assume that the rate coefficient is one - eighth of the size of the rate coefficient adopted for reaction [ tbh2a ] .
for three - body formation in which neutral helium is the third body , we follow @xcite and use a rate coefficient originally taken from @xcite . we assume , following @xcite , that the rate coefficients for the three - body formation of hd ( reactions 4345 ) are the same as those for the corresponding h@xmath2 formation reactions ( nos . 3032 ) .
we adopt values for the rate coefficients for collisional dissociation of h@xmath2 by h , h@xmath2 and he that are consistent with our choices for the three - body association rates in the sense that each pair of rate coefficients individually satisfies @xmath116 where @xmath117 is the rate coefficient for three - body association , @xmath118 is the rate coefficient for collisional dissociation , and where the equilibrium constant @xmath119 is given in all three cases by @xcite @xmath120 this procedure is necessary in order to ensure the correct chemical behaviour of the gas at high densities and temperatures , where the h@xmath2 formation and destruction timescales are both short , and most of the gas is close to chemical equilibrium .
llc 1 & @xmath121 & 1 + 2 & @xmath122 & 2 + 3 & @xmath123 & 3 + 4 & @xmath124 & 4 + 5 & @xmath125 & 2 + 6 & @xmath126 & 5 + 7 & @xmath127 & 6 + 8 & @xmath128 & 7 + 9 & @xmath129 & 8 + 10 & @xmath130 & 9 , 10 + 11 & @xmath131 & 11 + 12 & @xmath132 & 12 + 13 & @xmath133 & 13 + 14 & @xmath134 & 12 + 15 & @xmath135 & 12 + 16 & @xmath136 & 14 + 17 & @xmath137 & 12 + 18 & @xmath138 & 12 + 19 & @xmath139 & 15 , 16 + 20 & @xmath140 & 13 + 21 & @xmath141 & 17 + 22 & @xmath142 & 17 + 23 & @xmath143 & 18 + 24 & @xmath144 & 19 + 25 & @xmath145 & 19 + 26 & @xmath146 & 20 + 27 & @xmath147 & 21 + 28 & @xmath148 & 22 + 29 & @xmath149 & 23 + 30 & @xmath150 & 24 + 31 & @xmath151 & 25 + 32 & @xmath152 & 26 + 33 & @xmath153 & 27 + 34 & @xmath154 & 28 + 35 & @xmath155 & 28 + 36 & @xmath156 & 29 + 37 & @xmath157 & 30 + 38 & @xmath158 & 31 + 39 & @xmath159 & 30 + 40 & @xmath160 & 27 + 41 & @xmath161 & 32 + 42 & @xmath162 & 32 + 43 & @xmath163 & see text + 44 & @xmath164 & see text + 45 & @xmath165 & see text +
the dominant coolant in our simulations is molecular hydrogen . to model rotational and vibrational line emission from h@xmath2
, we use the detailed cooling function described in @xcite , that includes contributions from collisions of h@xmath2 with h , he , h@xmath2 , protons and electrons . at densities
@xmath166 , the strongest of the h@xmath2 lines become optically thick , reducing its effectiveness as a coolant . to model h@xmath2 cooling in this regime , we use an approach based on the sobolev approximation that was first used in models of primordial star formation by @xcite .
we write the h@xmath2 cooling rate in optically thick gas as @xmath167 where @xmath168 is the number density of hydrogen molecules in upper energy level @xmath169 , @xmath170 is the energy difference between this upper level and a lower level @xmath171 , @xmath172 is the spontaneous radiative transition rate for transitions between @xmath169 and @xmath171 , and @xmath173 is the escape probability associated with this transition , i.e. the probability that the emitted photon can escape from the region of interest . we take values for the level energies from the compilation made available by p. g. martin on his websitepgmartin / h2.html ] and for the radiative transition rates from @xcite , and we fix @xmath168 by assuming that the h@xmath2 level populations are in lte .
the problem of modelling optically thick h@xmath2 cooling thereby reduces to one of computing the escape probability for each transition .
we follow @xcite and write the escape probability for the transition @xmath174 as @xmath175 where we approximate @xmath176 as @xmath177 where @xmath178 is the line absorption coefficient and @xmath179 is the soloblev length . in the classical , one - dimensional spherically symmetric case , the sobolev length is given by @xmath180 where @xmath181 is the thermal velocity , and @xmath182 is the radial velocity gradient . in our inherently three - dimensional flow ,
we generalize this as @xmath183 following @xcite .
if the velocity dispersion of the gas is very small , then @xmath179 can become very large , much larger than the size of the collapsing core . to ensure that we do not reduce the h@xmath2 cooling rate in this case to an artificially low value
, we take as our actual length scale in equation [ tau ] the smallest of the soloblev length and the local jeans length , @xmath184 . since the line absorption coefficient @xmath178 is linearly proportional to the number density of h@xmath2 , we can write @xmath176 as @xmath185 where @xmath186 is an effective h@xmath2 column density , and where @xmath187 is a function only of temperature .
we therefore tabulate the cooling rate per h@xmath2 molecule in the optically thick limit as a function of two parameters : the gas temperature @xmath188 and the effective h@xmath2 column density @xmath189 , and compute cooling rates during the simulations by interpolation from a pre - generated look - up table . at densities
@xmath190 , a second form of h@xmath2 cooling becomes important , called collision - induced emission .
although h@xmath2 molecules have no electric dipole , the interacting pair in a collision of h@xmath2 with h , he or h@xmath2 briefly acts as a ` supermolecule ' with an non - zero electric dipole , and a hence a non - zero probability of emitting or absorbing a photon through a dipole transition . because the collision time is very short ,
the resulting collision - induced transition lines are very broad , effectively merging into a continuum ( for more details , see e.g. * ? ? ?
we include this process in our cooling function , using a rate taken from @xcite .
we crudely account for the reduction of the cie cooling rate by continuum absorption at very high number densities using the following prescription ( m. turk , private communication ) @xmath191 where @xmath192
however , we note that optical depth effects do not strongly affect the cie cooling rate for densities below our threshold for sink particle creation ( see [ sec : ics ] ) , and so the approximate nature of this opacity cutoff is unlikely to significantly affect our results .
in addition to h@xmath2 ro - vibrational line emission and cie cooling , our cooling function contains a number of other radiative processes : electronic excitation of h , he and he@xmath46 , cooling from the recombination of h@xmath46 and he@xmath46 , compton cooling and bremsstrahlung .
details of our treatment of these processes can be found in @xcite .
we also account for changes in the thermal energy of the gas due to changes in its chemical makeup .
specifically , we include the effects of cooling due to the collisional ionization of h , he and he@xmath46 , and due to the destruction of h@xmath2 by charge transfer and by collisional dissociation , as well as heating due to the three - body h@xmath2 formation .
the balance between cooling due to h@xmath2 collisional dissociation and heating due to three - body h@xmath2 formation plays a very important role in regulating the thermal evolution of the gas at densities @xmath193 .
given our somewhat arbitrary choice of initial conditions , it is prudent to ask whether they are applicable to primordial star formation , and in particular , whether the choice of the initial density profile affects the details of the collapse . in fig .
[ fig : profiles ] , we show , for the pop iii.1 clouds with 0.1 and 0.4 @xmath60 turbulence , the number density and enclosed mass as a function of radius and how the radial velocity and temperature varies as a function of the enclosed mass . in the case of the 0.1@xmath60 cloud ,
the radial profiles ( mass and number density ) are much like those presented for the standard pop iii studies @xcite , in which fully cosmological initial conditions were used . in particular , by the time that the first sink particle forms in this run , the gas has developed the same @xmath194 radial density profile as found in the fully cosmological runs , demonstrating that the density profile at this stage is insensitive to the shape of the initial density profile .
we do however see differences between the radial velocity and temperature profiles , compared to those published in the literature .
the infall velocities at this stage are somewhat higher than those seen in fully cosmological studies , but it should be noted that we do not include systemic rotation in our study , and so a major source of support is missing .
however , aside from the higher values of the radial velocity , the shape of the profile is also different . in the runs presented in this study , the infall velocities of the shells enclosing 1 and greater are systematically higher than those seen in the cosmological simulations published to date , with a peak in the radial velocity profile at about 10 - 100 ( depending on the level of turbulent support ) instead of around 0.3 .
some of this difference in the radial velocity profiles is reflected in the somewhat steeper than reported temperature profiles , since the gas will reacts to the increased @xmath195 heating .
however the temperature profiles are also affected by the differences in the chemistry , and in particular , our choice of the three - body h@xmath2 formation rate .
so the obvious question is , do our high infall velocities make fragmentation more likely than it would be reality ? to address this , we performed another pop iii.1 simulation with 0.4 @xmath60 turbulence , but this time we set the confining external pressure to a factor of 4 less than was used in the original simulations .
the radial and enclosed mass profiles at the point just before the first sink particle is formed are shown as the dashed lines in fig .
[ fig : profiles ] ( right - hand panel ) .
we see that radial velocity profile is affected by the change in boundary pressure , resulting in a factor of 4 reduction in the infall speed .
compared to the original run , which formed 31 sink particles , this simulation forms 22 sinks when 10 percent of the gas has been accreted .
while this suggests that the fragmentation we see is sensitive to the details of the collapse , and we are likely overestimating the number of fragments formed ( at least at the scale of our sink particles ) when one compares to the standard pop iii collapse profiles , it seems that this is not the main cause of the fragmentation .
the gas still fragments , and the processes leading up to the fragmentation are the same as those discussed in
[ sec : evol ] .
finally , we note that there is also new evidence to suggest that the radial collapse profiles seen in the literature so far , may not be representative of all pop iii star formation .
for example , while the radial density profiles from the studies by @xcite and @xcite are similar to the newer , higher - resolution , cosmological simulations of @xcite and @xcite , the radial ( or enclosed mass ) profiles of the other quantities in the newer calculations show considerable variation .
in particular , the velocity and temperature profiles in @xcite lie somewhere in between those found by @xcite and our 0.4 @xmath60 turbulent cloud .
a very high - resolution study of the onset of the initial gravitational instability in the baryonic component of dark matter minihalos , may help to establish whether these differences between the studies of , say , @xcite and @xcite are due to the increased resolution and improved physical treatment used in the latter study ( e.g. the inclusion of the effects of three - body h@xmath196 formation heating ) , or just reflect the effects of cosmic variance .
radial profiles from the pop iii.1 clouds with @xmath72 ( left panel ) and @xmath74 ( right panel ) .
the lines show the initial conditions ( solid line ) , the state of the gas just before the creation of the first sink ( dotted line ) . in the case of the @xmath74 cloud
, we also show the results of a collapse that has a factor of 4 less external pressure ( dashed line ) .
the dot - dashed line shows a slope of @xmath194 .
the crosses and circular points denote approximate values taken , respectively , from @xcite and @xcite , title="fig:",height=326 ] radial profiles from the pop iii.1 clouds with @xmath72 ( left panel ) and @xmath74 ( right panel ) .
the lines show the initial conditions ( solid line ) , the state of the gas just before the creation of the first sink ( dotted line ) . in the case of the @xmath74 cloud
, we also show the results of a collapse that has a factor of 4 less external pressure ( dashed line ) .
the dot - dashed line shows a slope of @xmath194 .
the crosses and circular points denote approximate values taken , respectively , from @xcite and @xcite , title="fig:",height=326 ] | we report results from numerical simulations of star formation in the early universe that focus on the dynamical behavior of metal - free gas under different initial and environmental conditions .
in particular we investigate the role of turbulence , which is thought to ubiquitously accompany the collapse of high - redshift halos .
we distinguish between two main cases : the birth of population iii.1 stars those which form in the pristine halos unaffected by prior star formation and the formation of population iii.2 stars those forming in halos where the gas has an increased ionization fraction .
we find that turbulent primordial gas is highly susceptible to fragmentation in both cases , even for turbulence in the subsonic regime , i.e. for rms velocity dispersions as low as 20 % of the sound speed .
fragmentation is more vigorous and more widespread in pristine halos compared to pre - ionized ones .
if such levels of turbulent motions were indeed present in star - forming minihalos , pop iii.1 stars would be on average of somewhat lower mass , and form in larger groups , than pop iii.2 stars .
we find that fragment masses cover over two orders of magnitude , suggesting that the population iii initial mass function may have been much broader than previously thought .
this prompts the need for a large , high - resolution study of the formation of dark matter minihalos that is capable of resolving the turbulent flows in the gas at the moment when the baryons become self - gravitating .
this would help to determine the applicability of our results to primordial star formation |
quantum phase transitions belong to the most notable manifestations of low - dimensional quantum spin systems , which can be achieved upon variation of some external force like for instance magnetic field , mechanical or chemical pressure ( doping).@xcite the quantum heisenberg model exhibits a great diversity of unconventional quantum orders such as topologically ordered haldane - type phases @xcite or dimerized states with an outstanding valence - bond - crystal order.@xcite among the most notable disordered quantum states without any local order parameter one could further mention resonating - valence - bond phases @xcite or other types ( e.g. tomonaga - luttinger ) of quantum spin liquids.@xcite a great variety of exotic quantum ground states can be found first of all in frustrated heisenberg spin models due to a mutual interplay between quantum effects with a geometric spin frustration .
@xcite the dimerized states with a valence - bond - crystal order are historically the most famous ground states of the frustrated spin-@xmath0 heisenberg models , which arise out from an effort of antiferromagnetically coupled spins to create a singlet dimer ( valence bond ) .
although the full exact solution of the heisenberg spin models is usually beyond the scope of the present knowledge , the variational principle provides an efficient tool for a rigorous determination of the dimerized ground states for a few paradigmatic examples such as the majumdar - ghosh model,@xcite the shastry - sutherland model,@xcite the frustrated ladder,@xcite etc . however , the usage of variational arguments is regrettably restricted mostly to the highly - frustrated parameter region and low magnetic fields , while the heisenberg spin systems often display spectacular quantum ground states also outside of this parameter space .
for instance , the spin-@xmath0 heisenberg diamond chain @xcite and diamond - like decorated planar lattices @xcite exhibit at moderate values of the spin frustration a peculiar tetramer - dimer ground state with spontaneously broken symmetry before they finally enter the monomer - dimer ground state predicted by the variational method in the highly - frustrated region .
on the other hand , in flat - band quantum spin systems the concept of localized magnons @xcite affords a powerful tool for a rigorous assignment of quantum ground states of the geometrically frustrated heisenberg spin models at sufficiently high magnetic fields ( see refs . for recent reviews and refs . for other flat - band systems ) .
this technique can be employed whenever destructive quantum interference traps magnon(s ) within cells with even number of bonds and hence , the frustrated quantum heisenberg model can be exactly mapped onto a classical lattice - gas model with a hard - core potential.@xcite using this approach , the microscopic nature of the last intermediate plateau in a zero - temperature magnetization curve of the quantum spin-@xmath0 heisenberg kagom lattice has been for instance elucidated along with a precise nature of a relevant second - order phase transition emerging at low but nonzero temperatures.@xcite the main advantage of the localized - magnon approach lies in that it also provides , besides an exact ground state , accurate description of low - temperature thermodynamics due to a proper counting of low - lying excited states.@xcite in the present work we will explore ground states , quantum phase transitions and low - temperature thermodynamics of the quantum spin-@xmath0 heisenberg octahedral chain , in which quantum spins placed at monomeric sites regularly alternate with the ones residing square plaquettes ( see fig . [ fig1 ] ) .
the proposed model belongs to a valuable class of the frustrated quantum heisenberg models , which satisfy a local conservation of the total spin on square plaquettes .
apart from a few exact results to be obtained within the variational and localized - magnon approaches , the spin-@xmath0 heisenberg octahedral chain can be consequently mapped onto effective mixed - spin heisenberg chains by following the approach developed previously by honecker , mila and troyer.@xcite the dmrg simulations of the effective mixed - spin heisenberg chains thus afford for the spin-@xmath0 heisenberg octahedral chain precise numerical results , which will be additionally corroborated through the ed calculations . it is worthwhile to remark that the polynuclear complexes , which involve quantum spin clusters with a geometric shape of octahedron as a magnetic core , constitute a relatively widespread family of compounds within an immense reservoir of coordination complexes . for illustration ,
let us quote a few specific examples of the hexanuclear complexes with an octahedral architecture of the magnetic core like cu@xmath1,@xcite v@xmath1,@xcite cr@xmath1,@xcite co@xmath1,@xcite fe@xmath1,@xcite mn@xmath1,@xcite mo@xmath1,@xcite w@xmath1,@xcite ru@xmath1,@xcite ir@xmath1,@xcite and ta@xmath1.@xcite although we are currently not aware of any experimental realization of the spin-@xmath0 heisenberg octahedral chain , we hope that our exciting theoretical findings presented hereafter could be inspiring for a tailored design of a one - dimensional polymeric chain of corner - sharing octahedra built out from discrete hexanuclear entities such as cu@xmath1@xcite or v@xmath1.@xcite the organization of this paper is as follows .
the quantum spin-@xmath0 heisenberg octahedral chain is introduced in sec .
[ sec : model ] along with basic steps of analytical and numerical methods used for its treatment .
the most interesting results for the ground - state phase diagram , zero - temperature magnetization process and low - temperature thermodynamics are discussed in sec .
[ sec : result ] . finally , several concluding remarks and future outlooks are mentioned in sec .
[ sec : conc ] .
heisenberg octahedral chain .
thick ( blue ) lines represent the heisenberg intra - plaquette coupling @xmath2 , while thin ( red ) lines correspond to the monomer - plaquette coupling @xmath3.,scaledwidth=45.0% ] let us consider a one - dimensional chain of corner - sharing octahedra schematically depicted in fig .
[ fig1 ] , which can be viewed as a generalization of the frustrated diamond chain @xcite and the double - tetrahedra chain.@xcite the hamiltonian of the quantum spin-@xmath0 heisenberg model defined upon the underlying octahedral chain is given by @xmath4.\end{aligned}\ ] ] above , @xmath5 denotes a standard spin-@xmath0 operator at a lattice site whose position is unambiguously determined by two subscripts , the former one specifies a position within the unit cell and the latter one the unit cell itself ( see fig . [ fig2 ] ) .
the coupling constant @xmath3 labels the heisenberg interaction between nearest - neighbor spins from monomeric and square - plaquette sites to be further referred to as the monomer - plaquette interaction , the coupling constant @xmath2 stands for the heisenberg interaction between nearest - neighbor spins from the same square plaquette to be referred to as the intra - plaquette interaction and the zeeman s term @xmath6 accounts for a magnetostatic energy of magnetic moments in an external magnetic field . for simplicity , the periodic boundary condition @xmath7 is imposed .
the hamiltonian ( [ ham ] ) can be attacked by making use of several complementary analytical and numerical approaches , which will be dealt with in what follows .
heisenberg octahedral chain , which is constituted by a five - spin cluster with the geometric shape of a square pyramid.,scaledwidth=10.0% ] an exact ground state of the spin-@xmath0 heisenberg octahedral chain can be rigorously found in the highly frustrated parameter region @xmath8 and low enough magnetic fields @xmath9 by exploiting the variational principle.@xcite to this end , let us decompose the total hamiltonian ( [ ham ] ) of the spin-@xmath0 heisenberg octahedral chain into a sum over the cell hamiltonians @xmath10 whereas the cell hamiltonian @xmath11 involves the interaction terms related to a five - spin cluster forming the unit cell with a geometric shape of a square pyramid ( see fig .
[ fig2 ] ) @xmath12 note that the factor @xmath0 at two latter interaction terms avoids a double counting of the intra - plaquette coupling @xmath2 and the zeeman s term @xmath13 , which are symmetrically split into two consecutive cell hamiltonians .
the variational procedure enables one to obtain the lower bound for the ground - state energy @xmath14 of the spin-@xmath0 heisenberg octahedral chain @xmath15 because the ground - state eigenvector @xmath16 can be alternatively viewed as a variational function for the five - spin heisenberg clusters ( fig .
[ fig2 ] ) .
it follows from eq .
( [ var ] ) that the relevant ground - state energy @xmath14 must be necessarily greater or equal to the sum of the lowest - energy eigenenergies of the five - spin clusters @xmath17 .
the energy spectrum of the five - spin heisenberg cluster ( fig .
[ fig2 ] ) with the geometric arrangement of a square pyramid can be expressed in terms of five quantum spin numbers @xmath18 , @xmath19 , @xmath20 , @xmath21 and @xmath22 @xmath23 \nonumber \\ \!\!&-&\!\ ! \frac{3}{8 } j_1 - h s_{t , j , k}^{z } , \label{spec}\end{aligned}\ ] ] which determine the total spin of the square pyramid @xmath18 and its @xmath24-component @xmath19 , the total spin of the square plaquette @xmath20 and the total spin of two spin pairs from opposite corners of a square plaquette @xmath21 and @xmath22 , respectively .
it can be easily checked from eq .
( [ spec ] ) that the lowest - energy eigenstate of the five - spin heisenberg cluster in the parameter space @xmath9 and @xmath8 is a doublet state , which can be characterized by the following quantum spin numbers @xmath25 , @xmath26 , @xmath27 and @xmath28 .
apparently , the four spins from each square plaquette are in a singlet - tetramer state given by the eigenvector @xmath29 and the spins from the monomeric sites are consequently decoupled from the other spins .
this lowest - energy eigenstate can be readily extended to the whole spin-@xmath0 heisenberg octahedral chain , which results in the monomer - tetramer ( mt ) ground state @xmath30\!. \nonumber \\ \label{mt}\end{aligned}\ ] ] to conclude this part , the spin-@xmath0 heisenberg octahedral chain exhibits the exact mt ground state in the low - field part @xmath9 of the highly frustrated parameter region @xmath8 , where four spins from each square plaquette form a singlet - tetramer state and the spins from the monomeric sites are fully aligned into the magnetic field or they are completely free to flip in a zero field . with regard to the perfect alignment of all monomeric spins , the mt ground state should manifest itself as the intermediate one - fifth plateau present in a zero - temperature magnetization curve within the field range @xmath31 .
it is quite clear that the lowest - energy eigenstate of the spin-@xmath0 heisenberg octahedral chain at sufficiently high magnetic fields exceeding the saturation value is the fully polarized ferromagnetic ( fm ) state @xmath32 with the following energy eigenvalue @xmath33 , @xmath34 is the respective zero - field energy .
it will be demonstrated hereafter that the concept of independent localized magnons @xcite can be employed in the highly frustrated region @xmath8 for a rigorous assignment of the saturation field and the exact ground state emerging below the saturation field .
the one - magnon eigenstates can be constructed within the orthonormal basis set @xmath35 ( @xmath36 , @xmath37 ) belonging to the sector @xmath38 with a single spin deviation from the fully polarized fm state .
applying the zero - field part of the hamiltonian ( [ ham ] ) within the given basis leads to the following set of equations @xmath39 which can be used for solving the eigenvalue problem @xmath40 in a zero field within the one - magnon sector by assuming @xmath41 .
the solution of the eigenvalue problem follows from the characteristic equation @xmath42 where @xmath43 labels an energy difference between the one - magnon state and the fully polarized fm state in a zero magnetic field .
the one - magnon energy spectrum of the spin-@xmath0 heisenberg octahedral chain in a zero magnetic field is composed of five energy bands @xmath44 which are for illustration depicted in fig .
[ fig3 ] for a few selected values of the interaction ratio @xmath45 .
it should be pointed out that three out of five one - magnon energy bands ( [ oms ] ) are completely dispersionless ( flat ) , which implies presence of the localized magnons within the flat bands.@xcite the flat band with the eigenenergy @xmath46 supports a single magnon trapped within a square plaquette @xmath47 and this one - magnon state corresponding to the quantum spin numbers @xmath48 has the lowest energy in the highly frustrated regime @xmath49 ( see fig .
[ fig3 ] ) . ) of the spin-@xmath0 heisenberg octahedral chain for four different values of the interaction ratio : ( a ) @xmath50 ; ( b ) @xmath51 ; ( c ) @xmath52 ; ( d ) @xmath53.,scaledwidth=50.0% ]
the many - magnon states of the spin-@xmath0 heisenberg octahedral chain can be subsequently obtained by filling in square plaquettes with the localized magnons of the type ( [ om1 ] ) .
the eigenstates involving @xmath54 independent one - magnon states ( [ om1 ] ) trapped on square plaquettes have the energy @xmath55 , which implies the following value of the saturation field @xmath56 in the highly frustrated region @xmath49 .
in addition , it can be easily verified that the lowest - energy state in the parameter space @xmath49 and @xmath57 is the many - magnon state with the highest possible number ( @xmath58 ) of the independent localized magnons ( [ om1 ] ) on the square plaquettes @xmath59 the localized - magnon ground state ( [ lm ] ) should manifest itself in a zero - temperature magnetization curve as the intermediate three - fifth plateau restricted to the field range @xmath60 . in the highly frustrated parameter space @xmath49
the concept of localized magnons @xcite can also be adapted for a rigorous description of low - temperature thermodynamics of the spin-@xmath0 heisenberg octahedral chain . under this circumstance , the many - magnon states constructed from the lowest - energy one - magnon state ( [ om1 ] ) are the most relevant low - lying states in the high - field region @xmath61 , while the many - magnon states including the localized two - magnon state ( [ st ] ) are the most important low - lying states in the low - field region @xmath62 .
the low - temperature thermodynamics of the spin-@xmath0 heisenberg octahedral chain can be accordingly reformulated as a two - component lattice - gas model , since each square plaquette can host at most one localized one - magnon state ( [ om1 ] ) represented by the first kind of particles with the chemical potential @xmath63 or one localized two - magnon state ( [ st ] ) represented by the second kind of particles with the chemical potential @xmath64 .
the chemical potentials @xmath65 and @xmath66 of two species of the particles are determined by an energy difference between the fully polarized ferromagnetic state ( vacuum ) and the respective localized magnon state ( either one - magnon or two - magnon state ) . by introducing the occupation numbers @xmath67 and @xmath68 for the two species of the particles
the overall energy of the many - magnon states is given by the classical lattice - gas hamiltonian @xmath69 the partition function of the spin-@xmath0 heisenberg octahedral chain accounting for all available many - magnon states constituted from the lowest - energy one - magnon ( [ om1 ] ) and two - magnon ( [ st ] ) states then follows from the formula @xmath70 where @xmath71 , @xmath72 is boltzmann s constant , @xmath73 is the absolute temperature and the prefactor @xmath74 establishes a hard - core constraint for both kinds of the particles as each square plaquette can host at most one one - magnon state ( [ om1 ] ) or one two - magnon state ( [ st ] ) or should be kept empty provided that a square plaquette is fully polarized .
the helmholtz free energy per spin can be calculated from the relation @xmath75 the helmholtz free energy ( [ lmgfe ] ) can be utilized for a calculation of the magnetization per spin @xmath76 and the specific heat per spin @xmath77 it is worthwhile to remark that the derived expressions for the helmholtz free energy ( [ lmgfe ] ) , magnetization ( [ lmmag ] ) and specific heat ( [ lmsh ] ) provide valuable description of low - temperature thermodynamics just in the highly frustrated parameter space @xmath49 .
one of the most essential features of the spin-@xmath0 heisenberg octahedral chain is being a local conservation of the total spin on square plaquettes , which directly follows from a validity of the commutation relation @xmath78 = 0 $ ] between the hamiltonian ( [ ham ] ) and the square of spin operator @xmath79 .
it is therefore convenient to rewrite the zero - field part of the hamiltonian ( [ ham ] ) in terms of the total spin operator @xmath80 of the square plaquette and two auxiliary spin operators @xmath81 and @xmath82 related to the spin pairs from opposite corners of square plaquettes ( see figs .
[ fig1 ] and [ fig2 ] ) @xmath83 the effective hamiltonian ( [ hamlcl ] ) evidently corresponds to the ferrimagnetic mixed spin-@xmath84 heisenberg chains with some shift of energy eigenvalues due to quantum spin numbers @xmath20 , @xmath21 and @xmath22 , whereas the quantum number determining the total spin on a square plaquette may achieve three different values @xmath26 , @xmath85 and @xmath86 .
hence , it follows that the ground state of the spin-@xmath0 heisenberg octahedral chain can be found from the lowest - energy eigenstates of the effective hamiltonian ( [ hamlcl ] ) by assuming all possible combinations of the involved quantum spin numbers . on assumption
that the translational period of a ground state is not broken one arrives just at three effective hamiltonians corresponding to the fragmentized ( paramagnetic ) mixed spin-@xmath0 and spin-@xmath87 system @xmath88 the ferrimagnetic mixed spin-@xmath89 heisenberg chain @xmath90 and the ferrimagnetic mixed spin-@xmath91 heisenberg chain @xmath92 the lowest - energy eigenvalues of the effective hamiltonians ( [ ham120])-([ham122 ] ) of the spin-@xmath0 heisenberg octahedral chain then readily follow from @xmath93 here , @xmath94 denotes the lowest - energy eigenvalue per spin of the mixed spin-@xmath95 heisenberg chain with the unit coupling constant and the total number of @xmath96 spins in each sector with the @xmath24-component of the total spin @xmath97 .
the lowest - energy eigenvalue @xmath98 apparently corresponds to the monomer - tetramer ground state ( [ mt ] ) with the paramagnetic character of the monomeric spins and all square plaquettes in the singlet - tetramer state ( [ st ] ) , which is the true ground state for @xmath8 and @xmath9 as exemplified by the variational method presented in sect .
[ vm ] .
however , one can not exclude neither the possibility that the period of a ground state is spontaneously broken and the quantum spin number @xmath20 determining the total spin of square plaquette varies along the effective mixed spin-@xmath84 heisenberg chain .
we have therefore took into account the possible doubling of unit cell by considering another effective hamiltonian of the ferrimagnetic mixed spin-@xmath99 heisenberg chain @xmath100 which assumes a regular alternation of the total spin @xmath101 and @xmath102 on odd and even square plaquettes .
another possible lowest - energy eigenvalue of the spin-@xmath0 heisenberg octahedral chain may thus follow from the formula @xmath103 where @xmath104 denotes the lowest - energy eigenvalue per spin of the mixed spin-@xmath99 heisenberg chain with the unit coupling constant and the total number of @xmath96 spins in each sector with the @xmath24-component of the total spin @xmath97 .
the lowest - energy eigenvalues @xmath105 , @xmath106 and @xmath104 of all aforedescribed effective mixed - spin heisenberg chains with the total number of 120 spins ( @xmath107 ) were calculated for all available sectors with the @xmath24-component of the total spin @xmath97 by means the numerical dmrg method when adapting the subroutine from algorithms and libraries for physics simulations ( alps ) project.@xcite it should be pointed out that the obtained numerical dmrg data correspond to the spin-@xmath0 heisenberg octahedral chain with @xmath107 unit cells , i.e. @xmath108 spins .
last but not least , the regular alternation of the total spin of square plaquettes @xmath109 and @xmath110 leads to the effective hamiltonian of the ferrimagnetic mixed spin-@xmath111 heisenberg chain , whose lowest - energy eigenstate can be found on analytical grounds because of a fragmentation at even square plaquettes in the singlet - tetramer state ( [ st ] ) with @xmath110 . owing to this fact
, the effective mixed spin-@xmath111 heisenberg chain decomposes into a set of the mixed spin-(@xmath112 ) heisenberg trimers separated from each other by the non - magnetic spin-@xmath87 monomers .
it can be easily verified that the lowest - energy eigenstate of the fragmentized mixed spin-@xmath111 heisenberg chain is the singlet tetramer - hexamer state @xmath113 in above , the former state vector refers to the spin-@xmath0 heisenberg octahedron in a singlet hexamer state @xmath114 while the latter state vector refers to the spin-@xmath0 heisenberg square in the singlet - tetramer state explicitly given by eq .
( [ st ] ) .
the singlet tetramer - hexamer state has a spontaneously broken symmetry on behalf of a regular alternation of singlets , which are being alternatively formed on octahedrons ( hexamers ) and square plaquettes ( tetramers ) as schematically illustrated in fig .
[ fig4 ] . ) .
thick ( black ) ovals represent singlet states of tetramers and hexamers given by eqs .
( [ st ] ) and ( [ sh ] ) , respectively.,scaledwidth=45.0% ] to avoid a danger of overlooking some higher - period ground state(s ) of the spin-@xmath0 heisenberg octahedral chain we have employed the numerical ed method based on the lanczos algorithm for the finite - size chains with @xmath115 spins ( @xmath116 unit cells ) imposing periodic boundary conditions .
the ed data can be regarded as a useful benchmark for the numerical data obtained from dmrg simulations of the effective ferrimagnetic mixed - spin heisenberg chains , because any substantial discrepancy between these numerical results would indicate that some higher - period quantum ground state was disregarded .
the ed data for various chain lengths show very small finite - size effects and fit well to the dmrg results , see below . beside this
, we have also performed the full ed of the finite - size spin-@xmath0 heisenberg octahedral chain with up to @xmath117 spins ( @xmath118 unit cells ) in order to verify reliability of the developed localized - magnon approach for a description of the low - temperature thermodynamics in the highly frustrated parameter space @xmath49 . to this end , we have adapted for the full ed calculations the subroutines from the spinpack project.@xcite
in this section , we will perform a comprehensive analysis of the most interesting results for the ground state , magnetization process and low - temperature thermodynamics of the spin-@xmath0 heisenberg octahedral chain .
let us begin with the analysis of ground state at zero magnetic field .
the zero - field ground - state phase diagram is schematically depicted in fig .
[ fig5 ] and it totally involves five different ground states ( three quantum ferrimagnetic states , a singlet tetramer - hexamer state and a monomer - tetramer state ) depending on a relative strength of two considered coupling constants . at small values of @xmath119 the ground state of the spin-@xmath0 heisenberg octahedral chain can be described as the quantum ferrimagnetic state of the effective mixed spin-(@xmath120 ) heisenberg chain @xcite with the energy eigenvalue ( [ ee122 ] ) .
another quantum ferrimagnetic ground state with a doubled period of the magnetic unit cell relates to the lowest - energy eigenstate ( [ eeeffd ] ) of the effective mixed spin-(@xmath121 ) heisenberg chain , which emerges just in a relatively narrow parameter region @xmath122 .
the last quantum ferrimagnetic ground state of the spin-@xmath0 heisenberg octahedral chain stems from the lowest - energy eigenstate ( [ ee121 ] ) of the effective mixed spin-(@xmath123 ) heisenberg chain , which has the lowest energy in the parameter region @xmath124 .
the bipartite nature of the effective mixed spin-(@xmath120 ) and spin-(@xmath123 ) heisenberg chains implies that two related quantum ground states can be identified with the conventional ferrimagnetic phases of lieb - mattis type.@xcite the same conclusion could be also inferred for the third quantum ferrimagnetic ground state even though the four - sublattice character of the effective mixed spin-(@xmath121 ) heisenberg chain precludes the simple argumentation on the grounds of lieb - mattis theorem.@xcite apart from the three aforementioned quantum ferrimagnetic ground states one also encounters two quantum ground states underlying fragmentation at square plaquettes in the singlet - tetramer state ( [ st ] ) .
the singlet tetramer - hexamer ground state ( [ sht ] ) with a spontaneously broken symmetry ( fig . [ fig4 ] ) is the lowest - energy eigenstate of the spin-@xmath0 heisenberg octahedral chain at moderate values of the coupling constants @xmath125
. finally , the ground state of the spin-@xmath0 heisenberg octahedral chain is the fully fragmentized monomer - tetramer state ( [ mt ] ) whenever the intra - plaquette coupling is at least twice as strong as the monomer - plaquette coupling @xmath126 . heisenberg octahedral chain .
the numbers in square brackets determine within a given ground state the total spin on monomeric sites and square plaquettes .
note that the ferrimagnetic phase ( 1/2 - 2 - 1/2 - 1 ) and the monomer - tetramer phase ( 1/2 - 1 - 1/2 - 0 ) break the translational symmetry.,scaledwidth=68.0% ] it is quite clear that the three quantum ferrimagnetic ground states related to the lowest - energy eigenstates of the effective mixed spin-(@xmath120 ) , spin-(@xmath121 ) and spin-(@xmath123 ) heisenberg chains should be manifested in zero - temperature magnetization curves as intermediate plateaus at @xmath127 , @xmath128 and @xmath129 of the saturation magnetization , respectively .
in addition , the singlet tetramer - hexamer ground state ( [ sht ] ) should be responsible for a zero magnetization plateau , while the monomer - tetramer state ( [ mt ] ) affords another @xmath129-plateau state due to a full polarization of the monomeric spins .
the overall ground - state phase diagram of the spin-@xmath0 heisenberg octahedral chain elucidating the effect of external magnetic field is displayed in fig .
[ fig6 ] in the @xmath130 plane .
it is evident that the highly - frustrated parameter region @xmath126 of the ground - state phase diagram is fully consistent with the rigorous theoretical predictions of the monomer - tetramer state ( [ mt ] ) and the localized - magnon state ( [ lm ] ) gained in sects .
[ vm ] and [ lmgs ] by making use of the variational procedure and localized - magnon approach .
moreover , one finds that the intermediate 3/5-plateau due to the localized - magnon state ( [ lm ] ) with a single magnon trapped on each square plaquette can alternatively be interpreted as the saturated state of the effective ferrimagnetic mixed spin-(@xmath123 ) heisenberg chain .
heisenberg octahedral chain in the @xmath130 plane .
a thick line schematically shows the quantum ferrimagnetic ground state , which originates from the effective mixed spin-(@xmath121 ) heisenberg chain and is stable in a very narrow interval of the magnetic fields.,scaledwidth=45.0% ] on the other hand , the ground - state phase diagram of the spin-@xmath0 heisenberg octahedral chain is much more intricate in the parameter region @xmath131 because of presence of two different spin - liquid ground states with short - range correlations but without any spontaneously broken symmetry.@xcite if the relative strength of the coupling constants is sufficiently weak , i.e. @xmath119 , the ground state of the spin-@xmath0 heisenberg octahedral chain entirely follows from the lowest - energy eigenstates ( [ ee122 ] ) of the effective mixed spin-(@xmath120 ) heisenberg chain .
consequently , the intermediate 3/5-plateau due to the lieb - mattis ferrimagnetic ( @xmath132 ) ground state breaks down at a field - driven quantum critical point towards the gapless spin - liquid ( @xmath132 ) ground state .
the similar scenario can be detected in the parameter space @xmath122 except that the tiny 2/5-plateau emerges at low enough magnetic fields due to the quantum ferrimagnetic ( @xmath133 ) ground state stemming from the lowest - energy eigenstate ( [ eeeffd ] ) of the effective mixed spin-(@xmath121 ) heisenberg chain . in the parameter region
@xmath124 the lieb - mattis ferrimagnetic ( @xmath134 ) ground state arising out from the effective mixed spin-(@xmath123 ) heisenberg chain is responsible for the extra 1/5-plateau at sufficiently low magnetic fields .
last but not least , the zero magnetization plateau reflecting the singlet tetramer - hexamer ground state ( [ sht ] ) emerges at moderate values of the coupling constants @xmath125 . within this parameter region , the 1/5-plateau emerging above the zero magnetization plateau either corresponds to the lieb - mattis quantum ferrimagnet ( @xmath134 ) or the monomer - tetramer phase ( [ mt ] ) depending on whether a relative size of the coupling constants is smaller or greater than the threshold value @xmath135 .
analogously , the intermediate 3/5-plateau is either due to the lieb - mattis quantum ferrimagnet ( @xmath132 ) or the localized - magnon ground state ( [ lm ] ) depending on whether a relative strength of the coupling constants is smaller or greater than the threshold value @xmath136 .
finally , another gapless spin - liquid ( @xmath134 ) ground state can be found in the parameter region @xmath137 and @xmath138 on behalf of the lowest - energy eigenstate of the effective mixed spin-(@xmath123 ) heisenberg chain .
it is worthwhile to remark that all displayed phase boundaries represent discontinuous quantum phase transitions with exception of three horizontal borders related to continuous quantum phase transitions . to illustrate the great diversity of possible scenarios of the magnetization process of the spin-@xmath0 heisenberg octahedral chain we have depicted in fig .
[ fig7 ] a few typical examples of zero - temperature magnetization curves .
the magnetization data obtained from the dmrg simulations of the effective mixed - spin heisenberg chains are shown in fig .
[ fig7 ] by thick solid lines and for a comparison , we have also plotted the magnetization data from ed method by thin broken lines . as one can see , there is in general perfect coincidence between the zero - temperature magnetization curves from dmrg and ed calculations as far as the height and width of all intermediate magnetization plateaus is concerned .
the only noticeable difference is thus attributable to a finite - size effect of the ed method , which is most clearly seen within both gapless spin - liquid ground states manifested through a continuous rise of the magnetization with the magnetic field .
it should be also stressed that the displayed magnetization curves corroborate correctness of the established ground - state phase diagram .
the magnetization curve shown in fig . [ fig7](a ) demonstrates two continuous quantum phase transitions , which occur at a rise and fall of the intermediate 3/5-plateau .
the magnetization curve plotted in fig .
[ fig7](b ) verifies uprise of 1/5- and 2/5-plateau , which is achieved upon small strengthening of the relative ratio @xmath45 .
the emergence of zero magnetization plateau on account of the singlet tetramer - hexamer ground state ( [ sht ] ) can be seen in the magnetization curve shown in fig .
[ fig7](c ) .
furthermore , fig .
[ fig7](d ) illustrates the magnetization curve with a change in character of 1/5-plateau as well as disappearance of the subtle 2/5-plateau .
the magnetization curve plotted in fig .
[ fig7](e ) demonstrates change over the character of 3/5-plateau as well as two emergent magnetization jumps towards the spin - liquid ground states .
finally , fig .
[ fig7](f ) displays typical magnetization curve in the highly frustrated region @xmath139 involving three abrupt magnetization jumps .
next , let us proceed to a discussion of the most interesting results for the low - temperature thermodynamics of the spin-@xmath0 heisenberg octahedral chain , which were obtained within the framework of the localized - magnon approach elaborated in sect .
it is worthwhile to recall , however , that the localized - magnon theory provides reasonable results only in the highly - frustrated parameter region @xmath126 , where the many - magnon states constructed from the lowest - energy one - magnon ( [ om1 ] ) and two - magnon ( [ st ] ) localized states are the most relevant ones for a proper description of the low - temperature thermodynamics . for the sake of comparison and verification , the magnetization and specific heat of the spin-@xmath0 heisenberg octahedral chain obtained from the localized - magnon approach
are plotted in fig .
[ fig8 ] along with the analogous data for the finite - size chain of @xmath140 spins obtained within the full ed method .
let us first emphasize two interesting features , namely ( i ) the extra low - temperature peak in the specific heat for magnetic fields slightly below the saturation field , cf .
[ fig8](d ) , and , ( ii ) the noticeable increase of the magnetization with growing temperature for @xmath141 and @xmath142 , cf . fig .
[ fig8](b ) .
both unconventional features are related to the flat bands in the one - magnon excitations , cf .
[ fig3 ] .
it can be understood from fig .
[ fig8 ] that the localized - magnon theory provides a plausible description of the low - temperature magnetization and thermodynamics whenever the external magnetic field drives the investigated spin system above a midpoint of the lowest intermediate 1/5-plateau appearing due to the monomer - tetramer ground state ( [ mt ] ) . under this circumstance ,
the magnetization calculated according to eq .
( [ lmmag ] ) exhibits for @xmath52 a perfect agreement with the ed data on assumption that the temperature is below @xmath143 . compared to this , the derived formula ( [ lmsh ] ) for the specific heat affords a reliable description of the numerical ed data for the same value of the interaction ratio @xmath52 just for much smaller temperatures @xmath144 .
it should be pointed out , however , that the height and position of low - temperature maximum of the heat capacity is in an excellent accordance with the numerical ed data provided that the magnetic field is fixed sufficiently close to the saturation field or the field - driven transition between @xmath129- and @xmath127-plateaus .
the present work deals with the ground - state phase diagram , the magnetization process and the low - temperature thermodynamics of the spin-@xmath0 heisenberg octahedral chain , which has been treated by means of various analytical and numerical techniques .
it has been demonstrated that the highly - frustrated parameter region of the ground - state phase diagram can be rigorously found on the grounds of the variational principle and the localized - magnon approach , which provide an exact evidence for the monomer - tetramer ( [ mt ] ) and localized - magnon ( [ lm ] ) ground states at low and high magnetic fields , respectively . on the other hand ,
the remaining part of the ground - state phase diagram was established through the numerical data gained from dmrg simulations of the effective mixed - spin heisenberg chains , which were additionally corroborated by ed data exploiting lanczos algorithm .
it has been verified that the spin-@xmath0 heisenberg octahedral chain exhibits in this parameter space an unexpected diversity of intriguing ground states including three different quantum ferrimagnetic phases , two spin - liquid phases as well as the unconventional singlet tetramer - hexamer phase ( [ sht ] ) .
the notable diversity of available quantum ground states can also be regarded as a primary reason for astonishing versatility of zero - temperature magnetization curves . as a matter of fact
, the magnetization curves of the spin-@xmath0 heisenberg octahedral chain may involve intermediate plateaus at zero , one - fifth , two - fifth and three - fifth of the saturation magnetization in addition to two gapless spin - liquid regimes with continuously varying magnetization .
the field - driven quantum phase transition between individual ground states can therefore have either character of a first - order phase transition accompanied with a discontinuous magnetization jump or of a second - order phase transition accompanied with a continuous rise of the magnetization .
last but not least , we have developed the modified localized - magnon theory accounting for the lowest - energy one - magnon and two - magnon states , which are essential for a design of the many - magnon states relevant for a proper description of low - temperature thermodynamics of the spin-@xmath0 heisenberg octahedral chain .
the validity and exactness of the developed localized - magnon approach in the highly - frustrated parameter region has been verified through a direct comparison with the full ed data .
it has been found that the localized - magnon theory gives plausible estimate of thermodynamic quantities at low up to moderate temperatures whenever the magnetic field drives the investigated quantum spin system above a midpoint of the intermediate 1/5-plateau due to the monomer - tetramer ( [ mt ] ) phase .
our future goal is to extend a validity of the localized - magnon approach down to zero field , because a steep variation of the magnetization at low magnetic fields is of great potential applicability for low - temperature magnetic refrigeration achieved through the enhanced magnetocaloric effect.@xcite this work was financially supported by the grant of the ministry of education , science , research and sport of the slovak republic under the contract no .
vega 1/0043/16 and by the grant of the slovak research and development agency under the contract no .
apvv-14 - 0073 .
o.d . and j.r .
acknowledge the support by the deutsche forschungsgemeinschaft ( project ri615/21 - 2 ) .
was partially supported by project ff-30f ( no .
0116u001539 ) from the ministry of education and science of ukraine .
100 s. sachdev , _ quantum phase transitions _ , cambridge university press , 2011 .
haldane , phys .
lett . a * 93 * ( 1983 ) 464 ; phys . rev
* 50 * , 1153 ( 1983 ) .
s. miyahara , k. ueda , j. phys . : condens .
matter * 15 * , r327 ( 2003 ) .
s. miyahara , in _ introduction to frustrated magnetism _ , edited by c. lacroix , ph .
mendels , f. mila ( springer , heidelberg , 2011 ) , pp . 513536 .
anderson , mater .
* 8 * , 153 ( 1973 ) . c. lhuillier , g. misguich , lect
. notes phys .
* 595 * , 161 ( 2002 ) .
l. balents , nature 464 , 199 ( 2010 ) .
g. misguich , _ quantum spin liquids and fractionalization _ , in introduction to frustrated magnetism , eds . : c. lacroix , p. mendels , f. mila , springer , 2011 , pp .
diep , _ frustrated spin systems _ , world scientific , singapore , 2004 . ,
lecture notes in physics * 645 * , edited by u. schollwck , j. richter , d. j. j. farnell , and r. f. bishop ( springer , berlin , 2004 ) . c. lacroix , f. mila , ph .
mendels , _ _ introduction to frustrated magneti__sm , springer , 2011 .
majumdar , d.k .
ghosh , j. math . phys . * 10 * , 1388 ( 1969 ) ; _ ibid . _ * 10 * , 1399 ( 1969 ) .
majumdar , j. phys .
c : solid state phys . * 3 * , 911 ( 1970 ) .
shastry , b. sutherland , physica b+c * 108 * , 1069 ( 1981 ) .
gelfand , phys .
b * 43 * , 8644 ( 1991 ) .
a. honecker , f. mila , m. troyer , eur .
j. b * 15 * , 227 ( 2000 ) .
v. ravi chandra , n. surendran , phys .
b * 74 * , 024421 ( 2006 ) .
a. honecker , s. wessel , r. kerkdyk , t. pruschke , f. mila , b. normand , phys .
b * 93 * , 054408 ( 2016 ) .
k. takano , k. kubo , h. sakamoto , j. phys . : condens .
matter * 8 * , 6405 ( 1996 ) .
k. morita , n. shibata , j. phys .
* 85 * , 033705 ( 2016 ) .
y. hirose , a. oguchi , y. fukumoto , j. phys .
* 85 * , 094002 ( 2016 ) .
y. hirose , a. oguchi , y. fukumoto , j. phys .
. jpn . * 86 * , 014002 ( 2017 ) .
j. schulenburg , a. honecker , j. schnack , j. richter , h .- j .
schmidt , phys .
lett . * 88 * , 167207 ( 2002 ) .
zhitomirsky , h. tsunetsugu , prog .
160 * , 361 ( 2005 ) .
o. derzhko , j. richter , eur .
j. b * 52 * , 23 ( 2006 ) .
o. derzhko , j. richter , m. maksymenko , int . j. mod .
b * 29 * , 1530007 ( 2015 ) .
h. tasaki , j. phys . : condens .
matter * 10 * , 4353 ( 1998 ) .
h. tasaki , prog .
* 99 * , 489 ( 1998 ) .
jo , j. guzman , c.k .
thomas , p. hosur , a. vishwanath , d.m .
stamper - kurn , phys .
* 108 * , 045305 ( 2012 ) .
s. mukherjee , r.r .
thomson , opt .
* 40 * , 5443 ( 2015 ) . s. weimann , l. morales - inostroza , b. real , c. cantillano , a. szameit , r.a .
vicencio , opt .
41 * , 2414 ( 2016 ) .
d. leykam , s. flach , o. bahat - treidel , a.s .
desyatnikov , phys .
b * 88 * , 224203 ( 2013 ) .
s. flach , d. leykam , j.d .
bodyfelt , p. matthies , a.s .
desyatnikov , epl * 105 * , 10001 ( 2014 ) .
liu , c .- m . hung , b. kumar santra , j .- ch .
wang , h .-
lin , inorg .
chem . * 42 * , 8551 ( 2003 ) . s .-
xiang , s .-
hu , j .- j .
zhang , x .-
wu , j .- q .
. j. inorg .
* 2005 * , 2706 ( 2005 ) .
zhao , j. wang , d .- x .
bao , s. xiang , d .- b .
luo , inorg .
commun . * 62 * , 77 ( 2015 ) .
x. gao , s. he , c. zhang , c. du , x. chen , w. xing , s. chen , a. clayborne , w. chen , adv .
sci . * 2016 * , 1600126 ( 2016 ) . c. daniel , h. hart , j. am .
soc . * 127 * , 13978 ( 2005 ) . c. daniel , h. hart , j. am .
soc . * 131 * , 5101 ( 2009 ) .
k. tsuge , h. imoto , t. saito , bull .
jpn . * 69 * , 627 ( 1996 ) .
m. hong , z. huang , x. lei , g. wei , b. kang , h. liu , polyhedron * 10 * , 927 ( 1991 ) .
brechin , m.j .
knapp , j.c .
huffman , d.n .
hendrickson , g. christou , inorg .
acta * 297 * , 389 ( 2000 ) .
m. murugesu , k.a .
abboud , g. christou , polyhedron * 23 * , 2779 ( 2004 ) .
g. arom , m.j .
knapp , j .-
claude , j.c .
huffman , d.n .
hendrickson , g. christou , j. am .
soc . * 121 * , 5489 ( 1999 ) .
stamatatos , k.v .
pringouri , k.a .
abboud , g. christou , polyhedron * 28 * , 1624 ( 2009 ) .
t. saito , n. yamamoto , t. nagase , t. tsuboi , k. kobayashi , t. yamagata , h. imoto , k. unoura , inorg .
* 29 * , 764 ( 1990 ) .
m. cindric , g. pavlovic , v. vrdoljak , b. kamenar , polyhedron * 19 * , 1471 ( 2000 ) . t.c .
zietlow , w.p .
schaefer , b. sadeghi , n. hua , h.b .
gray , inorg . chem . * 25 * , 2195 ( 1986 ) .
crawford , j.r .
long , inorg . chem . *
40 * , 3456 ( 2001 ) .
eady , p.f .
jackson , b.f.g .
johnson , j. lewis , m.c .
malatesta , m. mcpartlin , w.j.h .
nelson , j. chem .
soc . : dalton trans . *
1980 * , 383 ( 1980 ) .
pergola , l. garlaschelli , s. martinengo , f. demartin , m. manassero , n. masciocchi , j. chem .
soc . : dalton trans . *
1988 * , 2307 ( 1988 ) .
b. peric , d. jozic , p. planinic , n. brnicevic , g. giester , j. solid state chem .
* 182 * , 2443 ( 2009 ) .
m. mambrini , j. trbosc , f. mila , phys .
b * 59 * , 13806 ( 1999 ) .
o. rojas , f.c .
alcaraz , phys .
b * 67 * , 174401 ( 2003 ) .
m. maksymenko , o. derzhko , j. richter , eur .
j. b * 84 * , 397 ( 2011 ) .
b. bauer , l.d .
carr , h.g .
evertz , a. feiguin , j. freire , s. fuchs , l. gamper , j. gukelberger , e. gull , s. guertler , a. hehn , r. igarashi , s.v .
isakov , d. koop , p.n .
ma , p. mates , h. matsuo , o. parcollet , g. pawlowski , j.d .
picon , l. pollet , e. santos , v.w .
scarola , u. schollwck , c. silva , b. surer , s. todo , s. trebst , m. troyer , m.l .
wall , p. werner , s. wessel , j. stat .
: theor . exp .
* * 201**1 , p05001 ( 2011 ) .
j. richter , j. schulenburg , eur .
j. b * 73 * , 117 ( 2010 ) .
http://www-e.uni-magdeburg.de/jschulen/spin/ n.b .
ivanov , phys .
b * 57 * , 14024r ( 1998 ) .
e. lieb , d. mattis , j. math .
phys . * 3 * , 749 ( 1962 ) .
zhitomirsky , phys .
b * 67 * , 104421 ( 2003 ) . m.e .
zhitomirsky , a. honecker , j. stat .
mech . : theory exp .
p07012 ( 2004 ) .
j. schnack , r. schmidt , j. richter , phys .
b * 76 * , 054413 ( 2007 ) .
j. schnack , c. heesing , eur .
j. b * 86 * , 46 ( 2013 ) . | the spin-@xmath0 heisenberg octahedral chain with regularly alternating monomeric and square - plaquette sites is investigated using various analytical and numerical methods : variational technique , localized - magnon approach , exact diagonalization ( ed ) and density - matrix renormalization group ( dmrg ) method .
the model belongs to the class of flat - band systems and it has a rich ground - state phase diagram including phases with spontaneously broken translational symmetry .
moreover , it exhibits an anomalous low - temperature thermodynamics close to continuous or discontinuous field - driven quantum phase transitions between three quantum ferrimagnetic phases , tetramer - hexamer phase , monomer - tetramer phase , localized - magnon phase and two different spin - liquid phases .
if the intra - plaquette coupling is at least twice as strong as the monomer - plaquette coupling , the variational method furnishes a rigorous proof of the monomer - tetramer ground state in a low - field region and the localized - magnon approach provides an exact evidence of a single magnon trapped at each square plaquette in a high - field region . in the rest of parameter space
we have numerically studied the ground - state phase diagram and magnetization process using dmrg and ed methods .
it is shown that the zero - temperature magnetization curve may involve up to four intermediate plateaus at zero , one - fifth , two - fifth and three - fifth of the saturation magnetization , while the specific heat exhibits a striking low - temperature peak in a vicinity of discontinuous quantum phase transitions . |
it is nearly ten years since the first gravitational microlensing event was detected toward the galactic bulge ( udalski et al .
1993 ) . since then
, microlensing has proved to be a useful tool for many astrophysical applications ( for a review of galactic microlensing , see paczyski 1996 ) . in the past , however , studies of microlensing statistics have been limited by the small number of detected events .
this situation is improving as new microlensing projects begin operation , e.g. , the ogle - iii project , which is hoping to detect up to 1000 per year . two ways that microlensing statistics can be used to investigate galactic models are through the study of the observed microlensing optical depth and the distribution of event time - scales .
the optical depth toward the galactic bulge has been studied by many collaborations ( for example , alcock et al . 2000 and sumi et al .
2002 ) and the resulting values are significantly larger than predicted estimates ( see , for example , binney , bissantz & gerhard 2000 ; evans & belokurov 2002 ; klypin , zhao & somerville 2002 ) , although recent work has questioned the significance of this discrepancy ( popowski 2002 ; afonso et al .
2003 ) .
the distribution of event time - scales , which can be used to investigate the mass spectrum of lensing objects ( e.g. han & gould 1996 ; peale 1998 ) , also seems to disagree with predicted estimates ; it has long been suspected that microlensing searches appear to be identifying an unexpectedly large proportion of long - duration events .
an important early study into the distribution of microlensing time - scales was carried out by han & gould ( 1996 ; see also zhao , rich & spergel 1996 ) .
they found that in one - year of microlensing observations from the ogle and macho collaborations 8% of events had time - scale greater than 70 days , whereas the largest fraction predicted by their theoretical models was 2% .
bennett et al .
( 2002a ) analysed a more recent distribution of events timescales from macho data and found a similar excess of long duration events .
although no systematic analysis of event time - scales has been performed for the full 4-year ogle - ii catalogue , a preliminary investigation using the first 3-years of data also indicates the existence of an excess ( udalski et al .
2000 ) . to date
, two microlensing events have been identified with time - scales greater than one year ( ogle-1999-bul-32/macho-99-blg-22 , mao et al .
2002 , bennett et al . 2002b ; ogle-1999-bul-19 , smith et al .
we will introduce a third event , 5 , in this paper .
it has been proposed that some of these extreme long - duration events could be due to lensing by massive stellar remnants , e.g. black holes ( agol et al .
2002 ; bennett et al .
2002a ) , although it is unlikely that this hypothesis could account for every long - duration event , for example ogle-1999-bul-19 ( smith et al .
2002 ) .
in this paper we present the analysis of event 5 .
we begin by describing the observational data ( section [ data ] ) , before proceeding to fit the event with both the standard and parallax microlensing models ( section [ model ] ) .
we then investigate whether any constraints can be placed on the lens mass ; we first attempt to utilise finite source size effects ( section [ fss ] ) , before considering more general arguments based on the relative transverse velocity of the lens and limits on the lens brightness ( section [ lhood ] ) . in section [ fss ]
we also discuss the possible location and spectral type of the source star .
section [ disc ] contains a brief discussion regarding the nature of event 5 and the possibility that there are an excess of long duration microlensing events in the ogle - ii catalogue .
this section also mentions possible scientific returns from such long duration events .
we conclude with a summary ( section [ conc ] ) .
the event 5 was identified during the second phase of the ogle project ( in which it was named bul_sc5 244353 ; udalski et al .
2000 ) toward the galactic bulge .
the observations were carried out with the 1.3 m warsaw telescope at the las campanas observatory , chile , which is operated by the carnegie institution of washington .
the instrumentation of the telescope and ccd camera are described in detail by udalski , kubiak , & szymaski(1997 ) .
the position of the source is ra=17:50:36.09 and dec=@xmath230:01:46.6 ( j2000 ) , corresponding to galactic coordinates @xmath3 and @xmath4 .
this event is also included in the catalogue of woniak et al .
( 2001 ) , which employs the difference image analysis method of data reduction ( alard & lupton 1998 ) .
this method generally appears to result in greater accuracy compared to the classical point spread function approach ( schechter , mateo & saha 1993 ) , and so we have chosen to use the difference image analysis data in the following analysis .
unfortunately , this event was only detected after the peak in magnification had already occurred , which means that less than half of the full light curve is available for analysis .
the baseline magnitude for the source is given by @xmath5 mag mag , but the 3-year classical point spread function data have a baseline of @xmath6 mag .
however , the conclusions regarding the nature of event 5 remain unchanged if the classical point spread function data is used .
] , and the colour is @xmath7 mag . in fig .
[ fig : cmd ] we present the colour - magnitude diagram for the field around 5 . from this figure
it can be seen that the field around 5 is subject to a large amount of interstellar extinction , owing to the fact that the source is located very close to the galactic plane .
it appears that the lensed star is located in the main sequence branch of the colour - magnitude diagram .
we will return to the issue of the possible source location and spectral type in section [ fss ] .
we initially fit the event with the standard microlensing model , which assumes that the observer , lens and ( point ) source all move with constant velocities .
the magnification is given by ( see , for example , paczyski 1986 ) , [ amp ] a(t ) = u^2 + 2 u , u(t ) , where @xmath8 is the impact parameter ( in units of the einstein radius ) and , [ tau ] ( t ) = t - t_0 , with @xmath9 being the time of the closest approach ( i.e. , maximum , magnification ) , and @xmath10 the event time - scale .
the time - scale is defined such that , = v = _ e , where @xmath11 is the lens einstein radius , @xmath12 is the lens transverse velocity relative to the observer - source line of sight , and @xmath13 and @xmath14 are the values of these two quantities projected onto the observer plane .
therefore , @xmath10 corresponds to the time it takes for the lens to move a distance equal to its einstein radius , the einstein _ diameter _ crossing time , i.e. @xmath15 . ] .
the einstein radius projected onto the observer plane is related to the mass according to the following equation , [ eq : re ] _
e = , where @xmath16 is the lens mass , @xmath17 the distance to the source and @xmath18 is the ratio of the distance to the lens and the distance to the source .
this shows the well - known degeneracy inherent in the standard microlensing formalism ; the quantities @xmath14 , @xmath16 and @xmath19 can not be determined uniquely from a given microlensing light curve , even if the source distance is known .
the difference image analysis flux is given by , f(t ) = f_0 ( + 1 ) - f_ref , where @xmath20 is the total baseline flux from the source plus any blended star(s ) , if present , @xmath21 is the ratio of the baseline source flux to the total baseline flux ( i.e. a measure of the blending ) , and @xmath22 is the flux of the reference image .
all the fluxes here are in units of 10adu and can be converted into @xmath23-band magnitudes using the following transformation ( given in woniak et al .
2001 ) , i(t ) = i_ref - 2.5log_10 , where @xmath24 is the @xmath23-band magnitude of the reference image .
note that the reference image is brighter than the baseline magnitude for 5 .
we fit the light curve with the above five parameters for the standard model , i.e. , @xmath9 , @xmath10 , @xmath25 , @xmath20 , @xmath21 .
the best - fit parameters are given in table [ tb : sc5 ] , and the corresponding light curve is shown in fig .
[ lc : sc5 ] .
clearly this model is unable to provide a suitable fit for the data , with the best standard @xmath26 per degree of freedom greater than 5 .
since the duration of this event is longer than one year , the next logical step is to fit the light curve with a model that incorporates the parallax effect ( gould 1992 ) .
this effect , which arises when the earth s motion around the sun is considered , is described in detail in soszyski et al .
( 2001 ) ; see also alcock et al .
( 1995 ) , dominik ( 1998 ) .
it requires two additional parameters , the einstein radius projected onto the observer plane , @xmath27 , and an angle in the ecliptic plane , @xmath28 , describing the orientation of the lens trajectory ( given by the angle between the heliocentric ecliptic @xmath19-axis and the normal to the trajectory ) .
the parallax model produces a drastic reduction in @xmath26 .
the best - fit parameters are given in table [ tb : sc5 ] , and the corresponding light curve is shown in fig .
[ lc : sc5 ] .
the @xmath26 per degree of freedom is 1.23 , which indicates that the parallax model provides a reasonable fit to the data .
we also fit this event with a slight variation on the above parallax model . in the above model
we describe the geometrical properties in the ecliptic plane and then project these quantities into the lens plane . however , since the ecliptic plane intersects the galactic bulge , this can lead to the projection being almost singular .
therefore , to avoid this potential singularity , one can instead take the more conventional approach and describe the geometrical properties in the plane perpendicular to the line - of - sight ( see , for example , dominik 1998 , alcock et al .
1995 ) ; to describe the lens plane coordinate system we form a right - handed set with the @xmath19-axis chosen to correspond to the north ecliptic pole projected onto the lens plane and the @xmath29-axis chosen to be the observer - source line - of - sight , which implies that the @xmath30-axis corresponds to the intersection of the lens plane and the ecliptic plane .
we find that the best - fit parameters are practically identical and therefore we do not present them here .
however , for this coordinate system , the angle describing the relative lens - source trajectory is found to be @xmath31 degrees , where @xmath32 is the angle between the trajectory and the @xmath19-axis ( measured from the positive @xmath19-axis towards the positive @xmath30-axis ) . from table [ tb : sc5
] , the parallax parameters that provide information regarding the lens properties are , = 10.84^+ 0.44_-0.41 au , = 547.6^+22.6_-7.8 days , which implies that the transverse velocity of the lens relative to the source , projected onto the observer plane is , [ eq : v ] = = 34.2^+1.5_-1.9 km s^-1 .
if we convert the direction of @xmath14 from the lens plane ( @xmath33 degrees ) into the plane perpendicular to the line - of - sight to the galactic centre , we find that @xmath14 is directed almost parallel to the galactic plane in the direction of rotation ( @xmath34 degrees , where @xmath35 is measured from the galactic plane towards the north galactic pole ) . in this aspect 5
is similar to the strong parallax events presented in bennett et al .
( 2002a ) , all of which had @xmath36 .
this timescale of 548 days is the second longest time - scale ever identified ( after the event ogle-1999-bul-32/macho-99-blg-22 ; see mao et al .
2001 , bennett et al .
in addition , the value of @xmath27 is also unusually large , which implies that this event could be another black - hole microlens candidate ( c.f .
the three current black - hole microlens candidates , which have @xmath37 ; see agol et al .
2002 ) . using equation ( [ eq : re ] ) ,
the mass of the lens for this event is given by , [ eq : mass ] m = 7.21 m _ ( ) ^2 ( ) , .
a value of @xmath38 mas corresponds to a disk source with @xmath39 and @xmath40 .
there are various approaches that can be employed to constrain the location of the source and lens , and hence the lens mass ; these are considered in section [ mass ] .
to verify the viability of our parallax fit , we proceed to fit the event with a model that incorporates a constant acceleration term , instead of the earth s centripetal acceleration ( see smith , mao & paczyski 2003 ) .
this model is unable to provide a feasible fit for 5 , meaning that the parallactic nature of the deviations appears to be secure .
since the peak magnification is predicted to be greater than 40 , one may suspect that this event could be affected by finite - source effects ( gould 1994 ; witt & mao 1994 ; nemiroff & wickramasinghe 1994 ) .
such effects become apparent when the lens passes sufficiently close to the source , resulting in an invalidation of the assumption that the source is point - like .
however , there is one significant drawback for 5 , namely that there is no coverage for the peak of the light - curve , i.e. the point at which the lens and source are in closest alignment and therefore where the finite - source effects should be most prominent . to implement the finite - source model requires an additional parameter , @xmath41 , which denotes the source radius in units of the lens angular einstein radius . despite the lack of coverage around the peak
, constraints can be placed on @xmath41 lower - limit on @xmath41 ( @xmath42 ) , although we consider this to be unphysical since it would result in a highly unlikely value of @xmath43 .
] , [ 3sig ] _ 0.034 m _ , which corresponds to @xmath44
. therefore we are only able to place a weak lower - limit on the lens mass from finite - source considerations .
this outcome can be understood when one considers the likely values of the parameter @xmath41 . from equation ( [ eq : theta_e ] )
, we can obtain the following relation for @xmath41 , _
= 0.003 ( ) ( ) ( ) ^-1 ( ) ^-1 , where @xmath45 is the physical source radius , which we expect to be @xmath46 if the source for 5 is a main - sequence star .
gould ( 1994 ) showed that the detection of finite - source signatures is only possible provided @xmath47 , where @xmath48 is the peak magnification . for 5
we have @xmath49 , which implies that finite - source signatures are extremely unlikely to be detected .
in addition to the above weak constraint placed on the lens mass from the finite - source analysis , there is other information that we can use from the properties of the best - fit parallax model , i.e. independent of the above finite - source fit .
first , we shall consider the parallax velocity parameter ( @xmath14 ) , and then incorporate our knowledge of the limits on the blended flux . for parallax microlensing events the velocity parameter @xmath14 ( i.e. the transverse velocity of the lens relative to the source , projected onto the observer plane ) , must be sufficiently small so that the earth s orbital motion is able to affect the light curve . for 5 , this transverse velocity is @xmath50 directed almost parallel to the galactic plane in the direction of rotation ( @xmath51 degrees ; see section [ model ] ) .
the velocity vector @xmath52 is related to the transverse velocities of the observer ( i.e. the sun , @xmath53 ) , source ( @xmath54 ) and lens ( @xmath55 ) , where @xmath53 , @xmath54 and @xmath56 are the 2-dimensional velocities perpendicular to the line - of - sight ( i.e. in the lens plane ) , through the equation , [ eq : vproj ] = - _ .
previous studies of parallax microlensing events ( e.g. alcock et al .
1995 ) have used the parameter @xmath52 and equation ( [ eq : vproj ] ) to obtain a likelihood function for @xmath19 ( and hence , from equation [ [ eq : re ] ] , a likelihood function for the lens mass ) , [ eq : lhood ] l(x ; ) _
l(x ) || ( 1-x)^3 f_v , s(*v*_s ) f_v , l([1-x ] [ * v*_+]+x*v*_s)d*v*_s , where @xmath57 and @xmath58 are the lens and source velocity distributions , @xmath59 is the density of lenses at a distance @xmath19 , and all vectors are described in the plane perpendicular to the line - of - sight to the galactic centre ( i.e. two dimensional ) .
a more thorough analysis may be obtained by including a mass function prior ( see , for example , agol et al . 2002 ) .
however , we do not attempt such an analysis here ; for the purposes of this work we follow the approach of bennett et al .
( 2002b ) and assert that the likelihood function represents all of our knowledge about the lens mass and location ( i.e. we select a uniform prior ) , meaning that the above likelihood function can be interpreted as the probability distribution for @xmath19 .
we evaluate this function by assuming that both the lens and source reside in the disk , using density and mass distributions from belokurov & evans ( 2002 ) and taking the sun s two dimensional peculiar velocity to be 8.89 km / s in a direction @xmath60 ( dehnen & binney 1998 ) .
the results of this calculation are presented in figure [ fig : lhood ] . from this figure
, we can see that the distribution for @xmath19 is not very narrow , implying that the lens can take a wide range of masses .
for example , for @xmath61 kpc , this analysis suggests that @xmath62 and hence the lens mass @xmath63 ; moreover , the @xmath64 lower limit on the lens mass is @xmath65 , suggesting that a low mass lens is strongly disfavoured .
additional constraints can be placed on the lens nature by considering the blended flux parameter , @xmath21 .
the best - fit parallax model given in section [ model ] predicts that there is no blending ( i.e. , @xmath66 ) , meaning that all of the flux is being emitted by the lensed source .
this is important because it suggests that the lens may not be a main sequence star , since if this was true one would expect some blending due to the light from the lens , i.e. one would expect @xmath67 .
this possibility can be investigated by considering the limits on the @xmath21 parameter . from the best - fit parallax model
, we can say that the @xmath64 and @xmath68 limits on the lens brightness are @xmath69 and @xmath70 , respectively .
if we model the lens with a main - sequence mass - luminosity relation @xmath71 , we can use this to place a lower limit on @xmath19 through equation ( [ eq : re ] ) .
however , to do this we need to know the source distance , @xmath72 , and also how the extinction varies with distance .
if we again take @xmath73 kpc and assume the previous simple extinction model ( see section [ fss ] ) , we obtain @xmath64 and @xmath68 limits of @xmath74 and @xmath75 , respectively .
this conclusion depends only weakly on the assumptions regarding the extinction and mass - luminosity relation .
these constraints on @xmath19 appear to contradict the above conclusions from the @xmath14 likelihood analysis , which suggested that values of @xmath19 close to 1 are strongly disfavoured due to the low value of @xmath14 .
figure [ fig : lhood ] shows how the @xmath64 limits on the lens brightness compare to the likelihood distribution described in equation ( [ eq : lhood ] ) for three values of @xmath72 .
these constraints appear to be incompatible with the above likelihood analysis ( for example , for @xmath76 kpc , the likelihood analysis gives the probability that @xmath19 satisfies the @xmath64 constraints on the lens brightness to be @xmath77 ) . therefore , we tentatively conclude that the lens for 5 is unlikely to be a main - sequence star , which implies that this event could be a white dwarf ( although this seems unlikely , given that typical white dwarf masses of @xmath78 [ bergeron , ruiz , & leggett 1997 ] are disfavoured from our likelihood analysis ) , a neutron star , or possibly an example of lensing by a stellar mass black hole ( agol et al .
2002 ; bennett et al .
one approach that can be utilised to probe the nature of the lens for 5 is through proper motion analysis .
if the proper motion of the source can be measured , then this can be combined with the parallax model s prediction for the transverse velocity of the lens relative to the source ( i.e. @xmath14 ) to determine the proper motion of the lens .
the importance of this method is that it can be applied even in the case where the lens is not luminous . by analysing one ogle - ii field , sumi , eyer & woniak ( 2003 )
have shown that proper motion measurements can be determined for objects in the ogle - ii catalogue .
preliminary results from the analysis of all ogle - ii fields suggests that the source star for 5 may have a large proper motion ( sumi , private communication ) , although this requires verification as the source star is relatively faint . despite the fact that the light curve for 5 appears to be reasonably well fit by the parallax microlensing model ( see section [ model ] ) ,
it is not inconceivable that the variation in brightness could be due to some phenomena other than microlensing . whilst this ambiguity can affect any microlensing event , for 5 it is particularly severe since data only exist for the declining branch of the light curve
indeed , recently cieslinski et al . ( 2003 ) carried out a search of ogle - ii light curves in an attempt to identify new cataclysmic variables . in this work
they suggested that the shape of the light curve for 5 indicates that this event could be a candidate nova .
however , it is known that microlensing events , unlike many intrinsically variable stars , are expected to be achromatic ( this may not be true for heavily blended events , but for 5 the best - fit model predicts no blended flux ) . for 5
the ogle data only provide colour information for two epochs , namely @xmath79 days , and @xmath80 days .
reassuringly , the @xmath81 colour appears to stay roughly constant for these two epochs , with @xmath82 and @xmath83 , respectively , which supports our belief that 5 is a microlensing event .
an important aspect of 5 that is worth noting is that the above analysis is possible even though data exist for less than half of the microlensing light curve .
despite the fact that observations commenced after the peak , it is still possible to identify the parallactic signatures and gain well - constrained measurements of @xmath27 and @xmath32 .
obviously , this is only possible for 5 due to the high - quality data and the long duration of the event . in future ,
even with initially sparse sampling , long - duration events can provide useful information provided high - quality data are obtained for the latter part of the light curve .
this highlights the importance of real - time microlensing alert systems ( such as the ogle - iii early warning system or the moa transient alert page ) and also the importance of high - quality follow - up observations ( such as those performed by the planet collaboration ) . on the other hand ,
if real - time alert systems can detect events at a suitably early stage , then it could be possible to make direct determinations of the lens mass .
for example , once the eso very large telescope interferometer becomes fully operational , its high sensitivity may enable measurements to be made of the angular separation of the two microlensed images ( delplancke , grski & richichi 2001 ) .
coupled with a measurement of the parallax effect , this would completely break the lens degeneracy and unambiguously determine the mass of the lensing object . in future
this approach could prove to be very important for long - duration microlensing events .
another approach that would benefit from timely follow - up observations is the method of combining the parallax model with finite - source effects . as was shown in section [ fss ] , if firm measurements can be made of both parallax and finite - source effects ,
is it possible to completely break the lens degeneracy and determine the lens mass .
although this was not possible for 5 , it has recently been shown to be feasible ( for example , this approach was applied to the ogle - ii event sc26_2218 , resulting in a tentative lens mass determination [ smith , mao & woniak 2003 ] ) .
if real - time alert systems are able to identify high - magnification long - duration events suitably early , it may be possible to obtain good quality data for the peak of the light curve and hence determine the lens mass . with the addition of 5 , this brings the total number of known events in the ogle - ii catalogue with @xmath84 year to three : ogle-1999-bul-32 days ( bennett et al .
2002b ) . ]
( @xmath85 days ; mao et al . 2002 ) , 5 ( @xmath86 days ; see section [ model ] above ) , and ogle-1999-bul-19 ( @xmath87 days ; smith et al .
information regarding these three events is presented in table [ tb : excess ] .
[ cols="^,^,^,^,^,^,^,^,^",options="header " , ] table [ tb : excess ] shows that these three ogle - ii events are all highly magnified , with peak magnifications ranging from 10.2 to greater than 40 .
figure [ magcomp ] illustrates how these magnifications compare to a high - quality subset of the ogle - ii catalogue . by plotting the total baseline magnitude vs the change in magnitude at peak magnification
, it can clearly be seen that these three events do not appear to lie within the main cluster .
this could suggest that some , if not all , of these three long duration events are artifacts , i.e. the variability may not be due to microlensing .
however , we believe that this is more - likely due to a selection effect , which implies that there could be additional long - duration events residing in the main lower - magnification cluster , i.e. , events that have been omitted from existing ogle - ii microlensing catalogues due to insufficient magnification ( low signal - to - noise ratio ) . in addition , from this table it is interesting to note that that these events do not appear to be clustered in any one direction ( c.f .
popowski 2002 , where it is claimed that an abundance of long duration macho events appear to be clustered in one particular direction ) .
previous work has indicated the existence of an excess of long duration microlensing events toward the galactic bulge ( e.g. , han & gould 1996 ; zhao , rich & spergel 1996 ; bennett et al .
however , is it possible to quantify the extent to which these three longest - duration ogle - ii events constitute an excess ? if one navely extrapolates the behaviour of the timescale distribution of moderately long duration ogle - ii events to long timescales using a @xmath88 power - law ( mao & paczyski 1996 ) , this predicts far fewer than three events should have @xmath84 year .
however , to perform this calculation thoroughly requires the detection efficiency of the ogle - ii catalogue , which is currently undetermined .
we aim to investigate this important issue and report our findings in a future paper . one implication of this potential excess is that many long - duration events could easily be overlooked by microlensing search projects .
microlensing detection algorithms usually rely on the presence of a constant baseline to differentiate true events from variable stars
. however , such long - duration events may exhibit noticeable magnification over the course of many years , which can mean that a constant baseline is not observed .
for example , the detection algorithm of woniak et al .
( 2001 ) initially failed to identify the three longest - duration events ( i.e. , those presented in table [ tb : excess ] ) owing to the fact that they did not pass the constant - baseline criteria . these three events were only found when this ogle - ii catalogue was compared with a previous , independently constructed , catalogue of ogle - ii events .
also , as was mentioned above , it is conceivable that there are additional long - duration events in the ogle - ii catalogue that have been omitted due to having insufficient magnification ( i.e. , low signal - to - noise ratio ) .
however , it should be noted that if some long - duration events are being overlooked in microlensing catalogues then this would increase the observed optical depth , which is already significantly larger than current theoretical predictions ( see , for example , binney , bissantz & gerhard 2000 ; evans & belokurov 2002 ; klypin , zhao & somerville 2002 ) .
this paper has presented a new long - duration parallax event from the ogle - ii database , 5 , which has @xmath0 days .
we believe that both the lens and source reside in the galactic disk , making event 5 one of the first confirmed examples of so - called disk - disk lensing . in this aspect
it differs from the two other longest - duration events , which are both believed to have source stars located in the galactic bulge ( mao et al .
2002 , bennett et al . 2002b
; smith et al .
2002 ) .
the source star for 5 is likely to be located at a distance of @xmath89 kpc , although we are not able to provide a definitive value ( see section [ fss ] ) .
this is the major source of uncertainty in the analysis presented in section [ mass ] .
we strongly recommend that spectroscopic observations of the source are carried out in order to determine the spectral - type of the source and hence improve the estimate of @xmath72 . in section [ lhood ]
we showed that the transverse velocity of the lens relative to the source ( @xmath14 ) can be used to construct a likelihood function for @xmath90 and hence the lens mass .
this analysis suggests that lens is unlikely to be a main - sequence star , since the resulting luminosity of the lens would exceed the limits imposed on the blended flux from the parallax model .
therefore 5 could be another candidate for microlensing by a stellar mass black hole ( agol et al .
2002 ; bennett et al .
this work was supported by a pparc grant .
i would like to thank shude mao for many insightful comments and for detailed criticisms of draft versions of this manuscript .
i am also grateful to bohdan paczyski for helpful suggestions and comments , particularly concerning the importance of event 5 and fig .
[ magcomp ] , and to andy gould for his advice and guidance .
the full 4-year data for event 5 was generously provided by the ogle collaboration and i gratefully acknowledge the help of igor soszyski and przemysaw woniak for their assistance . in addition , i would like to thank takahiro sumi for providing preliminary data from his proper motion analysis .
after this paper was accepted for publication , it was discovered that the eros microlensing collaboration also observed the field of sc5_2859 during this event .
preliminary analysis of these data suggest that they are in good agreement with the ogle - ii data during the period when the observations overlapped . however , there are a number of eros data - points prior to the beginning of the ogle - ii observations ( i.e. for the period @xmath91 d ) , and these data appear to be incompatible with the microlensing model presented in this paper .
it is hoped that the definitive analysis of the eros data , when combined with relevant follow - up observations , will help to clarify the nature of this event .
the results of these findings will be presented at a later date .
agol e. , kamionkowski m. , koopmans l. , blandford r. 2002 , , 576 , l131 alard c. , lupton r. h. , 1998 , , 503 , 325 albrow m. d. , et al . , 2000 , , 534 , 894 alcock c. , et al . , 1995 , , 454 , l125 alcock c. , et al . , 2000 , , 541 , 734 ( erratum 557 , 1035 [ 2001 ] ) afonso c. et al .
( the eros collaboration ) , 2003 , a&a , 404 , 145 becker a.c . , 2000 ,
phd thesis , university of washington van belle g.t .
, 1999 , pasp , 111 , 1515 belokurov v. , evans n.w . , 2002 , mnras , 331 , 649 bennett d.p .
. 2002a , apj , 579 , 639 bennett d.p . ,
becker a.c . ,
calitz j.j . ,
johnson b.r . , laws c. , quinn j.l .
, rhie s.h . , sutherland w. , 2002b , (
astro - ph/0207006 ) bergeron p. , ruiz m. t. , leggett s. k. , 1997 , apjs , 108 , 339 bessell m. s. , 1979 , pasp , 91 , 589 bessell m. s. , brett j. m. , 1988 , pasp , 100 , 1134 binney j. , bissantz n. , gerhard o. , 2000 , apj , 537 , l99 cieslinski d. , diaz m.p .
, mennickent r.e . ,
pietrzyski g. , 2003 , pasp , 115 , 193 cox a. , 2000 , allen s astrophysical quantities ( 4th edition ) , springer - verlag , new york dehnen w. , binney j.j .
, 1998 , mnras , 298 , 387 delplancke f. , grski k. m. , richichi a. , 2001 , , 375 , 701 dominik m. , 1998 , , 329 , 361 drimmel r. , spergel d.n . , 2001 , , 556 , 181 evans n.w . , belokurov v. , 2002 , , 567 , l119 gould a. , 1992 , , 392 , 442 gould a. , 1994 , , 421 , l71 han c. , gould a. , 1996 , , 467 , 540 humphreys r.m .
, larsen j.a . , 1995 , aj , 110 , 2183 klypin a. , zhao h.s . , somerville r. , 2002 , apj , 573 , 597 mao s. , paczyski b. , 1996 , , 473 , 57 mao s. , smith m.c . , woniak p. , udalski a. , szymaski m. , kubiak m. , pietrzyski g. , soszyski i. , ebru k. , 2002 , mnras , 329 , 349 nemiroff r.j .
, wickramasinghe w.a.d.t . , 1994 , apj , 424 , l21 paczyski b. , 1986 , , 304 , 1 paczyski b. , 1996 , , 34 , 419 peale s.j . , 1998 , apj , 509 , 177 popowski , p. 2000 , apj , 528 , l9 popowski p. , 2002
, submitted to mnras ( astro - ph/0205044 ) rhie s.h . , becker a.c .
, bennett d.p . , fragile p.c .
, johnson b.r .
, king l.j . , peterson b.a . , quinn j. , 1999 , apj , 522 , 1037 schechter p.l . , mateo m. , saha a. , 1993 , pasp , 105 , 1342 smith m. c. et al .
2002 , mnras , 336 , 670 smith m.c . , mao s. , paczyski b. , 2003 , mnras , 339 , 925 smith m.c . , mao s. , woniak p. , 2003
, apjl , 585 , 65 soszyski i. et al . 2001 , apj , 552 , 731 sumi t. et al . , 2002 , submitted to ( astro - ph/0207604 )
sumi t. , eyer l. , woniak p. , 2003
, submitted to mnras ( astro - ph/0210381 ) udalski , a. 2002 , astro - ph/0210367 udalski a. , szymaski m. , kauny j. , kubiak m. , krzemiski w. , mateo m. , preston g.w . , 1993 , acta astron . , 43 , 289 udalski a. , kubiak m. & szymaski m. , 1997 , acta astron . , 47 , 319 udalski a. , ebru k. , szymaski m. , kubiak m. , pietrzyski g. , soszyski i. , woniak p. , 2000 , acta astron .
, 50 , 1 udalski a. , szymaski m. , kubiak m. , pietrzyski g. , soszyski i. , woniak p. , ebru k. , szewczyk o. , wyrzykowski . , 2002 , acta astron . , 52 , 217 woniak p. , udalski a. , szymaski m. , kubiak m. , pietrzyski g. , soszyski i. , ebru k. , 2001 , acta astron . , 51 , 175 witt h.j . , mao s. , 1994 , apj , 430 , 505 zhao h.s . , rich r.m .
, spergel d.n . , 1996 , mnras , 282 , 175 | we present a new long - duration parallax event from the ogle - ii database , 5 , which has the second longest time - scale ever identified ( @xmath0 days ) .
we argue that both the lens and source reside in the galactic disk , making event 5 one of the first confirmed examples of so - called disk - disk microlensing .
we find that the source star is most probably located at a distance of @xmath1 kpc , and from this we conclude that the lens is unlikely to be a main - sequence star due to the strict limits that can be placed on the lens brightness .
a simple likelihood analysis is carried out on the lens mass , which indicates that the lens could be another candidate stellar mass black hole .
we recommend that spectroscopic observations of the source be carried out in order to constrain the source distance , since this is the main source of uncertainty in our analysis .
in addition , we briefly discuss whether there appears to be an excess of long duration microlensing events in the ogle - ii catalogue . #
1= 0u_0 0m_i,0 5sc5_2859 epsf gravitational lensing - galaxy : bulge - galaxy : centre - galaxy : kinematics and dynamics - black hole physics . |
spins confined to self - assembled quantum dots are promising optically active qubits . compared to the electron
, there are strong reasons to believe that the spin of a heavy - hole @xcite may make a more robust qubit @xcite .
the main source of dephasing for the electron spin is the contact hyperfine interaction leading to typical dephasing times of @xmath0 , without the use of echo pulses @xcite .
however , for a heavy - hole with p - type bloch - function , the contact hyperfine interaction is zero , suppressing the effective hyperfine interaction by a factor of about 10 .
@xcite in addition , the hyperfine interaction is highly anisotropic , and for an in - plane magnetic field , as used in coherent control experiments @xcite the effect of fluctuations in an out - of - plane effective nuclear magnetic field are further suppressed @xcite .
however , reported measurements of @xmath1 from larmor precession of the hole - spin ( 3 - 20 ns @xcite)or coherence population trapping ( cpt ) ( @xmath2100 ns ) @xcite vary considerably .
the main source of extrinsic dephasing was attributed @xcite to charge noise acting on the e - field sensitive in - plane g - factor @xcite , a view supported by recent studies of spin noise @xcite and cpt @xcite .
methods to reduce the sensitivity of the hole zeeman energy to charge noise are therefore desirable .
recently , a number of experiments have demonstrated active stabilization of the emission frequency of a self - assembled quantum dot @xcite .
these are challenging experiments , where the bandwidth is limited by the signal strength .
another approach is to take advantage of the feedback between the overhauser field and the optical pumping of the nuclear spin bath .
this can result in the trion transition locking to a cw laser @xcite or the effective zeeman energy locking to the repetition rate of a pulsed laser @xcite . in this work ,
i draw inspiration from work on silicon quantum dots , where the qubit energy - splitting is engineered to be insensitive to the electric - field .
one approach , proposed in refs .
@xcite and recently demonstrated @xcite is to use a hybrid qubit composed of three electron spins in a dqd .
a second approach explored is to tune the spin - orbit coupling to make the larmor frequency of a hole spin qubit insensitive to electric - field @xcite .
yet another approach is to use a the ac - stark shift induced by a cw microwave field to compensate for charge noise induced shifts in the qubit energy - splitting @xcite .
also note the work of weiss _ et al _ @xcite where the insensitivity of a singlet - triplet qubit to electric and magnetic fields at a ` sweet - spot ' in an inas double quantum dot leads to long extrinsic coherence times . in this article
, we propose the use of an ac - stark shift induced by a cw laser to passively stabilize the hole zeeman energy against charge noise , and thereby increase the extrinsic dephasing time @xmath1 . a passive stabilization scheme benefits from experimental simplicity .
the bandwidth is not limited by a measurement time . in the following , the principle will be explained , and calculations using typical dot parameters are presented where a factor of 10 - 100 improvement in @xmath1 is shown to be possible .
consider a qubit encoded in the spin state of a heavy - hole confined to an ingaas / gaas quantum dot . to enable coherent control ,
a voigt geometry is used , where a magnetic field @xmath3 is applied in the sample plane , and the quantum dot is optically excited along the growth - axis .
the energy - splitting of the qubit is given by the hole zeeman energy , @xmath4 , where @xmath5 is the bohr magneton .
the in - plane hole g - factor @xmath6 is a result of light - heavy hole mixing .
it varies considerably from dot to dot @xcite , is anisotropic @xcite , and depends on the applied electric - field @xmath7 @xcite . as a result ,
fluctuations in the electric - field at the quantum dot due to charge noise , @xmath8 result in an extrinsic dephasing rate @xmath9 . to simplify the discussion , for now
, only fluctuations in the vertical electric - field are considered .
the effects of lateral electric - fields will be discussed in the appendix .
furthermore , the relatively weak e - field dependence of the in - plane electron g - factor @xcite , and interactions with the nuclear spins are neglected .
the g - factor has been measured to be linear with vertical electric - field @xcite , @xmath10 . in the following , quantum dot parameters measured in refs .
@xcite are used .
there a @xmath11 was measured at 4.7 t , for a quantum dot with @xmath12 , from which we infer that the e - field fluctuations responsible for the extrinsic dephasing is @xmath13 .
values of @xmath14 and @xmath15 are used .
our strategy is to use an ac - stark shift @xcite to reduce @xmath16 and hence the sensitivity to charge noise .
figure 1(a ) presents an energy level diagram of a 4-level heavy - hole/ positive trion system in a voigt geometry @xcite .
a magnetic field is applied in the sample plane , and the control laser is applied along the growth - axis .
the energy eigenstates of the heavy - hole and trion spins are aligned and anti - aligned with the magnetic field .
the selection rules are linearly polarized .
the case of an linearly x - polarized cw - laser is shown in fig .
[ fig : energylevels ] . a y - polarized laser couples the diagonal transitions : @xmath17 .
the laser of rabi energy @xmath18 is strongly detuned from all of the optical transitions to avoid populating the trion states . in this weak coupling regime , the laser induces an ac - stark shift on both of the heavy - hole states , without admixing the states .
the difference in the ac - stark shifts @xcite , results in a shift in the hole zeeman energy , @xmath19 , @xmath20 where @xmath18 is the rabi energy .
the label @xmath21 is used for a x(y ) linearly polarized laser , respectively .
the detuning between the laser , of photon energy @xmath22 and the mean of the optical transitions @xmath23 is @xmath24 the exciton transition energy @xmath23 depends on the electric - field due to an in - built dipole @xmath25 , and an induced electric dipole @xmath26 as @xcite @xmath27 @xmath28 is the zeeman splitting of the optical transitions addressed by the laser .
equation ( 1 ) treats the system as two independent two - level atoms , where the rabi energy is small compared to the laser detuning from both optical transitions : @xmath29 .
( right - axis ) at the turning point tp , the hole zeeman energy is relatively insensitive to e - field , leading to a drop in the extrinsic dephasing rate due to e - field fluctuations by 3 orders of magnitude .
( b ) close - up of turning point .
red - lines indicate factor of ( 10 , 100 , 1000 ) improvement in extrinsic dephasing time
. the top - axis indicates change in gate voltage required to change the electric - field , assuming a gate separation of 230 nm @xcite . ]
figure [ fig : energyvsf ] illustrates the passive stabilization scheme .
the red - trace shows the linear increase of the hole - zeeman energy with electric - field without an applied laser , at a magnetic field of 5 t. the blue trace presents the ac - stark shift to the zeeman energy , for a moderate rabi energy of @xmath30 and a detuning @xmath31 . due to the quantum confined stark effect
, the optical transitions move through the laser , tuning the ac - stark effect .
the black trace presents the total hole - zeeman energy @xmath32 . at an electric - field set - point of @xmath33
, there exists a turning - point in the energy where the hole - zeeman energy is insensitive to fluctuations in the electric - field .
the dashed - line provides an estimate of the extrinsic dephasing rate due to e - field fluctuations : @xmath34 at the ` sweet - spot ' the dephasing rate drops by about 3 orders of magnitude . to assess the robustness of the sweet - spot , fig .
2(b ) shows a close - up of the turning point . for a typical device with a gate separation of 230 nm ,
the gate voltage needs to be set within 2.2 ( 0.25 ) mv of the tp to achieve a factor of 100 ( 1000 ) improvement in the extrinsic dephasing time , respectively . alternatively
, the laser photon energy would need to be set within 5.3 ( 0.61 ) @xmath35 ) of the tp .
to summarize , the ac - stark effect can be used to create a turning point in the hole - zeeman energy versus electric - field where the hole - spin is insensitive to fluctuations in the e - field induced by charge noise .
versus laser detuning and rabi energy . ( a ) x - polarization ( b ) y - polarization ( c - e ) cross - sections at fixed rabi energy @xmath18 for y - polarization .
b = 5 t. power fluctuations of @xmath36 , @xmath37 are used . ]
however , there is a downside .
applying a laser introduces additional dephasing processes that need to be considered .
firstly , fluctuations in the laser power will introduce additional extrinsic dephasing : @xmath38 secondly , the laser will generate some population in the trion states . when the trion radiatively recombines , to either hole spin state , the hole spin is randomized and this will act as a decoherence process .
we estimate an upper limit on this dephasing to be @xcite : @xmath39 where @xmath40 is a typical radiative recombination rate . to calculate the net dephasing rate , i assume the fluctuations are independent , and add the contributions in quadrature : @xmath41 figure [ fig : dephasing](a , b ) present grayscale maps of the suppression factor , @xmath42 the ratio of the extrinsic dephasing rate with and without the control laser , against the laser detuning , @xmath43 , and rabi energy @xmath18 .
a dark value of @xmath44 indicates an improvement .
the magnetic field is set to @xmath45 , and the electric - field set - point to @xmath46 .
the power fluctuations are assumed to be 1% of total power , which is typical for a laser .
the black contour indicates the stable regime where an improvement of at least a factor of 10 , @xmath47 , is achieved .
the stabilization works better for y - polarization , fig .
3(b ) , than x - polarization , fig .
this is a bit counter - intuitive , since the zeeman splitting between the optical transitions is larger for x - polarization , @xmath48 , and the cancelation of the ac - stark shifts is smaller .
however , in the case of y - polarization , the ac - stark shift adds to the effective hole zeeman energy , rather than subtracts , resulting in a turning point further from the optical resonance .
this reduces the negative impact of @xmath49 due to trion generation .
the blue region indicates where the extrinsic dephasing is worse as a result of the control laser , mostly due to the trion generation .
the power fluctuations @xmath50 have little influence .
figures [ fig : dephasing](c , d , e ) present close - ups of the detuning dependence at fixed rabi energy , near the point where the dephasing is optimized .
the inclusion of trion generation reduces the suppression factor that can be achieved , and favors the use of low rabi energies of a few @xmath35 . increasing the rabi energy increases the robustness of the stabilization , at the expense of the optimum value of the dephasing rate . at @xmath51 ,
suppression factors of better than 2-orders of magnitude can be achieved over a detuning range of about 40 nev , equivalent to a change in gate voltage of @xmath52 for a device of 230-nm gate separation .
other possible contributions to dephasing introduced by the control laser that are neglected are heating and photo - generation of charge noise .
these are discussed in appendix b.
so far , i have considered a strong magnetic field of 5 t , where the extrinsic dephasing due to fluctuations in the g - factor due to charge noise is more of an issue .
figure 4 presents a calculation of the suppression factor @xmath53 for a moderate magnetic field of 1 t , for y - polarization .
the sweet - spot lies closer to the optical resonance as the strength of the ac - stark shift to the hole zeeman splitting , @xmath54 increases with the zeeman splitting between the optical transitions @xmath55 .
consequently , the gradient in @xmath54 is sharper and the optimum suppression factor is reduced to @xmath56 , corresponding to an extrinsic dephasing rate of @xmath57 , or an extrinsic dephasing time of @xmath58 .
the stabilization scheme will work much better in a faraday geometry , where the magnetic field is parallel to the laser .
there a circular polarized laser will couple to only one of the hole spin states , leading to a larger ac - stark shift for the hole zeeman energy , and much better performance at low magnetic fields .
to conclude , the use of a laser induced ac - stark effect to partially cancel the electric - field dependence of the hole - zeeman energy and suppress extrinsic dephasing of a quantum dot hole spin due to charge noise is proposed . calculations that consider additional dephasing induced by the laser , find that for optimized laser detuning and power for a typical quantum dot the extrinsic dephasing can be suppressed by a factor of @xmath59 .
the potential use of the ac - stark effect to stabilize an optical transition has also been considered . in the case of the hole spin , the optical transition energies are more sensitive to electric - field than the hole - zeeman energy , and a relatively low power can provide sufficient ac - stark shift to compensate the hole zeeman energy .
however , in the case of stabilizing an optical transition a much higher rabi energy , comparable to the laser detuning is needed , resulting in strong optical pumping , and a high @xmath60 .
this work was funded by hitachi europe ltd .
to simplify the discussion , only fluctuations in the vertical electric - field have so far been considered .
this approximation holds either if the system is isotropic , or if the system is highly anisotropic .
three factors need to be considered .
( i ) the electric - field dependence of the in - plane hole g - factor , @xmath61 .
( ii ) the electric - field dependence of the optical transition energy @xmath23 . (
iii ) the relative size of the fluctuations in the electric - field .
a typical ingaas / gaas self - assembled quantum dot has a truncated square pyramid shape of height 5 nm , and a base of 18 nm @xcite , with an indium rich top @xcite .
compared to the vertical direction , the quantum dot is relatively symmetric in the sample plane .
therefore the electric - field dependence of the in - plane hole g - factor has the form : @xmath62 the g - factor tuning arises from a displacement of the hole wavefunction changing the overlap with the indium rich regions of the quantum dot @xcite , therefore @xmath63 .
in addition , if a vertical electric - field @xmath33 is applied to the quantum dot , @xmath64 .
hence the effect of a lateral electric - field on the hole g - factor is relatively small . for many quantum dots ,
the g - factor is linear with applied vertical electric - field @xcite , and to the best of my knowledge , there are no reports on the lateral - field dependence of the hole g - factor
. if the hole g - factor is insensitive to the lateral electric fields , fluctuations in the optical transition energy @xmath65 will dephase the hole - spin due to fluctuations in the ac - stark shift of the hole zeeman energy . including lateral electric - fields
modifies eq .
( 3 ) : @xmath66 the contribution to the extrinsic dephasing rate due to lateral x - component of electric - fields can be estimated using eq .
( 1 ) as : @xmath67 where @xmath68 is the extrinsic dephasing rate due to the fluctuations in the hole zeeman splitting due to fluctuations in the vertical electric - field . at the sweet - spot , this is equal to the extrinsic dephasing rate without the control laser . in the case of isotropic fluctuations in the electric - field ,
@xmath69 , @xmath70 .
lateral field measurements of the polarizability of ingaas / gaas quantum dots give typical values for @xmath71 @xcite .
therefore an estimate of the contribution of the lateral fields to the suppression factor at the sweet - spot is @xmath72 , using @xmath73 .
this is negligible compared to the suppression factors of @xmath74 achieved at the sweet - spot .
a final consideration is the relative sizes of the fluctuations in the lateral and vertical electric - fields . in ref .
@xcite , shifts in the emission energy of quantum dots in a photodiode structure were attributed to charging of defect states at the algaas / gaas interface .
this is typically @xmath75 above the quantum dot .
this suggests that the strongest e - field fluctuations in atypical photodiode structure are mostly along the vertical direction .
the optimum rabi energies are a few @xmath35 .
this is small compared to rabi energies of @xmath76 reported in refs .
@xcite , where no evidence of heating was reported . hence , heating by the stabilization laser is neglected . in refs .
@xcite , under high excitation powers , an additional power dependent blue - shift in the exciton transitions , attributed to the photo - generation of charges within the device structure , was observed .
this will result in a power dependent shift in the set - point of the laser detuning needed to stabilize the hole zeeman energy .
more seriously , photo - generated charge noise may be significant , and the charge noise is power dependent , i.e. @xmath77 . the value of @xmath8 used in these calculations is inferred from an experiment where the power incident on the sample from the coherent control pulse sequence is about @xmath78 @xcite .
this is much larger than the 20-nw power needed to achieve a continuous wave rabi energy of @xmath79 @xcite needed to stabilize the hole spin .
therefore , the additional photo - generated charge noise due to the stabilization laser is likely to be small compared to the charge noise introduced by the laser pulses used to control the hole spin .
zhan shi , c. b. simmons , j. r. prance , j. k. gamble , teck seng koh , yun - pil shim , xuedong hu , d. e. savage , m. g. lagally , m. a. eriksson , m. friesen , and s. n. coppersmith , * 108 * 140503 ( 2012 ) .
a. laucht , j. t. muhonen , f. a. mohiyaddin , r. kalra , j. p. dehollain , s. freer , f. e. hudson , m. veldhorst , r. rahman , g. klimeck , k. m. itoh , d. n. jamieson , j. c. mccallum , a. s. dzurak , and a. morello , sci . adv . * 1 * e1500022 ( 2015 ) .
w. fry , i. e. itskevich , d. j. mowbray , m. s. skolnick , j. j. finley , j. a. barker , e. p. oreilly , l. r. wilson , i. a. larkin , p. a. maksym , m. hopkinson , m. al - khafaji , j. p. r. david , a. g. cullis , g. hill , and j. c. clark , * 84 * 733 ( 2000 ) . j. houel , a. v. kuhlmann , l. greuter , f. xue , m. poggio , b. d. gerardot , p. a. dalgarno , a. badolato , p. m. petroff , a. ludwig , d. reuter , a. d. wieck , and r. j. warburton , * 108 * 107401 ( 2012 ) . | the extrinsic dephasing of a hole spin confined to a self - assembled quantum dot is dominated by charge noise acting on an electric - field dependent g - factor . here
we propose the use of the optical stark effect to reduce the sensitivity of the effective hole zeeman energy to fluctuations in the local electric - field .
calculations using measured quantum dot parameters are presented , and demonstrate a factor of 10 - 100 reduction in the extrinsic dephasing .
compared to active stabilization methods , this technique should benefit from reduced experimental complexity . |
the purpose of this article is to present some new results that call attention to the presence and effects of particular types of heterogeneity observed to occur in diverse geophysical and ecological / biological spatial environments .
a few examples of interest to the authors are provided to highlight the occurrence of interfacial discontinuities , but one may expect that readers will easily be able to conceive of many others . apart from the motivating examples , theoretical results are provided to illustrate the interfacial effects in ways that may prove useful in analyzing and interpreting more complex data sets .
the most basic empirical considerations involve various temporal measurements , e.g. , breakthrough times , residence or occupation times , recurrence times and , as will be explained , a fundamental notion of _ local time_. mathematical local time is a quantity with a long history in the theory of stochastic processes .
it concerns a mathematical measure of the amount of time a stochastic process spends locally about a point , which we will refer to as _ mathematical local time _ in distinction with a modification introduced here to be referred to as _ natural local time_. starting with the celebrated theorem of trotter @xcite establishing the continuity of mathematical local time for standard brownian motion , the general problem of determining necessary and sufficient conditions for continuity of local time for stochastic processes is an old and well - studied problem at the foundations of probability theory ; see @xcite for recent progress and insight into the technical depth of the problem for a large class of markov processes .
while the present paper has little to offer to the mathematical foundations of the subject , observations from field and laboratory experiments are provided together with some theoretical results that point to new directions in this area from the perspective of applications and modeling of certain dispersive phenomena . at the most fundamental level ,
the issue involves the _ units _ of measurement .
in particular , the units of mathematical local time in the context of dispersion of particle concentrations are typically those of _ spatial length_. this is not unnatural in the mathematical context owing to the locally linear relationship between spatial variance and time when the dispersion coefficient is sufficiently smooth . however , as will be seen , this breaks down in the presence of an interface of discontinuity in the diffusion coefficient .
effort is made to make this article accessible to a diverse audience by including some basic mathematical definitions and background theory .
a complete and systematic treatment of essentially all of the underlying mathematical concepts can be found in @xcite .
we conclude this introductory section with an empirical example that serves to motivate and illustrate the role of a key mathematical construct , namely that of _ skew brownian motion_. this is followed by a section with additional empirical illustrations from the ecological and biological sciences .
the emphasis of the paper is on mathematical theory and modeling .
the main results demonstrate ( i ) the effects of general point interface transmission probabilities on breakthrough times and ( ii ) on residence times ; ( iii ) symmetry ( via martingale ) relationships between interfacial transmission probabilities , dispersion coefficients , and skewness parameters ; ( iv ) a special role for continuity of flux ( conservation of mass ) verses continuity of derivatives at an interface ; ( v ) a special role for ( natural ) local time continuity in the determination of the transmission parameter at the interface . .05 in * example a * ( _ * dispersion in porous media * _ ) .
the topic addressed in this paper originally initiated as a result of questions resulting from recent laboratory experiments designed to empirically test and understand advection - dispersion in the presence of sharp interfaces .
e.g. , experiments by @xcite , @xcite , @xcite .
such laboratory experiments have been rather sophisticated in the use of layers of sands and/or glass beads of different granularities and modern measurement technology ( see figure [ brianslab ] ) .
[ brianslab ] from an engineering science point of view , basic research in this area is aimed to improve understanding of the role of sharp discontinuities in the hydrologic parameters appearing in equations used for predictions of the spread of contaminants in saturated porous media .
there is a huge literature in the sciences and engineering pertaining to the general problem . for a perspective on the mathematical foundations
see @xcite , @xcite,@xcite , @xcite , @xcite , @xcite .
the former considers heterogeneity from a deterministic framework , while the latter involve randomness in the medium .
@xcite contains a survey of various approaches , together with a more exhaustive set of references to the geophysical and hydrologic literature . from a general mathematical point of view an _ interface _
is defined by a hypersurface across which the dispersion coefficient is discontinuous . as is well - known for the case of dilute suspensions in a homogeneous medium ( e.g. , water ) , perhaps flowing at a rate @xmath0 ,
the particle motion is that of a brownian motion with a constant diffusion coefficient @xmath1 and drift @xmath0 . when applied to colloidal suspensions , this is the famous classic theoretical result of albert einstein @xcite experimentally confirmed by jean perrin @xcite ;
see @xcite for a more contemporary experiment .
the essential physics of the problem , however , is relatively unchanged when one considers dissolved chemical species .
the basic phenomena of interest to us here is captured by the following : * question .
* suppose that a dilute solute is injected at a point @xmath2 units to the left of an interface at the origin and retrieved at a point @xmath2 units to the right of the interface .
let @xmath3 denote the ( constant ) dispersion coefficient to the left of the origin and @xmath4 that to the right , with say @xmath5 ( see figure [ inhomogeneous ] ) .
conversely , suppose the solute is injected at a point @xmath2 units to the right of the interface and retrieved at a point @xmath2 units to the left .
_ in which of these two symmetric arrangements will the immersed solute most rapidly breakthrough at the opposite end ? _ [ inhomogeneous ] in the experimental set - up , the coordinate in the direction of flow is one - dimensional flow across an interface . as will be explained , a localized point interface results in a skewness effect that explains much of the empirically observed results suggested above ; see @xcite , @xcite , @xcite , @xcite , @xcite , @xcite .
let us briefly recall the notion of _ skew brownian motion _ introduced by it and mckean in @xcite to identify a class of stochastic processes corresponding to feller s classification of one - dimensional diffusions . the simplest definition of skew brownian motion @xmath6 for a given parameter , referred to as the _ transmission probability _ , @xmath7 $ ] is as follows : let @xmath8 denote reflecting brownian motion starting at zero , and enumerate the excursion intervals away from zero by @xmath9
[ skewbm ] this is possible because brownian motion has continuous paths and , therefore , its set of zeros is a closed set ; the set s compliment is open and , hence , a countable disjoint union of open intervals .
let @xmath10 be an i.i.d .
sequence of bernoulli @xmath11 coin tossing random variables , independent of @xmath12 , with @xmath13 .
then , @xmath6 is obtained by changing the signs of the excursion over the intervals @xmath14 whenever to @xmath15 , for @xmath16 ( see figure [ skewbm ] ) .
that is , @xmath17 in particular , the path @xmath18 is continuous with probability one since the brownian paths are continuous . for positive parameters @xmath19 ,
consider a piecewise constant dispersion coefficient with interface at @xmath20 given by @xmath21 here @xmath22 denotes the indicator function of an interval @xmath23 defined by @xmath24 . in response to the question about breakthrough times raised above
, the following theorem provides an answer as a consequence of @xcite , @xcite . for convenience of mathematical notation
take @xmath25 for here .
just as standard brownian motion may be obtained as a limit of rescaled simple symmetric random walks , e.g. , @xcite , an alternative description of skew brownian motion with transmission probability @xmath26 may be obtained as a limit in distribution of of the rescaled _ skew random walk _ defined by @xmath11 displacement probabilities
, @xmath27 each , at nonzero lattice points , and having probabilities @xmath26 , @xmath28 for transitions from @xmath29 to @xmath30 and @xmath29 to @xmath31 , respectively , see @xcite . .15 in [ thmasympassagetime ]
let @xmath19 be arbitrary positive numbers , with @xmath5 .
define @xmath32 , where @xmath33 is skew brownian motion with transmission parameter @xmath34 , and @xmath35}(x ) , x\in { \mathbb r}$ ] .
let @xmath36 then , .05 in \(a ) for smooth initial data @xmath37 , @xmath38 , solves @xmath39 .05 in \(b ) for @xmath40 , @xmath41 .15 in observe , for example , that by integrating the complementary distribution functions in ( b ) , one obtains that the mean breakthrough time in fine to coarse media is _ smaller _ than that of the mean breakthrough time a coarse to fine media .
related phenomena and results on dispersion in this context , including the case @xmath42 , are also given in @xcite , @xcite , @xcite , @xcite , @xcite , @xcite .
the proof of part ( b ) of the theorem relies on a transformation to _ elastic skew brownian motion _ to eliminate the drift .
in addition , a recently obtained formula for the first passage time distribution for skew brownian motion is given in @xcite .
considerations of both local time and elastic standard brownian motion appear in @xcite to cite another biological context for their significance .
the identification of the stochastic particle motions in the presence of an interface for simulation can have utility in problems involving the computation of particle concentration at a single spatial location , e.g. , for so - called resident breakthrough , since other pde numerical schemes generally involve computation of the entire concentration curve .
some illustrative results pertaining to monte - carlo simulations of skew diffusions are described in @xcite and references therein .
in addition , the identification of skew diffusion answers the basic physics question of finding the particle motion that jean perrin would have reported had there been an interface ! from the macro - scale quantity of particle concentrations , the determination of @xmath43 in theorem [ thmasympassagetime ] may be viewed as the result of mass conservation .
the interface condition is simply continuity of flux at the interface .
however , as will be illustrated by examples in the next section , it is _ not _ necessary that continuity of flux always be obeyed in the presence of interfaces . moreover
, even the modeling may be at the scale of a single particle , e.g. , biological dispersion of individual animals , in which ` particle concentrations ' are not relevant , e.g. , see @xcite , @xcite , @xcite for such examples .
a stochastic particle model plays an essential role in such situations .
as illustrated by the examples in this section , the role of interfacial phenomena is of much broader interest than suggested by advection - dispersion experiments . however
the specific nature of the interface can vary , depending on the specific phenomena .
we briefly describe below three distinct classes of examples of phenomena from the biological / ecological sciences in which such interfaces naturally occur , together with compelling questions having substantial biological implications .
.05 in * example b * ( _ * coastal upwelling and fisheries * _ ) .
up - wellings , the movement of deep nutrient rich waters to the sun - lit ocean surface , occur in roughly one percent of the ocean but are responsible for nearly fifty - percent of the worlds fishing industry .
the up - welling along the malvinas current that occurs off of the coast of argentina ( see figure [ coastal ] ) is unusual in that it is the result of a very sharp break in the shelf , rather than being driven by winds .
the highlighted points in the figure represent a flotilla of fishing boats concentrated along the shelf where the up - welling occurs .
fishing vessels off the coast of argentina , http://subdude-site.com/webpics/webpicsearthlights/south-america-hilights-429x600.jpg courtesy of ricardo matano , coas , oregon state university ] the equation for the free surface elevation @xmath44 as a function of spatial variables @xmath45 is derived from principles of geostrophic balance and takes the form @xmath46 where @xmath47 and @xmath48 in the southern hemisphere , and @xmath49 is the depth of the ocean at a distance @xmath50 from the shore . in particular , the sharp break in the shelf makes @xmath51 a piecewise constant function with positive values @xmath52 .
the location of the interface coincides with the distance to the shelf - break .
in particular , if the spatial variable @xmath40 is viewed as a ` time ' parameter , then this is a skew - diffusion equation , but the physics imply continuity of the derivatives @xmath53 at the interface rather than ` flux ' ( see @xcite and references therein ) . .05 in * example c * ( _ * fender s blue butterfly * _ ) .
the fenders blue is an endangered species of butterfly found in the pacific northwestern united states .
the primary habitat patch is kinkaid s lupin flower ( figure [ butterfly ] ) .
[ butterfly ] quoting @xcite , _ given past research on the fender s blue , and the potential to investigate response to patch boundaries , we ask two central questions .
first , how do organisms respond to habitat edges ?
second , what are the implications of this behavior for residence times ?
_ sufficiently long residence ( occupation ) times in lupin patches are required for pollination , eggs , larvae and ultimate sustainability of the population .
empirical evidence points to a skewness in random walk models for butterfly movement at the path boundaries .
the determination of proper interface conditions is primarily a statistical problem in this application , however we will see that certain theoretical qualitative analysis may be possible for setting ranges on interface transmission parameters .
there is a rapidly growing literature on the statistical estimation of parameters for diffusions , @xcite , @xcite , @xcite .
however much ( though not all ) of this literature is motivated by applications to dispersive models in finance for which the coefficients are presumed smooth and the data is high frequency . .05 in * example d * ( _ * sustainability on a river network * _ ) .
the movement of larvae in a river system is often modeled by advective - dispersion equations in which the rates are determined by hydrologic / geomorphologic relationships in the form of the so - called _
horton laws_. in general river networks ( figure [ rivernetwork ] ) are modeled as directed binary tree graphs and each junction may be viewed as an interface .
there are also special relations known from geomorphology and hydrology that can be applied to narrow the class of graphs observed in natural river basins ; e.g. , see @xcite , @xcite , @xcite , and references therein .
[ rivernetwork ] conservation of mass leads to continuity of flux of larvae across each stream junction as the appropriate interface condition .
problems on sustainability in this context are generally formulated in terms of network size and characteristics relative to the production of larvae sufficient to prevent permanent downstream removal at low population sizes ; see @xcite for recent results in the case of a river network , and @xcite and @xcite for related mathematical considerations .
the following theorem provides a useful summary in one - dimension of the interplay between diffusion coefficients and broader classes of possible interfacial conditions . to set the stage we begin with brownian motion @xmath54 with constant diffusion coefficient @xmath1 and constant drift @xmath0 ;
i.e. , @xmath55 is standard brownian motion starting at zero , with unit diffusion coefficient and zero drift . the stochastic process @xmath56 has a number of properties characteristic of diffusion .
namely , the random path @xmath57 is continuous with probability one , and @xmath58 this makes @xmath56 a continuous semimartingale . in the case
@xmath25 , @xmath56 is in fact said to be a ( continuous ) martingale .
in general a continuous semimartingale is a process with continuous paths that differs from a ( local ) martingale by a ( unique ) continuous ( adapted ) process having finite total variation ( in the sense of functions of bounded variation ) .
such structure is a natural consequence of the interfaces discussed in this article .
this will be made clear at equation ( [ skewsde ] ) below .
the brownian motion @xmath56 is also a markov process with ( homogeneous ) transition probabilities @xmath59 satisfying the advection - dispersion equation @xmath60 with continuous derivatives of all orders everywhere . since the displacements @xmath61 are independent for disjoint intervals @xmath62 , it follows that the variance of @xmath63 is linear in @xmath64 and , in particular , @xmath65 the definition of _ mathematical local time at @xmath66 _ , denoted @xmath67 , for @xmath56 can be expressed as @xmath68 is the amount of time @xmath56 spends in the infinitesimal neighborhood @xmath69 prior to time @xmath64 .
more precisely , @xmath70 note the presence of the diffusion coefficient @xmath71 .
thus , in the context of dispersion of particles in a fluid , for example , @xmath72 = l^2/t$ ] and @xmath73 = [ x ] = l$ ] yields units @xmath74 = { 1\over l}\times{l^2\over t}\times t = l.$ ] the standard extension of the mathematical definition of local time for a continuous semimartingale @xmath56 exploits the _ quadratic variation _ @xmath75 of the process . if @xmath56 is a square - integrable martingale then @xmath76 is defined by the property that the process @xmath77 is a martingale ; see @xcite .
so , in the case of the brownian motion one may check that @xmath78 in the case of a continuous semimartingale , the quadratic variation is quadratic variation of its martingale component
. mathematical local time at @xmath66 can be defined as an increasing continuous stochastic process @xmath79 such that @xmath80 where @xmath81 and @xmath82 ( by convention ) . for purposes of calculation
it is often convenient to consider ( right and left ) one - sided versions defined by @xmath83 @xmath84}(x(s))d < x>_s , \qquad t\ge 0.\ ] ] then @xmath85 the utility of these quantities rests in the following two formulae : * it - tanaka formula : * if @xmath56 is a continuous semimartingale with local time @xmath86 , and @xmath87 is the difference of two convex functions then @xmath88 where @xmath89 is a positive measure corresponding to the second derivative of @xmath87 in the sense of distributions . .15 in * occupation time formula : * for a non - negative borel function @xmath90 and @xmath91 , one has a.e . that @xmath92 .15 in [ thmalphamart ] let @xmath19 be arbitrary positive numbers and let @xmath93 .
define @xmath94 , where @xmath6 is skew brownian motion with transmission parameter @xmath26 and @xmath35}(x ) , \
x\in { \mathbb r}$ ]
. then @xmath95 is a martingale for all @xmath96 if and only if @xmath97 .15 in in the case of skew brownian motion one has the following relationships between mathematical local time at zero and its one - sided variants , e.g. , see @xcite @xmath98 moreover , @xmath6 is the unique strong solution to the stochastic differential equation , see @xcite , @xmath99 in particular , considering the integrated version , @xmath12 is the martingale component and @xmath100 is the finite variation component in the view of @xmath6 as a semimartingale .
it is also straightforward to relate mathematical local times of @xmath101 and @xmath6 through the one - sided formulae as : @xmath102 applying the it - tanaka formula to the positive and negative parts of @xmath101 together with ( [ skewsde ] ) , one has @xmath103}db(t )
+ { 1\over 2}d\ell_t^{b^{(\alpha)},+}(0)\},\ ] ] and @xmath104}db(t ) - { ( 2\alpha-1)\over 2\alpha}d\ell_t^{b^{(\alpha)},+}(0)+ { 1\over 2}d\ell_t^{b^{(\alpha)},+}(0)\}.\ ] ] thus , since @xmath101 is the difference of its positive and negative parts and noting the cancellation , one has @xmath105 for a difference of convex functions @xmath106 , @xmath107 with these preliminaries , again use the it - tanaka theorem together with ( [ skwbmlocrel ] ) and ( [ natdiffsde ] ) , to get @xmath108 according to the mathematical occupation time formula relating mathematical occupation time and mathematical local time , and and noting that the quadratic variation is given by @xmath109 , it also follows that @xmath110 the asserted result now follows by subtracting this term from each side of the equation ( [ marteqn ] ) and noting the cancellation of local time terms leaving a stochastic integral with respect to brownian motion when and only when @xmath111 , i.e. , @xmath112\ell_t^{b^{(\alpha)},+}(0 ) = 0\end{aligned}\ ] ] if and only if @xmath113 . that this is enough follows from standard theory , e.g. , theorem 2.4 in @xcite .
some implications for the examples will be described below in terms of the stochastic particle evolution , however one may also note that one has the following consequence at the scale of kolmogorov s backward equation .
let @xmath19 be arbitrary positive numbers and let @xmath93 .
then for @xmath114 , the unique solution to @xmath115 is given by @xmath116 it is illuminating to consider theorem [ thmalphamart ] in the context of the examples .
* example a*. theorem [ thmalphamart ] provides a generalization of the results obtained in @xcite and @xcite for the case of advection - dispersion problems across an interface described in example a. one may check that @xmath117 follow from theorem [ thmalphamart ] for this application .
this coincides with the results of @xcite and @xcite obtained by other methods .
the following definition is made with reference to both the diffusion coefficient and the interface parameter in the context of this and the other examples .
[ natdiff ] with the choice of @xmath118 given by theorem [ thmalphamart ] , we refer to the process @xmath119 as the _ natural diffusion _ corresponding to the dispersion coefficients @xmath19 and interface parameter @xmath120 .
_ * additional nomenclature : * _ we sometimes refer to the natural diffusion corresponding to @xmath121 as the _ physical diffusion_.
we refer to the diffusion in the case @xmath122 as the _ stroock - varadhan diffusion _ since it is the solution to the particular martingale problem originating with these authors ; see @xcite .
.15 in * example b*. observe that in the application to the coastal up - welling problem one obtains @xmath123 the interface parameter provides continuity of the derivative , rather than of the flux , at the interface .
the natural diffusion for this example may be checked to coincide with the stroock - varadhan diffusion in this case .
note that the answer to the first passage time problem will be exactly opposite to that obtained for the advection - dispersion experiments of example a under this model .
the following modification of the usual notion of mathematical occupation time , where the integration is with respect to quadratic variation and in units of squared - length , provides a quantity in units of time that we refer to as natural occupation time
. demonstrates , the notion of mathematical local time and occupation time and their relationship is extremely powerful in singling out the special value of @xmath124 for given interface parameter @xmath120 and dispersion coefficients @xmath125 . ]
let @xmath56 be a continuous semimartingale .
the _ natural occupation time of @xmath126 by time @xmath64 _ , is defined by @xmath127 for an arbitrary borel subset @xmath126 of @xmath128 .
the following result illustrates another way in which the issue raised in example b relating interfacial conditions to residence times is indeed a sensitive problem .
the proof exploits the property of skew brownian motion that for any @xmath129 , @xmath130 this is easily checked from definition and , intuitively , reflects the property that the excursion interval @xmath131 of @xmath8 containing @xmath64 must result in a @xmath132 $ ] coin flip , an event with probability @xmath26 .
[ thmoccuptime ] let @xmath119 denote the natural diffusion for the dispersion coefficients @xmath19 and interface parameter @xmath120 . denote natural occupation time processes by @xmath133ds , \quad t\ge 0.\ ] ]
similarly let @xmath134 then , @xmath135 with equality when @xmath136 . using the definition of natural diffusion @xmath119 for the parameters @xmath137 and
the above noted property ( [ posprob ] ) of skew brownian motion , one has @xmath138ds\\ & = & \mathbb{e } \int_0^t1[\sqrt{d^+}b^{(\alpha(\lambda))}(s))>0]ds\\ & = & \int_0^tp(b^{(\alpha(\lambda))}(s ) > 0)ds\\ & = & t\alpha(\lambda).\end{aligned}\ ] ] similarly @xmath139 .
thus @xmath140 the assertion now follows .
.2 in * example a vs example c*. as noted previously , in the example c pertaining to insect movement , the determination of an interface condition is largely a statistical issue as there is no scientific rationale to apply mass conservation principles , or smooth fickean flux laws .
in fact , it is interesting to observe that under the mass conservation one arrives at the interface parameter @xmath141 for which the residence time is longer on average in the region with the faster dispersion rate ! while this is to be anticipated for physical experiments of dispersion in porous media of the type described in example a , theorem [ thmoccuptime ] shows that in fact the conservative interface condition ( defined by this choice of @xmath120 ) would not be appropriate for models of animal movement for which the faster dispersion occurs in more hostile environments ; e.g. , see @xcite , @xcite , @xcite for relevant considerations .
there is indeed something to be learned from data as to what exactly might apply , but such theoretical insights can provide a useful guide , and help to prevent mistaken assumptions when transferring more well - developed physical principles to biological / ecological phenomena . .15 in
we now discuss another issue pertaining to the definition of mathematical local time and the basis for the suggested modification to natural local time . in the often cited article @xcite , it was first observed that skew - brownian motion provides a non - trivial example of a continuous semimartingale on the interval @xmath142 having a discontinuous mathematical local time .
let us now examine this situation in the context of natural diffusions .
for the purposes of this discussion take the drift @xmath143 and consider a constant diffusion coefficient @xmath144 since the quadratic variation of skew brownian motion coincides with that of brownian motion , one has @xmath145 among this class of natural diffusions , one may ask what distinguishes the particular diffusion @xmath146 ?
of course , the answer is that @xmath56 is determined by @xmath147 .
if one views this choice in the context of the flux in particle concentration , then it provides continuity of flux . on the other hand , in view of the respective theorems of trotter @xcite and walsh @xcite , it is also the unique choice of @xmath26 from among all skew brownian motions to make local time continuous .
the latter may be viewed as a stochastic particle determination of the physical diffusion model , among natural diffusions , for constant diffusion coefficient @xmath71 .
the next theorem , a version of which was originally conceived in @xcite , extends this to the more general framework of the present paper , in particular to include the case @xmath148 .
however , it requires the following modification of the definition of mathematical local time , referred to here as natural local time .
[ natloctime ] let @xmath56 be a continuous semimartingale .
the _ natural local time at @xmath66 _
@xmath149 of @xmath56 is defined by @xmath150}(x(s))ds\ ] ] providing the indicated limits exist almost surely .
the units of natural local time are then @xmath151 = { t\over l}$ ] , appropriate to a measurement of ( occupation ) time in the vicinity of a spatial location @xmath66 .
while the purpose here is not to explore the generality for which natural local time exists among all continuous semimartingales , according to the following theorem it does exist for natural skew diffusion .
moreover , continuity has a special significance .
[ natloctimeofnatdiff ] let @xmath119 be the natural skew diffusion with parameters @xmath152 .
then the natural local time of @xmath119 at @xmath29 is continuous if and only if @xmath153 , i.e. , if and only if @xmath154 and thus @xmath155 is the physical diffusion . in view of ( [ skwbmlocrel ] )
observe that @xmath156 moreover , @xmath157ds\\ & = & { 1\over\sqrt{d^+}}\lim_{\epsilon\downarrow 0}{\sqrt{d^+}\over\epsilon}\int_0^t 1[0\le b^{(\alpha)}(s ) < { \epsilon\over\sqrt{d^+}}]ds\\ & = & { 1\over\sqrt{d^+}}\ell_t^{b^{(\alpha)},+}(0).\end{aligned}\ ] ] similarly , @xmath158 thus , @xmath159 now simply observe that this ratio is one if and only if @xmath160 , which establishes the assertion .
the authors are grateful to jorge ramirez for several technical comments that improved the content and exposition .
the first author was partially supported by an nsf - igert-0333257 graduate training grant in ecosystems informatics at oregon state university , and the remaining authors were partially supported by a grant dms-1122699 from the national science foundation .
the corresponding author is also grateful for support provided by the courant institute of mathematical sciences , new york university , during preparation of the final draft of this article .
appuhamillage , t. ( 2011 ) : the skew diffusion with drift : a new class of stochastic processes with applications to parabolic equations with piecewise smooth coefficients , _ phd thesis : oregon state university _ , corvallis , or .
appuhamillage , t. , v. bokil , e. thomann , e. waymire , b. wood ( 2011 ) occupation and local times for skew brownian motion with applications to dispersion across an interface , annals of appld .
probab , * 21*(1 ) 183214 .
[ correction : ann .
probab . , to appear , http://arxiv.org/abs/1009.5410 . ] appuhamillage , t. , v. bokil , e. thomann , e. waymire , b. wood ( 2009 ) solute transport across an interface : a fickian theory for skewness in breakthrough curve , water resour .
res , 46 , w07511 , doi:10.1029/2009wr008258 .
berkowitz , b. , a. cortis , i. dror , and h. scher ( 2009 ) , laboratory experiments on dispersive transport across interfaces : the role of flow direction , water resour .
, 45 , w02201 , doi:10.1029/2008wr007342 .
bhattacharya , r. n. ( 1999 ) : multiscale diffusion processes with periodic coefficients and an application to solute transport in porous media .
( special invited paper ) , annals of applied probability , * 9 * 951 - 1020 .
hoteit , h. , r. mose , a. younes , f. lehmann , ph .
ackerer ( 2002 ) , three - dimensional modeling of mass transfer in porous media using the mixed hybrid finite elements and random walk methods , _ mathematical geology _ , * 34*(4 ) , 435 - 456 .
le gall , j .- f .
one - dimensional stochastic differential equations involving the local times of the unknown process .
stochastic analysis and applications ( swansea , 1983 ) , 5182 , lecture notes in math .
, 1095 , springer , berlin , 1984 kuo , r. k. h. and irwin , n. c. and greenkorn , r. a. and cushman , j. h. ( 1999 ) , experimental investigation of mixing in aperiodic heterogeneous porous media : comparison with stochastic transport theory , _ transport in porous media , _ 37 , 169 - 182 .
labolle , e.m . , j. quastel , graham e. fogg , and j. gravner : diffusion processes in composite porous media and their numerical integration by random walks : generalized stochastic differential equations with discontinuous coefficients , _ water resour . res .
_ ( 2000 ) * 36*(3 ) , 651 - 662 .
g. papanicolaou , s. r. s. varadhan , boundary value problems with rapidly oscillating random coefficients , coll .
j?anos bolyai , * 27 * random fields , ( esztergom , 1979 ) , 835 - 873 north - holland , amsterdam 1982 .
ramirez , j. ( 2007 ) : skew brownian motion and branching processes applied to advective - diffusion in porous media and fluid flow , _ phd thesis _ ,
oregon state university .
ramirez , j. , e. thomann , e. waymire , r. haggerty , b. wood ( 2006 ) , a generalized taylor - aris formula and skew diffusion , _ siam multiscale modeling and simulation _ * 5 * 3 , 786 - 801 .
ramirez , j. m. , e. a. thomann , e. c. waymire , j. chastanet , and b. d. wood ( 2008 ) , a note on the theoretical foundations of particle tracking methods in heterogeneous porous media , water resour .
, 44 , w01501 , doi:10.1029/2007wr005914 .
sorensen , m. ( 2004 ) : martingale estimating functions for discretely observed stochastic differential equation models , in : _ interplay between @xmath161 semigroups and pdes : theory and application _ , eds . , romanelli , s. , r.m .
minnini , and s. lucente , aracne editrice , rome , 213 - 236 . | this article addresses a modification of local time for stochastic processes , to be referred to as _ natural local time_. it is prompted by theoretical developments arising in mathematical treatments of recent experiments and observations of phenomena in the geophysical and biological sciences pertaining to dispersion in the presence of an interface of discontinuity in dispersion coefficients .
the results illustrate new ways in which to use the theory of stochastic processes to infer macro scale parameters and behavior from micro scale observations in particular heterogeneous environments
. [ section ] [ theorem]lemma [ theorem]proposition [ theorem]corollary [ theorem]remark [ theorem]fact [ theorem]problem [ theorem]example [ theorem]question [ theorem]conjecture [ theorem]definition @addtoreset *p * l _ * keywords . * _ advection - dispersion , discontinuous diffusion , skew brownian motion , mathematical local time , natural local time _ * 2010 mathematics subject classification . * _ |
high pressure studies form an important area of research in solid state physics . over the years
such studies have provided useful insight into our understanding of the physical properties of solids@xcite .
pressure has basically two effects on the electrons in solids : increased kinetic energy and accompanying change in the effects of coulomb interaction among the electrons . in typical band structure calculations
the former effect is captured via increased overlap of the basis orbitals , leading to increased hopping matrix elements and band broadening .
the changes in the correlation effects are captured via self - consistency of charge and potential , as dictated by the density functional theory and its variants ( e.g. generalized gradient approximation ( gga ) , local density approximation with on - site coulomb interaction ( lda+u ) ) . both of these affect magnetic properties of materials such as exchange interaction , local moments , curie temperature and resistivity due to magnetic scattering .
of particular importance for metallic magnets are the change in the fermi surface and the hybridization between different orbitals , which influences the itinerant vs. localized nature of the charge carriers .
it is a useful exercise to explore how the existing theories of electronic and magnetic structure calculation fare in describing the pressure variation of these physical properties . in the present paper
we study the pressure - dependence of the curie temperature and the resistivity of some quaternary heusler alloys , which form an important class of magnetic materials , with potential industrial / technological applications@xcite .
the present work is a natural continuation of our previous paper @xcite in which we presented an extensive study of the magnetic and transport properties of these alloys at ambient pressure .
there exists a number of experimental studies of the variation of the curie temperature ( t@xmath4 ) under pressure for elemental ferromagnets , transition - metal alloys @xcite , as well as heusler alloys @xcite including co@xmath0tial , @xcite ni@xmath0mnsn , @xcite pd@xmath0mnsn and some others @xcite .
recently diluted magnetic semiconductors @xcite and some random heusler alloys @xcite have also been studied . however , the corresponding experimental studies of the pressure - dependence of the resistivity are rare : one example of such a study is the work by austin and mishra @xcite
. several studies of the ambient / equilibrium electronic and magnetic properties of such alloys have appeared recently@xcite .
theoretical studies of the pressure - dependence of the curie temperature of elemental transition metal ferromagnets include model - based studies @xcite as well as recent first - principles calculations .
@xcite turek @xcite have studied the pressure - dependence of the curie temperature in hcp gd and found a strong dependence of @xmath5 on the @xmath6 ratio , which , they suggested , can be related to the measured @xmath5 of thick epitaxial gd(0001 ) films on various transition - metal substrates .
similar calculations have also been carried out for the intermetallic compound gdal@xmath7 @xcite and for cr - based compounds in the zinc blende structure .
@xcite a systematic theoretical study of the pressure - dependence of @xmath5 in heusler alloys was carried out recently by sasioglu _
_ , @xcite where the experimentally observed increase of @xmath5 with pressure @xcite in ni@xmath0mnsn was reproduced correctly , albeit only qualitatively so . there exists some other studies of heusler alloys in which the dependence of the exchange integrals on volume contraction / expansion @xcite was explored .
the particular heusler systems considered in this work are the ( ni@xmath1,cu@xmath2)@xmath0mnsn and ( ni@xmath1,pd@xmath2)@xmath0mnsn alloys .
we employ the coherent potential approximation ( cpa ) to treat the disorder in the ni - sublattice due to doping with cu- or pd - atoms .
it is well known that cpa neglects the effects of short - range order , which are often present in random alloys . however , for the cases under study , such effects should be weak , as we need to describe the magnetic interactions on the non - random mn sublattice , which is only indirectly influenced by the randomness of the transition - metal sublattice via hybridization
. the most important effect of disorder is , therefore , the change in carrier concentration , which is well - described by the cpa . in a recent paper @xcite , it was demonstrated that the cpa provides a reasonably good description of the electronic structure for the case of disorder on the magnetic sublattice as well .
the electronic structure calculations are performed employing the tight - binding linear muffin - tin orbital ( tb - lmto ) basis @xcite and the density functional theory ( dft ) . the local spin - density approximation ( lsda ) for the exchange - correlation part of the potential
the effect of substitutional disorder among ( ni , cu ) or ( ni , pd ) atoms is described by the coherent potential approximation ( cpa ) as formulated in the framework of the tb - lmto green function method .
@xcite the calculations employ an @xmath8-basis , the same atomic radii are adopted for all atoms , and the vosko - wilk - nusair exchange - correlation potential @xcite is used .
the exchange interactions and curie temperatures are studied here by employing a classical heisenberg hamiltonian @xmath9 where @xmath10 denotes the exchange integral between mn atoms at sites @xmath11 and @xmath12 , and @xmath13 and @xmath14 are unit vectors in the directions of the local magnetization on sites @xmath11 and @xmath12 , respectively .
the parameters of the heisenberg hamiltonian are obtained using a two - step approach .
@xcite the reference state for our calculations is chosen to be the disordered local moment ( dlm ) state , @xcite which was successfully used in previous studies .
@xcite the exchange interactions @xmath10 can be expressed in the framework of the tb - lmto method as @xmath15 where @xmath16 characterizes the exchange splitting of the mn atom at the site @xmath11 , @xmath17 is the configurationally averaged green function describing the motion of an electron with spin @xmath18 , @xmath19 , between mn sites @xmath11 and @xmath12 , and the integration is done over the contour @xmath20 in the complex energy plane @xmath21 which starts below the valence band and ends at the fermi energy .
symbol @xmath22 denotes the trace over the basis orbitals of angular momentum symmetry @xmath23 . the effective exchange splitting @xmath16 is defined in the tb - lmto method in terms of the potential functions @xmath24 of mn atoms as @xmath25 .
further details on the exchange interactions evaluated in the dlm reference state can be found in ref . .
we have found that the mapping can be improved ( providing a better agreement of calculated @xmath5 with the experiment ) by including the electron correlation effect on ni - sites in the framework of the lsda+u method ( see fig . 1 in ref . ) .
the electron correlations on pd - sites are found to have negligible influence on the calculated @xmath5 .
we determine the curie temperature corresponding to the heisenberg hamiltonian using the rpa @xcite and by including all exchange interactions up to @xmath26 4 lattice constants .
the pressure is simulated by reduction of the lattice constants from their ambient pressure values , as reported in experiments ( see table 4 of ref . and table 2 of ref . ) .
in ref . an empirical pressure - volume relation was given for ni@xmath0mnsn .
although such a relation can be obtained entirely from theoretical calculations , in the present study , as in ref . , we simply reduce the lattice constant and estimate the corresponding pressure from the empirical pressure - volume relation . according to both the empirical pressure - volume relation@xcite and theoretical calculations@xcite , a 3 % reduction of the lattice
constant corresponds roughly to a pressure of 16 gpa in ni@xmath0mnsn . in heusler alloys ( ni@xmath1,t@xmath2)@xmath0mnsn ( t = cu , pd )
the effect of disorder among the ni- and t - atoms at the fermi energy @xmath27 is weak because the ni- and t - states lie well below @xmath27 .
@xcite consequently , the corresponding residual resistivity is small .
the resistivity due to phonon scattering @xcite is known to be small as well .
the dominant contribution to the resistivity , thus , is from the spin - disorder scattering , which increases with temperature , reaching its maximum value at @xmath5 . in ref .
we employed a simplified approach to estimate the temperature - dependent resistivity in ( ni@xmath1,pd@xmath2)@xmath0mnsn heusler alloys .
it was shown that the spin - disorder at and above @xmath5 can be described satisfactorily using the dlm model , and the corresponding resistivity can be described by the kubo - greenwood approach adapted to the tb - lmto method .
@xcite we have used the same approach to study the resistivity under pressure as well . for further details concerning computational techniques we refer the reader to our recent paper .
@xcite the method we have used is based on mapping the ( @xmath28 ) total energy to an effective heisenberg hamiltonian .
in fact it is just the band energy , not the total energy , which is considered in the mapping , by appealing to the so - called magnetic force theorem(see ref .
( ) and references therein ) .
for finite temperature ( @xmath29 ) calculations , as would be appropriate for the dlm reference state , one needs to consider the mapping of the free energy @xmath30 , where @xmath31 is the internal energy and @xmath32 is the entropy .
@xmath31 should include the electronic and the average vibrational energy and @xmath32 should include both electronic and vibrational entropy .
electronic energy should be calculated via finite temperature density functional theory .
this free energy can not , in principle , be mapped to an effective heisenberg form , which includes only the electronic parameter ( spin or magnetic moment ) .
such a scheme could be implemented for a supercell calculation . however , the advantage of being able to calculate exchange interactions up to a large distance with relative ease and reasonable accuracy and therefore predict results and trends for a group of materials will be gone . for finite temperatures , both electronic excitations and lattice vibrations will influence magnetic properties .
however , in the absence of an accurate scheme , we will refrain from speculating on these effects on our results .
in this section we present results for the pressure - dependence of the curie temperature and spin - disorder resistivity of ( ni@xmath1,t@xmath2)@xmath0mnsn ( t = cu , pd ) heusler alloys over a broad range of concentrations .
we start with a brief review of previous theoretical studies , both model - based and first - principles .
theoretical study of the pressure - dependence of @xmath5 on a model level has usually been based on a discussion of the indirect exchange interaction among spins @xcite or on the hubbard model of magnetism .
@xcite it is clear that the pressure - dependence of @xmath5 carries information about the volume - dependence of the electron wave functions @xcite ( bloch functions in the case of crystalline solids ) . in an rkky - type indirect exchange interaction model , the volume - dependence of @xmath5 can be related to the volume dependencies of the density of states at the fermi level @xmath33 and the exchange integral or the matrix element of the exchange interaction between the wave functions at the fermi energy @xcite(henceforth referred to as the bare exchange interaction @xmath34 , see eq.([eqjbare ] ) ) .
the density of states at the fermi level usually decreases with pressure ( unless , under pressure , new bands start crossing the fermi level ) , as the bands broaden due to increased overlap between the orbitals centered on neighboring atoms .
the bare exchange integral usually increases under pressure , due to increased overlap of the wave functions .
the above two effects compete with each other in determining the final change of @xmath35 under pressure . in a description based on the hubbard model ,
the band terms and the term involving the intra - atomic coulomb interaction @xmath31 are written separately . in this case
it is possible to divide the volume / pressure - dependence of @xmath5 into a term originating from the band broadening under pressure and another that relates to the volume - dependence of @xmath31 .
the importance of this latter term for the pressure - dependence of @xmath5 in ni and ni - cu alloys was first pointed out by lang and ehrenreich .
@xcite they argued that it is this term involving the intra - atomic coulomb interaction that is responsible for the increase in @xmath5 under pressure ( @xmath36 ) in ni and the decrease of @xmath37 with the increase of cu - concentration in ni - cu alloys .
in _ ab initio _ calculations based on dft , such as ours , the above effects are merged together and it is hard to delineate one from the other . the advantage , of course , is that the exchange interactions are realistically calculated for the system at hand , and no assumption related to the volume - dependence of any parameter needs to be made .
prototypical studies in this direction are refs . , addressing the pressure - dependence of the curie temperature of bcc iron .
in particular , a very recent first - principles study @xcite seems to agree with experiment , which gives essentially no pressure - dependence of @xmath5 .
the authors describe this result as a balance between the band structure effect which reduces the magnetization and an increase of the bare exchange integrals with decreasing volume , as mentioned above .
a somewhat different situation is encountered for fcc - ni .
this metal is a strong ferromagnet , its magnetic moment depends on pressure only weakly and an increase of exchange integrals under pressure thus dominates . as a result , @xmath35 of fcc - ni increases with pressure . in general
both effects are relevant and the observed change of @xmath35 under pressure is a result of the competition between them .
mnsn heusler alloy as functions of the relative interatomic distance @xmath38 where @xmath39 is the fcc lattice parameter : ( a ) the effective interactions @xmath10 calculated from eq .
( [ eqjij ] ) , ( b ) the bare interactions @xmath40 defined by eq .
( [ eqjbare ] ) .
the cases of ambient pressure ( open circles ) and the pressure corresponding to the 3% reduction of the lattice constant ( filled circles ) are shown .
, width=8 ] the above - mentioned competition of the two effects is illustrated in fig .
[ fx ] . here
we plot the effective exchange interactions @xmath10 ( eq.[eqjij ] ) , which control directly the curie temperature of the system ( fig .
[ fx]a ) , and the bare exchange interactions @xmath40 , defined as @xmath41 as a function of the distance between the mn atoms in ni@xmath0mnsn .
we consider the ambient pressure case , as well as elevated pressure simulated by reduction of the lattice constant . in eq.([eqjbare
] ) @xmath42 denotes the size of the local magnetic moment of mn atom at site @xmath11 ( fig .
this definition is motivated by the presence of the exchange splittings @xmath43 in eq .
( [ eqjij ] ) which are roughly proportional to the moment magnitudes @xmath42 .
the increase of the bare exchange interactions with pressure is clearly seen for the first five neighbors ( see fig .
it is also obvious that magnetic moments decrease with the pressure , as expected .
these two effects , i.e. , enhancement of @xmath40 and suppression of magnetic moments , both due to the volume decrease , compete with each other , resulting in smaller pressure - induced changes of the effective interactions @xmath10 ( fig .
the final result in the present case is an enhancement of @xmath5 under hydrostatic pressure .
the same qualitative explanation was put forward by csontos @xcite for the increase of @xmath5 with pressure in some diluted magnetic semiconductors . in this section
we present and compare results for ( ni , pd)@xmath0mnsn and ( ni , cu)@xmath0mnsn heusler alloys , with ni@xmath0mnsn considered as the reference case .
the heusler alloy ni@xmath0mnsn was studied theoretically in great detail in a recent paper , @xcite while corresponding experimental results can be found in refs . and .
the experiments give the rate of increase of @xmath5 with pressure around 0.62 k / kbar @xcite ( for pressures up to 10 kbar or 1 gpa ) and 0.744 k / kbar for higher pressures up to 9 gpa .
@xcite the authors of the theoretical study @xcite employ the frozen - magnon approach and inverse lattice fourier transform to construct the effective heisenberg hamiltonian , which is then analyzed in the framework of the multisublattice mfa .
they obtain an increase of @xmath5 by 38 k for pressures around 16 gpa , an increase from the ambient value @xmath5=362 k to @xmath5=400 k. our estimated values are @xmath5=322 ( 334 ) k for ambient pressure and @xmath5=375 ( 401 ) k ( an increase of 53 ( 67 ) k for 3% reduction of the lattice constant , roughly corresponding to the same pressure @xmath26 16 gpa ) .
the values in brackets correspond to the model assuming electron correlations on ni - sites ( the effective hubbard @xmath44 chosen is 2 ev ) . assuming linear relation between @xmath5 and pressure , the experimental increase is expected to be around 100@xmath45120 k. the theory thus correctly predicts a pressure - induced increase of @xmath5 , while the absolute increase is at least three times ( ref . ) to [email protected] ( present study ) too small .
the experimental data for pd@xmath0mnsn heusler alloys that we have come across in the literature @xcite also indicate an increase of @xmath5 with pressure at a rate of 0.75 k / kbar , the same @xcite or almost the same @xcite as that for ni@xmath0mnsn . we have correctly obtained an increase of @xmath5 for 3% reduction of the ambient pressure lattice constant , as in the case of ni@xmath0mnsn .
specifically , @xmath5 is increased by 40 k , a value smaller than but still comparable to that for ni@xmath0mnsn and in reasonable agreement with experiment .
we find that any additional ( i.e. , in addition to that given by lsda ) coulomb term @xmath31 on pd - sites has a negligible effect on @xmath35 or its pressure - dependence .
the hockey - stick - like variation of the composition dependence of @xmath35 for ni@xmath1,pd@xmath46)@xmath0mnsn alloys at ambient pressure is captured in the rpa , but not in the mfa .
the rpa is equivalent to including the reaction field ( or the onsager cavity field ) effect in the calculation of @xmath35 ( see , for example , cyrot@xcite and references therein ) .
our results point to the importance of including the reaction field effect in reproducing the correct ( experimentally observed ) composition - dependence of @xmath35 in this alloy system .
the temperature can influence the above results , as discussed at the end of section [ form ] .
increasing temperature leads to an expansion of the lattice , compensating somewhat for the pressure - caused contraction .
in essence , thus , the above reasoning should be applied to a net effective reduction of volume , considering both temperature and pressure effects .
lattice expansion effects and the effect of including the fermi - dirac distribution function in performing the exchange integrals in eq .
( 2 ) have been discussed recently by alling @xcite . however , a complete @xmath29 calculation , as described in section [ form ] , has not yet been done .
the results of the pressure - dependence of @xmath5 for ( ni , pd)@xmath0mnsn alloys are shown in fig . [ f1 ] .
we have already mentioned that the ratio by which @xmath5 increases with pressure for the end - point alloys is similar for both systems .
this is also reflected in the concentration - dependence of @xmath5 , namely a similar increase of @xmath5 under pressure over the whole concentration range .
the monotonic and essentially linear concentration - dependence of @xmath5 under pressure reflects a similar trend of exchange interactions found for the case of ambient pressure ( see ref . ) .
this similarity is due to the isoelectronic nature of ni- and pd - atoms coming from the same column , but different rows ( @xmath47- and @xmath48-@xmath3 , respectively ) of the periodic table .
, pd@xmath46)@xmath0mnsn alloys for ambient pressure and for the pressure corresponding to the reduction of the alloy lattice constant by 3% .
model assumes electron correlations on ni - sites treated in the framework of the lsda+u method.,width=245 ] results of a similar study for ( ni , cu)@xmath0mnsn alloy over a set of pressures realized by a linear reduction of the alloy lattice constant from the ambient value up to 3% are shown in fig .
[ f2 ] . , cu@xmath46)@xmath0mnsn alloys for ambient pressure and for set of pressures corresponding to indicated reductions of the alloy lattice constant
. model assumes electron correlations on ni - sites treated in the framework of the lsda+u method .
, width=245 ] for comparison with experiments , results for the ambient pressure are slightly improved for the present lsda+u model with respect to the lsda results of our previous study @xcite .
however , both lsda and lsda+u reveal the same general trend .
specifically , we observe two concentration regions , the first one for @xmath490.4 in which @xmath5 is essentially constant with a shallow minimum and that for @xmath500.4 where @xmath5 increases monotonically with @xmath51
. with increasing pressure the behavior in the first concentration region remains unchanged , but there is a gradual reduction of the slope in the other concentration region ( the cu - rich end ) .
such a development is clearly related to opposite trends in the pressure - dependence of @xmath5 in ni@xmath0mnsn and cu@xmath0mnsn , a reduction of @xmath5 with pressure for cu@xmath0mnsn and an increase for ni@xmath0mnsn .
the crossover from the positive derivative of the pressure - dependence of @xmath5 to the negative one takes place around @xmath51=0.7 .
thus , for the pressure corresponding to a 3% reduction of the lattice constant we predict only a weak dependence of @xmath52 on the alloy composition . in this subsection we will present a qualitative understanding of the composition - dependence of @xmath35 at ambient pressure as well as the effect of pressure on @xmath5 using the idea of anderson s superexchange @xcite interaction and stearns @xcite model of the indirect ( rkky - type ) interaction in ferromagnetic heusler alloys .
the mn - mn exchange interaction , as in most other cases , can be divided into three groups : direct , indirect and superexchange . because the mn - atoms are not the nearest neighbors of themselves and have a large separation ( @xmath53 ) , the direct interaction is not of any importance in heusler alloys .
the most important interaction turns out to be the rkky - type indirect interaction@xcite .
of somewhat minor importance is the superexchange interaction , introduced in a series of papers by anderson@xcite .
superexchange , which is also an indirect interaction , is antiferromagnetic , while the rkky - type interaction can be ferromagnetic or antiferromagnetic .
sasioglu @xcite recently used these two interactions to interpret the results for some heusler alloys .
superexchange is related to possible transitions from occupied states at the fermi level @xmath27 to unoccupied levels above , in the present case to the unoccupied minority mn - bands . because the superexchange interaction is negative , the stronger the superexchange part , the smaller is the total exchange integral ( and @xmath5 ) .
the strength of the superexchange interaction depends on the density of states ( dos ) at @xmath27 and the energy separation between @xmath27 and the unoccupied mn - bands
. the larger the dos at @xmath27 and/or the smaller the above energy separation , the stronger is the superexchange .
this lends a qualitative understanding of the increase of @xmath5 from pd@xmath0mnsn to ni@xmath0mnsn to cu@xmath0mnsn , if the dos mechanism dominates over the decreasing of energy separation in the above sequence of alloys ( fig.2 of ref . ) .
if we dope the ni - sublattices with cu - atoms in the ni@xmath0mnsn host , the observed hockey - stick - like composition - dependence of @xmath35 can be understood as a competition of above two effects : for low cu - content the decrease of @xmath5 with increasing cu - concentration is due to reduction of the above energy separation , causing an increase of the superexchange part of the interaction . for @xmath54
the dos - effect starts to dominate , with decreasing dos at @xmath27 and thus smaller superexchange and larger total exchange interaction .
this leads to higher @xmath5 of cu@xmath0mnsn as compared to ni@xmath0mnsn . in ( ni , pd)@xmath0mnsn alloy the dos - effect is the only important effect over the entire concentration range . because the ni- and pd - atoms have the same valency , the above mentioned energy separation does not change with concentration .
there is a small decrease in the dos at @xmath27 in pd@xmath0mnsn compared with ni@xmath0mnsn ( fig.2 of ref . ) , decreasing the superexchange and increasing @xmath5 . as mentioned above ,
although the superexchange mechanism seems to support the observed results , it can not be the driving mechanism for the heusler alloys under study .
this is because of relatively large separation of the unoccupied minority mn - peak above @xmath27 ( the mn - dos just above @xmath27 is small ) . of much - larger importance
is the indirect interaction and its key feature in ferromagnetic heusler alloys was captured and successfully described by stearns@xcite in a series of papers in the late seventies .
stearns asserts that the magnetism of heusler alloys ( as well as bcc fe@xcite ) can be understood by dividing the @xmath3-electrons into localized @xmath3-electrons ( @xmath55 ) associated with narrow @xmath3-subbands and itinerant @xmath3-electrons ( @xmath56 ) associated with ( usually ) one broader @xmath3-subband .
the interaction between these localized and itinerant @xmath3-electrons is similar in nature to rkky interaction , originally formulated for the interaction between localized moments in a free electron gas .
the @xmath3-band gets narrower across ( with increasing number of electrons ) a given series and broader with increasing row number ( e.g. from @xmath57 to @xmath58 ) .
thus stearns used her model to describe correctly the change of @xmath35 from ni@xmath7mnsn to cu@xmath7mnsn ( an increase in @xmath55 causing a rise of @xmath35 ) and from ni@xmath7mnsn to pd@xmath7mnsn ( a decrease of @xmath35 due to decreasing @xmath55 ) .
uhl@xcite used stearns model to explain the hockey - stick appearance of the composition - dependence of @xmath35 with increasing @xmath51 in ( ni@xmath1,cu@xmath46)@xmath0mnsn heusler alloys . with increasing cu - concentration , the lattice dilates .
initially , this dilation causes a reduction in the strength of the @xmath56-@xmath55 interaction , associated with increased mn - mn distance .
however , the addition of cu also causes an increase in the number of @xmath55 and thus in the @xmath56-@xmath55 interaction .
the resulting @xmath35 thus shows an initial decrease followed subsequently by an increase ( for further details , see uhl @xcite ) .
thus both superexchange and indirect @xmath56-@xmath55 exchange interaction give rise to the same trend for the composition - dependence of @xmath35 .
the pressure effect on @xmath35 can be understood in the light of the above explanation for the composition - dependence at ambient pressure .
pressure reduces the lattice parameter , an effect opposite to the dilation caused by cu - addition .
pressure also broadens bands , reducing @xmath55 and increasing @xmath56 , again opposite to what happens with increasing cu - content .
thus on the ni - rich side increased pressure increases @xmath35 ( opposite to the result with cu - addition ) , and beyond a certain critical cu - concentration increasing pressure results in decreasing @xmath35 ( opposite to the result with cu - addition ) .
similar reasoning explains the composition - dependence of ambient pressure @xmath35 and the pressure - dependence of @xmath35 in ( ni@xmath1,pd@xmath46)@xmath0mnsn alloys . doping ni - sublattice with pd - atoms results in dilation of the lattice .
in addition , d - electrons in pd are more delocalized than in ni
. both of these will lead to a decrease of @xmath35 with increasing pd concentration .
pressure , via contraction of the lattice , reverses this effect .
it appears that in this case the increasing strength of the interaction due to reduction in the mn - mn distance supersedes the effect of the pressure - induced delocalization of the electrons .
a couple of comments are in order .
stearns model is applicable to only ferromagnetic heusler alloys .
in several semi - heusler alloys , or in general in situations where the superexchange part of the interaction becomes important , stearns model may fail to predict the results correctly
. it would not be incorrect to point out that without an accurate first - principles study as presented here , it would be almost impossible to capture theoretically the details of the composition- or pressure - dependence of @xmath35 , although some qualitative understanding ( especially , _ a posteriori _ ) is possible . in this sense
the importance of first - principles studies like the present one is irrefutable . in magnetic alloys
there are three contributions to the resistivity : a temperature - independent residual resistivity due to the alloy disorder and other defects , and two temperature - dependent contributions , one due to electron - phonon and the other due to electron - magnon scatterings . in the present case ,
the residual resistivity is small because the states corresponding to ni- , cu- , and pd - atoms are well below the alloy fermi energy , @xcite and thus influence states at the fermi energy ( relevant for residual resistivity ) only weakly . the contribution due to phonons is well understood , and for heusler alloys around room temperature , @xcite this contribution is known to be small .
the dominant contribution is thus due to spin - disorder .
@xcite the spin - disorder induced resistivity is controlled by the spin - spin correlation function which is small at and above @xmath5 , so that a reasonable representation of the situation is given by the dlm state , as explained in detail elsewhere .
@xcite kasuya s @xmath59-@xmath3 interaction model@xcite shows that the temperature - dependence of spin - disorder resistivity is primarily quadratic and can be well - simulated by the @xmath60-law . in the present case
we determine the constant @xmath61 from first principles using @xmath62 , where @xmath63 is identified with the resistivity of the dlm state .
the dlm state is described in the framework of the cpa as an equiconcentration alloy with two types of mn atoms ( a and b ) with their moments pointing randomly up or down ( collinear disorder ) .
the corresponding resistivity can be properly calculated by the kubo - greenwood formula applied to this binary - mn alloy .
the results for ni@xmath0mnsn , ( ni@xmath64,pd@xmath64)mnsn , and pd@xmath0mnsn alloys are shown in fig .
[ f3 ] . , pd@xmath46)@xmath7mnsn alloys for ambient pressure ( full symbols ) and the pressure corresponding to the reduction of the alloy lattice constant by 3% ( open symbols ) : ( a ) @xmath65 , ( b ) @xmath66 , ( c ) @xmath67 . , width=245 ]
it should be noted that the residual resistivity of the ( ni@xmath64,pd@xmath64)mnsn alloy in the ferromagnetic state is about two orders of magnitude smaller than that in the dlm state .
@xcite the spin - disorder part of the resistivity is controlled by the constant @xmath61 . for the cases shown in fig .
[ f3 ] , @xmath5 increases under pressure .
due to band - broadening under pressure and accompanying delocalization of the electrons , @xmath68 decreases in general for a given @xmath69 less than the ambient pressure @xmath35 .
however , @xmath63 ( as well as @xmath68 for @xmath70 ambient pressure @xmath35 ) is higher under pressure because of the increase in @xmath35 . to our knowledge , there are no available experimental results for the systems studied in this paper , although the results of austin and mishra @xcite for related pd@xmath0mnsb heusler alloy do indicate , as in this work as well , an increase of the resistivity under pressure for temperatures above @xmath5 .
undoubtedly pressure has non - negligible influence on the part of the resistivity due to phonons .
this , however , is beyond the scope of the present article .
we have studied magnetic and transport properties of quaternary heusler alloys ( ni , t)@xmath0mnsn ( t = cu , pd ) under pressure by means of the first - principles density functional method .
in particular , we have investigated in detail the pressure - dependence of the curie temperature . in agreement with experiments ,
we obtain an increase of the curie temperature under pressure ( reduction of the lattice constant ) for ni@xmath0mnsn and pd@xmath0mnsn . on the other hand ,
cu@xmath0mnsn alloy exhibits a reduction of the curie temperature with applied pressure .
the results can be qualitatively understood as an interplay of two effects : an increase of the bare exchange integrals with the volume reduction and the decrease of magnetic moments with pressure due to band broadening .
the concentration - dependence of the curie temperature under pressure is simple in ( ni , pd)@xmath0mnsn heusler alloys where the increase of the curie temperature has an almost constant slope . on the other hand ,
we predict a dramatic change of the curie temperature with pressure from the positive for ni - rich ( ni , pd)@xmath0mnsn alloys to a negative one in cu - rich alloys , the crossover being around 70 % of cu .
we emphasize that such a complex behavior can not be obtained without an accurate calculation of the exchange interaction and a proper statistical treatment to compute @xmath35 ( i.e. , by going beyond the mfa ) . a simple explanation of the behavior of the calculated @xmath5 under pressure is given in the framework of anderson s superexchange interaction and the stearns model of the indirect exchange interaction between itinerant and localized @xmath3-electrons .
we have also investigated the pressure - dependence of the spin - disorder related part of the resistivity , which dominates at higher temperatures , the residual resistivity due to alloy disorder being small and important only for very low temperatures .
we have found that the resistivity at the curie temperature increases with pressure , but this result is due to the increase of the curie temperature with pressure itself . for temperatures below the ambient pressure
@xmath5 the resistivity decreases under pressure , as would be expected due to band broadening .
this work was supported by a grant from the natural sciences and engineering research council of canada .
j.k . and v.d .
acknowledge financial support from av0z 10100520 and the czech science foundation ( 202/09/0775 ) .
the work of i.t . has been supported by the ministry of education of the czech republic ( grant no .
msm 0021620834 ) and by the czech science foundation ( grant no .
202/09/0030 ) .
k. ozdogan , e. sasioglu , and i. galanakis , j. appl . phys .
* 103 * , 023503 ( 2008 ) .
o.k . andersen and o. jepsen , phys .
lett . * 53 * , 2571 ( 1984 ) .
i. turek , v. drchal , j. kudrnovsk , m. ob , and p. weinberger , _ electronic structure of disordered alloys , surfaces and interfaces _
( kluwer , boston , 1997 ) .
vosko , l. wilk , and m. nusair , can .
* 58 * , 1200 ( 1980 ) .
liechtenstein , m.i .
katsnelson , v.p .
antropov , and v.a .
gubanov , j. magn .
magn . mater . *
67 * , 65 ( 1987 ) .
i. turek , j. kudrnovsk , v. drchal , l. szunyogh , and p. weinberger , phys .
b * 65 * , 125101 ( 2002 ) .
m. cyrot , j. mag . mag . mater . * 45 * , 9 ( 1984 ) .
anderson , phys . rev . * 124 * , 41 ( 1961 ) .
stearns , j. appl . phys . * 50 * , 2060 ( 1979 ) ; see also j.r .
reitz and m.b .
stearns , _ ibid .
_ , 2066 ; m.b .
stearns , j. mag . mag .
15 - 18 * , 301 ( 1980 ) .
stearns , phys . today , april 1978 issue , pp .
t. kasuya , prog .
* 16 * , 58 ( 1956 ) .
t. kasuya , prog .
. phys . * 22 * , 227 ( 1959 ) . | using first - principles electronic structure calculations , we have studied the dependence of the curie temperature on external hydrostatic pressure for random ni@xmath0mnsn heusler alloys doped with cu- and pd - atoms , over the entire range of dopant concentrations .
the curie temperatures are calculated by applying random - phase approximation to the heisenberg hamiltonian whose parameters are determined using the linear response and multiple scattering methods , based on density - functional theory . in ( ni@xmath1,pd@xmath2)@xmath0mnsn alloys the curie temperature
is found to increase with applied pressure over the whole concentration range .
the crossover from the increase to the decrease of the curie temperature with pressure takes place for cu - concentrations larger than about 70% in ( ni@xmath1,cu@xmath2)@xmath0mnsn heusler alloys .
the results for the reference ni@xmath0mnsn heusler alloy agree well with a previous theoretical study of sasioglu _ et al . _ and also reasonably well with available experimental data .
results for the spin - disorder induced part of the resistivity in ( ni@xmath1,pd@xmath2)@xmath0mnsn heusler alloys , calculated by using the disordered local moment model , are also presented .
finally , a qualitative understanding of the results , based on anderson s superexchange interaction and stearn s model of the indirect exchange interaction between localized and itinerant @xmath3-electrons , is provided . |
in the present paper we will consider the relation between the deep inelastic structure function @xmath7 and the gluon density @xmath8 in the region of small @xmath2 . at very small
@xmath2 the main contribution to @xmath9 comes from the sea quarks and the value of dis cross section is given by the photon - gluon fusion subprocess fig .
1 . for large @xmath10 in the leading log @xmath10 approximation ( ll@xmath11a ) and in a physical ( say , axial ) gauge it is enough to deal only with the diagram fig .
however within the llog(@xmath12)a , which sums up all the terms of the type @xmath13 , one has to take into account the graph fig .
1b also . based on these graphs fig
. 1 one can extract the value of the gluon structure function @xmath14 from the experiment , where the dis structure function @xmath7 was measured .
the question we are going to discuss here in more detail is : what are the values of gluon kinematical variables @xmath15 and @xmath16 which correspond to the arguments @xmath17 and @xmath10 of the dis cross section ; i.e.@xmath9 ? in sect .
2 we will remind the ll@xmath18a formulae @xcite for the function @xmath9 ( and @xmath19 ) written in the framework of the high energy @xmath20-factorization scheme @xcite , where the gluon @xmath21 ( see fig .
1 ) is off mass - shell . then by the numerical computation in sect .
3 we will study the ratios of the deep inelastic ( bjorken ) variables ( @xmath22 and @xmath10 ) and the values of the gluon s momenta ( @xmath15 and @xmath16 correspondingly ) , which are essential in the integrals of eqs . ( 3 ) and ( 13 ) ; and how do these ratios depend on the form of the gluon structure function . to answer the last question we parametrize the gluon density by the power functions - @xmath23 and
plot the logarithms of the ratios @xmath24 as the functions of powers @xmath25 and @xmath26 .
even in the case of a strong scaling violation ( @xmath27 ) the average value @xmath28 turns out to be rather large .
it corresponds to @xmath29 gev@xmath5 for @xmath30 gev@xmath5 .
thus there appears a danger that an important part of @xmath7 is originated from the region of a small gluon s transverse momentum ( virtualities ) ( @xmath31 gev ) , where the perturbative qcd approach is not justified . to follow the role of the infrared ( small @xmath32 ) region we have changed also the mass of a light quark
was discussed in the previous paper [ 1].].the value of @xmath9 increases up to the factor of @xmath33 when instead of the constituent mass @xmath34 mev one puts the small ( close to the current ) mass @xmath35 mev . the situation becomes slightly better if we assume the saturation of gluon distributions at small @xmath2 and @xmath16 . at very small @xmath2 and
@xmath16 the cross section @xmath36 should tend to a constant .
thus at @xmath37 ( for example we put here @xmath38 gev@xmath5 ) the gluon density @xmath39 ; in other words at @xmath40 the anomalous dimension @xmath41 . in this case the rapid decrease of @xmath14 at @xmath42 suppresses the role of the infrared region and the result becomes more stable ( under the variation of @xmath43 ) .
this point will be discussed also in sect .
3 . finally we have to stress that the infrared ( @xmath44 gev ) region gives not too small contribution to the @xmath45 ratio and one has to remember about it , trying to extract the gluon structure function from the @xmath7 experimental data .
the cross section of heavy quarks as well as light quarks ( if @xmath10 value is large enough ) electroproduction is given schematically by the graphs in fig .
the main contribution to the cross section at small @xmath2 is known to come from gluons .
the lower ladder blocks presents the two - dimensional gluon distribution @xmath46 which is a function of the fraction @xmath2 of the longitudinal momentum of the initial hadron and the gluon virtuality .
its distribution over @xmath2 and transverse momenta @xmath47 in hadron is given in the semihard theory @xcite by function @xmath46 .
it differs from the usual gluon structure function @xmath48 : @xmath49 such definition of @xmath46 makes possible to treat correctly the effects arising from gluons virtualities . the exact expression for this function
can be obtained as a solution of the evolution equation which , contrary to the parton model case , is nonlinear due to interactions between the partons in small @xmath2 region . in what follows we use sudakov decomposition for the produced quarks momenta @xmath50 through the light - like momenta of colliding particles @xmath51 and @xmath52 and transverse ones @xmath53 : @xmath54 the differential cross sections of a quark pair electroproduction
denotes the virtual photon - proton transverse and longitudinal cross sections . ]
have the form to be equal to gluon virtuality , which is very close to the blm scheme@xcite;(see also @xcite ) . ] : @xmath55 @xmath56 here @xmath57 , @xmath21 is the gluon transverse momentum and @xmath58 is the quarks rapidity in the @xmath59 c.m.s .
frames , @xmath60 @xmath61 are the squares of the matrix elements@xcite for the cases of transverse and longitudinal electroproduction , respectively . in lla kinematic @xmath62 so @xmath63 ( the more accurate relation is @xmath64 but we are working in the kinematics where @xmath65 ) .
the matrix elements @xmath66 are calculated in the born order of qcd without standart simplifications of the parton model since in the small @xmath2 domain there are no grounds for neglecting the transverse momenta of the gluon @xmath67 in comparison with the quark masses and transverse momenta , especially in the case of light quark pair production . in the axial gauge @xmath68 the gluon propagator takes the form @xmath69 @xmath70 for the gluons in @xmath71-channel the main contribution comes from the so called nonsense polarization @xmath72 , which can be picked out by decomposing the numerator into longitudinal and transverse parts : @xmath73
the other contributions are suppressed by the powers of @xmath74 . since the sum of the diagrams in fig .
1 are gauge invariant in the lla , the transversality condition for the ends of gluon line enables one to replace @xmath75 by @xmath76 in expression for @xmath72 .
thus we get @xmath77 or @xmath78 if we do such a trick for vector @xmath79 too .
both these equations for @xmath80 can be used but for the form ( [ trans1 ] ) one has to modify the gluon vertex slightly ( to account for several ways of gluon emission
@xmath81.\ ] ] as a result the colliding gluons can be treated as aligned ones and their polarization vectors are directed along the transverse momenta .
ultimately , the nontrivial azimuthal correlations must arise between the transverse momenta @xmath82 and @xmath83 of the produced quarks . in the case of photoproduction the alignment of the gluon
makes the distribution of the difference of the transverse momenta of the quarks @xmath84 to be proportional to @xmath85 , where @xmath86 is the angle between the vectors @xmath87 and @xmath88 .
although the situation considered here seems to be quite opposite to the parton model there is a certain limit in which our formulae can be transformed into parton model ones .
let us consider the case of @xmath89 collisions and assume now that the characteristic values of quarks momenta @xmath82 and @xmath83 are many times larger than the values of gluons momenta , @xmath90 and @xmath91 @xmath92 and one can keep only lowest powers of @xmath93 .
it means that we can put @xmath94 everywhere in the matrix element @xmath66 except the vertices , as it was done in the parton model approach . on the other hand ,
the used assumption ( [ par1 ] ) is not fulfilled in a more or less realistic case .
the transverse momenta of produced light quarks should be of the order of gluon virtualities ( qcd scale values ) .
the function @xmath95 is unknown at small values of @xmath16
. therefore we rewrite the integrals over @xmath96 in eq .
( [ sel ] ) in the form @xmath97
@xmath98 @xmath99 where eq .
( [ xg ] ) is used . using eq .
( [ sel ] ) and eq .
( [ int ] ) we perform the numerical calculation of light quark production .
the results of computations are presented in figs .
2 - 8 . in figs . 2 and 3
we plot the values of @xmath7 coming from the sea @xmath100 quark contribution in the cases of mt(s - dis ) @xcite and grv ho @xcite gluon distributions , respectively .
the parameter @xmath101 in eq .
( 13 ) which separates the region of small and large gluon momenta @xmath20 was chosen to be equal to 1 gev ( figs .
2a and 3a ) and 2 gev ( figs . 2b and 3b ) .
it is seen that the results depend rather strongly on the value of @xmath101 as well as on the mass of light quark and in both cases are differed from the original mt(s - dis ) ( or grv ho ) predictions for @xmath100 sea quark contribution ( which , in our notations correspond to @xmath102 and @xmath103 ) and shown by solid curves in figs . 2 and 3 .
however the difference of our calculated @xmath7 values and the original ones is not large .
it means that both the procedure of the gluon density extraction from @xmath7 experimental data and the accuracy of our calculations are reasonably enough and there is no a large contribution of any ( say , nonperturbative ) corrections . the only explanations is the contribution of the small transverse momenta @xmath104 and @xmath105 to the final value of @xmath9 .
even for @xmath106 the difference between the cases of @xmath107 mev and 100 mev is of about 20% ( up to 30% for grv ho and @xmath108 ) at @xmath109 . to understand better what are the typical quark ( @xmath32 ) and gluon ( @xmath110 ) momenta , which are essential in the integrals eq .
( 13 ) , we have chose the parametrization of gluon distribution in the form @xmath111 and calculated the mean values of logarithms @xmath112 , @xmath113 and @xmath114 .
the advantage of the logarithms is that they does not destroy the main power structure of the integrand ( trying to estimate directly the value of @xmath115 or @xmath116 , one faces the divergence of the integral over @xmath117 at large ( small ) @xmath118 for the positive @xmath119 ) .
the ratio of the sea @xmath120-quark to the gluon densities are given in fig .
4 ( in terms of @xmath7 and @xmath8 structure functions ) for different values of @xmath26 ( 0.2 and 0.5 ) and @xmath25 ( 0.2 and 0.5 ) in eq . ( 14 ) . at very small @xmath121
this ratio tends to a constant and does not depend on @xmath2 .
however , even at large @xmath122 ( the value which corresponds to the asymptotics ( extremum ) of the bfkl solution ) when from the formal point of view the convergence of the integral in the infrared region should be good enough , the dependence of the result on the light quark mass is still not negligible ( @xmath12320% difference between the values of @xmath9 for @xmath43 = 350 mev and 10 mev at @xmath124 and @xmath10 = 16 gev@xmath5 ) .
the situation becomes slightly better ( see fig .
5 ) if one assumes that for small @xmath10 the gluon density falls down as @xmath125 ( say , @xmath26 = 1 for @xmath126 1 gev@xmath5 ) .
such a behaviour one might expect in the case of the gluon density saturation @xcite ( i.e. @xmath127 at @xmath128 ) .
unfortunately , the saturation hypothesis is also not enough to neglect completely the small @xmath32 contribution . increasing the quark mass from 10 mev to 350 mev one diminishs the cross section more than 15% ( for example , at @xmath129 = 16 gev@xmath5 , @xmath130 ; and 20% for @xmath131 , without the saturation ) .
this fact is confirmed by the @xmath26-dependence of the function @xmath132 ( fig .
7a ) . as it should be expected the value of @xmath132 depends slowly on @xmath25 but noticiably increases with @xmath26 . for a more heavy photon
( @xmath10 = 64 gev@xmath5 ) @xmath132 becomes smaller .
it means that the typical values of @xmath118 increases with @xmath10 not as quickly as @xmath10 ( more close to @xmath133 ) . in the best case ( @xmath130 for @xmath134 1 gev@xmath5 and @xmath135 for @xmath126 1 gev@xmath5 ) the value of @xmath132 ( @xmath136 at @xmath10 = 16 gev@xmath5 ) corresponds to @xmath137 , essential in the integral , of about 3 gev@xmath5 ( smaller than the usual cutoff @xmath138 gev@xmath5 for the mt or msr parametrizations @xcite of the structure functions ) . without the saturation at @xmath10 = 16 gev@xmath5 the @xmath139 ; i.e. typical @xmath140 0.8 gev@xmath5 . in order to estimate the values of the longitudinal gluon momenta ( @xmath15 )
we plot in fig .
8 the logarithms of the @xmath22 to @xmath15 ratios .
the last ratios increase with @xmath10 as for a small @xmath10 the transverse momenta of quarks sometimes becomes larger than the photon virtuality ( @xmath133 ) , see fig .
6 , and one needs a large @xmath15 of gluon to put these quarks on mass shell . in the region of @xmath141 16
gev@xmath5 the value @xmath142 ; it means that as an average the gluon @xmath15 is 3 times larger than the bjorken @xmath22 .
this result slightly differs from the prytz estimation @xcite where a gluon has @xmath143 .
the value of @xmath7 is the sum of transverse ( @xmath144 ) and longitudinal ( @xmath145 ) contributions .
as one can see from the table , at large @xmath10 the last contribution is more ifrared stable .
we hope that the ratio of @xmath9 to @xmath8 presented in figs .
4 , 5 as a functions of anomalous dimention @xmath26 and intercept @xmath25 may be used in future in order to extract , for example , the values of gluon density directly from the @xmath9 structure function ( in this case the values of @xmath26 and @xmath25 are also come from the @xmath9 data as @xmath146 and @xmath147 .
the results we have presented here are close to the standard gluon densities extracted from the derivative @xmath148 ( using the prytz method @xcite , or the modified one , taken into account the nlo corrections @xcite ) .
nevertheless , here we better understand the typical values of gluon momenta we deal with and can control the role of the small @xmath47 ( infrared ) region .
the mean logarithms @xmath149 and @xmath150 give us an estimate of the quark transverse momenta which are essential in the process , and the values of gluon virtuality ( @xmath16 ) and proton momentum fraction ( @xmath15 ) which correspond to the point @xmath22 and @xmath10 for the structure function @xmath9
. a lesson one has to extract from these calculations is the fact that an essential values of transverse momenta ( @xmath32 ) which do contribute to the next - to - leading order ( nlo ) corrections may be quite differ from the initial @xmath10 . at not too large @xmath10 the value of @xmath151 may be even larger than @xmath10 ( in fig .
6 @xmath152 for @xmath153 16 gev@xmath5 .
however the main danger comes from the cases where the virtuality in the internal loops is much smaller than @xmath10 .
for example when ( as it was done in @xcite ) one includes in the anomalous dimension @xmath154 the nlo corrections of the type @xmath155 in the first non - zero ( 4-loops ) gluon - gluon correction @xmath156 , he touches the region of the internal @xmath157 ; down to @xmath158 due to the large numerical coefficient @xmath159 in the lipatov @xcite diffusion - like distribution @xmath160 = \exp[-\ln^2(q^2/q^2)/(a\cdot n)]$ ] where @xmath161 and @xmath162 is the number of loops .
after this we use the value of the anomalous dimension @xmath154 ( including the nlo corrections ) as a local ( from the point of view of @xmath163 space ) contribution at the given point @xmath10 .
there are no problems if @xmath10 is very large but one can not to do this for not too large @xmath10 where the essential transverse momenta of intermidiate gluons may be rather small : close to the infrared region ( @xmath164 ) .
2 . @xmath166 sea quark contributions to @xmath7 as given by mt s - dis @xcite sea quark distributions ( solid curves ) and our calculations using mt s - dis gluon distributions for @xmath43 = 10 mev and 100 mev ( dashed and dash - dotted curves , respectively ) .
the values of @xmath101 parameter equal to 1 gev ( a ) and 2 gev ( b ) were used in eq .
( 13 ) 99 ryskin m.g .
, shabelski yu.m . , shuvaev a.g . : z.phys.c ( in print ) ; preprint santiago univ
. us - ft/11 - 95 ; hep - ph/9506338 .
catani s. , ciafaloni m. and hautmann f. : phys.lett .
b242 ( 1990 ) 97 ; nucl.phys .
b366 ( 1991 ) 135 ; phys.lett b307 ( 1993 ) 147 .
gribov l.v .
, levin e.m . and ryskin m.g .
: phys.rep . , 1983 , v.100 , p.1 ; levin e.m . and ryskin m.g .
: phys.rep . , 1990 , v.189 , p.267 .
brodsky s.j . , lepage g.p . and mackenzie p.b .
, 1983 , v.d28 , p.228 .
levin e.m . ,
ryskin m.g . , shabelski yu.m .
, shuvaev a.g . : yad.fiz .
1991 , v.54 , p.1420 .
morfin j.g . and
tung w .- k . : z.phys .
1991 , v.c52 , p.13 .
gluck m. , reya e. and vogt a. : z.phys .
1993 , v.c53 , p.127 .
martin a.d .
, stirling w.j .
and roberts r.g . :
preprint ral 94 - 055 , dtp/94/34 ( 1994 ) . prytz k. : phys.lett .
1993 , v.b311 , p.286 .
prytz k. : phys.lett .
1994 , v.b332 , p.393 .
ellis r.k .
, hautmann f. and webber b.r .
: preprint fnal - pub-95/006-t ( 1995 ) , cavendish - hep-94/18 ; ellis r.k .
, kunszt z. and levin e.m .
: nucl.phys . , 1994 , v.b420 , p.517 ; catani s. and hautmann f. : nucl.phys .
b427 ( 1994 ) 475 .
l.n.lipatov : sov.j.jetp 1986 , v.90 , p.1536 . | we consider the light quark electroproduction cross section and show that the contribution coming from the infrared ( @xmath0 ) region is not negligible .
it changes the @xmath1 ratio by 20 - 30% ( or even more ) for small @xmath2 and @xmath3 .
= 0.7 cm * us - ft/30 - 95 * .8 truecm * f*@xmath4*/xg(x , q*@xmath5 * ) ratio at small @xmath2 and the role of the contribution from infrared region * + m.g.ryskin and a.g.shuvaev + petersburg nuclear physics institute , + gatchina , st.petersburg 188350 russia + and + yu.m.shabelski@xmath6 + departamento de fisica de partculas , + universidade de santiago de compostela , + 15706santiago de compostela , spain e - mail [email protected] + e - mail [email protected] + e - mail [email protected] * ) permanent address : petersburg nuclear physics institute , + gatchina , sanct - petersburg 188350 , russia
. 1 .
truecm december 1995 + .1 truecm * us - ft/30 - 95 * |
in the last decade trapped yb^+^ ions have been a workhorse in the fields of quantum information processing @xcite , precision measurements @xcite and hybrid systems @xcite .
yb@xmath0 has a rich energy level structure because besides pure valence - electron excitations , also electrons from a closed f - shell can be excited .
this leads to unusual features such as the electric octupole transition @xmath2s@xmath11-@xmath2f@xmath12 , which is currently explored for the prospect of an ion based atomic clock , as it has a natural lifetime of several years @xcite . since the work of @xcite , a common feature of most trapped yb^+^ experiments is the continuous excitation of the ion within a four level system ( figure [ ybion_terms_spec ] ) .
the ion is cooled and detected on the @xmath2s@xmath11-@xmath2p@xmath3 transition near 369 nm .
a second decay channel to the metastable @xmath2d@xmath13 state necessitates a second laser to avoid interruptions of the fluorescence .
light near 935 nm repopulates the ground state via the @xmath5d@xmath6_{1/2}$ ] state . in this paper
, we extend the available laser manipulation toolbox for trapped yb^+^ ions by exciting the 4f@xmath16s @xmath2s@xmath3 - 4f@xmath45d6s @xmath5d@xmath6_{1/2}$ ] transition at 297 nm .
we perform laser spectroscopy and present frequency measurements for four isotopes .
the isotope shifts contain information about the change of the charge distribution within the nucleus and about the correlations of electrons , which are intricate because of a partially filled f - shell in yb^+^. therefore they can be used to test ab - initio calculations .
in addition , we show that the 297 nm light can be used as second the step in the photoionization process of neutral yb as well as for laser cooling of single yb^+^ ions
. the relatively narrow line width could potentially allow to reach lower laser cooling temperatures .
( branching ratios given in percentages ) .
@xmath14}$ ] , @xmath15 and branching ratio out of the @xmath2p@xmath3 state are taken from ref .
@xcite , @xmath14}$ ] from ref .
@xcite and the @xmath5d@xmath6_{1/2}$ ] branching ratio from ref . @xcite . ]
we trap single yb^+^ ions in a rf paul trap , which consists of two opposing endcap electrodes @xcite separated by 150 @xmath16 m .
the electrodes are formed from tungsten wire of 250 @xmath16 m diameter and are etched to a needle - like geometry @xcite . by applying a radio frequency signal of 20 mhz and a few hundred volts amplitude , the ion is trapped with secular trap frequencies of @xmath17 mhz and @xmath18 mhz .
single ions are loaded into the trap by two - step photoionization from a thermal atomic beam @xcite . a stainless steel tube filled with yb is resistively heated by a short current pulse ( typically 60 - 80 ms , 60 a ) and emits neutral yb atoms into the trap
. the atoms are resonantly excited on the @xmath19s@xmath20-@xmath19p@xmath21 transition with laser light at 398 nm and subsequently ionized by a laser at 369 nm .
the first excitation step is isotope selective .
after ionization , the ion is continuously laser cooled on the @xmath2s@xmath11-@xmath2p@xmath3 transition and the resulting fluorescence is collected by an in - vacuum objective and detected by a photomultiplier tube . the laser system for exciting the @xmath2s@xmath3-@xmath5d@xmath6_{1/2}$ ] transition at 297 nm is based on sum frequency generation ( sfg ) of light at 532 nm and 672 nm .
the sketch of the optical setup is shown in figure [ green_setup ] .
the green 532 nm laser is a commercial system ( coherent verdi v6 ) , whereas the red 672 nm diode laser with subsequent amplification stage is homebuilt . in order to supply an atomic frequency reference ,
the green laser is stabilized to the r(56)32 - 0 transition in molecular iodine @xcite .
as the study of isotope shifts in yb^+^ requires a large frequency scan range , we use the doppler broadened absorption line ( at room temperature : line width @xmath22 ghz ) and generate an error signal by fm spectroscopy @xcite .
we measured the rms frequency fluctuations of the green laser to be about 20 mhz .
we achieve type - i ( ooe ) sfg in a doubly resonant bow - tie cavity ( finesse : 75 , free spectral range : @xmath23 ghz ) .
the cavity length is locked to the 532 nm laser using the hnsch - couillaud stabilization technique @xcite , whereas the 672 nm laser follows the sfg cavity by the pound - drever - hall locking technique @xcite . by sweeping the green 532 nm laser we can scan the generated uv light continuously about 1 ghz .
input light intensities of 1 w ( 532 nm ) and 300 mw ( 672 nm ) yield about 10 mw of 297 nm light .
spectroscopy on the transition is performed in a double - resonance scheme .
the laser at 369 nm cools the ion continuously on the transition and yields fluorescence , which is detected .
simultaneously , the 297 nm probe laser excites the transition .
additional 935 nm light is necessary to deplete the population of the state . by sweeping the frequency of the 297 nm light ,
the resonance is observed as a drop in the 369 nm fluorescence ( figure [ typicalspec ] ) .
the ratio of the saturation parameters , @xmath24 and @xmath25 ( @xmath26 , with the ( saturation ) intensity @xmath27 ( @xmath28 ) ) is important for the strength of the observed signal . for @xmath29 the excitation into the @xmath5d@xmath6_{1/2}$ ] is much slower than the probing of the @xmath2s@xmath3 state due to the 369 nm laser , which effectively suppresses the transition to @xmath5d@xmath6_{1/2}$ ] in a quantum zeno like manner .
in contrast , the signal of the 369 nm @xmath2s@xmath11-@xmath2p@xmath3 fluorescence shows a significant drop for similar saturation parameters @xmath30 .
the steady state fluorescence is proportional to the population of the @xmath2p@xmath3 state , which is given by @xmath31 assuming fast repumping out of the @xmath2d@xmath13 state and neglecting coherences . here
@xmath32 is the detuning of the 369 nm laser and @xmath33 the natural line width of the @xmath2p@xmath3 state .
the term @xmath34 describes the coupling to the @xmath5d@xmath6_{1/2}$ ] state @xmath35 } } \right)^2 } , \ ] ] where @xmath36 is the detuning of the 297 nm laser and @xmath14}$ ] the natural line width of the @xmath5d@xmath6_{1/2}$ ] state . for typical experimental parameters of @xmath37 and @xmath38 ,
this results in a 20@xmath39 - 30% drop of the fluorescence , which is consistent with our measured spectra .
we typically scan the 297 nm light 400 mhz in 10 s and record the 369 nm fluorescence .
the line width of the 297 nm laser , residual micromotion and the continuous driving of the transition broaden the linewidth , which is measured to be 40 mhz .
the absolute frequency is determined by measuring the frequency of the 532 nm and the 672 nm laser with a wave meter ( highfinesse ws/07 ) .
the wave meter is calibrated to the d@xmath40 ( f@xmath41 , f@xmath42 ) crossover line in @xmath43rb and has a specified 3-@xmath44 accuracy of 60 mhz . in this manner we measured the frequencies for the isotopes @xmath7yb@xmath0 , @xmath8yb@xmath0 , @xmath9yb^+^ and @xmath10yb^+^ ( table [ tablefreq ] ) .
.measured frequencies of the @xmath2s@xmath3-@xmath5d@xmath6_{1/2}$ ] transition in yb@xmath0 ( 1-@xmath44 error ) . [ cols="<,<",options="header " , ] our results can be interpreted in comparison with the data of zinkstok et al .
@xcite , who investigated a similar line in neutral yb , namely the 4f@xmath16s@xmath2 @xmath19s@xmath45 to 4f@xmath45d6s@xmath2 @xmath19p@xmath46 transition .
they found a similarly large negative specific mass shift , which is explained by the strong coupling of d- and f - electrons @xcite .
in contrast , our measured field shift is about 50% higher .
this may reflect the fact that , in general the electron charge density at the nucleus is higher in ionic systems than in their neutral counterparts and thereby more sensitive to changes of the charge distribution within the nucleus .
photoionization of yb can be achieved by a photon at 398 nm ( @xmath19s@xmath20-@xmath19p@xmath21 transition ) and a second photon with a wavelength smaller than 394 nm . here
we demonstrate that light at 297 nm can be used for the second step . at the beginning of the loading sequence
the yb oven is heated for 75 ms and simultaneously the 398 nm , the 297 nm and the 935 nm laser lights are switched on .
after three seconds the light at 398 nm is turned off and one second later the 369 nm laser illuminates the ion and the resulting fluorescence is recorded ( figure [ loading ] ) .
we note that for reliable loading the oven heating time needs to be slightly longer compared to the standard loading with 369 nm light .
blue line : expected fluorescence of an 100 k hot ion @xcite .
black curve : sweep of @xmath47 from -200 mhz to -60 mhz in 3 s after heating the oven leads to a high initial level of 369 nm fluorescence indicating a cold ion . in both cases @xmath48 . ] the fluorescence curve after switching on the 369 nm light depends strongly on the kinetic energy of the ion due to the doppler shift and can be used to determine its temperature @xcite . in order to measure the onset of the fluorescence
, we interrupt the 935 nm light 10 ms before the 369 nm light is switched on .
this allows the 369 nm intensity controller to adjust without affecting the ion .
we switch the 935 nm light back on 100 ms later . only with both lights ,
935 nm and 369 nm , the excitation cycle is closed and the ion continuously scatters photons .
the inset in figure [ loading ] shows the onset of the fluorescence for two different loading sequences . in the first the 297 nm laser
is set 400 mhz below the resonance frequency during the loading ( yellow curve ) . because of the large detuning the ion scatters only few photons on the @xmath2s@xmath3-@xmath5d@xmath6_{1/2}$ ] transition and the cooling of the ion is very inefficient .
thus , the 369 nm fluorescence starts low due to doppler broadening and increases as the 369 nm laser cools the ion . in the second loading sequence , a smooth shifting of the 297 nm laser frequency towards the resonance ( from @xmath49 mhz to @xmath50 mhz ) in the first three seconds after photoionization yields a photon scatter rate , which starts immediately at its saturation value and hence indicates a cold ion ( black curve )
in conclusion , we have demonstrated laser spectroscopy , photoionization and laser cooling of a single yb@xmath0 ion on the 4f@xmath16s @xmath2s@xmath3 - 4f@xmath45d6s @xmath5d@xmath6_{1/2}$ ] transition .
we measure the isotope shifts and determine the specific mass shift as well as the field shifts .
the results on loading - and - cooling show that this transition complements the existing manipulation capabilities of yb@xmath0 ions and potentially could replace the @xmath2s@xmath3-@xmath2p@xmath3 transition . in the future , we plan to make use of the relatively narrow line width of this transition to facilitate doppler cooling to very low temperatures , potentially directly into the vibrational ground state .
moreover , the new availability of a @xmath51-system in yb@xmath0 will allow for new ways of quantum state manipulation and , for example , eit cooling @xcite .
we would like to thank sebastien garcia , neil , ben metcalf , christopher overstreet and the workshop of the cavendish laboratory . this work has been supported by epsrc ( ep / h005679/1 ) and erc ( grant no .
240335 ) .
25ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty ( , ) | we perform laser spectroscopy of yb@xmath0 ions on the 4f@xmath16s @xmath2s@xmath3 - 4f@xmath45d6s @xmath5d@xmath6_{1/2}$ ] transition at 297 nm .
the frequency measurements for @xmath7yb@xmath0 , @xmath8yb@xmath0 , @xmath9yb@xmath0 , and @xmath10yb@xmath0 reveal the specific mass shift as well as the field shifts .
in addition , we demonstrate laser cooling of yb@xmath0 ions using this transition and show that light at 297 nm can be used as the second step in the photoionization of neutral yb atoms . |
the equivalence ( mutual absolute continuity ) of probability distributions of solutions to infinite - dimensional stochastic equations has been extensively studied for equations driven by wiener process and is of importance in the investigation of large - time behaviour and ergodicity of solutions ( see e.g. @xcite for early works on stochastic reaction - diffusion equations , the book @xcite , and the references therein ) . indeed , by doob s theorem and its improvements @xcite it enables proofs of strong ergodicity and mixing of the markov semigroups . the first results in this direction for stochastic linear equations appeared in the pioneering monograph by da prato and zabczyk @xcite ( cf . also the earlier paper by j. zabczyk @xcite ) , where the following statement may be found .
consider the equation @xmath0 in a real separable hilbert space @xmath1 , where @xmath2 is a densely defined linear operator on @xmath1 generating a strongly continuous semigroup , @xmath3 is bounded on @xmath1 and @xmath4 is a standard cylindrical wiener process in @xmath1 . assuming the existence of a unique @xmath1-valued mild solution , for a given @xmath5 the probability laws of the solutions are equivalent for different values of the initial datum @xmath6 if and only if the corresponding deterministic controlled system @xmath7 is exactly null controllable at @xmath8 with controls from the set @xmath9 , i.e. , for each @xmath6 there exists a control @xmath10 such that @xmath11 .
this made it possible to utilize the well - developed deterministic controllability theory for proving equivalence of laws for the stochastic equation . in the non - markovian case
the problem of equivalence of laws was addressed in the paper @xcite in the so - called diagonal case ( when the operators @xmath2 and @xmath3 commute ) with a fractional brownian motion as the driving process . here
we extend the result of this paper by providing necessary and sufficient conditions for the equivalence of laws ( example 3.4 ) .
this toy model shows the interesting ( but natural ) feature that smoother paths of the driving process ( i.e. a bigger hurst parameter ) require more restrictive conditions for equivalence of laws ( which , of course , is just the converse with respect to regularity problems ) . in section [ sec:2 ] the first part of the paper we obtain an extension of the above - mentioned result from @xcite to the case of general gaussian noise .
the state space @xmath1 is a general banach space and the objective is to identify the suitable space of controls such that the above - described relation between equivalence of laws for and null controllability for holds true . in sections [ sec:3 ] and [ sec:4 ]
these results are applied to semilinear ( reaction - diffusion ) equations driven by a fractional brownian motion ; this part is based on the density formula established in @xcite .
the main result here is the equivalence of continuous dependence of laws of solutions to semilinear equations on the initial datum in topology of pointwise convergence of measures and null controllability for the corresponding control problem . as in the markov case
, we call this property the _ strong feller property_. the basic idea of the proof of this result is due to maslowski and seidler @xcite . in these results , for simplicity we assume that the space @xmath1 is hilbert . using the ideas of @xcite
, the results can be extended to ( type @xmath12 ) banach spaces , but our examples would not be improved much by this extension .
let @xmath13 be a real hilbert space with inner product @xmath14_{{\mathscr h}}$ ] , and let @xmath4 be an @xmath13-isonormal process on some probability space @xmath15 . by definition
, this means that @xmath4 is a bounded linear mapping from @xmath13 into @xmath16 such that the following two conditions are satisfied : 1 . for all @xmath17 ,
the random variable @xmath18 is centred gaussian ; 2 . for all @xmath19 , @xmath20_{{\mathscr h}}.\ ] ] let @xmath1 be a real banach space .
we denote by @xmath21 the completion of the algebraic tensor product @xmath22 with respect to the norm @xmath23 where @xmath24 is an orthonormal sequence in @xmath13 , @xmath25 is a sequence in @xmath1 , and @xmath26 is gaussian sequence , i.e. , a sequence of independent standard real - valued gaussian random variables .
the natural inclusion mapping @xmath27 extends to a continuous inclusion mapping @xmath28 ; this allows us to identify elements of @xmath21 with bounded operators acting from @xmath13 to @xmath1 .
the operators belonging to @xmath21 are called the @xmath29-_radonifying _ operators from @xmath13 to @xmath1 .
the mapping @xmath30 extends to an isometry , also denoted by @xmath4 , from @xmath21 into @xmath31 by putting @xmath32 moreover , @xmath33 for all @xmath34 and @xmath35 .
this extension can be viewed as an abstract stochastic integral , as can be seen by considering the case @xmath36 ( see @xcite ) : [ ex : bm ] if @xmath4 is an @xmath37-isonormal process , then @xmath38 defines a standard real - valued brownian motion @xmath39}$ ] and for all @xmath40 we have @xmath41 moreover , the extension @xmath42 coincides with the @xmath1-valued stochastic integral introduced in @xcite .
conversely , if @xmath39}$ ] is a standard real - valued brownian motion @xmath39}$ ] , then above identity defines an @xmath37-isonormal process @xmath4 .
thus , @xmath37-isonormal process are in one - to - one correspondence with standard real - valued brownian motions .
let @xmath43 be a real hilbert space and let @xmath44 .
an @xmath43-_cylindrical brownian motion _ is an @xmath45-isonormal process @xmath4 .
for any real banach space @xmath1 , the associated stochastic integral @xmath46 extends to an isometry @xmath4 from @xmath47 into @xmath31 ( see @xcite ) . in
what follows it will be useful to have an explicit description of the reproducing kernel hilbert space ( rkhs ) associated with the ( centred gaussian ) random variables @xmath48 associated with @xmath29-radonifying operators @xmath49 : [ prop : rkhs ] let @xmath4 be an @xmath13-isonormal process and let @xmath34 be a @xmath29-radonifying operator .
the reproducing kernel hilbert space @xmath50 associated with @xmath48 equals the range of @xmath49 .
let @xmath51 denote the covariance operator of @xmath48 .
for all @xmath52 we have @xmath53 hence @xmath54 where @xmath55 is the canonical embedding .
it follows that @xmath49 maps ran@xmath56 into ran@xmath57 . as an element of @xmath50
, @xmath58 equals @xmath59 , so @xmath60 it follows that @xmath49 extends to an isometry from @xmath61 onto @xmath62 . finally , if @xmath63 then @xmath64 .
from this point on , we fix a hilbert space @xmath43 and a finite time horizon @xmath65 , and write @xmath66 .
we fix a function @xmath67 which has the property that the adjoint orbits @xmath68 belong to @xmath45 . here , @xmath69 denotes the adjoint of @xmath70 defined via the riesz representation theorem . in order to discuss stochastic integrability of @xmath71 with respect to the @xmath13-isonormal process
@xmath4 , we need to connect the spaces @xmath13 and @xmath45 .
this will be done in the next two subsections , where we consider the situations where we have continuous dense embeddings @xmath72 and @xmath73 , respectively .
these embeddings can be interpreted as saying that @xmath4 is ` rougher ' , respectively ` smoother ' , than @xmath43-cylindrical motions .
this rough noise case is slightly subtler to deal with and will therefore be presented in detail ; the smooth noise case proceeds entirely analogous , with some slight simplifications .
the basic examples we have in mind are provided by @xmath43-cylindrical ( classical , liouville ) fractional brownian motions ; see section 3 . in this subsection
we make the following assumption .
[ ass1 ] the space @xmath13 is continuously and densely included in @xmath45 .
we then have continuous and dense embeddings @xmath74 where @xmath75 denotes the dual of @xmath13 relative to the @xmath45-duality .
thus , for all @xmath17 and @xmath76 we have @xmath77 for each @xmath78 we define the element @xmath79 by @xmath80_{{{\mathscr h}}}.\ ] ] by the riesz representation theorem , the correspondence @xmath81 sets up a bijective correspondence between @xmath13 and @xmath75 . for a mapping @xmath82 from a banach space @xmath83 into @xmath84
we denote by @xmath85 the adjoint , so that for all @xmath86 and @xmath87 we have @xmath88 the function @xmath67 is said to be _ stochastically integrable _ with respect to the @xmath13-isonormal process
@xmath4 if @xmath89 belongs to @xmath13 for all @xmath52 and there exists an operator @xmath90 such that @xmath91 for all @xmath52 .
the random variable @xmath92 is called the _ stochastic integral _ of @xmath71 with respect to @xmath4 , notation @xmath93 here , @xmath94 denotes the adjoint of @xmath95 defined via the riesz representation theorem .
[ prop : r ] let assumption [ ass1 ] hold and suppose @xmath71 is stochastically integrable with respect to @xmath4 .
define the bounded operator @xmath96 by @xmath97 then @xmath98 , and both operators map @xmath99 into @xmath1 .
for all @xmath76 we have @xmath100_h\,dt = { \langle}r f , x{^*}{\rangle}. \end{aligned}\ ] ] this proves that @xmath101 as operators from @xmath102 to @xmath103 . to prove that @xmath49 ( and hence also @xmath104 ) maps @xmath13 into @xmath1 it suffices to prove that for all @xmath17 we have @xmath105 in @xmath103 , and then to observe that @xmath106 takes values in @xmath1 . to prove this identity , note that for all @xmath52 we have @xmath107_{{\mathscr h}}= { \langle}r_\phi{^*}x{^ * } , \phi_h{\rangle}_{{\langle}{{\mathscr h}},{{\mathscr h}}^\star{\rangle } } = { \langle}x{^ * } , r_\phi^{*\star}\phi_h{\rangle}_{{\langle}e{^ * } , e^{**}{\rangle}}.\ ] ] it follows from the proposition that , under the stated assumptions , @xmath108 is pettis integrable and , for all @xmath76 , @xmath109 now let @xmath2 generate a @xmath110-semigroup @xmath111 on @xmath1 and let @xmath112 be a fixed operator .
we are interested in the control problem @xmath113 where @xmath114 is a ` rough ' control , that is , we assume that @xmath115 . for controls
@xmath76 the mild solution @xmath116 of the control problem at time @xmath8 is given by @xmath117 if we assume that @xmath118 satisfies the assumptions of proposition [ prop : r ] , the identity takes the form @xmath119 accordingly , for controls @xmath120 we define the mild solution of the problem at time @xmath8 to be @xmath121 where @xmath122 is the map of proposition [ prop : r ] .
[ thm : nc1 ] let assumption [ ass1 ] hold and let @xmath4 be an @xmath13-isonormal process .
suppose @xmath123 is stochastically integrable with respect to @xmath4 .
let @xmath124 be the rkhs associated with the stochastic integral @xmath125 .
the following assertions are equivalent : 1 .
@xmath126 ; 2 . for all @xmath127
the problem @xmath128 is null controllable in time @xmath8 with a control @xmath129 .
3 . for all @xmath127
the laws of the processes solving the stochastic evolution equation @xmath130 are mutually absolutely continuous .
\(i ) @xmath131 ( ii ) : the problem in ( ii ) is null controllable in time @xmath8 with control @xmath129 if and only if @xmath132 now , @xmath133 for some @xmath17 and @xmath134 furthermore , the range of @xmath106 equals the rkhs @xmath124 by proposition [ prop : rkhs ] .
combining things , we see that the problem in ( ii ) is null controllable in time @xmath8 if and only @xmath135 .
\(i ) @xmath131 ( iii ) : this is a special case of the feldman - hajek theorem on equivalent of gaussian measures .
next we consider the following dual assumption .
[ ass2 ] the space @xmath45 is is continuously and densely included in @xmath13 .
we then have continuous and dense embeddings @xmath136 .
the function @xmath67 is said to be _ stochastically integrable _ with respect to the @xmath13-isonormal process @xmath4 if there exists an operator @xmath90 such that @xmath91 for all @xmath52 .
the random variable @xmath92 is called the _ stochastic integral _ of @xmath71 with respect to @xmath4 , notation @xmath93 note that , since we are assuming that the dual orbits @xmath89 belong to @xmath45 , they automatically define elements of @xmath13 ( unlike in the rough noise case ) . [
prop : r - dual ] let assumption [ ass2 ] hold and suppose @xmath71 is stochastically integrable with respect to @xmath4 .
define the bounded operator @xmath96 by @xmath97 then @xmath98 , and both operators map @xmath99 into @xmath1 . the proof follows the lines of that of proposition [ prop : r ] _ verbatim _ , except that the identities hold only for elements @xmath129 .
now let @xmath2 generate a @xmath110-semigroup @xmath111 on @xmath1 and let @xmath112 be a fixed operator . in the present setting ,
no ambiguities with regard to the definition of a mild solution for the control problem arise and we have the following result . [ thm : nc2 ] let assumption [ ass1 ] hold and let @xmath4 be an @xmath13-isonormal process .
suppose @xmath123 is stochastically integrable with respect to @xmath4 .
let @xmath124 be the rkhs associated with the stochastic integral @xmath125 .
the following assertions are equivalent : 1 .
@xmath126 ; 2 . for all @xmath127
the problem @xmath137 is null controllable in time @xmath8 with a control @xmath129 .
3 . for all
@xmath127 the laws of the processes solving the stochastic evolution equation @xmath138 are mutually equivalent .
the proof follows the lines of that of theorem [ thm : nc1 ] _
in the present section the results of the preceding part are applied to the fractional ornstein - uhlenbeck process , i.e. to the linear stochastic evolution equation in which the driving process is a classical space - dependent fractional brownian motion ( fbm ) in time and white or , possibly , correlated in space . at
first we recall standard definitions of these objects and explain how they may be understood in the framework developed above .
we begin with the case of a scalar - valued fbm @xmath139}$ ] with hurst parameter @xmath140 . following the approach taken in @xcite we identify @xmath141 with an @xmath142-isonormal process @xmath143 , the inner product of the real hilbert space @xmath142 being given , for step functions @xmath144\to \r$ ] , by @xmath145_{{{\mathscr h}}_{\beta } } = \mathbb{e}w^\beta(\varphi_1)w^\beta(\varphi_2)= \mathbb{e}\int_0^t \!\varphi_1 \,db^\beta\int_0^t\!\varphi_2\,db^\beta = [ \mathscr{k}_{\beta}^*\varphi_1,\mathscr{k}_{\beta}^*\varphi_2]_{l^2(h)}.\ ] ] here , the operator @xmath146 is defined , for step functions @xmath147\to\r$ ] , by @xmath148 where @xmath149 is the real - valued kernel @xmath150 @xmath151 being a constant depending only on @xmath152 .
we conclude that @xmath153 is the completion of the linear space of step functions with respect to the norm @xmath154 to give a more specific description of this space it is convenient to distinguish two cases , corresponding to the `` rough '' and `` smooth '' noise cases considered above . for @xmath155 ( the rough noise case )
we have @xmath156 where @xmath157 and @xmath158 is the right - sided fractional derivative @xmath159 and @xmath160 is a constant depending only on @xmath161 .
it is not difficult to see that the space @xmath153 , as a set , may be identified as @xmath162 ( cf .
* proposition 6 ) ) . for @xmath163 ( the smooth noise case )
we have @xmath164 where @xmath165 is the right - sided riemann - liouville integral , @xmath166 in this case we have a continuous dense embedding @xmath167 and the operator @xmath168 restricts to a bounded operator on @xmath45 ( cf .
moreover , we have continuous dense embeddings @xmath169 where @xmath170 consists of those elements @xmath171 from @xmath172 such that @xmath173 with @xmath174 .
let @xmath43 be a real hilbert space .
an _ @xmath43-cylindrical fbm _ with hurst parameter @xmath161 is defined as a an @xmath175-isonormal process . here
, @xmath175 denotes the hilbert space tensor product of @xmath153 and @xmath43 . under the identification made at the beginning of the previous subsection ,
an @xmath176-cylindrical fbm is just a classical scalar - valued fbm ( with the same hurst parameter ) .
if @xmath43 has an orthonormal basis @xmath177 , one may think of an @xmath43-cylindrical fbm @xmath141 as a formal series @xmath178 where @xmath179 is a sequence of independent fbm s with hurst parameter @xmath152 .
let @xmath43 be a real hilbert space , @xmath1 a real banach space , and let @xmath143 an @xmath43-cylindrical fbm .
we consider the equation @xmath180 where @xmath2 generates a strongly continuous semigroup @xmath181 on @xmath1 , the operator @xmath3 is bounded and linear from @xmath43 into @xmath1 , and the initial datum @xmath116 belongs to @xmath1 .
the solution is understood in the mild sense , i.e. @xmath182,\ ] ] provided the stochastic integral is well - defined .
for liouville cylindrical fbm , a necessary and sufficient condition for this is given in @xcite ; for @xmath155 the same condition works for ( classical ) cylindrical fbm .
let @xmath183 denote the space of @xmath184-radonifying operators from @xmath43 into @xmath1 .
for later reference we recall that if @xmath1 is a hilbert space , then @xmath185 , the space of hilbert - schmidt operators from @xmath43 to @xmath1 , with equals norms .
a standard sufficient condition for existence and regularity of the so - called ornstein - uhlenbeck process @xmath186}$ ] is recalled in the following proposition .
we shall need it only in the case that @xmath1 is a hilbert space and refer to @xcite for a proof for the case @xmath187 . for reasons of completeness
we shall include a proof , which is an adaptation of the argument in ( * ? ? ?
* theorem 5.5 ) ( where more details are provided ) .
if the semigroup @xmath82 is analytic we may find @xmath188 sufficiently large such that the fractional powers of @xmath189 exist ( @xmath190 is fixed in the sequel ) and and we denote by @xmath191 is the domain of the fractional power @xmath192 equipped with the graph norm .
[ xe ] suppose @xmath82 is a strongly continuous analytic semigroup on the real hilbert space @xmath1 , let @xmath193 be bounded , and assume that for some @xmath194 and @xmath195 we have @xmath196,\ ] ] for some @xmath197 .
if @xmath198 the stochastic convolution process @xmath186}$ ] defined by is well - defined and has a modification with paths in @xmath199;e^\theta)$ ] .
fix @xmath200 . by the triangle inequality in @xmath31 , @xmath201b\,dw^{\beta}(r ) \big{\vert}_{e^\theta}^2\big)^\frac12
\\ & \qquad\qquad\quad\ + \big({{\mathbb e}}\big{\vert}\int_s^t s(t - r)b\,dw^{\beta}(r ) \big{\vert}_{e^\theta}^2\big)^\frac12 .
\end{aligned}\ ] ] we shall estimate both terms separately .
fix @xmath202 such that @xmath203 then ( with generic constants @xmath204 ) @xmath205 for the first term in we have , for any @xmath206 such that @xmath207 , @xmath208b\,dw^{\beta}(r ) \big{\vert}_{e^\theta}^2 \\ & \simeq { { \mathbb e}}\big{\vert}\int_0^s ( s - r)^{\delta+\tau+\theta+{\varepsilon}}(z_0-a)^{\delta+\tau+\theta } s(s - r ) \\ & \hskip3 cm \times ( s - r)^{-\delta-\tau-\theta-{\varepsilon } } [ s(t - s)-i](z_0-a)^{-\delta-\tau}b\,dw^\beta(r ) \big{\vert}^2 \\ & { \le } c^2 { { \mathbb e}}\big{\vert}\int_0^s ( s - r)^{-\delta-\tau-\theta-{\varepsilon } } [ s(t - s)-i](z_0-a)^{-\delta-\tau } b\,dw^{\beta}(r ) \big{\vert}^2 \\ & { = } c^2 { \vert}[s(t - s)-i](z_0-a)^{-\delta-\tau } b{\vert}_{{\mathscr{l}}_2(h , e)}^2 { { \mathbb e}}\big|\int_0^s ( s - r)^{-\delta-\tau-\theta-{\varepsilon}}\,dw^{\beta}(r)\big|^2 \\ & { = } c^2 s^{2{\beta}-2\delta-2\tau-2\theta-2{\varepsilon } } { \vert}[s(t - s)-i](z_0-a)^{-\delta}(z_0-a)^{-\tau}b{\vert}_{{\mathscr{l}}_2(h , e)}^2 \\ & { \le } c^2(t - s)^{2\delta}{\vert}(z_0-a)^{-\tau } b{\vert}_{{\mathscr{l}}_2(h , e)}^2,\end{aligned}\ ] ] where the numerical value of @xmath204 changes from line to line ( and is allowed to depend on @xmath8 ) .
similarly , @xmath209 combining these estimates , this gives @xmath210 the assertion now follows from a routine application of the kahane - khintchine inequalities ( to pass from moments of order @xmath12 to moments of order @xmath211 ) and the kolmogorov - chentsov continuity theorem .
now we are ready to formulate the main result for the fractional ornstein - uhlenbeck process .
[ thm : y1 ] suppose @xmath2 generates a strongly continuous analytic semigroup @xmath82 on @xmath1 .
suppose furthermore that @xmath193 is bounded and injective .
assume : * @xmath212 , for some @xmath213 , * @xmath214 for some @xmath215 .
then there exists a continuous mild solution @xmath216 to the equation , and for each @xmath5 the probability laws of @xmath217 , @xmath218 are equivalent . since @xmath3 is injective and @xmath219 maps @xmath1 into @xmath220 for all @xmath221 , the operators @xmath222 are well - defined , and by the closed graph theorem they are bounded . by ( ii ) , @xmath223 where the last step uses the analyticity of the semigroup @xmath82 .
\a ) _ the case @xmath224 . _ by proposition [ xe ] and theorem [ thm : nc1 ] it remains to show the null controllability of the equation @xmath225 on @xmath226 $ ] for the space of controls @xmath227 .
note that @xmath228 hence for @xmath229 we have @xmath230_{l^2(h)}| = \sup_{{\vert}g{\vert}_{l^2(h)}\le 1 } \big| \int^t_0 [ \varphi , ( \mathscr k_{\beta}^*)^{-1 } g]_h \,ds\big| \\ & = \sup_{{\vert}g{\vert}_{l^2(h)}\le
1 } \big| \int^t_0 [ \mathscr k_{\beta}^{-1 } \varphi , g]_h \ , ds \big| = \big ( \int^t_0 { \vert}\mathscr k_{\beta}^{-1 } \varphi{\vert}^2_h \,ds\big)^{\frac12},\end{aligned}\ ] ] taking into account that @xmath231 , where @xmath232 is a constant and @xmath233 is the left - sided fractional integral , @xmath234 moreover , by and the fact that the assumption on @xmath235 implies @xmath236 , @xmath237 where @xmath197 is a generic constant whose value is allowed to change from line to line . since by assumption @xmath238
, this shows that @xmath239 , and therefore the control @xmath240 steering @xmath116 to zero belongs to the space @xmath241 .
\b ) _ the case @xmath242 . _ as in the previous case we only need to show the null controllability of the equation @xmath243 on @xmath226 $ ] for the space of controls @xmath227 . using theorem
[ thm : nc2 ] and proceeding as in a ) it suffices to show that @xmath244 where @xmath245 , is fixed and @xmath246 and @xmath247 denotes the left - sided fractional derivative .
@xmath248 we have , for @xmath249 and @xmath250 large enough , @xmath251 \ , dt\\ & \le c\int^t_0 \big [ \frac1{t^{2\beta-1 + 2\mu}}+t^{2\beta-1 } \big ( \int^t_0 \frac{1}{(t - s)^{\beta+\frac12-\delta } s^{\delta+\mu+\beta-\frac12}}\,dr\big)^2\big ] \ , dt\end{aligned}\ ] ] for generic constants @xmath204 , where we used and analyticity of the semigroup @xmath82 .
it follows that @xmath252 \ , dt\\ & = c \int^t_0 \frac1{t^{2\beta -1 + 2\mu } } \ , dt,\end{aligned}\ ] ] which is finite since @xmath253 .
therefore , the control @xmath254 steering the solution from @xmath116 to zero belongs to @xmath227 as required .
[ @xmath255 order parabolic equation][y2 ] we consider the problem @xmath256 with the dirichlet boundary conditions @xmath257 where @xmath258 denotes the co - normal derivative , @xmath259 is a bounded domain with a smooth boundary and @xmath260 with @xmath261 is a uniformly elliptic operator on @xmath262 .
we let @xmath2 denote its realisation in @xmath263 .
the gaussian noise is fractional in time and is modelled as @xmath264 where @xmath3 is a given bounded injective operator on @xmath265 .
suppose that condition ( ii ) of the theorem holds with exponent @xmath266 ( where we may put @xmath267 ) , suppose that there exists @xmath268 $ ] such that @xmath269 and assume that @xmath270 then condition ( i ) of the theorem is satisfied as well .
indeed , as is well known , @xmath271 is hilbert - schmidt for any @xmath272 , so @xmath273 it follows that ( i ) is satisfied .
thus , equivalence of the laws of @xmath274 , @xmath6 , is obtained if and the condition ( ii ) of the theorem hold . note that is always satisfied if @xmath275 . for example , for the stochastic heat equation @xmath276 with the noise fractional in time and white in space @xmath277 ) the result holds if @xmath278 .
the condition ( ii ) of theorem [ thm : y1 ] is just sufficient and not necessary , as can be seen in the following example .
[ y3 ] in this example we take @xmath279 to be a separable real hilbert space with orthonormal basis @xmath280 and define the operators @xmath2 and @xmath3 by @xmath281 with @xmath282 and @xmath283 .
let @xmath284 denote the reproducing kernel hilbert space associated with the covariance operator @xmath285 of the gaussian random variable @xmath286 .
by the results of section [ sec:2 ] , equivalence of laws for @xmath274 , @xmath6 , holds if and only if the range of @xmath287 is contained in @xmath284 .
since @xmath287 is self - adjoint , this happens if and only if there is a constant @xmath197 such that @xmath288 under the above assumptions , this is equivalent to @xmath289 where @xmath290 .
it is easily seen that holds if and only if @xmath291 is bounded .
this follows from the fact that there exist some constants @xmath292 such that @xmath293 for @xmath294 , this was proved in @xcite .
if @xmath295 we have @xmath296 where @xmath297 . furthermore , @xmath298 where @xmath299 , and the conclusion easily follows . clearly , is implied by , but does not necessarily imply , condition ( ii ) of the theorem .
[ ex : s0 ] consider the 1d stochastic parabolic equation with inhomogeneous neumann boundary conditions of fractional noise type , formally written as @xmath300 where @xmath301 is a two - dimensional standard fbm with the hurst parameter @xmath302 , @xmath303 are real constants , the initial datum @xmath116 belongs to @xmath304 , @xmath305 $ ] is strictly positive , and @xmath306 $ ] .
it is standard to rewrite the formal equation in the infinite - dimensional form considered in the previous section ( see e.g. @xcite and references therein ) , @xmath307 , \\
x_0 & = x. \end{aligned } \right.\ ] ] here @xmath2 is the realisation in in the space @xmath308 of the partial differential operator @xmath309 with domain @xmath310 the operator @xmath2 is uniformly elliptic with homogeneous neumann boundary conditions and it is well known that it generates a strongly continuous and analytic semigroup @xmath82 on @xmath1 . in order to define the operator @xmath3 we set @xmath311 .
fix a constant @xmath312 and consider the second order boundary value problem @xmath313 for @xmath314 .
this problem has a unique solution for every @xmath314 .
the neumann map @xmath315 is an element of the space @xmath316 for arbitrary @xmath317 ( see @xcite for details ) . setting @xmath318
, @xmath319 is the isomorphic extension of the operator @xmath320 , we thus have @xmath321 .
now , the infinite - dimensional form ( the mild solution ) of the equation reads @xmath322,\ ] ] ( here the extension of the semigroup @xmath323 to the space @xmath324 is denoted again by @xmath219 ) which fits in the framework developed in section [ sec:2 ] with the spaces @xmath311 and @xmath325 ( for more detailed justification of this approach we refer to @xcite ) .
assume that @xmath326 and fix @xmath327 .
then by the analyticity of the semigroup @xmath82 , @xmath328,\ ] ] for some constants @xmath329 , so by @xcite ( or proposition [ xe ] ) the mild solution is a well - defined @xmath1-continuous process .
our aim is to investigate the equivalence of probability distributions @xmath330 of solutions @xmath331 corresponding to initial datum @xmath6 . to this end
, we use the null controllability result of fattorini and russel @xcite for the controlled equation @xmath332 with @xmath333 .
first , note that the operator @xmath2 is self - adjoint and possesses a sequence @xmath334 of real eigenvalues such that @xmath335 and @xmath336 .
moreover , there is a real constant @xmath337 such that @xmath338 where @xmath339 ( cf .
@xcite ) . denoting by @xmath280
the normalized eigenfunctions corresponding to @xmath340 , the solution to the equation may be expressed by the expansion @xmath341 in @xmath304 , where @xmath342 with @xmath343 and @xmath344 ( cf .
it is clear that is not controllable in any usual sense if @xmath345 .
assume that at least one of the constants @xmath303 is not zero ( for instance , @xmath346 ) .
set @xmath347 it is known that @xmath348 ( cf .
therefore , taking @xmath349 arbitrary : @xmath350 by .
hence by ( * ? ? ?
* corollary 3.2 ) , there exists a solution @xmath351 to the moment problem @xmath352 it follows that @xmath353 solves the moment problem @xmath354 and therefore , the control @xmath355 steers the point @xmath116 to zero at @xmath356 .
obviously , @xmath357 and it is easy to verify that @xmath358 for each value of the hurst parameter @xmath302 . summarizing , by theorems [ thm : nc1 ] and [ thm : nc2 ] we obtain that for each @xmath302 the measures @xmath330 , @xmath6 , are equivalent whenever @xmath359 .
in this section we present some applications of the general results from the previous part to stochastic semilinear equations with additive fractional noise .
it is shown that the null controllability of the deterministic equation @xmath360 in the appropriate sense is equivalent to the continuous dependence of probability laws of solutions to the corresponding stochastic semilinear equation on the initial datum . the latter property in the markovian case
is called the _
strong feller property_. consider the semilinear equation @xmath361 in the space @xmath1 , which is here for simplicity assumed to be a real separable hilbert space ( cf .
remark [ rem : r29 ] below ) .
the operators @xmath2 and @xmath3 and the driving noise @xmath362 are the same as in the linear case ( i.e. @xmath362 is an @xmath43-isonormal gaussian process representing the @xmath43-cylindrical fbm with the hurst parameter @xmath302 ) .
the operator @xmath363 is assumed to generate an analytic semigroup @xmath364 on @xmath1 and the condition of proposition [ xe ] is supposed to hold . under these assumptions , for each initial datum @xmath6 the mild solution @xmath365}$ ] of the linear equation exists and has a modification with paths in @xmath199,e)$ ] for all @xmath366 .
let us now in addition assume that @xmath367 is injective and that the range of the nonlinear function @xmath368 is contained in the range of @xmath3 .
this allows us to define the function @xmath369 , @xmath370 .
we impose the following conditions on @xmath371 : * the function @xmath369 is continuous and has at most linear growth , i.e. @xmath372 for some @xmath373 . if @xmath294 we make the additional hlder continuity assumption @xmath374 for some @xmath373 and @xmath375 define the integral operator @xmath376 induced by the kernel @xmath377 ( cf . ) , @xmath378 for @xmath379 .
by @xcite , the operator @xmath380 is a bijection and its inverse @xmath381 may be expressed , for @xmath382 , as @xmath383 for @xmath384 and @xmath385 for @xmath386 .
if moreover @xmath387 we have @xmath388 for @xmath384 ( here @xmath232 is a positive constant depending only on @xmath302 ) ( cf .
@xcite ) . by , the gaussian process @xmath389 defined as @xmath390 , where @xmath391 is the operator defined in @xmath392 , is isonormal on @xmath393 , i.e. it is the classical white noise ( see also @xcite ) .
the following result has been proved in @xcite .
[ prop : r9 ] assume that and @xmath394 are satisfied .
then the equation @xmath395 has a weak ( in the probabilistic sense ) solution @xmath396 satisfying @xmath397 which is weakly unique .
moreover , for each @xmath6 and @xmath5 the probability laws @xmath330 and @xmath398 are equivalent , where @xmath399 and @xmath400 , and the density is given by @xmath401 where @xmath402 is the expectation with respect to the probability space where the process @xmath403 is defined , @xmath404 is bounded borel measurable and @xmath405 let @xmath406 denote the @xmath407-algebra of borel sets in @xmath1 and @xmath408 the topology of pointwise convergence in the space of finite signed measures on @xmath406 .
thus a net @xmath409 converges to @xmath235 in @xmath408 if and only if @xmath410 for each @xmath411 . now we formulate the main result of this section . *
the controlled deterministic system @xmath413 is null controllable in time @xmath8 . *
the measures @xmath330 , @xmath6 , are equivalent . *
the measures @xmath398 , @xmath6 , are equivalent .
* @xmath414 in @xmath408 whenever @xmath415 in @xmath1 ( strong feller property ) . from the proofs of (
* theorems 3.3 and 3.4 ) it easily follows that @xmath417 hence it suffices to show that @xmath418 whenever we have @xmath419 in @xmath1 .
we will show separately for the cases @xmath384 and @xmath386 . first , consider the case @xmath420 .
by we have @xmath421 by continuity of @xmath371 we have @xmath422 for each @xmath423 $ ] @xmath424-almost surely , and yields @xmath425,e)})\ ] ] where @xmath426 is a constant independent of @xmath427 and @xmath428 $ ] .
hence we obtain by the dominated convergence theorem .
now , consider the case @xmath386 . by
it follows that @xmath429 as in the previous case , by the continuity of @xmath371 we have @xmath430 for each @xmath431 $ ] @xmath424-almost surely .
we aim at showing by the dominated convergence theorem . by we
immediately obtain @xmath432 furthermore , we have ( for a generic constant @xmath204 ) @xmath433 the first integral on the right - hand side of clearly tends to zero by the hlder continuity condition and the inequality @xmath434 again by and analyticity of the semigroup @xmath82 we have @xmath435,h)}}{r^{\alpha\varepsilon}(s - r)^{\beta+\frac12}}\end{aligned}\ ] ] where the generic constant @xmath197 does not depend on @xmath427 and @xmath436 , @xmath437 , @xmath438 , and @xmath439 is such that @xmath440 , @xmath441 and @xmath442 ( note that this choice is possible by and the fact that @xmath443,e))$ ] .
this gives us the convergent majorant for the second integral on the right - hand side of , and follows by dominated convergence .
[ rem : r20a ] note that a sequence @xmath444 of borel probability measures on @xmath1 is conditionally sequentially compact in @xmath408 if and only if it is equicontinuous , i.e. @xmath445 and therefore @xmath446 in @xmath408 provided and @xmath446 in the @xmath447-topology ( that is , weakly in probabilistic sense ) , cf .
* theorem 2.6 and lemma 3.15 ) .
the equivalence ( i ) @xmath131 ( ii ) has been proved in theorems [ thm : nc1 ] and [ thm : nc2 ] . by proposition [ prop : r9 ] , for each @xmath5 and @xmath6 the measures @xmath398 and @xmath330 are equivalent , so we have trivially ( ii ) @xmath131 ( iii ) . we prove ( iv ) @xmath448 ( ii ) by contradiction .
if ( ii ) is false there exist @xmath449 and @xmath5 such that @xmath450 ( gaussian measures must be singular unless they are equivalent ) . by the feldman - hjek theorem
, we then have @xmath451 where @xmath452 . by proposition [ prop : r9 ] @xmath453 which contradicts ( iv ) .
it remains to show ( ii ) @xmath448 ( iv ) .
the proof is based on the lemma [ lem : r13 ] above and follows the idea from @xcite ( where the proof is given for markov case ) . first , note that ( ii ) implies that @xmath454 in @xmath408 .
this easily follows from the cameron - martin formula ( for the density of @xmath455 with respect to @xmath330 , cf .
@xcite for a similar proof ) . also , by ( * ? ? ?
* theorem ii.21 ) and lemma [ lem : r13 ] we immediately obtain that @xmath456 and the sequence of densities @xmath457 is equiintegrable . by remark [ rem : r20a ]
it is sufficient to prove @xmath458 for arbitrary @xmath459 , @xmath460 , and @xmath461 for each @xmath462 .
we have @xmath463 } \rho_t ( x_n)\end{aligned}\ ] ] for arbitrary @xmath464 .
now , we have @xmath465 } \rho_t ( x_n ) \rightarrow 0 , \
k \rightarrow \infty\ ] ] by equiintegrability of @xmath457 and @xmath466 by remark [ rem : r20a ] applied to the linear equation , where we use the fact that @xmath467 in @xmath408 , and follows .
furthermore , @xmath468 } \rho_t ( x),\end{aligned}\ ] ] for arbitrary @xmath464 .
clearly , we have @xmath469 as @xmath470 , hence using we obtain , which concludes the proof of the theorem .
[ ex : r25 ] consider the 1d stochastic equation of reaction - diffusion type @xmath471 where @xmath472 and the noise @xmath473 is fractional in time with the hurst parameter @xmath140 and white in space .
for the linear case @xmath474 it is a particular case of the example [ y2 ] . the rigorous interpretation of is the equation where we put @xmath475 , @xmath476 , @xmath477 , @xmath478 and @xmath479 , @xmath480 , is the nemytskii operator .
we need to impose some conditions on @xmath114 so that the operator @xmath481 is well - defined a satisfies the assumptions of the present section . to this end
we assume that @xmath114 is continuous and of at most linear growth , latexmath:[\[\label{eq : r27 } for some @xmath373 and @xmath484 satisfying @xmath485 . using the fact that @xmath486 ( cf .
example [ y2 ] ) this easily implies that the hypothesis ( g ) is satisfied . as we have seen in the example [ y2 ]
, the correspond deterministic system is null controllable in this case and we may conclude that for each @xmath5 the probability laws @xmath398 , @xmath6 , are equivalent and the mapping @xmath487 is continuous in the topology of pointwise convergence ( the strong feller property holds ) .
[ rem : r29 ] ( i ) in the previous example , the conditions , ( which are clearly too strong for usual reaction - diffusion models ) may be replaced by @xmath488 and ( if @xmath489 ) @xmath490 for some @xmath491 , @xmath492 , and @xmath493 satisfying @xmath494 .
this may be shown in the same way as in the present example , but the proof of the analogue of lemma [ lem : r13 ] becomes technically more complicated ( and @xmath1 is no longer hilbert space ) .
\(ii ) using example [ ex : s0 ] we may also consider the semilinear equation with boundary noise @xmath495 where @xmath496 , @xmath497 , satisfy corresponding continuity and growth conditions .
unless both @xmath498 , @xmath497 , are zero we again obtain that the strong feller property holds for the above semilinear stochastic equation .
t. e. duncan , b. maslowski and b. pasik - duncan , _ linear stochastic equations in a hilbert space with a fractional brownian motion _ , in : `` stochastic processes , optimization , and control theory : applications in financial engineering , queueing networks , and manufacturing systems '' , h. yan , g. yin , and q. zhang , eds .
, springer - verlag , new york , 2006 , 201221 . | we obtain necessary and sufficient conditions for equivalence of law for linear stochastic evolution equations driven by a general gaussian noise by identifying the suitable space of controls for the corresponding deterministic control problem .
this result is applied to semilinear ( reaction - diffusion ) equations driven by a fractional brownian motion .
we establish the equivalence of continuous dependence of laws of solutions to semilinear equations on the initial datum in the topology of pointwise convergence of measures and null controllability for the corresponding deterministic control problem . |
the origin , strength and structure of intergalactic magnetic fields ( igmf ) remain a mystery up to the present day .
possible mechanisms to explain cosmic magnetogenesis may be divided into two main categories : cosmological scenarios predict that magnetic fields were generated through processes taking place in the early universe , such as inflation @xcite , electroweak @xcite or qcd phase transitions @xcite , and leptogenesis @xcite , among others ; in astrophysical scenarios the fields would be created during the later stages of evolution of the universe , for example during structure formation @xcite or even thereafter @xcite .
measurements of igmf are rather difficult due to their low magnitude .
common methods to estimate the strength of igmf are indirect and include the well - known faraday rotation measurements which yield upper limits of the order of a few ng @xcite .
lower bounds , @xmath4 , have been obtained by several authors using gamma - ray - induced electromagnetic cascades in the intergalactic space @xcite .
these lower bounds are controversial because of the claims @xcite that the development of the cascade is suppressed by plasma instabilities that arise from interactions with the intergalactic medium . on the other hand , recent direct observations of cascades @xcite
suggest that plasma instabilities are not operative and that the original bounds hold .
we expect that future analyses will clarify the role , if any , of plasma instabilities in the development of the electromagnetic cascade .
magnetic fields can carry helicity ( @xmath5 ) , which is defined as @xmath6 where @xmath7 is the magnetic vector potential and @xmath8 is the magnetic field . since magnetic helicity affects the dynamical evolution of magnetic fields ,
an indirect way to measure magnetic helicity is to measure the magnetic field power spectrum and compare it with the evolution seen in magnetohydrodynamical ( mhd ) simulations @xcite .
ther are also some proposals to _ directly _ measure magnetic helicity based on the propagation of cosmic rays @xcite .
more recently , it has been proposed that helicity can leave characteristic parity - odd imprints on the arrival directions of gamma rays that are the result of gamma - ray - induced electromagnetic cascades @xcite .
in particular , long & vachaspati @xcite have carried out a thorough analysis of the morphology of the arrival directions of gamma rays using a semi - analytical approach , but without including the stochasticity of the magnetic field or the cascade process .
hence , a full monte carlo approach and three - dimensional simulations are needed in order to confirm or refute their findings and provide a solid basis for further analyses .
the observation of helical primordial magnetic fields has profound implications for particle physics and the early universe .
scenarios in which the cosmological matter - antimatter asymmetry is generated dynamically are found to also produce helical magnetic fields @xcite .
the handedness of the field is related to details of the matter - genesis scenario @xcite .
if the observed magnetic fields are coherent on very large scales , they may have been produced at the initial epoch , perhaps during an inflation @xcite .
helicity on these scales would indicate the presence of certain parity violating interactions in the fundamental lagrangian @xcite . in the present work
we perform simulations of the propagation of gamma rays in both helical and non - helical igmf .
this paper is structured as follows : first , we discuss the theory and implementation of simulations of electromagnetic cascades in sec .
[ sec : emsim ] ; in sec .
[ sec : results ] we apply our approach to different magnetic field configurations , focusing in particular on the role of magnetic helicity ( sec .
[ sec : helmag1 ] - [ sec : helmag3 ] ) ; in sec .
[ sec : discussion ] we discuss the results , draw our conclusions and give a short outlook .
the basic physics underlying the development of electromagnetic cascades induced by high energy gamma rays from blazars is well - known @xcite .
a gamma ray emitted by a blazar interacts with photons from the diffuse extragalactic background radiation fields producing an electron - positron pair .
the electrons then upscatter photons of the cosmic microwave background to high energies in a process known as inverse compton scattering ( ics ) . the electrons continue to upscatter photons until their energy diminishes .
the upscattered photons can produce yet more electron - positron pairs until the energy of the photon drops below the threshold for pair production .
we should therefore observe the blazar source as well as gamma rays originating from the cascade process , unless magnetic fields bend the electron trajectories sufficiently away from the line of sight . to perform three - dimensional simulations of the development of gamma - ray - induced electromagnetic cascades in the igm ,
we have modified the crpropa 3 @xcite code , commonly used for ultra - high energy cosmic ray propagation .
taking advantage of the modular structure of the code and the flexibility to handle custom magnetic field configurations , we have implemented relevant interactions for gamma rays and electrons in the energy range of interest ( @xmath9 ) .
relevant interactions are pair production by gamma rays and inverse compton scattering by electrons .
adiabatic losses due to the expansion of the universe are also taken into account .
synchrotron losses , albeit small in this energy range , are considered as well , for the sake of completeness .
particles are propagated step - by - step . within each step
the probability of a given interaction to occur is computed using tabulated values for the interaction rate .
if the particle is charged , deflections due to magnetic fields are calculated by integrating the equations of motion . by doing so
, we are adopting a three - dimensional monte carlo approach for the propagation .
interaction rates for pair production and inverse compton scattering are calculated following the implementation used in the elmag code @xcite , and defined as the inverse of the mean free path @xmath10 .
they are tabulated for the cmb and various models of extragalactic background light ( ebl ) at different redshifts as follows . ]
@xcite : @xmath11 where @xmath12 is the energy of the interacting particle ( electron , positron or photon ) , @xmath13 is the comoving spectral density distribution of photons with energy @xmath14 at redshift @xmath15 , @xmath16 denotes the center of mass energy in the kinematic range @xmath17 , and @xmath18 is a function that depends on the interaction in question . in the case of pair production @xmath19
is @xmath20 where @xmath21 is the cross section for pair production , and @xmath22 , with @xmath23 being the angle between the gamma ray of energy @xmath12 and the background photon of energy @xmath14 .
the values of @xmath16 range from @xmath24 to @xmath25 , where @xmath26 is the cutoff energy for the photon field , assumed to be approximately 0.1 ev for the cmb and 15 ev for the ebl . for inverse compton scattering @xmath27 is given by @xmath28 with @xmath29 .
the center of mass energies in this case are @xmath30 , for @xmath17 , with @xmath31 and @xmath32 .
cross sections for these interactions are well - known ( see e.g. @xcite ) .
the spectral density distribution of the cosmic microwave background ( cmb ) can be described as a black - body .
the ebl is model - dependent . for this particular work
we adopt the lower limit ebl model of kneiske & dole .
synchrotron losses are given by @xmath33 following ref .
@xcite . here
@xmath34 is the electron mass , @xmath35 is @xmath36 with @xmath37 g , and @xmath38 the magnetic field vector acting on an electron with momentum @xmath39 .
adiabatic losses due to the expansion of the universe are given by @xmath40 with @xmath41km / s / mpc designating the hubble constant at present time , @xmath42 being the density of matter , and @xmath43 being the density of dark energy , assuming the standard @xmath44cdm cosmological model .
in our simulations we consider a monochromatic source and all emitted gamma rays are assumed to have an energy of @xmath45 .
photons from the source with energies much smaller than this will be below the threshold for creating a cascade , while photons with much higher energies will have a diminished flux . in order to run a simulation for a given magnetic field scenario or , more specifically , for a given magnetic field ( and magnetic helicity ) spectrum
, one has to sample a magnetic field grid which then may be used as input .
this procedure is explained in the following using the formalism of @xcite .
the aim is to decompose the magnetic field into modes of the divergence - free eigenfunctions @xmath46 of the laplace operator which for a specific wave vector @xmath47 are given by @xmath48 where @xmath49 is a right - handed orthonormal system of real unit vectors with @xmath50 . in order to obtain @xmath51 and @xmath52
we chose a fixed arbitrary vector @xmath53 with which we calculate @xmath54 with these definitions the @xmath46 fullfil the following relations @xcite : @xmath55 considering these relations the magnetic field with @xmath56 or , in fourier space , @xmath57 may be decomposed as @xmath58 \frac{{\rm d}^{3}k}{(2 \pi)^{3 } } \,,\ ] ] for which , in order for @xmath59 to be real , the condition @xmath60 must hold .
a possible realization of this condition is @xmath61 which can be fulfilled by setting @xmath62 which we are going to use in the following .
together with ( [ qpm ] ) and ( [ breal ] ) this leads to @xmath63 and thus @xmath64 the @xmath65 may be obtained from the given spectra using the relations @xcite
@xmath66 \frac{k^{2 } \ , { \rm d } k}{16 \pi^{3 } } \\ \\
& \equiv \int e_{b}(k ) \ , { \rm d } \ln k \end{aligned}\ ] ] and @xmath67 \frac{k
{ \rm d } k}{2 \pi^{2 } } \\ & \equiv \int h_{b}(k ) \ , { \rm d } \ln k\ , , \end{aligned}\ ] ] where @xmath7 is the vector potential and @xmath68 and @xmath69 are the spectra of the magnetic energy density and the magnetic helicity density , respectively . @xmath68 and @xmath69 are related to each other through the inequality @xcite @xmath70 which may be also expressed as @xmath71 with @xmath72 .
numerical and analytical analyses @xcite show that @xmath68 is a power - law for small @xmath73 , _ i.e. _ @xmath74 with @xmath75 .
this power - law behavior for @xmath76 is also known as the batchelor spectrum . in our numerical analysis with stochastic magnetic fields of sec .
[ sec : helmag1 ] , we will use magnetic fields with the spectrum @xmath77 where , for a correlation length @xmath78 , @xmath79 is the cutoff scale ( cf .
( [ lcdef ] ) below ) . finally , solving ( [ eb(k ) ] ) and ( [ helicitymodes ] ) for @xmath80
gives @xmath81 \\ & = \left(\frac{2 \pi}{k}\right)^{3 } \left [ 1 \pm f_{h}(k ) \right ] e_{b}(k)\ , .
\end{split}\ ] ] with these considerations the procedure for sampling a magnetic field for given spectra @xmath68 and @xmath69 on a grid in @xmath82-space is the following : first , for each @xmath47 in the fourier - transformed @xmath73-space a value for the norm of @xmath83 is generated from a normal distribution with mean value @xmath84 and standard deviation @xmath85 as follows from ( [ bsquaredeb ] ) with @xmath86 .
next , we include a random phase factor @xmath87,\ ] ] where @xmath88 are random phases distributed uniformly on @xmath89 .
once we have @xmath90 , we use eq .
( [ bcondition ] ) to find @xmath91 .
these @xmath92 can then be plugged into ( [ bx ] ) to obtain the value for @xmath59 at a given @xmath93 .
as sometimes it is more convenient to have @xmath94 given in terms of the real and imaginary parts , we use eq .
( [ qpm ] ) to write it down in the form @xmath95 \\ & + i \left [ \left ( \left| \tilde{b}^{+}(\mathbf{k } ) \right| \sin\theta^{+ } + \left| \tilde{b}^{-}(\mathbf{k } ) \right| \sin\theta^{-}\right ) \mathbf{e}_{1 } + \left ( \left| \tilde{b}^{+}(\mathbf{k } ) \right| \cos\theta^{+ } - \left| \tilde{b}^{-}(\mathbf{k } ) \right| \cos\theta^{-}\right ) \mathbf{e}_{2 } \right ] \big\}\ , .
\end{split}\ ] ]
in this section we present the results .
some preliminary considerations regarding the setup of simulations should first be made .
we use the large sphere observer approach which is a computationally efficient method for studying cosmic and gamma rays from a single source @xcite .
it is defined by the fact that this source is located in the center of a sphere which has a radius equal to @xmath96 , the distance from the source to the observer .
hence , if a particle crosses the sphere from the inside to the outside , it is flagged ` detected ' .
this will henceforth be called a ` hit ' and it corresponds to the particle reaching the observer .
the source can emit gamma rays either within a jet or isotropically .
due to the choice of a large sphere as an observer , all events above a given energy threshold ( here we use @xmath97 ) are detected .
moreover , we can easily select a subset of the events and consider an arbitrary emission pattern , such as a jet of arbitrary half - opening angle @xmath98 , or an emission around an arbitrary direction tilted with respect to the line of sight .
simple geometrical considerations allow us to correct the arrival directions on the large sphere to mimic earth s field of view . in the sky maps presented in this work , for each hit ,
the corresponding coordinate system of the observer is placed such that its origin is located at the position of the hit while the @xmath15 axis points towards the source , _ i.e. _ in the direction of the center of the sphere . in order to determine the directions of the @xmath82 and @xmath99 axes , we take a `` global '' reference frame at a fixed point of the sphere and parallel - transport it along a geodesic to the location of the hit .
then the spherical angles of the event are measured in the local frame located at the hit point . while the `` large sphere observer '' method is economical
as no photons are wasted , one possible concern is that in a realistic set - up most photons would indeed be wasted and the actual halo morphology would be sensitive to the absent photons .
however , our results in the test case of a uniform magnetic field correlate well with analytic simulations @xcite , giving us confidence in the method .
the magnetic field ( except for the uniform case ) is sampled in a grid with @xmath100 cells , where each cell has a size of @xmath101 . for a gamma ray emitted at tev energy @xmath102 and observed at an energy @xmath103 , originating from a source ( blazar in our case ) located at redshift @xmath104 and distance @xmath96 ,
traversing a magnetic field of strength @xmath105 , the expected average angular arrival direction is @xcite @xmath106 this formula is only a rough estimate where @xmath107 is a factor close to unity , @xmath108 , which varies slightly with the ebl model chosen .
furthermore , this equation is only valid if the coherence length ( @xmath109 ) of the field is much larger than the propagation length of electrons before they upscatter photons via inverse compton .
this is always true in our simulations because the propagation length is of the order of 30 kpc , whereas the minimum coherence length is 10 mpc . in order to compare our results with eq .
( [ thetaebd ] ) , we simulate the propagation of gamma rays with initial energies @xmath110 , distance @xmath111 ( @xmath112 ) , emitted in a collimated jet along the line of sight assuming stochastic magnetic fields with strength of @xmath113 and @xmath114 .
the maps containing the arrival directions are shown in the top panel of fig .
[ fig : analytcomp ] .
we have compared the deflections obtained from the simulations with the theoretical prediction of eq .
( [ thetaebd ] ) .
this is shown in the bottom panel of fig .
[ fig : analytcomp ] .
the results show a good agreement with the expected deflections .
differences are due to the nature of the analytic formula itself , which has been derived in @xcite using various simplifying assumptions .
furthermore , as has been pointed out in ref .
@xcite , the deflection angle is highly sensitive to the particular ebl model used . in particular , for the ebl model used here ( kneiske & dole ) , we expect @xmath115 . as pointed out in ref .
@xcite , @xmath116 for typical ebl models found in the literature .
we now consider a simple scenario with a uniform magnetic field . by definition ,
a uniform magnetic field has a preferred direction , and therefore one has to distinguish among three general cases depending on the orientation of the magnetic field with respect to the axis of the jet : parallel , perpendicular , and intermediate orientation .
the jet direction is assumed to be along the line of sight . in fig .
[ fig : homfield ] these different cases are shown for a magnetic field of strength @xmath117 , assuming that the gamma rays are emitted in a jet with a half - opening angle of @xmath118 and with energy @xmath119 .
the results for the three cases with a specific focus on their energy dependence are shown in fig .
[ fig : homfield ] .
the results for the first case , in which the magnetic field is parallel to the jet axis , are rather intuitive and are shown in the upper left panel of fig .
[ fig : homfield ] .
one can see that there is only one possible arrival direction , face - on , _
i.e. _ @xmath120 , which means that only electrons created with momenta parallel to the magnetic field lines , and thus not influenced by the lorentz force , can reach the observer .
any electron that deviates from the line of sight will have a trajectory that leaves the plane spanned by the line of sight and the velocity direction of the initial tev photon and will not reach the observer .
the second case , shown in the top right panel of fig .
[ fig : homfield ] , has a magnetic field perpendicular to the line of sight . here only photons arriving in a plane perpendicular to the magnetic field are detected .
this means that the parent - electrons of these photons describe circular motion in this same plane .
if an electron has a velocity component _ parallel _ to the magnetic field , it is initially directed away from the line of sight , and there is no component of the lorentz force that can bend it back towards the observer . in the case of an intermediate orientation of the magnetic field , illustrated through the bottom panel of fig .
[ fig : homfield ] for a tilt angle of @xmath121 ( left ) and @xmath122 ( right ) , we obtain results between the two extreme cases previously discussed , as expected .
it is interesting to notice that the patterns are now smeared out since electrons from a range of directions can be directed towards the observer .
still , the dilution of the signal is small compared to the actual deflection , and hence this can be observed .
therefore , relevant information can still be extracted from sky maps by using the morphology of the arrival directions .
now we introduce magnetic helicity to the simulations .
the source is assumed to have a half opening angle @xmath123 .
we take the field to be stochastic with a batchelor spectrum as in eq . ( [ eb120 ] ) . as we are assuming the maximal helical case , _
@xmath124 , this also fixes the spectrum of @xmath125 according to eq .
( [ hbeb ] ) .
the field has an average field strength of @xmath126 and a correlation length of @xmath127 . here
, @xmath128 can be extracted from eq .
( [ eb(k ) ] ) by setting @xmath129 while @xmath109 is defined by @xcite @xmath130 such that for the @xmath68 defined in ( [ eb120 ] ) we have @xmath131 , where @xmath132 is the cutoff scale .
we have simulated the propagation of gamma rays with initial energy @xmath119 in the presence of stochastic magnetic fields with maximally negative ( @xmath133 ) , zero ( @xmath134 ) and maximally positive ( @xmath135 ) helicities gamma ray distribution by a parity reversal of the @xmath136 plot .
however , we simulate the two cases independently to show two different stochastic realizations . ] . to simulate @xmath137 photons in our standard scenario described above , _ i.e. _ with @xmath111 and @xmath114 , the current version of the code takes @xmath138 hours on @xmath139 cores at @xmath140 . the actual values for helicities for the whole simulation box as well as along the line sight
are shown in fig .
[ fig : helgrid ] in order to illustrate to which extent statistics play a role .
as one can see , both for the whole grid as well as just along the line of sight , which is more important to judge about the statistical significance for a given case , the helicity distribution corresponds to the sign it has been assigned .
furthermore , from the panel on the right , one can see that for these particular realizations , for @xmath141 the absolute magnitude of @xmath142 is high close to the source and low close to the observer , while for @xmath143 it stays roughly equal along the propagation path .
we can understand the qualitatively similar patterns for @xmath144 , _
i.e. _ both patterns are spirals with similar twist , by noting that pair production on average happens close to the source , and both cases have similar helicity measures in that region . as a confirmation of this interpretation we found that in simulations in which the absolute value of @xmath142 is small close to the source , the spiral - like structures tend to be less distinct .
the sky map containing the arrival directions of gamma rays are shown in fig .
[ fig : test5 g ] .
we consider both the case for which the jet is directed along the line of sight ( left column ) and for which it is tilted by @xmath118 ( right column ) . for the former one can see the impact of magnetic helicity by comparing the top ( @xmath133 ) and bottom ( @xmath135 ) panels .
a remarkable spiral - like pattern is clearly visible , being left- or right - handed depending on whether the helicity is negative or positive , respectively . for zero helicity ( @xmath134 , middle panels ) ,
on the other hand , no clear orientation can be seen .
we show here the results for @xmath145 .
for lower coherence lengths ( @xmath146 ) and @xmath147 we find that the arrival direction pattern is washed out , and it is not possible to infer the presence of helicity , thus confirming the analytical predictions of ref .
@xcite for this combination of parameters using simulations .
this can be seen in fig .
[ fig : difflc ] where the results for different @xmath109 and @xmath148 are shown .
while for @xmath149 a clear characteristic spiral in the arrival directions can be seen , it becomes less visible for @xmath150 and disappears for @xmath151 .
therefore , @xmath78 is a reasonable choice in order to show the effects of helicity discussed below .
it is also a valid value in certain magnetogenesis scenarios @xcite . to understand the dependence of the spiral pattern on the coherence scale , we note that , for small coherence lengths , the spirals become too tight to be resolved , _
i.e. _ their angular size becomes too small compared to the overall halo @xcite .
it seems , however , that the quality of the spiral might be highly sensitive to the specific values of the parameters of the setting such as @xmath105 , @xmath96 and @xmath109 which we will further investigate in the future . on the other hand , for larger coherence lengths
the spirals tend to a straight line , similarly to the top right panel of fig .
[ fig : homfield ] , approaching the case of a simple uniform magnetic field .
this , again , is rather intuitive , since if @xmath152 , the stochastic magnetic field will effectively be uniform on the length scales in question . on the right hand side of fig .
[ fig : test5 g ] we show the same scenario described above , but this time the direction of the jet is tilted by @xmath118 with respect to the line of sight . as one can see in the figure , this reduces the effective area of arrival directions and also the symmetry of the pattern . in our example , for instance , one of the `` arms '' of the spiral pattern or a part of it is removed .
this enables us to apply the @xmath0-statistics @xcite ( discussed below ) to relate the helicity of the field with the arrival directions of gamma rays .
it should be noted that all findings of this and the previous sections are in good agreement with the analytic predictions of ref .
@xcite .
one possibility to quantify the role of magnetic helicity is to use the @xmath0-statistics , introduced in refs .
the key elements here are the observed energies and the arrival directions of gamma rays at earth . for sets of photons with energies
@xmath153 , @xmath154 and @xmath155 with @xmath156 , the @xmath0-statistics is given by @xcite @xmath157 \cdot \mathbf{n}_{j}(e_{3})\ , , \ ] ] where @xmath158 is the arrival direction of the @xmath159-th photon with energy @xmath160 , @xmath161 is the total number of photons of energy @xmath162 , and @xmath163 is given by @xmath164 where @xmath165 represents the set of photons of energy @xmath162 that are located in a disk of radius @xmath166 centered on @xmath167 .
essentially , the @xmath0-statistics is the average value of the triple product of photon arrival vectors of energies @xmath168 that lie within an angle @xmath166 of the highest energy photon ( @xmath155 ) . as has been shown in refs .
@xcite , the calculation of the parity - odd statistics , or @xmath0-statistics , should enable us , depending on the sign and general shape of the @xmath0-factors for different values of @xmath153 , @xmath154 , @xmath155 , and @xmath166 , to draw conclusions about the helicity of the intervening helical magnetic field .
we now use eq .
( [ qstat ] ) to calculate the @xmath0-factors for the three helicity scenarios analyzed ( @xmath169 , @xmath170 and @xmath171 ) .
we display the results for the case of tilted jets ( _ i.e. _ the scenario shown in the right panel of fig .
[ fig : test5 g ] ) in fig .
[ fig : qtest5 g ] .
we consider triplets of energies @xmath172 as needed for eq .
( [ qstat ] ) , where each energy @xmath173 corresponds to an interval @xmath174 $ ] .
the reason we consider the scenario of _ tilted _ jets is that this is the most probable case
it is very unlikely for the blazar jet to be directed exactly along the line of sight . as discussed in ref .
@xcite , the function @xmath175 is expected to start at the origin since the angular deflections are small for small @xmath166 . for larger @xmath166
, the magnetic helicity causes @xmath0 to grow , and at much larger @xmath166 , @xmath0 will approach a constant value ( @xmath176 ) as there are no more photons to include at such large @xmath166 .
the large @xmath166 behavior gets modified in a realistic setting where , in addition to the blazar photons , we also observe background photons from other sources .
then , for large @xmath166 , the blazar contribution gets diluted by the background noise and @xmath0 decreases to zero . in this case
, we would see a peak in @xmath175 whose position is set by the relative number of blazar to background photons . in our simulations , however , we do not include background photons and indeed find @xmath177 at large @xmath166 . as the last part of our results we present a new alternative way to quantify the pattern of gamma ray arrival directions and thus , indirectly , the helicity orientation .
the idea underlying idea of this new method that we denote @xmath1-statistics ( for `` spiral '' ) is that a gamma ray from a cascade that has a greater deflection away from the source direction due to the magnetic field will also have a greater azimuthal deflection if the magnetic field is helical .
the pattern of observed gamma rays will have a spiral structure that can be measured by finding the average deflection of gamma rays , @xmath178,@xmath103 ) , as a function of the azimuthal angle @xmath179 and the considered gamma ray energy @xmath103 .
we assume that there is at least one angle @xmath180 for which @xmath181 has a well - defined and significant maximum , _
i.e. _ a maximum from an average in a bin which has a reasonable number of photons _ and _ is statistically significant .
we consider events inside a band around @xmath180 with width @xmath182 , _
i.e. _ events with @xmath183 . for a right - handed spiral
there should be higher values of @xmath181 for @xmath184 than for @xmath185 inside the band , while for a left - handed pattern @xmath181 should be smaller for @xmath184 than for @xmath185 .
in other words , the peak of the function @xmath186 should be skewed to the right or to the left depending on whether the spiral is right- or left - handed , respectively . by finding a measure for this asymmetry or skew of the maximum
one can deduce the orientation and subsequently the magnetic helicity .
more concretely , the calculation is performed in the following way : first , we subdivide the interval on which @xmath179 is defined , _
@xmath187 , in @xmath188 bins , such that each of the bins has a width @xmath189 .
the @xmath159th bin , which corresponds to the interval @xmath190 , @xmath191 , will be labeled @xmath192 . for each bin
we calculate @xmath193 by @xmath194 where @xmath195 are the coordinates of the @xmath196th event in the set @xmath197 and @xmath198 is the total number of events in this set . if @xmath197 is empty , we set @xmath199
. furthermore , for real data it might be necessary to restrict the analysis to events with @xmath200 smaller than a certain value @xmath201 as for @xmath202 background photons might dominate and result in a false signal . in this set of @xmath203 one
now has to identify the relevant and significant maxima as well as the corresponding bin number @xmath204 and calculate the quantities @xmath205 where @xmath206 is the number of bins we need to consider in order to include the width of the peak .
here one has to assume periodicity , _
i.e. _ @xmath207 .
essentially @xmath208 corresponds to the average value of @xmath209 to the left of the peak and @xmath210 to the right of the peak .
the final step is to define the @xmath1-statistics that measures the skewness of the peak @xmath211 for a right - handed spiral @xmath1 will be positive , whereas for a left - handed spiral it will be negative .
we performed this computation for the data shown in fig .
[ fig : test5 g ] . the plots for @xmath181 , shown for different energies ,
are presented in fig .
[ fig : thetabar ] . even without any further analysis one can see in this figure that the peaks for opposite orientations indeed show opposite skews while on the left panel
higher angles are achieved for @xmath212 , on the right panel , even more clearly , that is the case for @xmath185 . for the central panel , however ,
peaks of either skewness are found . in order to support these qualitative considerations ,
one has to look at the @xmath1-values which have been calculated and are presented for the three cases in tabs .
[ tab : stab-1][tab : stab+1 ] .
the most clear case here is the one for @xmath143 , where for all energy ranges we obtain @xmath213 with two values even going as low as @xmath214 .
this means that the morphology of the arrival directions is solely right - handed which is also clearly seen in the top panel of fig .
[ fig : test5 g ] . for the case of @xmath141
the situation is less clear as there is only one value being as high as @xmath215 .
nevertheless , since @xmath1 is positive for all energy ranges with three exceptions for which , however the absolute value of @xmath1 is close to zero , this is strong evidence for a left - handed orientation .
finally , no clear statement can be made regarding the case with no helicity ( @xmath216 ) here one does not find significant negative or positive values for @xmath1 .
as the last step , we need to connect the handedness of the arrival direction pattern with the sign of helicity . their correlation has been found in ref .
@xcite , where an analysis has been carried out for a homogeneous magnetic field . in this reference
the authors indeed find that for a positive helicity one expects a right - handed orientation , whereas for negative helicities a left - handed orientation should be observed , thus confirming our results for stochastic fields . as a concluding remark
it should be stated that for the results presented above we used simulations containing approximately @xmath217 photons arriving at earth in the energy range @xmath218 , which provided clear patterns with satisfactory statistical significance . the upcoming cherenkov telescope array ( cta )
@xcite might be able to detect this kind of signature in the energy range @xmath219 with @xmath220 of observation .
fewer photons would distort the picture since , for example in the case of the @xmath1-statistics , the peaks would become less visible , such that a reliable calculation would no longer be possible .
this is the case for high energy photons as their contribution to the total flux is rather small . on the other hand , for
the lowest energies ( @xmath221a few gev ) , even with as few as @xmath222 photons relevant peaks can be seen , which , however , might be more difficult to construe in a more realistic case considering additionally diffuse gamma ray radiation and multiple sources .
.table for @xmath1 for maximal negative helicity ( @xmath143 ) . [
cols="^,^,^,^,^,^",options="header " , ] [ tab : stab+1 ]
we have performed three - dimensional monte carlo studies of the development of gamma - ray - induced electromagnetic cascades in the intergalactic medium in the presence of magnetic fields .
we have used the `` large sphere observer '' method for improved computational performance . in this case
all cascade photons hitting the surface of the sphere are detected by the the observer . with a standard three - dimensional monte carlo simulation
most cascade photons would not reach earth , resulting in wasted computation and very low statistics .
a simplification made in our treatment is that the magnetic field evolves adiabatically with redshift as @xmath223 .
this is justified because the cascade development we have discussed occurs in cosmic voids where mhd amplification and contamination by sources is minimal . also , the sources are at redshifts @xmath224 .
we first compared our computational setup with analytical approximations and then validated it in simple scenarios containing a uniform magnetic field oriented parallelly and perpendicularly to the line of sight of the blazar jet .
as expected , for a magnetic field parallel to the direction of the jet of half - opening angle @xmath98 , assumed to be pointing toward earth , effects of the field were not observed . for a magnetic field perpendicular to the direction of the jet , deflections were non - zero and in the expected direction .
similar results were obtained for stronger and weaker magnetic fields and other orientations .
these results are in accordance with ref .
@xcite and also with the predictions of eq .
( [ thetaebd ] ) .
we have also studied the particular case of a magnetic field with a batchelor power spectrum with and without helicity .
the effects of helicity can be clearly seen in fig .
[ fig : test5 g ] , where arrival directions follow right- or left - handed spirals , depending on the sign of the helicity . for stochastic fields , in general ,
the results tend to converge toward the case of a uniform magnetic field in the limit of large coherence lengths .
we have considered only large values of correlation length ( @xmath127 ) since for much smaller coherence lengths , with the other parameters being held fixed , no clear signature of helicity can be seen , as shown in fig .
[ fig : difflc ] .
nevertheless , one should bear in mind that the current upper limits of coherence length of magnetic fields in voids range between a few and hundreds of mpc @xcite , placing the chosen value of 120 mpc well within the allowed bounds .
we have deployed the so - called @xmath0-statistics , a powerful analysis tool that makes it possible to determine the properties of magnetic helicity directly from the observables of gamma rays measured at earth . in this
work we for the first time applied @xmath0-statistics to realistic three - dimensional simulations of electromagnetic cascades .
our results for @xmath0 are shown in fig .
[ fig : qtest5 g ] .
the plots do not show a strong correlation between @xmath0 and the existence and sign of the helicity . at the moment we can not clearly state whether averaging over several objects will show a stronger correlation .
we plan on investigating this issue in a future work .
it is important to stress the fact that @xmath0-statistics might not be the final method to quantify magnetic helicity , however it is a good initial approach and has been used in several works ( refs .
@xcite ) with satisfactory results . in this work
we have , for the first time , introduced the @xmath1-statistics , which is a direct measure of the handedness of a pattern with respect to the line of sight .
we have shown that the orientation , represented by the sign of @xmath1 , is directly correlated with the sign of helicity .
this shows that the @xmath1 measure is also a powerful tool to be used in the analysis of helicity of igmf .
backgrounds at the @xmath22510 - 100 gev energy range are expected due to secondary photons from agn halos whose jet opening angles do not encompass the earth .
other astrophysical sources of photons in this energy range also exist and have to be taken into account . in this first work we have neglected these backgrounds , which will be included in future studies .
we found that it is probably necessary to analyze various sources in order to make a definite statement about the sign of the helicity , since a clear signature can not always be seen . in the future
we will extend our simulations to the case of multiple sources and diffuse gamma rays .
we expect to be able to reproduce actual detections and consequently retrieve more precise information about igmf , which can be used to infer their origin and evolution .
in addition , we will extend the analysis by further exploring the parameter space as varying quantities such as the magnetic field strength @xmath226 , the magnetic correlation length @xmath109 and source parameters such as its distance from the observer , its energy spectrum or its cutoff energy , as they may be important in order to obtain a complete picture of their influence as discussed above and to explain actual observations .
r. a. b. acknowledges the financial support from the john templeton foundation . the work of a. s. has been supported by the daad funded by the bmbf and the eu marie curie actions .
a. s. would like to thank the arizona state university for the hospitality during his stay .
furthermore , a. s. is grateful to the `` helmholtz alliance for astroparticle physics '' ( hap ) and the collaborative research center sfb 676 `` particles , strings and the early universe '' for providing generous travel funds which were important for the collaboration leading to this work .
t. v. is supported by doe , office of high energy physics , under award number # de - sc0013605 at asu .
t. v. is grateful to the institute for advanced study , princeton for hospitality while this work was being done .
special thanks go to andrew j. long for his ideas and discussions which helped to complete this work and to nordita in stockholm for organizing the workshop `` origin , evolution , and signatures of cosmological magnetic fields '' during which important aspects of this work have been discussed . | we determine the effect of intergalactic magnetic fields on the distribution of high energy gamma rays by performing three - dimensional monte carlo simulations of the development of gamma - ray - induced electromagnetic cascades in the magnetized intergalactic medium .
we employ the so - called `` large sphere observer '' method to efficiently simulate blazar gamma ray halos .
we study magnetic fields with a batchelor spectrum and with maximal left- and right - handed helicities .
we also consider the case of sources whose jets are tilted with respect to the line of sight .
we verify the formation of extended gamma ray halos around the source direction , and observe spiral - like patterns if the magnetic field is helical .
we apply the @xmath0-statistics to the simulated halos to extract their spiral nature and also propose an alternative method , the @xmath1-statistics .
both methods provide a quantative way to infer the helicity of the intervening magnetic fields from the morphology of individual blazar halos for magnetic field strengths @xmath2 and magnetic coherence lengths @xmath3 .
we show that the @xmath1-statistics has a better performance than the @xmath0-statistics when assessing magnetic helicity from the simulated halos . |
since the first observation of bose - einstein condensation ( bec ) in a dilute atomic gas , we have been witnessing dramatic progress in both theoretical and experimental studies of degenenerate quantum gases @xcite . a variety of static and dynamic properties of the condensed , trapped atoms have been exploited @xcite .
interference studies with bose - condensates have experimentally demonstrated the coherence properties expected for such a ` giant matter wave ' @xcite .
more recently , the mott - insulator phase transition has been observed by loading a magnetically trapped condensate into a steep optical lattice @xcite .
the standard approach to produce a bose - einstein condensate involves a laser cooling of the atomic sample to a phase space density of @xmath2 , after which atoms in weak - field seeking states are transferred into a magnetic trap @xcite .
further cooling proceeds by a radiofrequency field selectively transferring the highest energetic atoms into untrapped states , which evaporatively cools the atoms to quantum degeneracy .
a drawback of magnetic traps is that they only confine weak field seeking spins states , which e.g. prevents one from studying spinor condensates directly in these traps @xcite .
moreover , atoms in field independent @xmath1 spin projections , as are of large interest for many precision experiments @xcite , can not be confined in magnetic traps .
far detuned optical dipole traps can confine atoms in arbitrary spin states .
dipole traps furthermore allow for very variable trapping geometries , as e.g. periodic lattice geometries .
experiments aiming at an achieving of quantum degeneracy in optical dipole traps have been first carried out by the stanford group @xcite . in far detuned dipole traps , a variety of refined optical and evaporative cooling techniques
have been implemented , which has allowed for atomic phase space densities very close to quantum degeneracy @xcite .
friebel et al .
found that in quasistatic dipole traps , as can be realized with a co@xmath0-laser , polarization gradient cooling alone can accumulated atoms to a phase space density three orders of magnitude above that of a conventional mot @xcite .
chapman and coworkers have reported on the observation of bose - einstein condensation by direct evaporative cooling in a crossed co@xmath0-laser dipole trap @xcite .
all - optical cooling techniques have also enabled the first observations of quantum degeneracy of @xmath3 , cesium and ytterbium atoms @xcite .
we here describe experiments creating rubidium bose - einstein condensates by direct evaporative cooling in both a crossed and a single beam quasistatic dipole trapping geometry @xcite . due to the choice of a tight trapping laser focus , stable evaporation to a bose - einstein condensate is possible in both geometries even with moderate initial number of atoms . especially the single dipole trapping geometry allows for a considerable experimental simplification of methods to produce bose - condensed atoms .
note that previous experiments realizing quantum degeneracy in dipole traps have either required a more alignment sensitive crossed beam geometry or feshbach resonances to enhance the collisional rate @xcite . by applying a magnetic field gradient to the single beam configuration , we realized a trap stable for atoms in a field insensitive @xmath1 spin projection only .
in a recent work , this has enabled us to realize an all - optical atom laser @xcite .
the chemical potential of the generated condensate is first order insensitive to fluctuations from stray magnetic fields .
optical dipole traps for atoms are based on the force which arises from the coherent interaction between the induced atomic dipole moment and the laser electric field @xcite . when the laser frequency is tuned to the red side of an atomic resonance ,
the dipole force pulls atoms towards the intensity maxima of the field . for optical fields with frequency far below all electric dipole resonances of ground state atoms ( i.e. ` quasistatic ' fields ) ,
the trapping potential is given by @xmath4 where @xmath5 denotes the static atomic polarizability of the electronic ground state and @xmath6 is the electric field of the optical wave .
for the ground state and the first excited state of the rubidium atom , the scalar static polarizability is @xmath7 and @xmath8 , respectively @xcite .
since the frequency of the co@xmath0-laser radiation ( @xmath9 is near @xmath10 ) used in our experiment is approximately one order of magnitude below that of a typical electric dipole transition of ground state alkaline atoms , the photon scattering rate is very small . in particular ,
the rayleigh photon scattering rate in this quasistatic regime is given by the expression @xmath11 where @xmath12 is the beam waist of a laser beam with power p , @xmath13 the classical electron radius , @xmath14 the electron mass and @xmath15 denotes the optical frequency of the laser light . in our experiments ,
the rayleigh scattering rate of trapped rubidium atoms is of the order of @xmath16 .
this results in a small coupling of the atoms to the environment , so that the effect of decoherence can be kept small .
let us note that for alkali atoms trapped in an extremely far - detuned laser field , the detuning from resonance is much larger than both the hyperfine and fine structure splitting .
this condition is met in our quasistatic trapping field .
the trapping potential for ground - state alkali atoms here corresponds to that of a @xmath17 to @xmath18 transition , and the ac stark shift for all ground - state sublevels is almost identical .
this state - independent confinement of extremely far - detuned optical dipole traps for many experiments represent an advantage over magnetic traps .
it is however clear that evaporative cooling can not be achieved by selectively transferring atoms into untrapped states , as usually done in magnetic traps .
such a cooling in optical traps can be achieved by lowering the intensity of the trapping beams in a controlled way , which decreases the depth of trapping potential @xcite .
a scheme of our experimental setup used for direct bose - einstein condensation of rubidium atoms in a dipole trap is shown in fig .
[ fig:1 ] .
a commercial , rf - excited single mode co@xmath0-laser generates up to @xmath19 mid - infrared radiation near @xmath10 .
the light passes two acoustooptic modulators ( aoms ) .
the first one allows us to regulate the laser beam intensity via controlling of the rf - drive power .
the light diffracted in first order from this modulator passes a second aom , whose function is that of a beam divider for the crossed beam experiments .
note that the zero s and first order diffracted beams differ in frequency , which eliminates unwanted standing - wave effects .
the mid - infrared beams enter a vacuum chamber through znse windows . for each beam
, an adjustable , spherically corrected znse lens ( @xmath20 ) placed inside the vacuum chamber focuses the laser beams to a minimum waist size of @xmath21 .
the beam focus can be varied with a telescope located outside of the vacuum chamber . for the crossed trap geometry ,
a beam waist of @xmath22 is chosen , whereas for a single running wave geometry we use a beam size of @xmath23 in the trapping region .
for the former geometry , the trapping beams are directed horizontally and vertically respectively and cross each other with a 90 degree angle . for the single beam trapping geometry , the second aom is omitted and the transmitted beam travels through the vacuum can in horizontal direction .
atoms of the isotope @xmath24 are initially collected and pre - cooled in a magneto - optical trap ( mot ) , which is loaded from the thermal gas emitted by heated rubidium dispensers .
the measured background gas pressure inside the vacuum chamber is @xmath25 .
the near - resonant radiation for cooling and trapping atoms in the mot is generated by grating stabilized diode lasers .
a cooling laser operates with variable red detuning from the closed @xmath26 , @xmath27 to @xmath28 , @xmath29 hyperfine component of rubidium d2-line .
its radiation is amplified by a second injection locked free - running diode laser . in order to repump rubidium atoms into the cooling cycle ,
a second grating - stabilized diode laser locked to the @xmath26 , @xmath30 to @xmath28 , @xmath31 transition is employed . both cooling and repumping light beams
pass acoustooptic modulators and are spatially filtered .
these near - resonant beams are expanded to a beam diameter of @xmath32 , spatially overlapped and then directed to the vacuum chamber to provide three retroreflected mot beams .
the total optical power of the cooling light after spatial filtering is @xmath33 , whereas that of the repumping light is @xmath34 .
we use a pair of magnetic coils oriented in anti - helmholtz configuration to generate a magnetic quadrupole field with a @xmath35 field gradient , whose center is close to the intersection of the two co@xmath0-laser beams .
the coil axes is oriented at an angle of 45 degree with respect to the vertical and horizontal co@xmath0-laser beams .
this quadrupole field is e.g. used to operate the mot .
an experimental cycle begins by loading the magneto - optical trap .
during the course of the experiment , we have used different mot loading times . a typical value for e.g. the earlier crossed beam dipole trapping experiments is @xmath36 trapped atoms in the mot accumulated during a 5 s loading time .
during this mot loading phase , the cooling laser operates with a red detuning of @xmath37 to the red of the cycling transition .
subsequently , we increase the detuning of the cooling laser to a value of @xmath38 and simultaneously reduce the repumping laser intensity by a factor 100 for a @xmath39 long time interval . in this temporal dark mot phase ( d - mot)@xcite
, the atoms are pumped into the lower hyperfine state ( @xmath30 ) , which reduces hyperfine changing collisions .
these quoted parameters are optimized by maximizing the number of atoms transferred to the dipole trap . throughout this cycle
the co@xmath0-laser radiation is left on , allowing for rubidium atoms to accumulate in the dipole trapping region . at the end of the d - mot phase
, the repumping light is extinguished and after a delay of @xmath40 also the cooling light .
simultaneously , the magnetic quadrupole field is switched off . by this time
, the atoms are confined by the quasistatic dipole trap alone .
all near - resonant beams are carefully extinguished by both aoms and mechanical shutters . the small delay between the extinguishing of repumping and
cooling light minimizes losses in the dipole trap due to hyperfine changing collisions . to analyze the properties of trapped atomic clouds , we employ the technique of absorption imaging . at the end of an experiment cycle ,
the trapped atoms are released by switching off the co@xmath0-laser beams . after a variable expansion time
( typically 5@xmath41 ) , cold atoms are irradiated with a pulse of light tuned resonantly to the @xmath27 , @xmath29 component of the rubidium d2 line . for this measurement
, we use a spatially filtered laser beam with a @xmath42 beam diameter
. the beam is then imaged onto a slow - scan ccd camera , where shadow images of the atomic cloud are recorded .
from these images , both number of atoms and atomic temperatures are extracted .
the pulse length is @xmath43 and the intensity ( @xmath44 ) is chosen to be clearly below saturation intensity . during this detection time , repumping light is provided by the mot beams .
in initial experiments , we studied bec in a crossed beam geometry , similarly as done in work by chapman et al .
@xcite . to reach bose - einstein condensation by direct evaporative cooling in our dipole traps , the initial phase space density of the laser cooled clouds is first optimized .
we have characterized the trap lifetime , trap frequencies and the temperature of the trapped atomic cloud .
the vibrational frequencies are measured by periodically modulating the optical confining power with the first aom , which allows us to study parametrically excitation of the trapped atoms @xcite .
significant trap loss is observed when the modulation frequency is close to twice a trap vibrational frequency . with typical parameters of @xmath45 optical power in each of the co@xmath0-laser trapping beams and a @xmath22 waist radius ,
we observe parametric resonance corresponding to a vibrational frequency of @xmath46 .
for this crossed beam geometry , the oscillation frequencies in the central trap region in all three spatial directions are similar .
the trap lifetime is extracted by monitoring the number of atoms detected after different trapping times for a constant potential depth .
a typical result of such a measurement is shown in fig .
[ fig:2]a .
the data points are taken after a minimum dipole trapping time of 70 ms , since for shorter delay times atoms not being transferred to the dipole trap have not yet been removed completely from the detection region by gravity .
the data can be described with a two time constants exponential decay , which involves a fast and a slow decay with time constants near @xmath47 and @xmath48 respectively .
the fast decay in the beginning of the dipole trapping phase is attributed to a plain atomic evaporation of the highest energetic atoms .
in contrast , the slow decay is attributed to losses due to collisions with background gas , which ultimately limit the available trapping time .
[ fig:2]b shows the dependence of the atomic temperature on trapping time .
the temperature clearly decreases with time , which we attribute to the plain evaporation .
this process is effective only for small trapping times when the loss rate is large .
the measured atomic density after a trapping time of @xmath49 is @xmath50 .
together with the measured temperature of @xmath51 we derive a phase space product @xmath52 , which corresponding to a @xmath53 phase space density if we assume equal distribution of spin projections .
these results are comparable to previous results observed in co@xmath0-laser optical lattice and crossed dipole geometries @xcite .
we also infer a collisional rate of @xmath54 at this time , clearly representing a favorable starting point for further , forced evaporative cooling to lower temperatures . to induce such a forced evaporation ,
the optical potential is lowered in a controlled way .
experimentally , this is achieved by reducing the trapping laser beam intensity via a decreasing the rf - drive power of the first aom .
mandatory for the success of evaporative cooling is to maintain a relatively high atomic collision rate , so that the remaining atoms can rethermalize to a distribution of lower temperature .
we reduce the potential depth with time according to the formula @xmath55 where @xmath56 denotes the value of the initial potential depth .
it was shown in ref .
10 that this ramp form maintains a nearly constant ratio @xmath57 between potential depth and the atomic temperature . following a 100 ms long plain evaporation period ,
the mid - infrared power is reduced in a @xmath58 long ramp time from initially @xmath45 to a final value of @xmath59 in each of the trapping beams . in these crossed dipole experiments ,
we find optimum cooling when choosing the parameters of eq .
( 3 ) to be @xmath60 and @xmath61 .
we have experimentally verified that the cutoff parameter @xmath62 was nearly constant throughout the evaporation ramp . for a characterization of the cooling
, we have analyzed shadow images of the expanded atomic cloud at the end of the evaporation process .
[ fig:3]a shows a cross - section of such an image for a final trapping power of 150 mw , corresponding to an average trap vibrational frequency of 350 hz .
it is clear that here the ensemble is still purely thermal .
the phase space density is further increased when we ramp to lower power values .
[ fig:3]b shows data recorded for a final beam power of @xmath63 , where a typical bimodal distribution corresponding to an atomic cloud near the transition point is observed .
the central feature corresponds to atoms in the bose - condensate , while the wings represent thermal atoms .
the critical temperature here is @xmath64 . for final power of @xmath59 ,
an almost pure condensate is obtained as shown in fig .
[ fig:3]c .
the condensate contains about 10000 atoms distributed among the three @xmath65 states of the electronic hyperfine ground state @xmath66 .
in other measurements , we have applied a magnetic quadrupole field during the course of the evaporation by leading on the mot coils .
this leads to a condensate with about 70 percent population in the @xmath67 and 30 percent in the @xmath68 spin projection . for these trap parameters , the magnetic field gradient predominantly populates one field - sensitive state . in earlier work in a crossed dipole trap
it was demonstrated that with different field gradients also other spin states can be populated , although it was noted that by the time of writing the origin of the obtained magnetization still remained to be determined @xcite . in this chapter , we describe our experiments successfully creating a rubidium bose - einstein - condensate in a single beam co@xmath0-laser dipole trap .
this is the simplest possible dipole trapping geometry .
the use of single running wave allows for a compact setup and a very reliable production of the bec .
previous experiments creating quantum degeneracy in dipole traps have either required the use of a more alignment - sensitive crossed dipole trap geometry @xcite or feshbach resonances @xcite to enhance the collisional rate during the course of the forced evaporation . as in the case of the crossed dipole trap configuration , evaporative cooling towards bec
requires a ramping down of the trapping potential .
since the average of the trap vibrational frequencies for a given power and beam waist is smaller than in the crossed beam geometry , it is _ a priori _ not clear that the rethermalization rate is sufficiently high throughout the evaporation ramp .
obviously , the final stages of evaporation here are most critical , since the collisional rate decreases with trap power .
it is easy to show that the trap vibrational frequencies in a single dipole trap orthogonal to the beam axis scale as @xmath69 , while along the beam axes scale as @xmath70 , where @xmath9 is the laser wavelength .
our experiments demonstrate , that by choosing a sufficiently small beam waist , the collisional rate can be high enough throughout the ramp , and bec can be reached . for the single beam geometry ,
the focus of the ( horizontally oriented ) trapping beam is reduced to a @xmath23 beam waist . in principle
, one could work with an even tighter focus to fully compensate for the smaller compression in this geometry . in our experiments , we have instead started with a higher number of atoms and also used a slightly longer evaporation time .
the weak confinement along the beam axis of the single dipole trap geometry offers the additional benefit that it allows us to remove atoms in field sensitive spin projections with a moderate magnetic field gradient .
we note that an alternative to such a state selection is to achieve field - insensitivity by using of atoms with a spin singlet ground state , as e.g. atomic ytterbium . very recently in an impressive experiment , evaporative cooling in a crossed dipole trap has allowed for the production of a bose - einstein condensate of these atoms @xcite . for our single dipole trap experiments
, we increase the mot loading time to @xmath71 , during which @xmath72 are captured . at the end of the mot loading
, we apply a temporal dark - mot phase as described in the above section .
subsequently , all near - resonant optical beams are extinguished and the atoms are trapped in the quasistatic dipole trapping potential alone . typically , @xmath73 atoms are captured in the single beam trap .
we have characterized the vibrational frequencies of the trap . at full trapping laser power ( @xmath74 ) ,
we measure @xmath75 and @xmath76 , corresponding to vibrations orthogonal and collinear respectively to the beam axes . in this single beam geometry
, there is no spherical symmetry and the longitudinal trap frequency is about 13 times smaller than the radial one .
we again have first analyzed the trapping of atoms while maintaining full co@xmath0-laser trapping power . following a 100ms long plain evaporation phase
, we measure an atomic temperature of @xmath77 and an atomic density of @xmath78 .
this corresponds to a product @xmath79 .
although this value is below best values obtained in quasistatic crossed dipole trap configurations or optical lattices , it is clearly above results achieved in magnetic traps prior to forced evaporative cooling . from the above values , we derive a high collision rate of @xmath80 . this has encouraged us to proceed with forced evaporative cooling in this single beam dipole trapping geometry . to cool the atoms towards lower temperatures , we apply a @xmath81 long forced evaporation ramp , during which the power of the dipole trapping beam is reduced from its initial value of @xmath74 to a final value of @xmath82 .
this ramp is applied directly after the dark mot phase , as for this geometry no improved cooling is observed when an initial plain evaporation phase is added .
we again use a ramp form as described by eq . [ 3 ] .
optimum cooling in the single dipole trap was observed when using parameters @xmath83 and @xmath84 near @xmath85 and 1.4 respectively .
we have recorded time - of - flight ( tof ) shadow images of the atomic cloud for different free expansion times .
for this measurement , the mot quadrupole field was switched off after the d - mot phase .
[ fig:4]a shows an image recorded directly after switching of the trap laser , i.e. with no free expansion .
the trapped cloud is cigar shaped and strongly elongated along the weakly confining beam axis .
[ fig:4]b and [ fig:4]c give time of flight images recorded after a 8ms and 15ms respectively long free expansion phase .
in the former image the cloud is almost symmetric , while in the latter image the symmetry axis is inverted .
we interpret this series of images as a clear evidence for bose - einstein condensation in the single beam optical trap . in this asymmetric trap geometry
, the cloud expands faster along the radial than along the axial direction due to the anisotropic release of mean field energy .
the formed condensate has spinor nature . to analyze the distribution of zeeman components ,
we apply a stern - gerlach magnetic field gradient during the free expansion phase , which spatially separates the different zeeman components .
[ fig:5]a shows a typical measurement , where a separation into clouds with different spin projections is clearly visible .
we produce condensates with typically 12000 atoms distributed among the @xmath86 zeeman components of the @xmath66 ground state .
as stated above , one of the main advantages of optical dipole traps is their state independent trapping , which allows for the formation and confinement of spinor condensates . in subsequent experiments ,
we have condensed atoms in the @xmath68 spin projection only . for this measurement , the magnetic quadrupole field generated by the mot coils
was left on throughout the experimental cycle , i.e. also during both dipole trapping and detection phases . for our experimental parameters , stable trapping of atoms during the final phase of the evaporative cooling
is then only possible for the field insensitive ( @xmath68 ) spin projection .
atoms in spin projections @xmath87 are removed during the course of evaporative cooling by the field gradient ( note that the dipole trap does in general not exactly overlap with the center of the quadrupole field ) .
[ fig:5]b shows a typical obtained time - of - flight shadow image of the condensed cloud .
the measurement shows that only the @xmath68 zeeman state is populated .
we typically obtain 7000 condensed atoms in this first order field - insensitive state .
note that this number is larger than the obtained @xmath68 fraction of the above discussed spinor condensate , which we attribute to sympathetic cooling with atoms in @xmath87 spin projections during condensate formation .
the critical temperature of the @xmath68 condensate is @xmath88 , and the peak - density is @xmath89 .
for both the spinor and the @xmath68 condensate we measure condensate lifetimes near @xmath90 .
this value is shorter than the inverse loss rate due to background gas collisions , and we attribute this lifetime to be mainly limited by three - body losses . within our experimental uncertainty , we do not observe any difference between the lifetimes of a spinor and the @xmath68 condensate .
further , we do not observe transfer of atoms from @xmath68 state into field - sensitive states . to our belief , spin changing collisions here
are suppressed by the quadratic zeeman shift in the inhomogeneous magnetic field @xcite .
we have investigated bose - einstein condensation of rubidium atoms by direct evaporative cooling in tightly confining , quasistatic dipole traps . the quantum degenerate regime has been reached both in a crossed beam and a single beam co@xmath0-laser dipole trapping geometry .
the former configuration allows cooling to quantum degeneracy with a somewhat larger trap focus . on the other hand ,
the newly developed single beam approach towards bose - einstein condensation provides a simpler setup and is less alignment sensitive .
the scheme provides a very stable route towards the achieving of quantum degeneracy .
finally , we demonstrated that in the single beam setup a moderate magnetic field gradient yields a trapping configuration , in which stable confinement is possible only for atoms in a @xmath68 spin projection .
the chemical potential of the produced bose - einstein condensate is first order insensitive to stray magnetic fields . for the future
, we believe that the demonstrated single beam method for producing bose - einstein condensates due to its technical simplicity opens up wide applications in the physics with quantum gases . further research could investigate an extension to the bose - einstein condensation of other atomic species , as e.g. alkaline - earth atoms .
further , we anticipate the described method to produce condensates in a field - insensitive spin state can have impact on the application of quantum degenerate atoms to precision measurements , as e.g. atomic clocks or atom interferometry experiments .
one can furthermore envision experiments with two - component quantum degenerate gases driven by exciting the microwave clock transition between the hyperfine ground states @xmath66 , @xmath91 and @xmath92 , @xmath91 .
it clearly remains important to study the stability of different spin projections against spin changing collisions in more detail .
quasistatic optical dipole traps are of large interest for quantum computation . in a co@xmath0-laser lattice geometry
the lattice spacing is so large that individual sites can be optically resolved , which allows for an individual addressing of qubits @xcite . by inducing a mott insulator phase transition ,
unity occupation of atoms in lattice sites can be achieved .
an entanglement of atoms in the lattice is e.g. possible by cold , controlled collisions @xcite .
a. j. kerman , v. vuleti@xmath93 , c. chin , and s. chu , phys .
lett . * 84 * , 439 ( 1999 ) ; d. j. han , s. wolf , s. oliver , c. mc.cormick , m. t. depue , and d. s. weiss , phys . rev .
lett * 85 * , 724 ( 2000 ) ; t. ido , y. isoya , and h. katori , phys .
a * 61 * , 061403 ( r ) ( 2000 ) .
barrett , m. s. chang , c. hamely , k. fortier , j. a. sauer , and m. s. chapman , in `` proceedings of the xviii international conference on atomic physics '' , h.r .
sageghpour , e.j .
heller , and d.e .
pritchard ( eds . ) ( world scientific , 2003 ) . ) .
( a ) image taken with nearly no free expansion , showing thus the cigar - shaped spatial distribution of the trapped cloud .
( b ) and ( c ) : recorded after a @xmath95 and @xmath96 respectively of free expansion . especially at the latter time , the asymmetry of the cloud is inverted . ] .
( a ) stern - gerlach magnetic field applied only during the free expansion phase .
the three components @xmath87 , 0 of a spinor condensate have separated into three spatially resolved clouds .
( b ) stern - gerlach field activated during the evaporative cooling phase . here , a pure @xmath68 condensate is generated . ] | we report on the achieving of bose - einstein condensation of a dilute atomic gas based on trapping atoms in tightly confining co@xmath0-laser dipole potentials .
quantum degeneracy of rubidium atoms is reached by direct evaporative cooling in both crossed and single beam trapping geometries . at the heart of these all - optical condensation experiments is the ability to obtain high initial atomic densities in quasistatic dipole traps by laser cooling techniques .
finally , we demonstrate the formation of a condensate in a field insensitive @xmath1 spin projection only .
this suppresses fluctuations of the chemical potential from stray magnetic fields . |
single molecule electronics@xcite has been mostly investigated in the high temperature and strong contact to the electrode regime .
the opposite limit of low temperature and weakly coupled molecular junctions poses a challenge to the currently available experimental techniques .
still the possibility to probe the spectroscopy of single molecule junctions via a lateral gate could offer new insights to the peculiar coupling of the electrical and mechanical degrees of freedom at the nanoscale . in order to be able to establish the transport mechanisms governing such molecular junctions , a technique which could tackle on one hand single electron charging effects and , on the other hand ,
the inclusion of the electron - vibron coupling is of extreme importance . in the last ten years
, the nonequilibrium green function ( negf ) formalism has been successfully employed to describe transport observables on the base of a density functional theory ( dft ) description of the electronic structure@xcite and model hamiltonian approaches,@xcite recently it is applied for the influence of the vibron dynamics onto a molecular transistor , and lots of excellent results are obtained.@xcite however , when coming to the cb regime , the negf method has been usually substituted with master equation approaches ( me ) , which prevents dft from describing cb effects .
last year , we proposed an ansatz to combine negf with equation of motion ( eom ) method.@xcite with help of the ansatz above , in non - equilibrium case , the coulomb blockade can be completely described just within single - particle space simply , and the result fully agrees with the one from me which is performed in many - particle space . for dft to describe coulomb blockade,@xcite the double subspace ( spin - up and spin - down ) is usually employed .
however , it is hard to obtain the coulomb blockade effects correctly in non - equilibrium case . our purpose is to introduce coulomb blockade effect to non - equilibrium dft calculation by this ansatz . in this paper , with the help of the ansatz in ref .
@xcite , by the model hamiltonian and eom approach , we propose a self energy to describe cb effects in non - equilibrium case . electronic occupation number , electronic current and differential conductance are calculated .
the charging - induced steps and the ratio of 2/3 : 1/3 in the step heights of the occupation number and the current , which are the important characteristics of cb by me,@xcite is obtained correctly .
the comparison with the complete cb results@xcite also is done . for the occupation number and the current ,
the difference is very small . for the conductance , only in the peak height ,
the difference is very clear , while there is no difference in the peak position .
the cb stability diagram is also shown in the paper .
we can see that the self energy can describe cb characteristics .
the more important is that it is very convenient to introduce this self energy to dft code even in non - equilibrium case . a scheme to perform cb correction in dft calculation
is suggested with double counting correction ( dcc ) .
then it is realized in gdftb.@xcite a _ toy _ molecule is taken for testing , and the cb characteristics are shown in the results correctly .
[ toy_mol ] the paper is organized as follows : firstly a self energy for cb is proposed by the model hamiltonian and eom approach ( sec .
ii ) ; secondly , based on the self energy above , a scheme with dcc is proposed to introduce cb correction to dft code ( sec .
iiia ) ; finally , the calculation on a _ toy _ molecule is performed in weak coupling regime , and the cb characteristics are shown in the results correctly ( sec .
[ n - i - v - comp ] v , @xmath1v , @xmath2v , @xmath3v , @xmath4 .
@xmath5 . the red dash curves are from approximation in eq .
[ gf_r / a_2 ] , the black solid curves from the truncation in ref .
@xcite which fully agree with the results from me , @xcite while the blue dot curves are for the case @xmath6v in which there is no cb . in the red dash curves and the black solid curves ,
the new steps appear for charging effect , and the heights of them are in the ratio of 2/3 : 1/3 , which are the important characteristics of cb obtained by me . @xcite , title="fig : " ] v , @xmath1v , @xmath2v , @xmath3v , @xmath4 .
the red dash curves are from approximation in eq .
[ gf_r / a_2 ] , the black solid curves from the truncation in ref .
@xcite which fully agree with the results from me , @xcite while the blue dot curves are for the case @xmath6v in which there is no cb . in the red dash curves and the black solid curves ,
the new steps appear for charging effect , and the heights of them are in the ratio of 2/3 : 1/3 , which are the important characteristics of cb obtained by me .
@xcite , title="fig : " ] in non - equilibrium dft calculation here , the cb correction will be introduced to each level of the molecular fragment @xmath7 ( see fig . 1 ) .
@xcite therefore , the multi - level anderson impurity model is read as follows , @xmath8 with @xmath9 where @xmath10 and @xmath11 are the operators for electrons on the dot and on the left ( @xmath12 ) and the right ( @xmath13 ) lead , @xmath14 is the charging energy of level @xmath15 , @xmath16 is the ( @xmath15 , @xmath17 ) level of the quantum dot , while @xmath18 is the spin @xmath17 level of lead @xmath19 in @xmath20 space , @xmath21 . with the help of the eom and the truncation approximation ,
we can obtain a closed set of equations for the retarded and advanced gfs @xmath22 , @xmath23 where @xmath24 and @xmath25 are the electron self - energies from leads with @xmath26 .
re - arranging eqs .
( [ gf_r / a_1 ] ) and ( [ gf_r / a_2 ] ) , we can obtain the retarded gf as follows , @xmath27 with @xmath28 @xmath29 is a diagonal matrix composed of @xmath30 , @xmath31 is the one of @xmath32 , and @xmath33 is of @xmath14 , while @xmath34 is the self - energy matrix from eq .
( [ self - e ] ) . from eq .
( [ gf - r ] ) , we can see that the coulomb interaction is just included in @xmath35 .
therefore , with the help of eq .
( [ gf - u - r ] ) , the retarded cb self energy @xmath36 can be obtained by the relation @xmath37^{-1}&=&\left\{\frac{1}{\omega - h_{0}+\textrm{i}\eta } + \frac{1}{\omega - h_{0}+\textrm{i}\eta}\right.\nonumber\\ & & \left.\cdot\sigma^{r}_{\textrm{h}}\cdot\frac{1}{\omega - h_{0}-u+\textrm{i}\eta}\right\}^{-1}\nonumber\\ & \equiv&\left\{\omega - h_{0}-\sigma^{(\textrm{cb})r}\right\}^{-1}.\end{aligned}\ ] ] the result is as follows , @xmath38^{-1}.\end{aligned}\ ] ] [ g - v - comp ] = 1.0 by the ansatz in ref .
@xcite , the lesser gf can be written out directly , @xmath39 with @xmath40 , and @xmath41 , @xmath42 , while @xmath43 are the diaganol matrix composed of @xmath44 , and @xmath45 is the equilibrium fermi function . after the re - arrangement by the way in ref .
@xcite , we can obtain , @xmath46 by the helps of @xmath47 ( see appendix [ app : sigma - cb - l ] ) , we could get @xmath48 if @xmath49 . therefore , here , the current @xmath50 can be calculated simply by the landaur formula , @xcite @xmath51,\end{aligned}\ ] ] instead of the complicated formula in ref .
@xcite .
[ g - v - v ] , with @xmath52v , @xmath53v , @xmath2v , @xmath54v.,width=302 ] in the case of double levels ( @xmath55 and @xmath56 ) , the numerical calculation is performed , which contributes the direct comparison between the truncation in eq .
( [ gf_r / a_2 ] ) and the one in ref .
@xcite by which cb results fully in agreement with the ones from me can be obtained .
@xcite the comparisons of the electronic occupation number @xmath57 and the current @xmath50 as a function of the bias voltage @xmath58 at fixed gate voltage @xmath59 are shown in fig . 2 .
firstly , the difference of the results by the two methods is very small .
secondly , the cb characteristics are very clear in them : the steps appear for charging - induced level - split , and the step heights are in the ratio of 2/3 : 1/3 , which are also obtained by me as the important cb characteristics .
@xcite only in the comparison of differential conductance ( @xmath60 ) ( see fig .
3 ) , the difference in the heights of the peaks ( @xmath61-axis ) is very clear , while no difference appears in the positions of the peaks ( @xmath58-axis ) . from the cb stability diagram in the case of @xmath55 and @xmath62 ( fig . 4 )
, it can be seen that in this case , the approximation in eq .
( [ gf_r / a_2 ] ) _ almost _ includes all cb characteristics ( the _ complete _ cb stability diagram by negf is shown in ref . @xcite ) .
it should be noted that although the approximation above can just include _ some _ cb characteristics correctly , @xcite it is very convenient to be introduced to dft calculation ( which will be shown in the next chapter ) .
for cb calculation , the system is partitioned as follows ( shown in fig .
1 ) : the molecular fragment @xmath7 is the cb part , the fragments @xmath63 and @xmath64 are contacting area , and the @xmath65 and @xmath66 are the leads .
then we just introduce cb correction in fragment @xmath7 , while the non - equilibrium calculation should be performed within the fragment @xmath67 . in dft , the ks equation @xmath68 can be re - written as , @xmath69 with @xmath70 is the overlap matrix .
then an effective hamiltonian matrix@xcite for model - hamiltonian calculation can be obtained from ks one , @xmath71 after performing the transformation on @xmath72 from atomic basis to the fragment basis ( which are from the eignvectors of molecular fragment @xmath67 and are orthonormal ) , we take , @xmath73 with @xmath74 are the index for the eigenvectors of the fragment @xmath67 , and @xmath75 are the the element of effective hamiltonian matrix @xmath76 . within the fragment basis , from equation ( [ eq : key-1 ] ) , we can obtain the self energy for cb correction in dft as follows , @xmath77 where @xmath78 , @xmath14 is the hubbard energy of the fragment orbital _
m _ , and @xmath79 if @xmath80 . the occupation number of electrons @xmath81 can be obtained by the transformation on the density matrix from atomic basis @xmath82 to fragment basis @xmath83 . considering dcc along
the idea similar to the case in lda+u , @xcite we can get @xmath84 .
the way to calculate @xmath14 within dftb / gdftb is shown in appendix [ app : u ] .
finally the cb - correction self energy @xmath85 will be transformed back to the atomic basis @xmath86 from the fragment basis @xmath87 . by the help of eqs .
( [ gf - r ] ) , ( [ eq : key-0 ] ) , ( [ gf - l ] ) and ( [ sigma - l ] ) , introducing the self energy @xmath88 from leads , we can calculate the gfs as follows , @xmath89 and @xmath90 it should be noted that the eigenvectors of the fragment @xmath67 is updated in every cycle according to the updated ks hamiltonian .
[ fig : tunneling ] = 0.93 [ fig : occ - cur - gdftb ] of lumo as a function of the bias voltage .
( b ) the current @xmath50 as a function of the bias voltage . in them , the cb characteristics ( the charging - induced steps and the ratio of 2/3 : 1/3 in the heights of the steps ) are very clear . ,
title="fig : " ] of lumo as a function of the bias voltage .
( b ) the current @xmath50 as a function of the bias voltage . in them ,
the cb characteristics ( the charging - induced steps and the ratio of 2/3 : 1/3 in the heights of the steps ) are very clear . , title="fig : " ] the scheme above is realized within gdftb.@xcite a _ toy _ molecule ( @xmath91 ) is taken for the calculation ( see fig . 1 ) .
the @xmath92 is the cb part , while @xmath93 is the linker .
the bond length of c - c is @xmath94 , and the one of s - c is @xmath95 . in the testing calculation of this chapter , for clarity within the level structure and the cb characteristics , the contribution from fragments
@xmath63 and @xmath64 is ignored , and @xmath96 is taken instead of the @xmath14 calculated by the method in appendix [ app : u ] ( which are about @xmath97 ) . also for simplicity in the calculation ,
the fictitious golden leads are used,@xcite and the minimal basis is taken .
the transmission function in spectral space is shown in fig .
5 . in the case without cb correction , we can see that there are seven levels close to the fermi energy , while the homo and the lumo+1 are double - degenerate , respectively .
there is no level - split for charging , though the electronic occupation numbers of the seven levels above are : 2.00 , 1.89 , 1.99 , 1.99 , 0.15 , 0.00 , 0.00 , ( it is clear that lumo and homo-2 are not completely empty / occupied ) .
then , introducing cb correction , the lumo and the homo-2 will split into @xmath98 and @xmath99 for the charging effect in the open shell , respectively .
the electronic occupation number of lumo and the current as a function of bias voltage @xmath58 are shown in fig .
the charging - induced steps and the correct ratio of 2/3 : 1/3 in the heights of the steps are very clear , which is consistent with the results in fig
the physics under it can be understood with the help of the charging - induced level - split , which is shown in the following .
1 ) when @xmath100 , the lumo level is almost empty .
2 ) then the positive bias voltage is added , and when the lumo level @xmath101 is coming into the fermi windows ( @xmath102 ) , there are two channels ( @xmath103 ) to be opened for current . as leads to the 2/3 in the ratio .
3 ) after that , the electronic occupation number of lumo comes to be 0.66 ( see fig .
6(a ) and fig .
2(a ) ) , and for the charging effect , the degenerate levels ( @xmath104 ) will split into two ( @xmath105 and @xmath106 ) .
4 ) at the time that the level @xmath107(@xmath108 ) enters the fermi windows , there will be only one channel to be opened , which is the reason of the 1/3 in the ratio .
in this paper , we have shown a new way to introduce cb correction to dft calculation for the non - equilibrium case .
the main elements of the approach are the following .
\1 ) with the help of the ansatz in ref .
@xcite , a self - energy is proposed for the ( non - equilibrium ) cb by the model hamiltonian and eom approach . from the comparison with the complete cb results , we can see that it can include the characteristics of cb correctly .
further , the more important is that , from the view of dft - based quantum transport calculation _ in non - equilibrium case _ , this self - energy is very convenient for programming more than the truncation in ref .
@xcite and me approaches .
\2 ) based on this self energy , a scheme with dcc is proposed to introduce cb correction to _ non - equilibrium _ dft calculation . as is then realized within gdftb .
@xcite by the new code above , the quantum - transport properties of a toy molecule is calculated in cb regime . in the results of the electronic occupation number and the current as a function of bias voltage ,
the cb characteristics ( the charging - induced steps and the ratio of 2/3 : 1/3 in the step heights ) appear correctly .
i thank a lot prof .
gianaurelio cuniberti for great supports , prof .
marcus elstner for helpful discussions , and prof . thomas frauenheim and dr .
alessandro pecchia for kindly gdftb code .
for simplicity , here , only a double - level ( @xmath55 with spin - up and spin - down ) case is taken to describe the quantum dot .
then the index @xmath15 is ignored in this part , and from eq .
( [ h - d ] ) , the hamiltonian of the dot can be re - written as , @xmath110 by eq .
( [ gf - u - r ] ) , with the help of the ansatz in ref .
@xcite , we can directly write out @xmath111 as follows , @xmath112 where @xmath113 is _ some kind of pseudo_-fermi function in non - equilibrium case , and @xmath114 assuming that there exists the relation @xmath115 , and @xmath111 can be re - written in the form , @xmath116 then with the help of eq .
( [ gf - u - r ] ) , we will obtain @xmath117 comparing eqs .
( [ eq:1 ] ) and ( [ eq:2 ] ) , we can get @xmath118 with @xmath119 therefore , it is clear that @xmath120 , since @xmath114 and @xmath121 never comes to be infinite .
in dftb / gdftb,@xcite the tight - binding hamiltonian is written as follows , @xmath122 with @xmath123 where @xmath124 and @xmath125 .
@xmath126 is the index for the eigenstate of molecule ( or molecular fragment ) , @xmath127 are the index of atomic basis , while @xmath128 indicates atoms .
@xmath129 is the charge of atom @xmath19 , and @xmath130 is the electron occupation number of eigenstate @xmath126 .
@xmath131 is a function of @xmath132 , @xmath133 and @xmath134 , @xcite with @xmath132 ( @xmath133 ) is the hubbard @xmath33 of atom @xmath19 ( atom @xmath135 ) , and @xmath134 is the distance between the two atoms .
@xmath136 is the project of the eigenvector @xmath126 on the atomic basis @xmath137 , while @xmath138 are the elements of overlap matrix in atomic basis . with the help of linear combination of atomic orbitals(lcao ) ansatz ,
@xcite we can transform the hamiltonian matrix ( [ dftb_h ] ) from atomic basis to molecular basis , and get the eigenvalue of the eigenstate @xmath126 as follows , @xmath139 then , according to the defination of hubbard @xmath33 , @xcite we will obtain , @xmath140 ignoring the contribution from @xmath141 and @xmath142 , with the condition @xmath143 , we can get @xmath144 with @xmath124 , @xmath125 .
the calculation of hubbard @xmath33 by the approximations ( [ u - approx ] ) is performed in four examples : 1 ) guanine - cytosine base pair ( gc ) , 2 ) adenine - thymine base pair ( at ) , 3 ) benzene ( @xmath145 ) , 4 ) double carbon ( @xmath92 ) in the case that the bond length is 1.2 @xmath146 .
the results are shown in table i. the error from the approximations is less than @xmath147 , which is acceptable within dftb / gdftb . | quantum transport through single molecules is very sensitive to the strength of the molecule - electrode contact .
when a molecular junction weakly coupled to external electrodes , charging effects do play an important role ( coulomb blockade regime ) . in this regime , the non - equilibrium green function is usually substituted with master equation approaches , which prevents the density functional theory from describing coulomb blockade in non - equilibrium case .
last year , we proposed an ansatz to combine the non - equilibrium green function technique with the equation of motion method . with help of it ,
coulomb blockade was obtained by non - equilibrium green function , and completely agrees with the master equation results [ phys .
rev .
b * 76 * , 045408 ( 2007 ) ] . here , by the ansatz , we show a new way to introduce coulomb blockade correction to dft calculation in non - equilibrium case . and the characteristics of coulomb blockade are obtained in the calculation of a @xmath0 molecule correctly . |
although there is no universally agreed upon central engine powering gamma - ray bursts ( grbs ) , growing evidence supports the association of least the long - soft grbs ( all those whose counterparts have been localized ) with the death of massive stars .
this evidence includes the association of grbs with regions of massive star formation ( bloom , kulkarni , & djorgovski 2002 ) and `` bumps '' in the optical afterglows of several grbs that have been related to the light curves of type i supernovae ( e.g. , bloom et al . 2002 ; garnavich et al .
in addition , grb 980425 has been associated with sn 1998bw ( e.g. , iwamoto et al .
1998 ; woosley , eastman , & schmidt 1999 ) .
frail et al .
( 2001 ) , panaitescu & kumar ( 2001 ) have studied beaming angles and energies of a number of grbs and have found that the central engines of grbs release supernova - like energies . given these discoveries , the currently favored models are all based upon the collapse of massive stars and their byproducts , one of which is a relativistic jet . among those models involving massive stars , the collapsar model ( woosley 1993 ;
macfadyen & woosley 1999 ) has become a favorite .
a `` collapsar '' is a rotating massive star whose iron core has collapsed and formed a black hole and an accretion disk . in this model , the central black hole and the disk use neutrinos or magnetic fields to extract part of the gravitational potential or rotational energy and form powerful relativistic jets along the polar axes .
jets from collapsars have been studied numerically in both newtonian ( macfadyen & woosley 1999 ; macfadyen , woosley , & heger 2001 ) and relativistic simulations ( aloy et al .
2000 ; zhang , woosley , & macfadyen 2002 ) and it has been shown that the collapsar model is able to explain many of the observed characteristics of grbs .
these previous studies have also raised issues which require further examination , especially with higher resolution .
for instance , the emergence of the jet and its interaction with the material at the stellar surface and the stellar wind will definitely lead to some sort of `` precursor '' activity .
the long term dynamics of the jet is critical in producing the observed gamma - rays and afterglows . we have recently carried out multi - dimensional calculations to address some of these issues .
a collapsar is formed when the core of a massive star collapses to a black hole and an accretion disk .
the interaction of this disk with the hole , through processes that are still poorly understood , produces jets with a high energy to mass ratio . for our simulations , we are concerned primarily with the propagation of these relativistic jets , their interactions with the star and the stellar wind , and the observational implications , and not so much with how they are born .
we model the propagation of relativistic jets inside and just outside collapsars using a multi - dimensional relativistic hydrodynamics code that has been used previously to study relativistic jets in the collapsar environment ( zhang et al .
briefly , our code employs an explicit eulerian godunov - type method developed by aloy et al .
three - dimensional numerical simulations of relativistic jets in collapsars are still rather expensive , so the present study consists of a series of two - dimensional calculations with high resolution .
our initial model is a @xmath1 helium star calculated by heger & woosley ( 2002 ) .
the helium star has been evolved to iron core collapse .
this presupernova model is then remapped into our two - dimensional cylindrical grid @xmath2 , which consists of 1500 zones in @xmath3 and 2275 zones in @xmath4 . in one of the models
, we used a larger grid in @xmath5-direction , 2375 zones for @xmath6 . in each case , the zoning is nonuniform with higher resolution in the inner region .
the radius of the initial helium star is about @xmath7 and the surface of the star is very finely zoned . outside the star , the background density , which comes from the stellar wind ,
is assumed to be @xmath8 , and the density at @xmath9 is set to @xmath10 , here @xmath11 is the distance to the center of the star .
this corresponds to a mass loss rate of @xmath12 and a velocity of @xmath13 at @xmath14 .
we presume that a highly relativistic jet has already formed inside the inner boundary of our computational grid and study its evolution after it enters the grid .
in particular , an axisymmetric jet is injected along the rotation axis through the inner boundary within a radius , @xmath15 , here taken to be a free parameter .
the initial jet is additionally defined by its total energy ( excluding rest mass energy ) flux per jet , @xmath16 , initial lorentz factor , @xmath17 , and the ratio of its kinetic energy to total energy , @xmath18 .
we have run a series of calculations : ( a ) @xmath19 , @xmath20 , @xmath21 , @xmath22 ; ( b ) @xmath23 , @xmath24 , @xmath25 , @xmath26 ; and ( c ) @xmath19 , @xmath27 , @xmath21 , @xmath22 .
there presently exists no rigorous calculation of the jet formation process .
however , a total energy of order @xmath28 ergs is reasonable .
the typical duration of a long grb is @xmath29 seconds . in our calculations ,
the jet is left on for 20 seconds , and then gradually shut off ( linearly with time ) . we know from observations that the grb outflows are narrowly beamed . in our calculations ,
the radius of the jet is chosen to have a half - opening angle of @xmath30 at @xmath31 .
it is more difficult to estimate the lorentz factor and the ratio of kinetic energy to total energy .
as our previous studies showed , a jet that starts with a high lorentz factor @xmath32 at @xmath33 will be shocked and its lorentz factor will become @xmath34 at @xmath31 ( zhang et al .
for the parameters chosen , the jet starts with a lorentz factor of 5 or 10 at @xmath31 , and would have a final lorentz factor of @xmath35 if all internal energy is converted into kinetic energy . in all models , the jet propagates along the rotational axis .
after a short time , the jet consists of a supersonic jet beam ; a cocoon made of shocked jet material and shocked medium ; a terminal bow shock ; a contact discontinuity ; and backflows .
a snapshot of model b is shown in figure 1 . in agreement with our previous studies
( zhang et al . 2002 ) , the jet in both models is narrowly collimated and its core , very thin .
interaction of the jet with the star and the cocoon imprints considerable time structure on the flow . as the jet passes through the star
, it also explodes it .
eventually , the jet breaks out the star .
the average velocities of the head of the jet just before breakout are @xmath36 , @xmath37 , and @xmath38 , for models a , b , and c , respectively .
as expected , the jet accelerates as it moves outwards because the density and pressure of the star decrease quickly , especially at the surface of the star .
as the jet passes through the star , an over - pressurized cocoon is formed .
this cocoon escapes from the star along with the jet beam and accelerates as its internal energy is converted into kinetic energy by adiabatic expansion ( figure 2 ; see also ramirez - ruiz , celotti , & rees 2002 ) . after the breakout into the low density stellar wind , the cocoon is no longer confined .
the evolution of the jet and its cocoon in the stellar wind will be further discussed in 3.2 .
the velocity of the head of the jet in the star is subrelativistic , and the jet beam is highly relativistic .
shocked medium and shocked jet material piles up near the contact discontinuity and feeds the cocoon .
this forms a `` plug '' at the head of the jet . after breakout
, the initially subrelativistic plug will be moving at a lorentz factor of @xmath39 since it is accelerated by the highly relativistic jet and the stellar wind can not hinder its movement .
the interaction of the `` plug '' and the jet and its implications for grbs will be discussed in 3.2 and 4
( see also waxman & mszros 2002 ) .
after it breaks out the star , both the jet and its cocoon are loaded with a lot of internal energy .
although its current lorentz factor is only about 10 , the final lorentz factor of the jet can be @xmath40 if it can expand adiabatically to gain its terminal lorentz factor .
the lorentz factor and density at the end of our simulation for model a is shown in figure 3 . in all models ,
the average lorentz factor in the jets becomes more than 100 at the end of the simulations ( t= 70s ) , and there are still some internal energy remaining in the jets at that moment .
when they are viewed on the axis , the jets have an isotropic equivalent total energy of @xmath41 , @xmath42 , and @xmath43 , for models a , b , and c , respectively .
the isotropic equivalent energy is derived by assuming an isotropic fireball instead of a jet ] .
this energy should be enough to power a normal grb even if the efficiency for making @xmath0-rays is very small .
after the jet breaks out the star , its cocoon expands both outwards and sideways ( figure 2 ) .
an interesting effect of the cocoon is that it prevents the jet beam from expanding sideways . without the cocoon ,
a bare `` hot '' jet with a lorentz of about 10 will inevitably expand sideways ( zhang et al . 2002 ) .
the cocoon emerges out the star with an energy close to the total energy injected at the base of the jet before the jet emerges from the star ( ramirez - ruiz et al .
the conversion of its internal energy into kinetic will give the cocoon a lorentz factor of @xmath44 and the cocoon can reach up to @xmath45 .
for instance , the isotropic equivalent energy of relativistic material at @xmath45 is about @xmath46 for model c. in models a and b it is an order of magnitude greater .
the jet breaks out the star with a plug in front of it .
the plug is made of the shocked stellar and jet material .
after breakout , the plug has comparable density to that of the jet , and a lorentz factor of about 10 - 20 ( see e.g. , figure 3 ) . in model
a , the total rest mass in the plug at @xmath47 , is about @xmath48 , which is @xmath49 of that of the jet .
they are @xmath50 and @xmath51 , about half of that of the jet , for models b and c , respectively .
these massive plugs will surely alter the structure of the grb jets .
the interaction of the jet and the plug will accelerate the plug .
meanwhile , the jet will be slowed down .
part of its kinetic energy will be converted into its internal energy .
this interaction may have important implications for observations .
our special relativistic calculations show that a jet originating near the center of a collapsing massive will emerge while still carrying almost all the energy injected at the center .
after it breaks out of the star , the jet will move forward almost freely , no more cocoon will be generated .
the standard fireball model requires not only high lorentz factor , but also the lorentz factor to vary rapidly .
our simulations show promise in satisfying these constraints . however , due to the low resolution ( @xmath52 ) outside the star , it is hard to resolve those small scale variabilities and the numerical viscosity gradually smooths the time structure in the jet .
we have found that the cocoon can expand sideways up to @xmath53 after it emerges from the star .
the deceleration of the cocoon by external shocks will produce @xmath0-rays and x - rays ( see also ramirez - ruiz et al .
the `` anomalous '' grb 980425 might be a cocoon viewed from a large angle .
it might also contribute to the afterglows and precursors of a normal grb and may be the origin of recently discovered hard x - ray flashes ( heise et al . 2001
; kippen et al 2002 ) .
these flashes have many properties of grbs ( energy , isotropy , approximate duration , distribution with redshift ) , but have a much softer spectrum ( and so far no optical afterglow ) .
the association of this kind of softer transient with gamma - ray bursts seen off - axis has been predicted by our group for many years ( woosley et al 1999 ; woosley & macfadyen 1999 ; woosley 2000 , 2001 ) . behind the bow shock ,
there is a plug in front of the jet beam . after breakout
, the main jet beam continues to push and accelerate this plug .
meanwhile , the jet beam is slowed down as its energy is transfered to the plug and part of its kinetic energy is converted into its internal energy .
the basic picture of a fast moving jet beam pushing a `` slowly '' moving plug is that there is a reverse shock , which moves backward and slows down the jet beam , a contact discontinuity , and a forward shock .
since the plug has a limited length , the forward shock will reflect at the end of the plug .
this further complicates the dynamics ( e.g. , waxman & mszros 2002 ) . at the end of our simulation
, the jet and the plug are moving relativistically and the stellar wind has not been able to decelerate the jet , so we can make some analytic estimates treating the jet and the plug as a spherical symmetric fireball . where will the jet and the plug become optically thin ? in our case , the opacity in the jet and the plug is dominated by thomson scattering .
the optical depth is @xmath54 , here , @xmath55 , and @xmath56 is the initial optical depth at @xmath15 . from our results at @xmath57 ,
the plug will become optically thin at @xmath58 , @xmath59 , and @xmath60 , for models a , b , and c , respectively . and
the jet will become optically thin at @xmath61 , @xmath62 , and @xmath63 , for models a , b , and c , respectively .
when it becomes optically thin , the interaction of the plug with the jet , will produce hard emission .
in fact , it is possible that this emission could be a short - hard grb ( as defined in fishman & meegan 1995 ) . where will the jet catch up the plug and be decelerated by the plug ?
this is a very critical question . during the catch - up period ,
some of the kinetic energy of the jet is converted into internal energy .
if the deceleration happens in the optically thick regime , the increased internal energy can still be converted back into kinetic energy .
if , however , it happens in the optically thin regime , the increased internal energy will become radiation energy via this special kind of `` internal shock '' and escape from the jet .
there may not be enough kinetic energy left and high enough lorentz factor to make @xmath0-rays via `` normal '' internal shocks . in this case
, the short hard precursors from the plug may dominate at x - ray and @xmath0-ray wavelengths , and a short hard grb is likely to be seen .
numerical simulations on the catch - up process are currently underway .
we analytically estimate the radius , @xmath64 , at which the catch - up process will happen .
for simplicity we assume that the jet moves at a lorentz factor of 100 , the plug moves at a lorentz factor of 20 , and the length of the jet is about @xmath65 , then we get @xmath66 .
this radius is comparable to the radius where the jet becomes optically thin .
our simulations show a tendency for the energy in the plug relative to that in the jet behind the plug to be larger when the total energy is less ( see 3.2 ) .
this suggests that less energetic jets may be more likely to make short - hard grbs .
the near coincidence of the masses of the plug and jet and of the radii where the two share their energy with the gamma - ray photosphere suggests that there may be cases where the most prominent display comes from the plug and others where it still comes from internal shocks in the jet ( or at the jet - plug interface ) . where will the jet be decelerated by the stellar wind ?
the deceleration by the stellar wind happens when the jet sweeps up @xmath67 of its rest mass , here @xmath68 is the lorentz factor of the jet . assuming a stellar wind that has a mass loss rate of @xmath12 and a velocity of @xmath13 at @xmath14 , the deceleration radius , @xmath69 , is @xmath70 for model a. and they are @xmath71 and @xmath72 , for models b and c , respectively .
the deceleration by external shocks with the stellar wind always happens after the above events .
this justifies our one - dimensional analytic calculations because the effects of the sideways expansion are negligible when the jets are highly relativistic . in our numerical simulations , we have followed the propagation of jets to @xmath73 . however , many interesting events which are directly related to observations happen at large radius .
so the long time evolution of the jets in the stellar wind is very critical .
it will help us make many testable predictions .
numerical calculations which follow the long time evolution of the jets are under way .
hopefully , they will further strengthen our knowledge on relativistic outflows of grbs .
this research has been supported by nasa ( nag5 - 8128 , nag5 - 12036 , and mit-292701 ) and the doe program for scientific discovery through advanced computing ( scidac ; de - fc02 - 01er41176 ) .
we are indebted to alex heger for providing the precollapse model used in this calculation and to illuminating conversations with chris matzner , martin rees , and enrico ramirez - ruiz . | growing observational evidence supports the proposition that gamma - ray bursts ( grbs ) are powered by relativistic jets from massive helium stars whose cores have collapsed to black holes and an accretion disk ( collapsars ) .
we model the propagation of relativistic jets through the stellar progenitor and its wind using a two - dimensional special relativistic hydrodynamics code based on the ppm formalism .
the jet emerges from the star with a plug in front and a cocoon surrounding it . during its propagation outside the star
, the jet gains high lorentz factor as its internal energy is converted into kinetic energy while the cocoon expands both outwards and sideways .
external shocks between the cocoon and the stellar wind can produce @xmath0-ray and hard x - ray transients .
the interaction of the jet beam and the plug will also affect both of them substantially , and may lead to short - hard grbs .
internal shocks in the jet itself may make long - soft grbs . |
micro- and nano - mechanical resonators have been an interesting research topic due to their broad application in technology and fundamental physics @xcite .
this includes studies of ultrahigh precision displacement detection @xcite , mass detection @xcite , gravitational - wave detectors @xcite , and attempts to observe quantum behavior of mechanical motion @xcite .
many of the applications are fundamentally limited by thermal fluctuations , and in order to reduce their effects , it is desirable to cool the mechanical oscillators . recently , various schemes like the laser sideband cooling schemes developed for trapped ions and atoms @xcite , have been proposed for significantly cooling a mechanical resonator ( mr ) coupled to a cooper - pair box @xcite , a flux qubit @xcite , a superconducting single - electron transistor @xcite , quantum dots @xcite , trapped ions @xcite , and optical cavities @xcite . on the experimental side , optomechanical cooling schemes have been shown to be promising @xcite : the mr was cooled to ultra - low temperatures via either photothermal forces or radiation pressure by coupling it to a driven cavity .
there are two main optomechanical cooling schemes .
the first one involves an active feedback loop @xcite , and the second one works via passive back - action cooling ( also called self - cooling ) @xcite .
a fully quantum - mechanical description of cavity - assisted cooling schemes for optomechanical systems has been given in refs .
@xcite ( for a review , see @xcite ) .
ground - state cooling of a mechanical resonator via passive cooling schemes based on radiation pressure has also been investigated theoretically @xcite .
recently , other optomechanical - like cooling schemes have been proposed to replace the optical cavity by a radio - frequency ( rf ) circuit @xcite or a one - dimensional transmission line resonator ( tlr ) @xcite . however , the theoretical understanding of the cooling schemes via a rf circuit in refs .
@xcite or via a tlr in ref .
@xcite is based on a classical description of the motion of the mr .
a quantum - mechanical description of the motion of the mr , in a similar system consisting of a mechanical resonator capacitively coupled to a superconducting coplanar waveguide ( which is an example of a tlr ) , was discussed recently in ref .
@xcite , which focused on studying the entanglement between the mr and the tlr without considering the cooling of mr .
most recently , teufel _ et al . _
@xcite considered the cooling of a mr by applying directly the theoretical analysis of the cavity - assisted back - action cooling scheme @xcite to a superconducting microwave resonator .
they also presented experimental data about the cooling effect on the mr due to the microwave radiation field .
the quantum - mechanical description of tlr - assisted cooling of a mr has also been investigated in ref .
@xcite via embedding a squid @xcite , which allows to control the coupling strength between mr and tlr by controlling the flux through the squid .
there are some practical advantages @xcite in the microwave tlr schemes .
the tlr is realized in a thin on - chip superconducting film and is easily pre - cooled by standard dilution refrigeration techniques .
it is ready to be integrated with quantum circuits containing josephson junctions which may offer a sensitive measurement and a connection with quantum information processing .
in addition , the size of the mechanical resonator could be much smaller than the wavelength of the radiation in the tlr , unlike in optical cavity experiments that work with reflection . in this paper , we present a quantum - mechanical description and use it to investigate the motion of the mr when it is coupled capacitively to a driven tlr as in ref .
@xcite where the calculation was carried out in a semi - classical framework .
the hamiltonian of our tlr - assisted model is also studied in refs .
@xcite , and is very similar to that of a mr coupled to a driven optical cavity via radiation pressure coupling @xcite .
we study the tlr - assisted passive back - action cooling of a mr in detail by using a quantum langevin description ( without taking into account quantum entanglement @xcite between the mr and the tlr ) .
one of the main results of our work is to show that the mr can be cooled close to its ground state using realistic parameters : final effective mean phonon numbers below @xmath0 can be reached assuming an initial temperature of @xmath1 mk which can be achieved using a dilution refrigerator .
we discuss in detail how such a ground state cooling of the mr can be obtained for all kind of parameter choices in practice .
the system that we consider is shown schematically in fig .
[ fig1 ] : a mr is fixed on both ends ( or a cantilever fixed on one end ) located at the center of the tlr and is coupled capacitively to the central conductor of the tlr @xcite .
we restrict the description to the fundamental flexural mode of oscillation of the mr which is modeled as a harmonic oscillator of frequency @xmath2 and effective mass @xmath3 .
the tlr is driven by an external microwave at a frequency @xmath4 and can be modeled as a single mode lc resonator with frequency @xmath5 ( the second mode of the tlr @xcite ) , where @xmath6 is the inductance and @xmath7 the capacitance of the tlr .
the hamiltonian of the system reads @xmath8 the first line describes the free hamiltonian of the tlr and the mr , respectively , with lowering ( rising ) operator of the tlr mode @xmath9 ( @xmath10 ) , and the position ( momentum ) operator of the mr @xmath11 ( @xmath12 ) which satisfy @xmath13=1 $ ] and @xmath14=i\hbar$ ] .
the first term in the second line is the capacitive coupling between the tlr and the mr .
actually , it describes the capacitive energy between them .
the mr and the tlr are assumed to form a capacitor with the capacitance @xmath15 ( for small displacement ) depending on the position of the mr along the @xmath11-direction ( @xmath16 is the initial equilibrium distance without the coupling and @xmath17 the corresponding initial capacitance ; typically @xmath18 @xmath19 m @xcite ) .
the capacitor is assumed to be placed in the center of the structure , i.e. , its voltage is given by the antinode voltage of the second mode : @xmath20 ( where @xmath21 is the rms voltage @xcite ) , since the length of the mr is usually much shorter than that of the tlr : @xmath22 cm @xmath23 @xmath19 m .
the last term in eq .
( [ ham0 ] ) describes the input driving of the tlr by an external microwave field with the coupling strength @xcite @xmath24 , where @xmath25 is the decay rate of the tlr , @xmath26 is the input external microwave drive power . here ,
the non - rotating wave terms like @xmath27 and @xmath28 have been ignored since we keep @xmath29 .
usually , the fundamental oscillation frequency is of the order of @xmath30 ( @xmath31 - @xmath32 ) hz for micromechanical resonators and @xmath30 ( @xmath33 - @xmath34 ) hz for nanomechanical resonators ; the tlr frequency can be made to be of the order of @xmath35 ghz . here
we will focus on the case of a micro - mr for which the condition @xmath2
@xmath36 @xmath37 is satisfied . in the interaction picture with respect to @xmath38 and neglecting the rapidly - oscillating terms
, the hamiltonian reads @xmath39 where @xmath40 is a real coupling constant ; @xmath41 is the detuning , and @xmath42 is the modified frequency of the tlr shifted by the coupling between tlr and mr .
this hamiltonian resembles that used in cavity - assisted cooling schemes @xcite .
this suggests that the capacitive - coupling scheme in a microwave tlr can be used to cool the mr like in the case of radiation - pressure cooling in an optical cavity .
the dynamics is also determined by fluctuation - dissipation processes that affect both the tlr and the mechanical mode .
they are taken into account in a fully consistent way by the quantum langevin equations @xcite : [ nonlinlang1 ] @xmath43 here @xmath44 ( @xmath45 ) is the noise operator due to the external microwave drive , and @xmath46 denotes the quantum brownian force that the resonator is subject to .
they satisfy @xcite @xmath47 where @xmath48 $ ] is the mean number of thermal microwave photons of the tlr , @xmath49 is the boltzmann constant , @xmath50 the temperature of the environment , and @xmath51 the damping rate of the mr . for simplicity , we have assumed that both the bath correlated to the external microwave drive field and the one connected to the mr have the same temperature @xcite .
we now perform a similar calculation as that given in refs .
the steady - state solution of the quantum langevin equations ( [ nonlinlang1 ] ) can be obtained by first replacing the operators by their average and then setting @xmath52 .
hence we can get the steady - state values as @xmath53 where @xmath54 is the effective detuning . in eq .
( [ nonlinlang3 - 1 ] ) , we can also take @xmath55 , since we will focus on the case @xmath56 which can be achieved by controlling the input power of the external microwave drive .
rewriting each operator as a @xmath57-number steady - state value plus an additional fluctuation operator , and neglecting the nonlinear terms ( since we have chosen @xmath58 ) , we obtain a set of linear quantum langevin equations ( see eq .
( [ nonlinlang42 ] ) ) and then solve for the spectrum of the position and momentum of the mr as in refs .
@xcite , see appendix a. using the fluctuation spectra of the mr as given in eqs .
( [ spectr - x],[spectr - p ] ) , we can define the final mean phonon number in the steady state @xcite as @xmath59 where the variances of position and momentum are @xmath60 this allows us to define the effective temperature @xmath61 as @xmath62 in the next section , we will consider the cooling of the mr by discussing its final effective mean phonon number ( or equivalently , its effective temperature ) in detail .
the final effective mean phonon number of the mr can be calculated directly by evaluating the integral in eq .
( [ integral ] ) numerically and using eq .
( [ n_eff ] ) .
alternatively , instead of being evaluated directly , eq .
( [ integral ] ) can also be evaluated analytically using the approximation scheme described in the following .
the effective mechanical damping rate @xmath63 can be significantly increased , @xmath64 , when @xmath65 is very large , see eq .
( [ gamma_ca ] ) .
let us consider the most interesting regime when the significantly increased effective mechanical damping rate is less than the mechanical frequency : @xmath66 , ( that is , the effective quality factor @xmath67 ) , and also less than the decay rate of tlr : @xmath68 @xcite . in this regime ,
the effective frequency is unchanged @xmath69 @xcite according to eq .
( [ omega_eff ] ) , and the effective susceptibility is peaked around the points @xmath70 .
then one can get an approximate expression for the variance @xmath71 where the effective thermal noise spectrum @xmath72 and the induced noise spectrum @xmath73 are the symmetrized parts of @xmath74 and @xmath75 , respectively : @xmath76 similarly , one can obtain @xmath77 in units of @xmath78 as a function of effective detuning @xmath79 .
the dashed lines are obtained by numerically evaluating the integral in eq .
( [ integral ] ) , the solid lines by using the approximate expressions eqs .
( [ approximate - x],[approximate - p ] ) . here , @xmath80 , @xmath81 , @xmath82 , @xmath83 , @xmath84 , and @xmath85 ( upper lines ) or @xmath86 ( lower lines).,width=7 ] figure [ fig2 ] shows the variance of position @xmath87 as a function of the effective detuning @xmath79 .
the dashed lines correspond to a numerical evaluation of the integral in eq .
( [ integral ] ) .
the solid lines describe the approximate results obtained through eq .
( [ approximate - x ] ) which can be seen to agree perfectly with the exact numerical evaluation .
we checked that this is also the case for the variance of the momentum @xmath88 . in eq .
( [ approximate - x ] ) , the induced noise spectrum @xmath89 increases ( heats ) the motion of the mr . on the other hand , when the effective damping rate is enhanced : @xmath90 , the mechanical motion will reduce , that means cooling . mathematically , the cooling effect would dominate the heating effect when the effective damping rate is sufficiently increased .
actually , this is right when the significantly increased effective damping rate is positive for positive detuning .
however , it is not the case when @xmath91 is negative and @xmath92 ( for negative detuning @xmath93 ) .
that is because the stability conditions , derived using ref .
@xcite , are satisfied only for positive detuning @xcite . in fact , a negative effective
damping means the amplitude motion of the mr will be amplified which will lead to an instability @xcite . in what follows
, we will focus on the case of positive detuning @xmath94 . according to eqs .
( [ n_eff],[approximate - x],[approximate - p ] ) , one has @xmath95 where @xmath96 is the initial mean thermal excitation phonon number of the mr ; @xmath97 is the induced mean phonon number due to the capacitive coupling between the mr and the tlr . vs. effective detuning @xmath79 and decay rate @xmath25 of the tlr ( @xmath98 , for the other parameters see fig .
[ fig2]).,width=264 ] as discussed above , the significant reduced value of @xmath99 in eq .
( [ nb - eff ] ) can only be obtained when the additional damping rate @xmath100 ( or effective damping rate @xmath101 ) is positive and much larger than the original one ( but still less than the decay rate of tlr and less than the frequency of the mr as discussed before ) : @xmath102 this can be satisfied by enhancing the value of @xmath103 , that is , by controlling the capacitive coupling strength @xmath104 and increasing the external microwave drive power @xmath26 to make @xmath105 large ( equivalently , @xmath106 will be large ) . in practice ,
the capacitive coupling strength @xmath104 would be limited by the realistic system , and the external microwave drive strength @xmath107 would also be limited according to the validity of the rotating wave approach as we mentioned before .
here we put the length of the mr @xmath108 as large as @xmath109 @xmath19 m and the distance between the mr and tlr @xmath16 as small as @xmath0 @xmath19 m ( see fig .
[ fig1 ] ) in order to get a large capacitance @xmath17 which will lead to large @xmath104 , and fix @xmath110 for all the numerical calculations .
then for a significantly - increased effective damping rate , the final mean phonon number reduces to @xmath111 in order to get the ground state cooling , that is , @xmath112 , both @xmath113 and @xmath114 should be much less than @xmath0 .
especially , if the @xmath100 is significantly increased enough to make @xmath115 then @xmath99 approaches the limit @xmath113 : @xmath116 now we discuss the possible minimal value of @xmath113 by discussing all kinds of parameters , e.g. , @xmath25 , @xmath79 , and @xmath117 . from eq .
( [ nca ] ) , it is obvious that the optimal value of @xmath25 satisfies the high - quality cavity limit @xmath118 and the optimal detuning satisfies @xmath119 .
then the corresponding induced mean phonon number @xmath113 is @xmath120 the optimal @xmath117 needs a sufficiently low initial temperature of the bath which is limited in practice to the experimental dilute refrigerator temperatures . for the superconducting tlr scheme ,
its microwave frequency is of the order of @xmath121 hz . for the initial temperature @xmath122
k , @xmath123 and @xmath124 , the ground - state cooling of the mr is not possible .
therefore , initial temperature less than @xmath125 mk are required to achieve ground - state cooling .
our result on the limiting value in eq .
( [ general limit ] ) is consistent with that in other optical schemes except the limit of initial temperatures . in the optical cavity case
@xmath126 ( even at room temperature ) and therefore @xmath127 , the optimal value of @xmath113 becomes @xmath128 which is just the case of resolved sideband cooling as discussed in the optomechanical cooling schemes @xcite .
we would like to point out that these references also mention another cooling limit : the doppler cooling limit , which is realized in our system when @xmath127 , @xmath129 , and @xmath130 in eq .
( [ nca ] ) : @xmath131 on the other hand , if @xmath132 , the induced mean phonon number @xmath113 in eq .
( [ general limit ] ) becomes @xmath133 . in the classical limit when the initial temperature is so high that @xmath134 , one has @xmath135 which is also given in ref .
the doppler cooling limit in eq .
( [ dopper - limit ] ) and the classical cooling limit in eq .
( [ classical limit ] ) preclude ground state cooling .
we will focus on the resolved sideband cooling in this paper .
the final mean phonon number @xmath99 is plotted as a function of the effective detuning @xmath79 and the decay rate @xmath25 of the tlr in fig . [ fig3 ] .
it is clear that one can obtain a significant suppression of the mechanical motion of the mr in the positive detuning range @xmath136 .
the optimal cooling is obtained for @xmath137 , which agrees with both the above analysis and that in other treatments of radiation - pressure cooling @xcite .
physically , as discussed in the back - action optomechanical cooling schemes in optical cavities ( refs .
@xcite ) , the external driving microwave is scattered by the tlr + mr " system mostly to the first stokes sideband ( @xmath138 ) and the first anti - stokes sideband ( @xmath139 ) .
the generation of an anti - stokes photon will cool the mr by taking away a phonon of the mr . on the contrary
, the generation of a stokes photon will heat the mr by creating a phonon . when the effective detuning @xmath94 , the microwave field of the tlr ( with the frequency @xmath140 @xmath141 ) interacts with the first anti - stokes sideband ( @xmath142 ) more than it interacts with the first stokes sideband ( @xmath143 ) , and
cooling will occur .
this is the physical reason why the positive effective detuning ( @xmath94 ) will lead to cooling . in the high - quality cavity limit @xmath144 , the anti - stokes ( stokes ) sideband
is resolved , and the corresponding cooling ( heating ) process is prominent .
especially , for the optimal effective detuning @xmath145 , the frequency of the tlr is resonant with that of the anti - stokes sideband , which will apparently lead to optimal cooling .
this physical discussion is consistent with the calculation presented above .
figure [ fig3 ] suggests there is a finite optimal value of @xmath25 for a fixed effective detuning .
that is because one should have both a small value of @xmath113 and a large effective damping rate @xmath146 in order to get strong cooling
: @xmath25 should not be too large , since @xmath113 depends somewhat on the value of @xmath147 ; @xmath25 should not be too small , since @xmath148 @xmath149 when @xmath150 @xcite . in the cooling process ,
the thermal energy of the mr is mainly first transferred to the tlr , and then leaks out of the tlr through the bath coupled to the tlr .
when the decay rate of the tlr is too small : @xmath151 , the energy leakage out of the tlr is too weak , and one could not obtain a strong cooling . )
( dashed line ) or using the approximate expression in eqs .
( [ approximate - x],[approximate - p ] ) ( solid line ) as a function of @xmath25 at the optimal effective detuning @xmath152 .
( b ) logarithm of the ratio of the corresponding effective damping rate @xmath153 to @xmath25 as a function of @xmath25 . here , @xmath154 .
for the other parameters see fig .
[ fig2].,width=226 ] we would like to emphasize that the results shown in fig .
[ fig3 ] are based on the approximate expressions eqs .
( [ approximate - x],[approximate - p ] ) , where the condition of the so - called weak - coupling limit @xcite has been assumed , that is , the effective damping rate of the mr should be always less than the decay rate of the cavity and less than the frequency of the mr @xmath155 .
normally the weak - coupling is satisfied but not in some special cases . in fig .
[ new - fig4](b ) , the weak - coupling condition is violated when @xmath156 at the optimal effective detuning @xmath152 . beyond the weak coupling limit , fig .
[ new - fig4](a ) shows that the approximate treatment through eqs .
( [ approximate - x],[approximate - p ] ) ceases to be valid .
then one should discuss the cooling , e.g. , effective mean phonon number in eq .
( [ n_eff ] ) , by using the numerical evaluation of the integral in eq .
( [ integral ] ) . but going beyond the weak - coupling limit , the contribution from the position variance is not equivalent to that from the momentum variance any more @xcite .
in other words , the energy equipartition is not satisfied .
that means it is hard to define an effective temperature since it is not in a strict thermal state .
according to the above analysis , both the high - quality cavity and weak - coupling limit should be satisfied , so the optimal decay rate of the tlr is better taken to be @xmath157 for the typical parameters in fig .
[ new - fig4 ] .
the weak - coupling condition depends only weakly on the initial temperature @xmath50 and the original damping rate of the mr @xmath51 in the cooling process . in what follows
, we will consider the optimal decay rate at @xmath157 for different parameters @xmath50 and @xmath51 , for which the weak coupling limit is always satisfied . in fig .
[ new - fig5 ] , the ratio of final effective temperature @xmath61 to bath temperature @xmath50 is plotted as a function of the effective detuning @xmath79 for the optimal @xmath157 . apparently , here the weak coupling limit is satisfied ( according to the above analysis in fig .
[ new - fig4 ] ) . for initial temperatures @xmath158 , @xmath159 , and @xmath1 mk , the corresponding initial mean
phonon numbers are @xmath160 $ ] @xmath161 @xmath162 , @xmath163 , and @xmath164 , with the final mean phonon number @xmath165 , @xmath166 , and @xmath167 , respectively .
it is obvious that a significant cooling of the mr is obtained and lower initial temperatures will generally lead to better cooling . for an initial temperature @xmath168 mk , which can be realized experimentally by using a dilution refrigerator , the mr ( with the frequency @xmath169 mhz ) can be cooled close to the ground state since the final mean phonon number @xmath170 . for three initial temperatures : @xmath168 mk ( dotted lines ) , @xmath171 mk ( solid lines ) , @xmath158 mk ( dot - dashed lines )
here , @xmath172 , @xmath173 kg , @xmath174 mhz , @xmath175 ( equivalently , @xmath176 ) . for the other parameters see fig .
[ fig2 ] .,width=264 ] for an initial temperature of @xmath168 mk as in fig .
[ new - fig5 ] , the final mean phonon number would in principle be @xmath177 ( in eq .
( [ nca ] ) ) , which is much less than that obtained in fig .
[ new - fig5 ] : @xmath178 @xmath179 @xmath180 .
this is because @xmath99 @xmath181 only when the condition in eq .
( [ limit - condition ] ) is satisfied .
unfortunately , it is not the case for the parameters in fig .
[ new - fig5 ] . a possible way to approach this condition is to increase the quality factor of the mr . in fig .
[ new - fig6 ] , the final effective mean phonon number is plotted as a function of the effective detuning @xmath79 for different quality factors of mr @xmath182 ( @xmath183 ) : @xmath184 ( typically , see ref .
@xcite ) ; @xmath185 , @xmath186 , @xmath32 ( expected in the near future ) .
the corresponding minimal @xmath187 , @xmath188 , @xmath189 , @xmath190 .
one can find the cooling is better for a higher quality factor of the mr .
is plotted as a function of @xmath79 for different quality factors of the mr : @xmath191 , @xmath192 , @xmath193 , @xmath32 ( from up to down ) , the corresponding damping rate @xmath194 , @xmath195 , @xmath1 , @xmath196 hz . here , @xmath168 mk .
for the other parameters see fig .
[ new - fig5].,width=264 ] the cooling discussed above can be measured by a homodyne detection method like that given in the scheme of cavity - assisted radiation - pressure cooling of a mr @xcite .
the motion of the mr can be detected by monitoring the output microwave signal ( e.g. , the field phase quadrature ) of the tlr ( as seen in fig .
[ fig1 ] ) since the measurement of the output spectrum corresponds to a faithful measurement of the mr motion @xcite .
we have found that a mr with frequency @xmath197 hz can be cooled close to its ground state when it is coupled to a typical tlr ( @xmath198 hz ) .
actually , by considering the optimal parameters in this scheme , that is , assuming the high - quality cavity limit ( @xmath199 ) , a positive optimal effective detuning ( @xmath145 ) , a low initial temperature ( e.g. , @xmath168 mk in order that @xmath200 ) , a high quality factor of the mr ( @xmath201 ) , and both strong external input microwave drive power @xmath26 and strong capacitive coupling strength @xmath104 to get the significantly increased positive effective damping rate ( @xmath202 ) , we find that resolved sideband cooling of the mr occurs .
the possible minimal value of the final effective phonon number could approach the induced mean phonon number : @xmath203 @xmath204 .
moreover , one should also consider the condition of weak - coupling limit , that is , the significantly increased effective damping rate @xmath205 should be less than both @xmath2 and @xmath25 ( in the high - quality cavity limit , one only needs @xmath206 ) .
this condition requires that @xmath25 must not be too small though lower @xmath25 will lead to lower @xmath113 . as shown in fig .
[ fig2 ] and its discussion , there will be an optimal range of @xmath25 .
for the typical parameters in this cooling scheme , we take @xmath207 ( though this is not the optimal result in general ) .
we find that a mr with @xmath174 mhz can be cooled close to its ground state with the final effective mean phonon number in the steady state : @xmath208 ( for a typical quality factor @xmath209 ) or @xmath210 ( for a high quality factor @xmath211 ) by using the resolved sideband cooling scheme when it is coupled to a driven tlr ( with the frequency @xmath212 hz ) .
we would like to stress the condition in eq .
( [ limit - condition ] ) , which can lead to @xmath203 . as pointed out in the discussion of figs .
( [ new - fig5 ] , [ new - fig6 ] ) , this is not always satisfied . a possible way to approach
this condition is to increase the quality factor of the mr .
for example , if the quality factor of the mr is high enough ( e.g. , @xmath213 ) , one would have the optimal @xmath214 , for which the mr is cooled much closer to the ground state .
the back - action self - cooling scheme presented here is similar to the optical - cavity - assisted cooling scheme @xcite . in both cases ,
the mr can be cooled close to its ground state using resolved sideband cooling which is possible in the limit of a high - quality cavity .
but it seems that this limit ( e.g. , @xmath172 , for @xmath215 hz ) is easier to reach in the microwave tlr than that in the optical cavity . in the case of a tlr ( typically @xmath212 hz ) ,
quality factors of @xmath216 have been seen in experiments @xcite .
however , in the case of an optical cavity ( @xmath217 hz ) , the corresponding quality factor should be @xmath218 , which is hard to achieve since the typical cavity quality factor is @xmath219 @xcite . to conclude , we have studied the self - cooling of a mechanical resonator that is capacitively coupled to a transmission - line resonator .
the discussion was based on a linearized quantum langevin equation .
the cooling method presented here is similar to the self - cooling of a mr coupled to an optical cavity by radiation pressure . by using the optimal parameters discussed above
, the mr can be cooled close to its ground state in the high - quality cavity and weak - coupling limit .
we would like thank c.b .
doiron and i. wilson - rae for helpful discussions .
this work was supported by the swiss nsf , the nccr nanoscience , the ec ist - fet project eurosquip , and partially supported by the nsfc through grant no .
y.d.w . also acknowledges support by the jsps kakenhi no .
18201018 and mext - kakenhi no .
f.x . was supported in part at the technion by an aly kaufman fellowship .
the linearized quantum langevin equations read to solve these equations , we define the fourier transform for an operator @xmath221 ( @xmath222 , @xmath223 , @xmath224 , @xmath44 , @xmath225 ) @xmath226 and for its hermitian conjugate @xmath227 ( if any ) @xmath228 which lead to @xmath229 according to eqs .
( [ nonlinlang51 ] ) , the spectra of the mr can be written as @xmath237 , \label{spectr - x}\\ s_{p}(\omega ) & = \left ( \omega m\right ) ^{2}s_{x}(\omega ) , \label{spectr - p}\end{aligned}\ ] ] where @xmath238\ ] ] is the thermal noise spectrum due to the brownian motion of the mr ; and @xmath239 is the induced noise spectrum due to the capacitive coupling to the driven tlr the effective susceptibility is defined as @xmath240 /b(\omega)$ ] and can be simplified to @xmath241 } , \ ] ] where the effective frequency of the mr is according to the definition of the additional damping rate in eq .
( [ gamma_ca ] ) , the induced noise spectrum @xmath75 in eq .
( [ s - ca ] ) can also expressed as @xmath245 \gamma_{ca}(\omega ) .
\label{s - ca2}\ ] ] c. m. caves , k. s. thorne , r. w. p. drever , v. d. sandberg , and m. zimmermann , rev .
mod . phys . * 52 * , 341 ( 1980 ) ; m. f. bocko and r. onofrio , rev .
* 68 * , 755 ( 1996 ) ; m. d. lahaye , o. buu , b. camarota , and k. c. schwab , science * 304 * , 74 ( 2004 ) .
f. xue , l. zhong , y. li , and c. p. sun , phys .
rev.b * 75 * , 033407 ( 2007 ) ; f. xue , y .- x .
liu , c.p .
sun , and f. nori , phys .
b * 76 * , 064305 ( 2007 ) ; f. xue , y. d. wang , c. p. sun , h. okamoto , h. yamaguchi , and k. semba , new j. of phys . * 9 * , 35 ( 2007 ) . here
we assumed the mechanical resonator to be fabricated at the center of the tlr . as a result ,
the qubit is coupled to the second mode of the tlr , which as an antinode of the voltage in its center as given in ref .
@xcite .
m. pinard , a. dantan , d. vitali , o. arcizet , t. briant and a. heidmann , europhys . lett . * 72 * , 747 ( 2005 ) . strictly speaking ,
@xmath246 is not always satisfied for arbitrary parameters , but this is the case for the parameters considered in the following discussion .
it is right that @xmath247 @xmath149 when @xmath150 except the scope nearly - close to @xmath248 where @xmath247 @xmath249 when @xmath150 .
but we will show in fig .
[ new - fig4 ] that some correction should be done since it may be beyond the weak - coupling limit . | we propose a quantum description of the cooling of a micromechanical flexural oscillator by a one - dimensional transmission line resonator via a force that resembles cavity radiation pressure .
the mechanical oscillator is capacitively coupled to the central conductor of the transmission line resonator . at the optimal point
, the micromechanical oscillator can be cooled close to the ground state , and the cooling can be measured by homodyne detection of the output microwave signal . |
in order to determine the neutrino oscillation parameters there is a lot of future experimental projects which would have ability of precision measurement .
it means that we may have sensitivity to not only the standard oscillation parameters but also some new physics effect on neutrino sector . in this talk
, we discuss about probing the non - standard neutrino physics with future long baseline experiment based on the works@xcite@xcite . due to limitations of space
, we would like to concentrate mainly on the work of non - standard neutrino interaction with neutrino factory .
since early time , it is suggested the possible existence of non - standard interaction(nsi ) of neutrinos with matter@xcite@xcite@xcite . in this section
we discuss about probing such a nsi with future long baseline experiments . introducing the lagrangian for nsi : @xmath1 where @xmath2 is the fermi constant , @xmath3 for the species of fermions in the earth e , u , d , and @xmath4 . in this notation , @xmath5 is the effective coupling normalized by week interaction .
concentrating on the effect of nsi in neutrino propagation , the evolution equation is written as , @xmath6 ~ \left ( \begin{array}{c } \nu_e \\ \nu_\mu \\
\nu_\tau \end{array } \right ) \label{nsi - evolution}\end{aligned}\ ] ] where u is the maki - nakagawa - sakata flavor mixing matrix@xcite , @xmath7 , @xmath8 is the electron number density in the earth , and @xmath9 describes the total effect of nsi coming from charged fermion e , u , d. though off - diagonal @xmath10 parameters may have complex phase , we assume in this talk that all @xmath9 s are real . in order to improve the sensitivity to nsi , we examine the power of golden channel at neutrino factory@xcite .
neutrino factory has a great sensitivity to @xmath11 . considering that @xmath9 produces non - standard flavor changing effect
, it is natural to expect that we have a high sensitivity also to nsi parameters .
in fact , the contribution to oscillation probability @xmath12 of @xmath13 and @xmath14 are same order to the one of @xmath15 .
though this feature helps us to probe the nsi , but at the same time possible existence of nsi may destroy the precision measurement of @xmath0 and cp - phase @xmath16 , the so called confusion problem @xcite@xcite . see the upper panels of figure [ fig_th13del_const ] .
this is the allowed region of standard oscillation parameters @xmath11 and @xmath16 on neutrino factory with a detector at 3000 km obtained by marginalizing certain combinations of two @xmath9 s .
as noted above , the sensitivity to the standard oscillation parameters become significantly worse . in particular
we can not determine the nonzero value of @xmath0 even at the input value of @xmath17 for the case with @xmath18 . and cp - phase @xmath16 with use of the @xmath19 and @xmath20 channel for various combination of nsi elements @xmath9 .
the upper(middle ) panels are for a detector at 3000(7000)km , and the bottom panels for the combined one .
the input values of @xmath9 are zero , @xmath21 , and @xmath22 .
the figure complements those in @xcite with @xmath23 and @xmath24 . ] in order to solve this confusion problem , we consider the 2nd detector at 7000 km .
this is nearby so called magic baseline@xcite@xcite , where @xmath25 km for constant earth matter density @xmath26g / cc .
the magic baseline is the special distance because @xmath12 is independent of the solar oscillation parameters and @xmath16 . at the magic baseline ,
oscillation probability with @xmath14 is written as@xcite : @xmath27 where @xmath28 , @xmath29 , and @xmath30 . with the information of complex phase @xmath16 at 3000 km , we can determine a certain combination of @xmath0 and @xmath14 at the magic baseline .
this advantage helps us solve the confusion problem .
bottom panels in figure [ fig_th13del_const ] are the allowed region of @xmath31 and @xmath16 with the combined information of two detectors .
the sensitivity becomes dramatically higher .
the synergy effect is also powerful for probing the nsi .
figure [ fig_eps_const ] presented the allowed region of the nsi parameters .
though it is very hard to determine the @xmath32 with a individual detector , we can determine the nsi parameters very accurately if we combine the two . .
] in this work@xcite , the sensitivity reach for @xmath9 we have obtained is : @xmath33 in more recent work with information of @xmath34 disappearance channel@xcite , the sensitivity reach to nsi parameters in @xmath35 sector is reported as @xmath36 they confirmed that the two detector setting at @xmath37 km and @xmath38 km , which is similar to ours , is close to optimal even if nsi in matter are considerd .
notice that it is nothing but the standard ids - nf setup@xcite@xcite .
finally we would like to point out that even though the true value of @xmath0 vanishes or extremely small , in which case we can not measure the cp - phase @xmath16 neither , we would have a reasonable sensitivity to non - standard neutrino interaction .
it is because the contribution of @xmath13 and @xmath14 to @xmath12 remains as @xmath39 and @xmath40 terms in the perturbative formula .
in the relatively near future experiment , we have the sensitivity to nsi on the @xmath41 sector @xmath42 and @xmath43 .
we briefly mention hare the potential of t2kk@xcite for searching the nsi using @xmath34 disappearance channel@xcite . for small @xmath9 ,
@xmath34 disappearance channel can be written by two flavor approximation . in such system , the effects of nsi to oscillation probability are large with @xmath44 , @xmath45 and , @xmath46 .
so concentrating these @xmath10 s , the sensitivity at 2 @xmath47 cl .
may be reached by t2kk setup : @xmath48 where @xmath49 and @xmath45 is assumed zero because only the difference can be measured .
we discussed the power of the two detector setting at different baseline distance for probing non - standard neutrino physics .
neutrino factory , especially with two detectors at 3000 and 7000 km , is a powerful setting for solving the confusion problem that make the precision measurement of standard oscillation parameters untenable if we consider the possible existence of nsi .
it is also useful to search the nsi even for extremely small value of nsi parameters .
i am very grateful to my collaborators , hisakazu minakata , hiroshi nunokawa , renata zukanovich funchal , and nei cipriano ribeiro for valuable collaboration .
osamu yasuda gave me a fruitful comment .
this work was supported in part by grant - in - aid for jsps fellows no .
209677 .
n. cipriano ribeiro , h. minakata , h. nunokawa , s. uchinami and r. zukanovich - funchal , jhep * 0712 * , 002 ( 2007 ) [ arxiv:0709.1980 [ hep - ph ] ] . n. cipriano ribeiro , t. kajita , p. ko , h. minakata , s. nakayama , and h. nunokawa , phys . rev .
d * 77 * , 073007 ( 2008 ) [ arxiv:0712.4314 [ hep - ph ] ] .
l. wolfenstein , phys . rev .
d * 17 * , 2369 ( 1978 ) .
j. w. f. valle , phys .
b * 199 * ( 1987 ) 432 .
m. m. guzzo , a. masiero and s. t. petcov , phys .
b * 260 * , 154 ( 1991 ) .
z. maki , m. nakagawa and s. sakata , prog .
* 28 * , 870 ( 1962 ) .
s. geer , phys .
d * 57 * , 6989 ( 1998 ) [ erratum - ibid .
d * 59 * , 039903 ( 1999 ) ] [ arxiv : hep - ph/9712290 ] .
p. huber , t. schwetz and j. w. f. valle , phys .
lett . * 88 * , 101804 ( 2002 ) [ arxiv : hep - ph/0111224 ] .
p. huber , t. schwetz and j. w. f. valle , phys .
d * 66 * , 013006 ( 2002 ) [ arxiv : hep - ph/0202048 ] .
p. huber and w. winter , phys .
d * 68 * , 037301 ( 2003 ) [ arxiv : hep - ph/0301257 ] .
a. y. smirnov , arxiv : hep - ph/0610198 .
j. kopp , t. ota and w. winter , arxiv:0804.2261 [ hep - ph ] .
a. bandyopadhyay _ et al .
_ [ iss physics working group ] , arxiv:0710.4947 [ hep - ph ] . | we discuss the possibility of constraining or discovering the non - standard neutrino physics beyond the standard model with future long baseline neutrino oscillation experiments . among very many possible
experimental set up we discuss neutrino factory and t2kk , in which two detector setting may be useful to achieve the goal . in particular , neutrino factory with two detectors at baselines of 3000 km and 7000 km have a great sensitivity to non - standard interaction ( nsi ) and solve the @xmath0-nsi confusion problem . |
information is just noise if we do not know it exists , or how to use it .
the smithsonian / nasa astrophysics data system , in collaboration with astronomy s data centers and publishers , has been working for somewhat more than a decade towards making the information of astronomy increasingly discoverable , available , and useful .
the test of an organization which provides its users with information is whether it has provided the information which the users need , and at what cost .
a cost benefit analysis of the information system in astronomy , consisting of the ads , the cds , and the journals ( the data archives were not included in the analysis ) shows a benefit of about 7% of the total cost of the entire discipline can be ascribed to the joint effors of these organizations in creating the current on - line system , compared with the paper system of a decade ago ( kurtz , et al , 2005 ) .
this ten times the total cost of running these organizations , and about a factor of 100 times the differential cost of making the services on - line .
astronomy is currently creating another layer of on - line information services , beginning with the archives , as opposed to the journals .
this effort , known as the virtual observatory , has likely much in common with the efforts which preceeded it , and which now collaborate with it .
the greatest lesson to be learned from the recent past is that massive success is possible ; however it is not inevitable .
concentrated , complex organized information is the basis of intellegence .
before an information system can provide sophisticated tools to facilitate the search and retrieval of complex information , it must actually have sufficient information about the information ( meta - data ) to allow the information to be organized with a degree of sophistication equal to the capabilities of the search
. from its inception the ads has put a substantial fraction of its total resources toward the task of collecting , creating , and organizing meta - data .
grant , et al ( 2000 ) list hundreds of sources of these data ; this is but a partial list . very often a single source can require several years of patient negotiation with different entities ( e.g. the copyright holder and the publisher ) .
collecting data from existing data providers consumes substantial effort .
there is no widely used standard for meta - data transmission ; the oai - pmh system ( www.openarchives.org ) , while promising in concept , actually is only used by a fraction ( @xmath0 ) of our sources , and even with these each individual provider has unique extensions .
this means that the responsibility for parsing the meta - data and putting it into a standard format rests with the ads staff .
this is different than for some other information services , where the data sources are required to adhere to a standard format .
examples are citeseer ( citeseer.ist.psu.edu ) and citebase ( www.citebase.org ) .
the difference resulting from the two approches is large .
ads has been a mature research tool since its inception ; researchers have been able to count on it to be up to date and complete for the past decade .
the services which require provider compliance are still quite incomplete ; they function primarily as research tools for computer scientists interested in text retrieval.a754-1.ps this is also different than for google scholar ( scholar.google.com ) which , like ads , negotiates access to meta - data from providers , but unlike ads is not able to devote the resources to understanding , and individually parsing , the peculiarities of each individual source . for the areas ads covers its abilities far exceed those of google scholar .
there is a lesson here for other information services : just do it ! waiting for all the information to be included in the service to be available in just the right form delays the full implementation of the service by a substantial factor .
if a service is valuable enough to offer , it is valuable enough for the service provider to make work .
this is not a totally overwhelming task ; the three major data services in astronomy today , ads , cds , and ned all function with relativly small staffs ( the ads staff size is six ) , and all are able to put substantial effort into data ingest .
information retrieval is successful if you find the information you need , not , as mick jagger might say , the information you want .
it is also important that this be able to be done quickly , as the most valuable commodity in any research endevour is the time of the best researchers .
there are many different techniques for achieving this , some as simple as making the send button on the web site a bright color . the ads has been developing and implementing these techniques , often in advance of any other group , throughout its entire history , begining with our use of combination of evidence ( belkin , et al 1995 ) ; partial match , natural language techniques , astronomy specific ontologies , and hyperlinks from the outset ( kurtz , et al 1993 ) . the connection with cds / simbad ( grant , et al 1993 , wenger , et al 2000 ) was another early achievement . linking directly to data
has been a feature of the ads since its inception , with links to the _ a&a _ data tables ( ochsenbein & lequeux 1995 ) and to the archival data - sets in the adil ( plante , et al 1996 ) being early examples .
much of the talk at the conference described the ir advances made by ads , the interested reader is referred to kurtz , et al ( 2000 ) and kurtz , et al ( 2005 ) for historical summaries .
second order database operators ( kurtz , 1992 ) are database functions which , instead of returning the data corresponding to a query , performs some function on data associated with those data .
for example a query might return a list of papers concerning redshift surveys ; these papers have lists of papers they reference .
a @xmath1 order operator could take these lists , collate and sort them , then return a list of those papers which are most cited by papers concerning redshift suveys .
kurtz , et al ( 2002 , 2005 ) discuss how using these operators within ads one can obtain the most recent , most popular , most useful , and most educational papers on a complex topic of interest . here
we demonstrate how to use ads , and @xmath1 order queries , to discovery the most useful on - line data sets for a particular topic .
we begin , in figure 1 , by asking for all articles with linked data - sets ( this requirement is not shown , it is futher down the form ) on the x - ray properties of the nearby cluster of galaxies abell 754 .
the result of this simple ads / simbad query is shown in figure 2 , there are 79 articles which meet this description . b now we are ready to use a @xmath1a754 - 1.ps order operator .
we assert that , on average , people who have read the articles in the list in figure 2 are people interested in data about the x - ray properties of abell 754 .
the articles these people read the most , will , on average , be important to us , as we are interested in data concerning the x - ray properties of abell 754 .
next we go to the bottom of that list , shown in figure 3 , we select all 79 articles from the list , and request the also - read articles for that group . the result of that query is shown in figure 4 .
the top five articles are compendium articles on the x - ray properties of nearby clusters from the chandra , rosat , xmm , asca , and einstein satellites , including abell 754 .
this confirms our assertion above .
if we follow the d
link for the top article , we go to the associated page in the chandra archive , shown in figure 5 . from here one can download a substantial fraction of the chandra data on nearby x - ray clusters , and one can also access the software tools necessary to reduce and analyze it .
the basic data used by this service is currently being created by the archives , establishing links between data in the archives and published papers .
recently the astrophysical journal has established a method of doing this directly by authors .
since late 2003 ( kurtz , et al 2003 ) the ads has provided a notification service , myads , which uses the @xmath1 order operators to give our users a sophisticated and powerful means of keeping abreast of the current literature in their sub fields of astronomy and physics .
beginning in the spring of 2005 we have extended this service with the collaboration of the arxiv ( www.arxiv.org ) to include the arxiv e - print service .
we now have three different , complimentary services : myads - astronomy , myads - physics , and myads - arxiv .
myads - arxiv is facilitated by a tight collaboration between ads and arxiv , this includes sharing of readership information , parsing the reference lists , and merging the preprints with the journal articles for the same article .
once an article appears in the final published source ads adopts that version as definitive .
myads - arxiv is a fully customized ( to each individual user ) , open access virtual journal covering the most important papers of the past week in physics and astronomy . between a fifth and a quarter of
all working astronomers already subscribe to myads .
figure 6 shows the custom version belonging to the first author , for the week preceeding the conference .
it is also possible to make a myads page which is not associated with an individual ; for example the link http://adsabs.harvard.edu/myads/cache/[email protected] points to a myads - arxiv page covering virtual observatory matters .
this can be viewed as an automatically generated newsletter . )
szalay ( 2005 ) has presented the challenges and goals of a mode of astronomy based on very large , homogeneous data - sets , such as the sdss , or in the future the lsst .
the scientific case for these developments is exciting and compelling .
it seems likely , however , that for at least the next twenty years most of the money and effort of astronomy will be devoted to pointed observations with very large and expensive instruments .
currently hst , xmm , keck , vlt , vla and many other large pointeed instruments dominate top level astronomy . in the future ,
in addition to the mega survey instruments , owl , gmt , celt , con - x , jwst , alma and other very costly pointed instruments will play a major role . with the cost of dark time on a $ 1,000,000,000 telescope running around $ 50,000 per hour the re - use of data in the telescope archives would seem a highly cost effective means of doing research , as is now the case with many current archives . in order for astronomers to be able to find relevant data - sets from pointed observations the ads will index the abstracts of observing proposals , and will point to the page in the archive devoted to that project .
we have begun by indexing the proposals from several leading archives . | since it was first announced at adass 2 the smithsonian / nasa astrophysics system abstract service ( ads ) has played a central role in the information seeking behavior of astronomers .
central to the ability of the ads to act as a search and discovery tool is its role as metadata agregator . over the past 13 years the ads
has introduced many new techniques to facilitate information retrieval , broadly defined .
we discuss some of these developments ; with particular attention to how the ads might interact with the virtual observatory , and to the new myads - arxiv customized open access virtual journal .
the ads is at http://ads.harvard.edu |
low - dimensional quantum - magnetism has received considerable interest recently due to the discovery of numerous novel materials with spin-@xmath0 moments arranged in chain , ladder , and depleted planar structures .
many of these materials exhibit unconventional magnetic phases due to dimerization and frustration of antiferromagnetic exchange .
particular effort@xcite has been devoted to systems like srcu@xmath1(bo@xmath2)@xmath1@xcite , which display a complete frustration of the magnetic exchange as in the two - dimensional shastry - sutherland model@xcite . in one dimension complete frustration
can occur in two - leg spin - ladders if an additional cross - wise exchange is included as depicted in fig .
[ fig3 ] which resembles a chain of edge sharing tetrahedra . for @xmath3
such tetrahedral ladders have been investigated in the past@xcite .
very recently , tellurates of type cu@xmath1te@xmath1o@xmath4x@xmath1 with x = cl , br have been identified as a new class of spin-1/2 tetrahedral - cluster compounds@xcite .
bulk thermodynamic data have been analyzed in the limit of isolated tetrahedra@xcite .
raman spectroscopy , however indicates a substantial inter - tetrahedral coupling along the c - axis direction of cu@xmath1te@xmath1o@xmath4x@xmath1@xcite . in this direction
the exchange topology is likely to be analogous to that of fig .
[ fig3 ] with @xmath5 . from a materials perspective
it is an open question if the magnetism of the tellurates can be understood in terms of a dimerized tetrahedral spin - ladder . from a theoretical point of view
, however , the magnetic properties of this model are an interesting issue which forms the motivation for this work .
the paper is organized as follows . in the remainder of this section
we discuss the basic properties of the tetrahedral chain hamiltonian . in section two
the quantum phase diagram is analyzed . in section three a bond - operator method
is applied to the tetrahedral chain and in section four the magnetic raman cross - section is evaluated .
the hamiltonian of the tetrahedral chain can be written in terms of the total edge - spin operators @xmath6 and the dimensionless couplings @xmath7 and @xmath8 @xmath9 \label{bb1}\end{aligned}\ ] ] this model displays infinitely many local conservation laws : @xmath10=0;\,\forall\,l , i=1,2 $ ] .
therefore , the hilbert space decomposes into sectors of fixed distributions of edge - spin eigenvalues @xmath11 or @xmath12 , each corresponding to a sequences of spin-1 chain - segments intermitted by chain - segments of _ localized _ singlets . if @xmath5 the spin-1 chain - segments are dimerized . in the infinite length ,
dimerized @xmath13-chain sector , i.e. for @xmath11 @xmath14 , the model simplifies to @xmath15 + \frac{a}{2}d \label{w3}\end{aligned}\ ] ] where @xmath16 refer to spin-1 operators and @xmath17 is the number of tetrahedra ( dimers ) .
we have checked @xmath20 to increase monotonously with @xmath21 by evaluating the ground - state energy per site of a dimerized @xmath13 chain with _
open _ boundary conditions as a function of system size . in principle
two - particle irreducible vertices can be constructed to second order in @xmath23 from @xmath24 . in the singlet channel , however , because of the symmetries of the spin indices in ( [ wr3 ] ) w.r.t .
those of the levi - civita symbol in @xmath24 these vanish identically . using a different technique
an expression very similar to the denominator of ( [ wr11 ] ) has been obtained in a study of phonon - assisted optical two - triplet - absorption on _ spin-@xmath25 two - leg ladders_@xcite .
in contrast to the dimerized @xmath13 chain case , however , @xmath26 has been found for the ladder in that case . in that work
it was predicted , that the singlet bound - state of the ladder should be observable in optical absorbtion .
recent work has confirmed this prediction@xcite . | we discuss the magnetic properties of a dimerized and completely frustrated tetrahedral spin-1/2 chain . using a combination of exact diagonalization and bond - operator theory the quantum phase diagram
is shown to incorporate a singlet - product , a dimer , and a haldane phase .
in addition we consider one- , and two - triplet excitations in the dimer phase and evaluate the magnetic raman cross section which is found to be strongly renormalized by the presence of a two - triplet bound state .
the link to a novel tellurate materials is clarified .
-2.8 cm psfig.sty |
network coding allows senders or intermediate nodes of a network to mix different data packets / flows and can enhance the throughput of many network setups @xcite . but
this is often at the price of large decoding delay , because mixed data needs to be network decoded before delivery to higher layers @xcite . understanding
the tradeoff between throughput and decoding delay in network coded systems has been the subject of research in recent years @xcite , where a packet - level network coding model is particularly suitable for such studies . in this paper , we are primarily concerned with the achievable throughput - delay tradeoff of packet - level network coding in wireless broadcast scenarios , where a sender wishes to broadcast a block of data packets to some receivers through wireless channels with packet erasures .
we first review two classic packet - level network coding techniques under this scenario : random linear network coding ( rlnc ) @xcite and instantly decodable network coding ( idnc ) @xcite .
the primary advantage of rlnc is its optimality in terms of block completion time @xcite , which is the time it takes to successfully broadcast all data packets in a block to all receivers and is a fundamental measure of throughput .
rlnc achieves its optimality by sending random linear combinations of all data packets as coded packets , with coefficients randomly chosen from a finite field @xmath0 where usually @xmath1 .
another advantage of rlnc is that it only requires one ack feedback from each receiver upon successful decoding of all data packets in a block .
however , rlnc can suffer from large decoding delays , since a receiver needs to collect enough linear combinations to perform block - wise decoding .
this also incurs heavy decoding computational load , since gaussian eliminations on a coefficient matrix under large field size @xmath2 are involved @xcite . on the contrary , idnc aims at minimizing decoding delay . by carefully choosing data packets
to be coded together , it guarantees a subset of ( or if possible , all ) receivers to instantly decode one of their wanted data packets upon successful reception of a coded packet @xcite .
this means that data packets can be potentially delivered to higher layers much faster than rlnc .
another advantage of idnc is its simple xor - based encoding and decoding , i.e. , coding coefficients are chosen from the binary field @xmath3 .
however , idnc is generally not optimal in terms of block completion time , as there might exist a subset of receivers who can not instantly decode any of their wanted data packets from the received coded packet .
moreover , idnc requires feedback from all receivers at a proper frequency for making coding decisions .
rlnc and idnc have many contrasting features .
they generally trade off decoding delay or throughput for one another , respectively .
they also differ in practical issues such as feedback frequency and decoding complexity
. moreover , since rlnc uses a larger finite field than idnc , any adaptive choice and switching between idnc and rlnc is impossible during the broadcast of one block . in this paper , we are interested in investigating the performance spectrum between idnc and rlnc .
this study will help us develop coding schemes that offer moderate throughput - delay tradeoffs , or equivalently , more balanced throughput and decoding delay performance compared with rlnc and idnc .
we are also interested in the connection between the coding mechanisms of idnc and rlnc .
this study will help us manage feedback frequency , decoding complexity , and even in - block performance adaption .
the first question we ask is : * _ is there a coding scheme which provides more balanced throughput and decoding delay performance than idnc and rlnc ? _ at the beginning of section[sec : partition_motivation ] , we will show the existence of such a scheme through an example , in which the decoding delay can be reduced without trading off the throughput .
the key idea is to appropriately partition a partially - received block of data packets into _ sub - generations _ based on their reception states at the receivers after sending them uncoded once first .
a similar partitioning has been studied in the literature @xcite , where sub - generations with equal number of data packets are generated and transmitted separately to reduce decoding delay in peer - to - peer scenarios . however , as we will prove in section[sec : partition_motivation ] , such partitioning may generally yield a prohibitively large block completion time compared with rlnc and idnc in our broadcast scenario . hence , the following question arises : * _ is there a better way of packet partitioning to achieve more balanced throughput and decoding delay performance during block transmission ? _
our answer to this question is through using the concept of _ non - conflicting data packets _ @xcite , which is defined as data packets not jointly wanted by any receiver . by properly grouping non - conflicting data packets together into coding sets ,
we obtain an idnc solution @xcite . then by appropriately partitioning coding sets in an idnc solution into sub - generations , we obtain a desirable balance between throughput and decoding delay performance , which lies between that of rlnc and idnc . consequently , we redefine the concepts of sub - generation and its size , which is the core of this work .
based on these new concepts , we achieve the following contributions : 1 .
we fill the throughput and decoding delay performance gap between idnc and rlnc by proposing new adaptive coding schemes .
they offer more balanced throughout and decoding delay performance than idnc and rlnc ; 2 .
the new coding schemes , together with idnc and rlnc , constitute a general coding framework .
this coding framework unifies idnc and rlnc , because they become its two specific cases ; 3 .
we study the throughput and decoding delay properties of this coding framework .
the results provide a good insight from the perspective of the proposed sub - generation size ; 4 .
we develop various implementations of this coding framework , which can further improve throughput and decoding delay performance without sacrificing on one another .
they also enable in - block performance adaption by allowing switching between idnc and rlnc .
plenty of hands - on examples are designed to demonstrate the proposed concepts and algorithms .
extensive simulations are also provided to evaluate the performance of the proposed coding schemes .
we consider wireless broadcast of @xmath4 data packets , denoted by @xmath5 , from a sender to @xmath6 receivers , denoted by @xmath7 .
time is slotted . in each time slot , a packet ( either original data or coded ) is sent .
wireless channels between the sender and the receivers are subject to packet erasures . for the ease of simulations ,
we assume i.i.d .
memoryless erasure channels to all receivers with equal probability @xmath8 .
however , this is not strictly needed and can be easily extended to more general cases . in this scenario ,
@xmath4 data packets are first sent uncoded once using @xmath4 time slots .
this phase is known as a _ systematic transmission phase _ @xcite .
we justify the application of this phase on idnc and rlnc in the following remark .
we will review this again under the proposed coding framework .
[ remark : systematic ] for idnc , a systematic transmission phase is compulsory @xcite , since coding opportunities are unavailable at the beginning .
rlnc with a systematic transmission phase is known as systematic rlnc @xcite .
it reserves the throughput optimality of the traditional rlnc .
it also introduces extra benefits such as 1 ) smaller packet decoding delay , and 2 ) smaller field size @xmath2 and thus lower coding / decoding complexity . in the rest of this paper
, the term rlnc always refers to systematic rlnc unless otherwise specified . during the systematic transmission phase a receiver might miss any data packet due to packet erasures .
the receivers then send lossless feedback to the sender about their packet reception states .
feedback information can be expressed by an @xmath9 binary state feedback matrix ( sfm ) @xcite , denoted by @xmath10 $ ] , where @xmath11 means receiver @xmath12 still wants data packet @xmath13 and @xmath14 otherwise . here
@xmath15 is the number of receivers who have not received all @xmath4 data packets and @xmath16 is the number of data packets that have not been received by all @xmath6 receivers .
the set of these @xmath16 partially - received data packets is denoted by @xmath17 .
an example of a @xmath18 sfm is given in fig .
[ fig : sfm ] .
the subset of @xmath17 wanted by receiver @xmath12 is called the
_ wants _ set of this receiver , denoted by @xmath19 .
its size is denoted by @xmath20 and the largest @xmath20 across all receivers is denoted by @xmath21 .
the subset of receivers that want @xmath13 is called the _
target _ set of @xmath13 , denoted by @xmath22 .
its size is denoted by @xmath23 .
based on the sfm , the sender initiates a _ coded transmission phase _ which is also subject to erasures . in this phase , the sender sends network coded packets to efficiently complete the broadcast of the data block .
this two - phase broadcast is illustrated in fig .
[ fig : phases ] .
the throughput and decoding delay performance of the network coding schemes can be measured by the minimum block completion time and minimum average packet decoding delay , respectively .
we first define the minimum block completion time : given an sfm and a certain coding scheme , the minimum block completion time , or equivalently the minimum number of coded transmissions , is the smallest possible number of coded transmissions by the sender that is needed to satisfy the demands of all receivers in the absence of any packet erasures in the coded transmission phase .
this number is denoted by @xmath24 .
the minimum block completion time @xmath24 reflects the best throughput performance of a certain coding scheme , calculated as @xmath25 .
such measure of throughput is important in the sense that it disentangles the effect of channel - induced packet erasures and algorithm - induced network coded packet design on the throughput . in reality ,
due to packet erasures or suboptimal choices of coded packet transmissions , we may need more than @xmath24 transmissions to complete a data block . in this case
, @xmath25 serves as an upper bound on throughput and will still serve as a useful measure of performance .
we now define the minimum average packet decoding delay : given an sfm and a certain coding scheme , assume that a sequence of coded packets that can achieve the corresponding minimum block completion time are transmitted without erasures .
also assume that coded packets are transmitted in the decreasing order of their targeted receivers to minimize decoding delay .
denote by @xmath26 the first time slot when original data packet @xmath13 can be decoded by receiver @xmath12 , and let @xmath27 if @xmath14 .
we have @xmath28 $ ] .
then the minimum average packet decoding delay , @xmath29 , is defined as : @xmath30 it is noted that @xmath29 refers to the minimum possible delay under a certain coding scheme achieving its minimum block completion time , but not refer to _ universal minimum delay _ under all coding schemes , which is still an opten question to the best of our knowledge .
we now review idnc and rlnc schemes in the coded transmission phase and discuss their throughput and decoding delay performance .
an idnc coded packet takes the form of @xmath31 where @xmath32 is the time index , @xmath33 and the summation is bit - wise xor @xmath34 .
we denote by @xmath35 the set of original data packets that have non - zero coefficients in @xmath36 , namely , @xmath37 .
@xmath35 fully represents @xmath36 and is called a coding set .
it is instantly decodable if it satisfies the following _ idnc constraint _
@xcite : [ def : coding_set ] an idnc coding set @xmath35 contains at most one data packet from the wants set , @xmath19 , of any receiver @xmath12 .
according to this constraint , we have two important concepts called conflicting and non - conflicting data packets @xcite : [ def : conf_packet ] we say that two data packets @xmath38 and @xmath39 conflict with each other if there exists at least one receiver who wants both packets .
that is , @xmath38 and @xmath39 conflict if both belong to the wants set , @xmath19 , of at least one receiver such as @xmath12 .
we say that two data packets @xmath38 and @xmath39 do not conflict with each other if there is no receiver who wants both packets .
it is clear that to avoid non - instantly decodable coded packets , two conflicting data packets @xmath38 and @xmath39 can not be coded together .
conflict states among all @xmath16 data packets can be represented by an undirected idnc graph @xmath40 , where vertex @xmath41 represents packet @xmath13 , and edge @xmath42 exists if @xmath38 does not conflict with @xmath39 @xcite .
below is an example : consider the sfm in fig .
[ fig : sfm ] .
@xmath43 and @xmath44 conflict with each other because @xmath45 wants both packets .
hence , in the idnc graph @xmath46 in fig .
[ fig : graph ] , @xmath43 and @xmath44 are not connected .
if @xmath47 were sent , @xmath45 would not be able to instantly decode @xmath43 or @xmath44 .
on the other hand , @xmath43 and @xmath48 do not conflict with each other because there is no single receiver who wants both packets .
thus @xmath43 and @xmath48 are connected in @xmath46 . by receiving @xmath49
, @xmath45 can instantly decode @xmath43 as @xmath50 , since @xmath45 already has @xmath48 .
similarly , the other three receivers can also instantly decode one data packet .
since data packets in the same coding set @xmath51 do not conflict with each other , the vertices representing these packets are connected to each other in @xmath46 .
these vertices form a clique @xcite of @xmath46 . furthermore
, a coding set @xmath51 is said to be _ maximal _ if its corresponding clique is maximal , i.e. , is not a subset of a larger clique .
we then have the following lemma @xcite , which identifies the minimum block completion time of idnc : the minimum block completion time of idnc , denoted by @xmath52 , is equal to the chromatic number of the complementary idnc graph @xmath53 . here
, @xmath53 denotes the complementary of @xmath46 , which has the same vertices as @xmath46 , but with opposite vertex connectivities .
we then have the concepts of an idnc solution and optimal idnc solution : [ def : idnc_solution ] an idnc solution is a collection of idnc coding sets which satisfy the following two conditions : 1 ) they jointly cover all @xmath16 data packets ; and 2 ) the union of any of them is not an idnc coding set .
[ def : optimal_idnc_solution ] while the first condition ensures completeness of the solution , the second condition prevents redundant idnc coding sets which can be absorbed into other coding sets to reduce the cardinality of the solution .
an idnc solution is optimal if : 1 ) its cardinality is @xmath52 ; and 2 ) all its coding sets are maximal .
an optimal idnc solution is denoted by @xmath54 . the minimum average packet coding delay of the optimal idnc solution can be computed using and is denoted by @xmath55 .
below is an example .
[ exmp : sfm_idnc ] the optimal idnc solution for the sfm in fig .
[ fig : sfm ] is @xmath56 , @xmath57,@xmath58 , @xmath59 , @xmath60 .
thus @xmath61 .
one can easily verify that every coding set in @xmath62 is maximal . by sending @xmath63 , @xmath64 , @xmath65 , and @xmath66 using four time slots ,
the minimum average packet decoding delay is computed using to be @xmath67 .
coded packets in rlnc are random linear combinations of _ all _ @xmath16 data packets : @xmath68 where coding coefficients @xmath69 are randomly chosen from a finite field @xmath0 and usually @xmath1 . in order to decode @xmath20 data packets
, a receiver @xmath12 needs to receive @xmath20 linearly independent coded packets .
hence , to satisfy the demands of all @xmath15 receivers , at least @xmath21 linearly independent coded packets need to be broadcast . the minimum block completion time using rlnc is thus equal to @xmath21 and is denoted by @xmath70 .
the minimum average packet decoding delay of rlnc , denoted by @xmath71 , can be calculated as : @xmath72 where we ignore early decoding chances of rlnc and assume that decoding of @xmath20 wanted packets by receiver @xmath12 is only possible after @xmath20 coded packets are received . for the sfm given in fig .
[ fig : sfm ] , if rlnc is applied , @xmath73 because @xmath74 .
the minimum average packet decoding delay is @xmath75 , which is almost @xmath76 larger than idnc with @xmath67 .
[ remark : rlnc_vs_idnc ] rlnc has a better throughput performance than idnc , i.e. , @xmath77 .
this relationship is always valid .
actually , @xmath70 is the benchmark for any network coding technique under the scenario considered here . on the other hand
, idnc is expected to have better decoding delay performance than rlnc , i.e. , @xmath78 . however , this relationship is not always valid , as we may find some instances of sfm where @xmath79 .
below is one such example .
detailed comparisons can be found in @xcite . for which @xmath80 results in @xmath79 . ]
consider the sfm in fig .
[ fig : sfm_2 ] .
assume rlnc is applied , then @xmath81 and @xmath82 .
if idnc is applied , because all four data packet conflict with each other , they must be sent separately .
thus , @xmath61 and @xmath83 .
the question that motivates our following work is whether there exist coding scheme(s ) that can provide a block completion time between @xmath70 and @xmath52 and an average packet decoding delay between @xmath55 and @xmath71 .
if one can find such schemes then moderate throughput - delay tradeoffs compared with idnc and rlnc can be achieved .
we will show that this is possible by packet partitioning .
the initial motivation of this work is to investigate whether the decoding delay of rlnc can be reduced without sacrificing on throughput .
rlnc suffers from large decoding delay mainly because it encodes all @xmath16 partially - received data packets in a block @xmath17 together .
hence , one may infer that decoding delay could be reduced by partitioning these @xmath16 data packets into several _
smaller generations _ and applying rlnc to these small generations separately . by doing so
, data packets in early small generations can be decoded sooner .
the idea of applying small generations has been studied in the literature for traditional rlnc , e.g. , @xcite for reducing decoding complexity and @xcite for reducing decoding delay , where throughput loss has been reported as inevitable .
we now show that such partitioning does not generally work well for systematic rlnc either .
in this paper we use the term _ sub - generations _ to denote smaller generations after partitioning of a partially - received data block @xmath17 of original size @xmath16 .
a sub - generation , by classic definition of generation @xcite , is a set of @xmath84 consecutive data packets , where @xmath84 is called the sub - generation size and ranges from 1 to @xmath16 here . according to this definition ,
sub - generations are generated by consecutively and evenly partitioning @xmath16 data packets in @xmath17 into @xmath85 partitions .
denote by @xmath86 the @xmath87-th sub - generation , we have @xmath88 .
its coded packet is : @xmath89 where coefficients @xmath90 are randomly chosen from an appropriate finite field @xmath0 .
let @xmath91 be the largest number of data packets in @xmath86 wanted by any one receiver across all receivers .
@xmath86 thus requires a minimum of @xmath91 coded transmissions .
then the minimum block completion time under partitioning with sub - generation size @xmath84 , denoted by @xmath92 , is calculated as : @xmath93 we further denote by @xmath94 the minimum average packet decoding delay under sub - generation size @xmath84 . its calculation follows .
we now demonstrate via an example that this classic partitioning can result in large @xmath92 and @xmath94 and hence is undesirable in terms of both throughput and delay .
[ exmp : bad_partition ] consider the sfm given in fig .
[ fig : sfm ] .
@xmath95 data packets can be partitioned into @xmath96 sub - generations under @xmath97 : the first sub - generation is @xmath98 and the second sub - generations is @xmath99 . because @xmath100 and @xmath101 , the minimum block completion time is @xmath102 , which is much greater than both @xmath52 and @xmath70 .
the minimum average packet decoding delay is @xmath103 , which is also much greater than both @xmath55 and @xmath71 studied in example [ exmp : sfm_idnc ] and below .
however , by changing the content of each sub - generation , we can obtain better performance : [ exmp : good_partition ] consider the sfm given in fig .
[ fig : sfm ] . we still partition the data packets into two sub - generations , but the first sub - generation is @xmath104 and the second sub - generation is @xmath105 .
the two sub - generations have @xmath106 and thus @xmath107 , which is the same as @xmath70 .
however , since data packets in @xmath108 can be decoded after only 2 coded transmissions , @xmath94 is only 2.7 , which is smaller than @xmath75 calculated earlier .
example [ exmp : bad_partition ] demonstrates that the classic partitioning does not provide convincing throughput and decoding delay performance .
whereas example [ exmp : good_partition ] shows that there may be better ways of partitioning .
explicitly , under classic partitioning , the minimum block completion time @xmath92 has the following property : if data packets are partitioned into classic sub - generations , @xmath92 is lower bounded as : @xmath109 it is obvious that @xmath110 . furthermore , because @xmath111 for @xmath112 $ ] , @xmath113 according to ( [ eq : u_calculation ] ) .
this lemma indicates that , when a small sub - generation size @xmath84 is applied in an attempt to reduce decoding delay , the resulted @xmath114 and @xmath92 can be much larger than @xmath70 and even @xmath52 .
a large @xmath92 implies large decoding delays for data packets sent in the last few sub - generations .
this will in turn increase @xmath94 , making such partitioning pointless in terms of both throughput and decoding delay .
in example [ exmp : bad_partition ] , due to the fact that @xmath102 , the decoding delay of data packets @xmath115 is as large as 7 .
in conclusion , following the classic partitioning we are unable to fill the performance gap between idnc and rlnc . a better way of partitioning is needed and will be developed next , based on which the concept of sub - generations will be redefined and a new coding framework will be proposed . the partitioning in example [ exmp : bad_partition ] failed to reduce either @xmath92 or @xmath94 because all data packets in the same sub - generation , e.g. , @xmath99 , are jointly wanted by at least one receiver , yielding a minimum of @xmath101 coded transmissions to complete this sub - generation .
in contrast , as shown in example [ exmp : good_partition ] , sub - generations of @xmath116 and @xmath105 only requires @xmath106 coded transmissions because all receivers want at most 2 data packets from them .
these two examples motivate the key to a better partitioning , that is , _ to avoid as much as possible partitioning data packets that are jointly wanted by any receiver into the same sub - generation .
_ by doing so , @xmath117 of the sub - generations are reduced .
imagine an extreme sub - generation in which any two data packets are not jointly wanted by any receiver .
then the broadcast of this sub - generation only requires one coded transmission .
such a sub - generation , recalling definition [ def : coding_set ] in section [ sec : system_idnc ] , is exactly an idnc coding set .
the key step of partitioning then becomes clear : instead of partitioning packets in the packet set @xmath17 based on their consecutive index , we partition a set of idnc coding sets which together cover all @xmath16 data packets . in other words , we partition an idnc solution . has the smallest cardinality ( @xmath52 ) , it is the desired object for partitioning .
although we will develop the new coding framework based on the optimal idnc solution , the proposed definitions , properties , and implementations are not restricted to the optimal idnc solution .
they can be applied to any idnc solution satisfying definition [ def : idnc_solution ] , such as those heuristically found in @xcite . ] accordingly , the concept of sub - generations is redefined : a sub - generation @xmath118 is a collection of idnc coding sets in an idnc solution .
the definition of sub - generation size is also changed : sub - generation size @xmath84 is the number of coding sets in a sub - generation .
the above definitions enable a new coding framework . given the optimal idnc solution @xmath54 , the @xmath52 maximal coding sets are partitioned into @xmath119 sub - generations , where @xmath120 $ ] .
the simplest partitioning method is a consecutive one : @xmath121 , @xmath122 , @xmath123 .
a schematic of this partitioning is plotted in fig.[fig : partition ] .
more efficient partitioning algorithms will be developed in section [ sec : implementations ] .
for each sub - generation @xmath86 , its coded packets are generated using as in classic partitioning .
linear independency among the coded packets are promoted by randomly choosing coding coefficients from a sufficiently large finite field @xmath0 .
the field size @xmath2 can be reduced with decreasing sub - generation size @xmath84 . in fact , @xmath124 when @xmath125 , since all receivers want at most one data packet from a sub - generation of size @xmath125 .
the definition of @xmath91 of a sub - generation @xmath86 , where @xmath112 $ ] , is still the largest number of data packets in @xmath86 wanted by any one receiver across all receivers .
since @xmath86 needs a minimum of @xmath91 coded transmissions , the relationship between @xmath92 and @xmath91 is still as in .
below is an example of the proposed coding framework .
consider the sfm in fig .
[ fig : sfm ] , whose optimal idnc solution is : @xmath126 if @xmath127 , there will be @xmath128 sub - generations , where @xmath104 and @xmath105 .
this result is exactly the same as that in example [ exmp : good_partition ] and explains the success of the partitioning in example [ exmp : good_partition ] compared to the classic partitioning in example [ exmp : bad_partition ] .
interestingly , there are two extreme cases of this coding framework , taking place when @xmath125 and @xmath129 .
when @xmath125 , since @xmath124 , coding within a sub - generation is through xor and there are @xmath52 such sub - generations .
hence , the coding scheme becomes idnc with @xmath130 .
when @xmath129 , there will be only one sub - generation , which contains all coding sets and thus all @xmath16 data packets .
hence , the coding scheme becomes rlnc with @xmath131 .
therefore , the proposed coding framework successfully unifies the coding mechanisms of idnc and rlnc .
it also enables the coding schemes in the spectrum between idnc and rlnc with @xmath132 $ ] . in the next section
, we will study throughput and decoding delay properties of the proposed coding framework under all values of sub - generation size @xmath84 .
the minimum block completion time @xmath92 is determined by @xmath117 according to ( [ eq : u_calculation ] ) .
@xmath117 has the following property : when sub - generation size satisfies @xmath133 , @xmath134 $ ] for all sub - generations @xmath112 $ ] . 1 .
@xmath86 is the superset of all @xmath84 idnc coding sets in it .
if all receivers want at most one data packet in @xmath86 , i.e. , if @xmath135 , then @xmath86 itself is an idnc coding set .
this contradicts with the fact that the union of any coding sets in an idnc solution is not a coding set according to definition [ def : idnc_solution ] .
thus @xmath136 ; 2 . if there is a receiver who wants @xmath137 data packets in @xmath86 , these @xmath91 data packets conflict with each other and at least two of them must belong to the same coding set .
this contradicts the idnc constraint stated in definition [ def : coding_set ] .
thus , @xmath138 . then , by considering the relationship between @xmath91 and @xmath92 in ( [ eq : u_calculation ] ) and noting that @xmath119 , the above lemma yields an important corollary : [ theo : u_bounds ] for all values of sub - generation size @xmath84 , the minimum block completion time is bounded between @xmath70 and @xmath52 : @xmath139\ ] ] where @xmath140 holds when @xmath129 and @xmath141 holds when @xmath142 ( because @xmath143 when @xmath127 ) .
explicitly , @xmath92 approaches @xmath70 when the sub - generation size @xmath84 increases gradually from @xmath144 to @xmath52 and approaches @xmath52 the other way around .
the above lower and upper bounds on the minimum block completion time generally hold under the proposed coding framework .
they are also useful benchmarks for the throughput performance of packet - level network coded wireless broadcast schemes in general .
while @xmath70 is the lower bound for any such scheme , the upper bound @xmath52 is an indicator of throughput efficiency , because any scheme requiring a minimum block completion time of greater than @xmath52 can be treated as inefficient .
one such inefficient scheme is the rlnc with classic partitioning in example [ exmp : bad_partition ] . for the minimum average packet decoding delay , @xmath94 ,
the situation is more complicated .
it is clear that @xmath94 lies between @xmath55 and @xmath71 .
however , as we have discussed in remark [ remark : rlnc_vs_idnc ] , there is no guarantee that @xmath145 .
hence , instead of @xmath55 or @xmath71 , we propose an alternative upper bound on @xmath94 in terms of sub - generation size @xmath84 : the minimum average decoding delay under sub - generation size @xmath84 satisfies @xmath146 denote the largest decoding delay of data packets in @xmath86 by @xmath147 .
it is equal to @xmath148 .
when @xmath149 is maximized , i.e. , when @xmath150 for all @xmath151 $ ] , @xmath147 is maximized with a value of @xmath152 . since we can always send @xmath86 with more targeted receivers first
, the largest @xmath94 happens when all @xmath86 have the same number of target receivers , denoted by @xmath153 .
then , as a variation of ( [ eq : d_definition ] ) , the largest @xmath94 is calculated as : @xmath154 this bound also justifies the application of the systematic transmission phase : at the beginning of the broadcast of @xmath4 data packets , we have an all - one sfm @xmath155 of size @xmath156 .
thus , while @xmath157 , ( [ eq : d_bound ] ) indicates that using @xmath125 offers the smallest packet decoding delay , which requires all @xmath4 data packets to be sent separately uncoded .
in summary , in this section we showed that throughput and decoding delay performance of all coding schemes in the proposed coding framework is well bounded between that of idnc and rlnc .
therefore , our coding framework is a general one with idnc and rlnc identified as two extreme cases with @xmath125 and @xmath158 , respectively .
it successfully fills the performance gap between idnc and rlnc and enables a series of coding schemes with a range of more balanced throughput and decoding delay performance . in the next section
, we will turn to implementations of the proposed coding framework to further improve its throughput and decoding delay performance .
in the last section , we showed that throughput and decoding delay performance of the proposed coding framework is well bounded between idnc and rlnc for all sub - generation sizes @xmath120 $ ] .
although we can not further improve the performance of idnc and rlnc , we can do so for the coding schemes with @xmath132 $ ] .
after the systematic transmission phase , there are two steps in the coded transmission phase : 1 ) partitioning idnc coding sets into sub - generations ; and 2 ) broadcasting these sub - generations following a transmission strategy .
we will first optimize these two steps and then further explore the potentials of this coding framework .
we denote by @xmath159 the number of targeted receivers of a coding set @xmath35 in the optimal idnc solution @xmath160 @xmath161 @xmath162 @xmath163 @xmath164 , where : @xmath165 without loss of generality we also assume that @xmath166 , i.e. , @xmath167 is wanted by the most receivers , followed by @xmath168 , and so on .
the simplest algorithm is to consecutively partition the @xmath52 coding sets .
that is , @xmath121 , @xmath169 , @xmath123 .
this algorithm is referred to as _ direct partitioning _ ( dp ) .
dp offers good decoding delay performance , because coding sets with more targeted receivers are partitioned into earlier sub - generations and are broadcast earlier .
however , @xmath91 may not be uniform across coding sets in @xmath86 .
dp overlooks this fact and thus does not necessarily minimize @xmath117 and @xmath92 . below is an example . when @xmath170 .
] _ in the matrix in fig .
[ fig : algorithmc_partition ] , an entry of one at row @xmath171 and column @xmath172 means receiver @xmath12 wants one data packet in coding set @xmath173 .
suppose the sub - generation size is @xmath170 .
then , according to dp , we have @xmath174 , and thus @xmath100 .
however , if we use @xmath175 , we have @xmath176 .
_ hence , to strike a balance between throughput and decoding delay , we design a heuristic algorithm which aims at 1 ) reducing @xmath92 by exploiting the opportunities of reducing @xmath91 for all @xmath87 , and 2 ) preserving low decoding delay by partitioning the most wanted coding sets into earlier sub - generations .
this algorithm , which we refer to as _ smart partitioning _
( sp ) , fills sub - generations sequentially , i.e. , the @xmath84 coding sets for @xmath108 are first determined , followed by the @xmath84 coding sets for @xmath177 , etc .
details of this algorithm are presented in algorithm [ alg : partition ] .
initialize : an idnc solution @xmath62 , and @xmath114 empty sub - generations , @xmath178 ; sort the coding sets in @xmath62 in a descending order in terms of their number of targeted receivers ; find the coding sets in @xmath62 that do not increase @xmath91 by one . denote the collection of such coding sets by @xmath179 ; add the coding set in @xmath179 with the smallest index ( _ and thus wanted by most receivers _ ) to @xmath86 ; add the coding set in @xmath62 with the smallest index to @xmath86 ; remove the chosen coding set from @xmath62 ; terminate the algorithm ; compared with dp , sp costs light extra computations .
however , as will be numerically compared later , the performance of sp is better than dp in a wide range of system settings .
the reason why sp outperforms dp in terms of decoding delay is that , by reducing @xmath92 in sp we also reduce the worst packet decoding delay , which will in turn reduce the average packet decoding delay .
with sub - generations generated , we are ready to broadcast them through erasure - prone channels .
in this subsection , we present two coded transmission strategies , called sequential and semi - online , respectively . they are superior to each other in different aspects
. given all @xmath114 sub - generations , the simplest strategy for the coded transmission phase is to segment this phase into @xmath114 rounds . in each round ,
coded packets of a sub - generation are broadcast until all its targeted receivers have decoded it and informed the sender .
we name this strategy _
sequential_. in sequential strategy , the decoding of different sub - generations are independent of each other .
this property leads to the primary advantage of this strategy , namely , tunable throughput and decoding delay even within the broadcast of a data block , because we can apply different sub - generation sizes in different rounds without affecting the decoding of sub - generations in other rounds .
for example , when decoding delay becomes the primary concern , the system can easily switch to idnc by setting @xmath125 in all remaining rounds .
such switching was impossible because rlnc generally applies a larger field size than idnc .
another example is schematically shown in fig.[fig : change_g ] . in this example
, there are three rounds , with sub - generation sizes of 2 , 1 , and @xmath180 , respectively .
sequential strategy requires at most @xmath114 ack feedback from every receiver .
the sub - generation size can also be adjusted to fit the system s specific feedback frequency if there is any .
the main disadvantage of this strategy is its large decoding delay , because the broadcast of a sub - generation @xmath86 can not be started until the broadcast of @xmath181 is completed .
even if only one targeted receiver of @xmath181 experiences a bad channel , the round for @xmath181 must continue and thus all targeted receivers of @xmath86 have to wait . to overcome this drawback
, we propose another coded transmission strategy called _ semi - online _ strategy . during the broadcast using sequential strategy . ] in semi - online strategy , the coded transmission phase is also segmented into rounds , which is , though , different from the rounds in sequential strategy . here in each round , all @xmath114 sub - generations are broadcast , where for @xmath86 , @xmath91 coded packets are broadcast . thus in each round , there are @xmath182 coded transmissions .
after each round , the sender collects feedback from the receivers about how many more coded packets of each sub - generation they still want .
@xmath117 is thus updated accordingly before the next round starts .
below is a simple example .
assume there are three receivers , @xmath45 to @xmath183 .
they want two , three , and four data packets from sub - generation @xmath108 , respectively .
therefore , @xmath184 coded packets of @xmath108 are broadcast in the first round , along with coded packets of other sub - generations .
assume that after the first round , @xmath45 to @xmath183 have received two , one , and three coded packets of @xmath108 , respectively
. then they still want zero , two , and one coded packets to decode @xmath108 , respectively .
@xmath185 is thus updated to two . in the second round ,
two coded packets of @xmath108 will be broadcast , along with coded packets of other sub - generations .
semi - online strategy overcomes the main drawback of sequential strategy , and thus significantly improves the decoding delay performance . on the other hand , from a statistical point of view
, semi - online strategy has the same throughput performance as sequential strategy , since we only `` swap '' the coded transmissions for each sub - generation .
the main drawback of semi - online strategy is that throughput and decoding delay performance is no longer tunable .
moreover , since the number of rounds in semi - online strategy increases with decreasing channel quality , the amount of feedback in this strategy can not be predetermined .
however , as will be presented in section [ sec : merge ] , semi - online strategy offers the opportunity to algorithmically merge sub - generations together , which can further improve both throughput and decoding delay performance . therefore , there is no clear winner between sequential and semi - online strategies .
which one to adopt depends on the application . in the next subsection
, we will discuss the concept and impacts of _ packet diversity _ on these two strategies .
the diversity of a data packet in the proposed coding framework is defined as follows : the diversity of a data packet is the number of sub - generations in which it appears . in the proposed coding framework ,
data packets might have diversities of greater than one because the sub - generations are generated from maximal idnc coding sets , whose intersections are usually not empty .
our motivation of studying packet diversity is the fact that , we could possibly reduce @xmath91 of a sub - generation @xmath86 by removing a data packet from it . below is an example .
assume that there are two sub - generations .
the first sub - generation @xmath108 contains @xmath186,@xmath44,@xmath187 and some other data packets .
the second sub - generation @xmath177 contains @xmath188 and some other data packets .
here data packet @xmath188 has a diversity of 2 .
also assume that receiver @xmath45 wants @xmath186,@xmath44,@xmath187 , which means @xmath185 is at least 3 .
now let us remove @xmath188 from @xmath108 , which reduces the diversity of @xmath188 to 1 .
the targeted receivers of @xmath188 can still decode it from @xmath177 , while @xmath185 can be possibly reduced to 2 , which reduces the minimum block completion time @xmath92 by 1 .
inspired by this fact , we investigate the benefits and problems that a packet diversity of greater than one brings to sequential and semi - online coded transmission strategies , and then decide whether it should be reduced to one or not . a packet diversity of greater than one is redundant in sequential strategy regardless of the sub - generation size .
suppose data packet @xmath43 is included in @xmath108 and @xmath177 . by the end of the round for @xmath108 , all receivers who want
@xmath43 will have received it , indicating that @xmath43 does not need to be included in @xmath177 . unlike sequential strategy
, the impacts of packet diversity in semi - online strategy is much more complex .
since all data packets are broadcast in every round , a higher packet diversity can be translated into a higher probability of being received and decoded , and thus reduces the number of coded transmissions in the next round . however , a high packet diversity incurs complicated coding and decoding decision makings .
this drawback can be illustrated by the following example .
assume that receiver @xmath45 wants @xmath189 from @xmath108 , and wants @xmath190 from @xmath177 .
here @xmath44 has a diversity of 2 .
imagine a case that , after the first coded transmission round , @xmath45 has received one coded packet of @xmath108 and one coded packet of @xmath177 .
in the next round , @xmath45 only needs one coded packet of either @xmath108 or @xmath177 to decode all three data packets .
thus the sender needs to decide to send a coded packet of @xmath108 or @xmath177 or both .
this case is referred to as case-1 .
imagine another case that , after the first coded transmission round , @xmath45 has received one coded packet of @xmath108 and two coded packets of @xmath177 .
@xmath45 can thus directly decode @xmath190 from @xmath177 .
after that , @xmath45 can substitute @xmath44 into the received coded packet of @xmath108 to decode @xmath43 .
this case is referred to as case-2 . when case-1 in the above example is extended to all receivers ,
decision making by the sender will become very complicated .
when case-2 in the above example is extended to all sub - generations , decoding by the receiver will become complicated , because once it has decoded some data packets from a sub - generation , it has to look up all other sub - generations for more decoding opportunities .
the only exception happens when @xmath125 , i.e. , idnc . in this case , since there is no linear equations to solve , such search is unnecessary . in conclusion ,
we suggest to reduce the diversities of all data packets to one for both sequential and semi - online strategies unless @xmath125 .
this reduction is applied to the optimal idnc solution by removing data packets from a maximal coding set if they have already been covered by the previous coding sets .
the resulted solution is denoted by @xmath191 and also has a length of @xmath52 .
any of the partitioning algorithms discussed before can then be applied to @xmath191 before the coded transmission phase .
it is noted that our theoretical analysis and partitioning algorithms are not affected by the diversity reduction .
throughput properties are not affected because both @xmath52 and @xmath70 remain the same and the relationship between @xmath92 and @xmath117 always holds .
decoding delay properties are not affected because , by its definition in , we only consider the first time slot that a data packet can be decoded .
its reception in later time slots due to diversity is not considered .
partitioning algorithms are not affected because they work for any valid idnc solution . in this subsection , we reduce packet diversities primarily for reducing @xmath117 at the beginning of the coded transmission phase . in the next subsection , we will introduce an operation called _ sub - generation merging _ , which could reduce @xmath117 in the second and further coded transmission rounds in the semi - online strategy .
sub - generation merging is an operation that can only be applied under semi - online strategy .
we use a simple example to introduce it .
imagine that after a semi - online coded transmission round , there are two uncompleted sub - generations and two receivers . as shown in fig .
[ fig : dof_merge ] , @xmath45 still wants two coded packets of @xmath108 and @xmath192 still wants three coded packets of @xmath177 . in this case ,
@xmath176 and @xmath193 , thus @xmath194 .
@xmath45 can decode after two coded transmissions , and @xmath192 can decode after five coded transmissions . alternatively , let us merge @xmath108 and @xmath177 together , that is , combine data packets in @xmath108 and @xmath177 together to form a new sub - generation @xmath195 .
@xmath45 wants two and @xmath192 wants three coded packets of @xmath195 , respectively .
thus @xmath196 and @xmath92 is reduced from 5 to 3 , i.e. , throughput is improved . moreover , while @xmath45 can still decode after two coded transmissions , @xmath192 can decode after only three coded transmissions rather than five , thus decoding delay is also reduced . in this example ,
the two sub - generations @xmath108 and @xmath177 are not jointly wanted by any receiver .
compared with the definition of non - conflicting data packets in section [ sec : system ] , we define such sub - generations as _ non - conflicting sub - generations_. it is straightforward that non - conflicting sub - generations can be merged together so that both the throughput and decoding delay performance can be improved . the way of deciding which sub - generations to merge together is the same as finding an idnc solution from its graph representation . since the number of sub - generations is small
, such decision making is not computationally expensive and can be heuristically found using the methods introduced in @xcite .
a flow chart is presented in fig.[fig : implementations ] to summarize proposed implementations of our coding framework . with their advantages and drawbacks .
] the proposed coding framework and its implementations can be easily adapted to existing linear network coding systems , because at its core , the proposed coding framework generates linear combinations of the data packets as coded packets .
the only modification is that the sender requires the receivers to provide feedback by the last coded transmission of each round .
four sets of simulations are carried out in this section . in the first simulation
, we numerically demonstrate the well - bounded throughput and decoding delay performance of the proposed coding schemes compared with idnc and rlnc . in the second to fourth simulations ,
we verify the effectiveness of the proposed implementations , including partitioning algorithms , coded transmission strategies , and sub - generation merging , respectively .
we simulate broadcast of @xmath197 data packets to @xmath198 $ ] receivers .
wireless channels between the sender and the receivers are subject to i.i.d . memoryless packet erasures with a probability of @xmath199 .
since the first two simulations are concerned with the minimum block completion time and the minimum average packet decoding delay , packet erasures in the coded transmission phase are not considered in these two simulations , but will be incorporated in the third and fourth simulations . in this simulation , we evaluate the block completion time ( @xmath92 ) and decoding delay ( @xmath94 ) performance of the proposed coding schemes with various values of sub - generation size @xmath84 , including @xmath125 ( idnc scheme ) , @xmath200 , and @xmath158 ( rlnc scheme ) .
direct partitioning is applied .
the results are plotted in fig.[fig : tradeoff ] .
as we can see , throughput and decoding delay performance of the coding schemes with @xmath120 $ ] is well bounded between the performance of idnc and rlnc .
they fill the performance gap between idnc and rlnc , and thus offer moderate throughput - delay tradeoffs compared with idnc and rlnc .
another observation is that the throughput performance under @xmath127 is always the same as under @xmath125 , i.e. , @xmath201 .
this result matches our upper bound on @xmath92 in corollary [ theo : u_bounds ] . on the other hand ,
their decoding delay performance is generally different due to their different sub - generation sizes .
fig.[fig : tradeoff ] also provides an example where rlnc outperforms idnc in terms of both throughput and decoding delay performance .
it takes place when @xmath52 becomes much larger than @xmath70 ( 11.3 and 8 , respectively , at @xmath202 ) , which means the worst packet decoding delay of idnc is much larger than rlnc .
data packets . ] in this simulation , we evaluate the direct and smart partitioning algorithms .
we apply two sub - generation sizes , @xmath203 and @xmath204 .
simulation results are shown in fig .
[ sim : partition ] . from this figure , we observe that the performance of smart partitioning is equal to or better than direct partitioning for all values of @xmath6 and for both values of sub - generation size @xmath84 .
in this simulation , we evaluate sequential and semi - online coded transmission strategies .
we use smart partitioning and apply three sub - generation sizes , @xmath205 , @xmath204 and @xmath129 .
simulation results are shown in fig .
[ sim : straight_vs_semi ] .
our first observation is that the two strategies always share the same throughput performance .
this result matches our statistical claim in section [ sec : trans_schemes ] .
the second observation is that the decoding delay performance of the semi - online strategy successfully outperforms the sequential strategy when @xmath206 .
when @xmath129 , since there is only one sub - generation , both strategies become equivalent to rlnc and thus have the same decoding delay performance . in this simulation , we compare the performance of semi - online strategy with and without sub - generation merging .
the results are shown in fig .
[ sim : merge ] , from which it is clear that sub - generation merging improves both throughput and decoding delay under all parameter settings . at a high level , our simulations show that throughput and decoding delay performance do not have to be improved by trading each other off , but they can have coordination in our coding framework . by applying proposed implementations and choosing a proper sub - generation size , a large range of throughput - delay tradeoffs can be achieved .
the best configuration in terms of performance is smart partitioning combined with semi - online coded transmission strategy with sub - generation merging .
for wireless network coded broadcast , we showed that it is possible to build upon an idnc solution to obtain a series of more general linear network coded solutions with varying throughput and decoding delay performance , as well as varying implementation complexity and feedback frequency .
the core of our work was introducing a novel way of partitioning a partially - received data block into sub - generations based on the coding sets in a given idnc solution .
consequently , when an idnc solution is used unaltered for coded transmissions , we are at one end of the spectrum , namely idnc with sub - generation size @xmath207 . when all idnc coding sets are combined , we reach the other end of the spectrum , namely rlnc with sub - generation size @xmath208 .
the primary advantage of our coding framework is that the throughput and decoding delay of all intermediate coding schemes with different sub - generation sizes are guaranteed to lie between those of idnc and rlnc . with this crucial advantage
, we showed that we are able to focus on further improvement of throughput and delay for intermediate coding schemes by more advanced partitioning and transmission strategies such as smart partitioning combined with semi - online transmission strategy and sub - generation merging , or even smart partitioning with sequential transmission strategy that enables in - block switching between idnc and rlnc for performance adaptation .
b. swapna , a. eryilmaz , and n. shroff , `` throughput - delay analysis of random linear network coding for wireless broadcasting , '' in _ proc .
symposium on network coding ( netcod ) _
, jun . 2010 , pp . 16 .
p. sadeghi , r. shams , and d. traskov , `` an optimal adaptive network coding scheme for minimizing decoding delay in broadcast erasure channels , '' _ eurasip j. on wireless commun . and netw .
_ , pp . 114 , jan . 2010 .
m. yu , p. sadeghi , and n. aboutorab , `` on the throughput and decoding delay performance of instantly decodable network coding : , '' _ ieee / acm trans .
_ , ( submitted to ) , 2013 .
[ online ] .
available : \url{http://arxiv.org / abs/1309.0607}[\url\{http://arxiv.org / abs/1309.0607 } ] j. k. sundararajan , p. sadeghi , and m. mdard , `` a feedback - based adaptive broadcast coding scheme for reducing in - order delivery delay , '' in _ proc .
workshop on network coding , theory , and applications _
, 2009 , pp .
n. aboutorab , s. sorour , and p. sadeghi , `` o2-gidnc : beyond instantly decodable network coding , '' in _ network coding ( netcod ) , 2013 international symposium on_.1em plus 0.5em minus 0.4emieee , 2013 , pp . | our primary goal in this paper is to traverse the performance gap between two linear network coding schemes : random linear network coding ( rlnc ) and instantly decodable network coding ( idnc ) in terms of throughput and decoding delay .
we first redefine the concept of packet generation and use it to partition a block of partially - received data packets in a novel way , based on the coding sets in an idnc solution . by varying the generation size
, we obtain a general coding framework which consists of a series of coding schemes , with rlnc and idnc identified as two extreme cases .
we then prove that the throughput and decoding delay performance of all coding schemes in this coding framework are bounded between the performance of rlnc and idnc and hence throughput - delay tradeoff becomes possible .
we also propose implementations of this coding framework to further improve its throughput and decoding delay performance , to manage feedback frequency and coding complexity , or to achieve in - block performance adaption .
extensive simulations are then provided to verify the performance of the proposed coding schemes and their implementations . |
in recent years , the concept of topological band theory has been extended to nonelectronic systems such as magnons @xcite and phonons @xcite . in the former
, the dmi @xcite is the primary source of topological magnon bands and magnon edge states @xcite , as well as thermal magnon hall effect @xcite .
these systems are dubbed topological magnon insulators @xcite in analogy to topological insulators in fermionic systems with spin - orbit coupling ( soc ) @xcite . however , as magnons are charge - neutral quasiparticles , the perfectly conducting edge states are believed to be useful for dissipationless transport applicable to magnon spintronics .
our conception of dmi being the primary source of topological effects in magnonic systems has been firmly established by recent experimental realization of topological magnon insulator in collinear kagom ferromagnet cu(1 - 3 , bdc ) @xcite . in these systems
the dmi comes naturally because the kagom lattice lacks an inversion center . in reality , however , there are more frustrated kagom magnets than collinear ferromagnets .
the physics associated with the former has no analogy with the latter .
the former are considered as a candidate for quantum spin liquid physics due to an extensive classical degeneracy that prevent magnetic ordering to lowest accessible temperatures . however , in recent experimental synthesis it has been shown that the effects of soc or dmi are not negligible in most kagom antiferromagnets .
the dmi appears as a perturbation to the heisenberg spin exchange .
one of the striking features of this perturbation in frustrated kagom systems is that it induces magnetic ordering with a @xmath5 propagation wavevector .
thus , it suppresses the spin liquid phase of kagom antiferromagnets up to a critical value @xcite .
various experimentally accessible frustrated kagom antiferromagnets show evidence of coplanar / noncollinear @xmath5 magnetic ordering at specific temperatures .
the famous one is iron jarosite kfe@xmath1(oh)@xmath2(so@xmath3)@xmath4 @xcite .
the unanswered question is how does topological effects arise in these systems ? from the experimental perspective , it has been previously shown that iron jarosite posseses a finite spin scalar chirality induced by an out - of - plane magnetic field but no topological properties were measured @xcite .
in a recent letter @xcite , we have provided evidence of magnon hall effect and thermal hall conductivity @xmath6 for this kagom material , which originates from non - coplanar spin texture and survives in the absence of dmi and magnetic ordering .
topological magnon insulator with magnon edge modes are another area of recent development @xcite . in this report , we complete our analysis of topological magnon effects in geometrically frustrated kagom antiferromagnets by providing evidence of topological magnon edge modes .
the main purpose of this report is to relate finite thermal hall conductivity to topologically protected magnon edge states .
we consider three different models , viz : @xmath7 bilayer kagom ferromagnets coupled antiferromagnetically or layer antiferromagnets .
@xmath8 the model hamiltonian for iron jarosite kfe@xmath1(oh)@xmath2(so@xmath3)@xmath4 with dmi and second nearest neighbour interaction @xcite . @xmath9 the xxz model for kagom antiferromagnet without the dmi .
the arrangement of this paper is as follows . in sec .
ii we introduce the three models , analyze the magnon tight binding models , and present the distinctive magnon bands . sec .
iii introduces the concept of topologically protected magnon edge modes and fictitious chern numbers .
we relate these concepts to experimentally accessible thermal magnon hall effect . in sec .
iv we conclude and discuss potential methods to realize topological magnon bands in geometrically frustrated kagom antiferromagnets with / without dmi .
in various frustrated kagom magnets , a strong out - of - plane magnetic field is sufficient to circumvent frustrated interactions and leads to magnetic ordering .
a good example is the bilayer frustrated kagom magnet ca@xmath10cr@xmath11o@xmath12 @xcite , which shows evidence of ferromagnetic alignment at a magnetic field of magnitude @xmath13 .
the frustrated kagom volborthite cu@xmath1v@xmath4o@xmath11(oh)@xmath4 @xmath142h@xmath4o also shows evidence of magnetic ordering at several field values @xcite . in the ordered regimes ,
the magnetic excitations are definitely magnons .
owing to the fact that many kagom magnets come naturally in bilayer forms , we first consider bilayer kagom magnets with non - negligible interlayer coupling .
we assume that the top layer is placed right above the bottom layer forming aaa - stacked pattern .
the hamiltonian is given by @xmath15 where @xmath16 is the spin moment at site @xmath17 , @xmath18 labels the top ( @xmath19 ) and bottom ( @xmath20 ) layers respectively , @xmath21 is an external magnetic field in units of @xmath22 .
we consider the case of ferromagnetic intra - layer exchange @xmath23 and antiferromagnetic interlayer coupling @xmath24 with an out - of - plane magnetic field @xmath21 . at zero magnetic field ,
the spins on the top and bottom layers lie in opposite directions on the @xmath25-@xmath26 kagom planes , and interlayer coupling is antiferromagnetic .
a nonzero magnetic field is expected to introduce canting along the direction of the field .
hence , the ground state is no longer the collinear ferromagnets . in the classical limit
, the spin operators can be represented as classical vectors , written as @xmath27 , where @xmath28 is a unit vector .
the classical energy gives @xmath29 where @xmath30 , @xmath31 is the number of sites per unit cell , and the magnetic field is rescaled in units of @xmath32 . minimizing the classical energy yields the canting angle @xmath33 .
we see that both the out - of - plane dm vector ( @xmath34 ) and the in - plane dm vector ( @xmath35 ) does not contribute to the classical energy due to ferromagnetic ordering on each layer . for the magnon excitations above the classical ground state , the basic procedure involves rotating the spins from laboratory frame to local frame by the spin oriented angles @xmath36 about the @xmath37-axis . due to the field - induced canting
, we perform another rotation about the @xmath26-axis with the canting angle @xmath38 .
the total rotation matrix is given by @xmath39 where @xmath36 labels the spin oriented angles on each layer with @xmath40 for the top layer , @xmath41 for the bottom layer , and @xmath38 is the field canting angle .
i=_z(_)_y()s_i^,which can be written explicitly as @xmath42 where @xmath43 applies to the layers @xmath44 respectively .
this rotation does not affect the ferromagnetic @xmath45-term on each layer .
a crucial distinguishing feature of this system is that both the out - of - plane dmi ( @xmath34 ) and the in - plane dmi ( @xmath35 ) contribute in the present system .
this is as a result of field - induced canting of the system .
it should be noted that this is not the case in previously studied kagom ferromagnets @xcite . to linear order in @xmath46 expansion we have @xmath47 where @xmath48 for top and bottom layers respectively .
we see that the latter case has opposite signs on both layers .
next , we study the excitations above the classical ground state by using the holstein - primakoff spin bosonic operators for the rotated prime coordinates : @xmath49 , @xmath50 , and @xmath51 . the hamiltonian maps to a magnon tight binding model@xmath52,\end{aligned}\ ] ] where @xmath53 is the occupation number , @xmath54 , and @xmath55 with @xmath56 , @xmath57 .
the fictitious magnetic flux on each triangle of the kagom lattice is given by @xmath58 .
using the vectors @xmath59 , with @xmath60 , the momentum space hamiltonian is given by @xmath61 where @xmath62 @xmath63 and @xmath64 , @xmath65 @xmath66 where @xmath67 is a @xmath68 identity matrix , @xmath69 , @xmath70 , @xmath71 . at the saturation field @xmath72 , we obtain ferromagnetically coupled layers applicable to cu(1 - 3 , bdc)@xcite assuming strong interlayer coupling . at two values of magnetic field . @xmath73 .
, width=288 ] at two values of magnetic field . @xmath73 . , width=288 ] plotted in figs .
[ band ] and [ band1 ] are the magnon bands along the brillouin zone paths in fig .
[ neel](c ) , with the parameter values of ca@xmath10cr@xmath11o@xmath12 , @xmath74 @xcite and the dm value @xmath75 . for @xmath34 ,
the dmi does not contribute at zero field because the spins are along the @xmath25-@xmath26 kagom plane .
the resulting magnon bands are doubly degenerate between @xmath76 . at finite magnetic field increases ,
each spin has a component along the @xmath37-axis , hence the degeneracy of the bands between @xmath77 is lifted and the effects of the dmi results in a gap opening at @xmath78 .
at the saturation point @xmath79 ( not shown ) each layer is fully polarized along the field @xmath37-direction , again the dmi leads to gapped non - degenerate magnon bands . for the in - plane dmi @xmath35
the situation is different .
the degeneracy persists at zero field but since the spins are along the @xmath80 plane the dmi leads to gap magnon bands .
the degeneracy is always lifted at nonzero magnetic field , but in this case the there is a staggered flux emanating from both layers and the bands cross each other at @xmath81 . at the saturation point @xmath79 the spin are fully polarized along the @xmath37-axis and the in - plane dmi does not contribute and the non - degenerate magnon bands are gapless ( not shown ) . the zero field coplanar @xmath82 nel order on the kagom lattice corresponding to the @xmath5 ground state of kagom antiferromagnets .
solid lines connect nn sites and dash lines connect nnn sites .
@xmath83 field - induced non - coplanar out - of - plane spin canting with nonzero spin scalar chirality @xmath84 , where @xmath85 is the field - induced fictitious flux . , width=288 ] in the previous section , we studied one of the possible field - induced magnetically ordered phases in geometrically frustrated bilayer kagom magnets . in this section
, we study another magnetically ordered phase which has been realized experimentally in various frustrated kagom magnets . without loss of generality , we focus on the ideal kagom material kfe@xmath1(oh)@xmath2(so@xmath3)@xmath4 @xcite . the hamiltonian is given by @xmath86 where @xmath87 and @xmath88 are the isotropic antiferromagnetic couplings for nearest - neighbour ( nn ) and next - nearest - neighbour ( nnn ) sites respectively , @xmath89 , where @xmath90 denotes the directions of the out - of - plane dmi in the up / down triangles of the kagom lattice , and @xmath21 is the out - of - plane magnetic field in units of @xmath22 . for the iron jarosite kfe@xmath1(oh)@xmath2(so@xmath3)@xmath4 the interlayer exchange interaction can be neglected for two reasons .
first , it is very small compare to @xmath91 and @xmath92 .
second , a single crystal of iron jarosite can be synthesized @xcite .
we have also neglected the in - plane dm vector because it neither stabilizes magnetic ordering nor induces topological effects .
therefore , it can not change the results obtain here .
( oh)@xmath2(so@xmath3)@xmath4 with @xmath93 at two values of magnetic field .
@xmath94.,width=288 ] .,width=288 ] the alternating out - of - plane dmi between the up and down triangles of the kagom lattice is fictitious in that only one ground state is selected for each sign with positive chirality @xmath95 or negative chirality @xmath96 . at zero magnetic field , the out - of - plane dmi induces a coplanar 120@xmath97 nel order on the @xmath25-@xmath26 kagom plane with positive chirality @xcite as shown in fig .
[ neel](a ) . for nonzero out - of - plane field , inelastic neutron scattering experiment on the ideal kagom material kfe@xmath1(oh)@xmath2(so@xmath3)@xmath4 @xcite has uncovered a non - coplanar spin texture with nonzero spin chirality @xmath98 @xcite as shown in fig .
[ neel](b ) .
however , the topological effects of this non - collinear system have not been studied both theoretically and experimentally . in a recent letter @xcite
, we showed for the first time that this material possesses a finite thermal hall conductivity in this non - collinear regime .
thus , the iron jarosite kfe@xmath1(oh)@xmath2(so@xmath3)@xmath4 is a possible candidate for investigating thermal hall effect of magnon , which is accessible by using inelastic neutron scattering . in the present study
, we analyze the finite thermal hall conductivity in terms of topological magnon edge states .
we consider the @xmath5 ground state with positive chirality @xmath99 and @xmath100 .
the basic procedure is similar to the bilayer system studied above .
the rotation matrix is same as eq .
[ rot ] , however , with the oriented angles @xmath101 .
the classical energy is given by @xmath102 where @xmath103 , @xmath104 , and the magnetic field is rescaled in units of @xmath32 .
minimizing @xmath105 yields the canting angle @xmath106 , with @xmath107 .
we see that the dmi depends on the classical energy as it contributes to the stability of the @xmath5 ground state .
besides , @xmath88 can also stabilize the coplanar magnetic structure in the absence of the dmi .
the topological excitations above the classical ground state is mediated by a field - induced scalar chirality @xmath108 defined as the solid angle subtended by three neighbouring spins .
as mentioned previously , the scalar spin chirality originates from non - coplanar spin texture and does not need the presence of dmi or magnetic ordering .
it is also the basis of chiral spin liquid states which suggests that topological effects may persist in the spin liquid regime of frustrated kagom magnets .
this model does not have an analogy to collinear ferromagnets @xcite or bilayer collinear in model i. it also differs significantly from field - induced topological magnons in bipartite frustrated honeycomb lattice because @xmath109 @xcite .
in fact , the bipartite honeycomb antiferromagnets fall into the class of model i as they are doubly degenerate at zero magnetic field and require an explicit dmi .
the magnon tight binding hamiltonian for model ii is given by @xmath110 where @xmath111 and the coefficients are given by @xmath112 with @xmath113 .
@xmath114 where @xmath115 and @xmath116 .
the fictitious magnetic fluxes are given by @xmath117 , and @xmath118 , where @xmath119 using the vector notation ^_= ( b_a^,b_b^,b_c^ , b_-a , b_-b , b_-c ) , the hamiltonian can be written as h_sw=_0 + s _ ^ _
( ) _ , where @xmath120 and @xmath121 is a constant .
@xmath122 is an identity @xmath123 matrix and @xmath124 is a pauli matrix .
@xmath125 , the same for @xmath126 .
the eigenvalues of this hamiltonian can not be obtained analytically as opposed to the zero field case , @xmath127 , with coplanar 120@xmath97 nel order on the @xmath25-@xmath26 kagom plane .
it is important to note that both fluxes do not vanish at zero dmi .
this means that the dmi does not provide topological effects in stark contrast to ferromagnets @xcite . as shown in fig .
[ band2 ] the magnon bands are not topological at zero magnetic field even in the presence of dmi .
the final model we shall consider is the xxz kagom antiferromagnet without dmi subject to an out - of - plane magnetic field . the hamiltonian is governed by @xmath128 where @xmath129 and @xmath130 is the easy - plane anisotropy , and @xmath21 is the magnetic field in units of @xmath22 . at zero field ,
the easy - plane anisotropy favours the positive chirality @xmath5 ground state @xcite .
the canting angle is determined from the classical energy @xmath131-h\cos\chi,\end{aligned}\ ] ] where @xmath106 and @xmath132 . in this system ,
topological magnon bands originate from @xmath133 as in model ii .
the fictitious magnetic flux encountered by propagating magnon is given by @xmath134 where @xmath135 , and @xmath136,\\&\delta_m=-\frac{\sqrt{3 } } { 2}\cos\chi,~ \delta^\prime= \frac{\sin^2\chi}{2}\lb\frac{3}{2}-\delta\rb .
\end{aligned}\ ] ] in fig .
[ band3 ] we show the magnon bands for @xmath137 . at zero field ,
the magnon bands are very similar to model ii .
we see that the flat mode acquires a small dispersion at nonzero field and the magnon bands are gapped at all points in the brillouin zone . as mentioned in the text
, the dmi does not provide any topological effects for the noncollinear @xmath5 spin configuration on the kagom lattice .
this means that topological effects persist at zero dmi and nonzero out - of - plane magnetic field via an induced spin scalar chirality @xmath138 , where @xmath139 with @xmath140 . however , dmi is usually present on the kagom lattice due to lack of inversion center . in this section ,
we show that anomalous magnon hall effect is present at zero dmi in contrast to collinear ferromagnets .
figures [ band4 ] and [ n_thd ] show the magnon bands and the corresponding @xmath6 respectively for zero dmi .
kagom antiferromagnet with zero dmi at three field values .
the parameters are @xmath141.,width=360 ] for the bands in fig .
[ band4 ] at three field values.,width=288 ]
the hamiltonians for insulating antiferromagnets are diagonalized by the generalized bogoliubov transformation @xmath142 , where @xmath143 is a @xmath144 paraunitary matrix and @xmath145 with @xmath146 being the quasiparticle operators .
the matrix @xmath143 satisfies the relations , @xmath147 where @xmath148 , @xmath149 , and @xmath150 are the energy eigenvalues . from eq .
[ eqna ] we get @xmath151 , and eq . [ eqn1 ] is equivalent to saying that we need to diagonalize the hamiltonian @xmath152 whose eigenvalues are given by @xmath153 and the columns of @xmath154 are the corresponding eigenvectors .
the paraunitary operator defines a berry curvature given by @xmath155_{\alpha\alpha } , \label{bc1 } \end{aligned}\ ] ] with @xmath156 and @xmath157 are the columns of @xmath158 .
this form of the berry curvature simply extracts the diagonal components which are the most important .
suppose the explicit analytical form of @xmath157 is known as in honeycomb - lattice hardcore bosons @xcite , the berry curvature can be computed directly from eq .
[ bc1 ] . unfortunately
, this is not the case in the present models . from eq .
[ eqn1 ] the berry curvature can be written alternatively as @xmath159}{\lb\epsilon_{\bo\alpha}-\epsilon_{\bo\alpha^\prime}\rb^2 } , \label{chern2}\end{aligned}\ ] ] where @xmath160 defines the velocity operators .
the present form can be computed once the eigenvalues and eigenvectors of the hamiltonian are obtained numerically .
similar to fermionic systems , the chern number can still be defined for bosonic systems as the integration of the berry curvature over the first brillouin zone , @xmath161 indeed , topologically protected magnon edge states are characterized by nonzero chern numbers
. however , the chern numbers are fictitious because the notion of completely filled bands and fermi energy do not apply to bosonic ( magnonic ) systems . .
the parameters are @xmath73.,width=288 ] .
the parameters are @xmath73.,width=288 ] ( oh)@xmath2(so@xmath3)@xmath4 @xmath162 . ] . ] experimentally accessible magnon hall effect has been previously analyzed in ref .
now , we complete this study by providing evidence of topological magnon insulator with chern number protected magnon edge modes .
we have solved for the magnon edge modes using a strip geometry with open boundary conditions along the @xmath26 direction and infinite along @xmath25 direction @xcite .
first , let us consider model i with @xmath34 shown in fig .
[ mo1 ] . in this case
the magnon bulk bands are degenerate between @xmath163 sectors at @xmath164 and the dmi vanishes in the noninteracting limit as the spins are along the @xmath25-@xmath26 kagom plane at zero field .
this results in gapless magnon bulk bands at @xmath165 and @xmath166 with a single edge mode connecting these points and @xmath6 vanishes as expected . for @xmath167 the spins
are non - collinear and the degeneracy between @xmath163 is lifted because each spin has a component along the @xmath37-axis .
the dmi opens a small gap between the bands .
we see that pairs of gapless magnon edge states appear in the vicinity of bulk gap signifying the strong topology of the system , yielding nonzero @xmath6 @xcite . at the saturation field
@xmath79 the spins are collinear along the @xmath37-axis corresponding to bilayer ferromagnet coupled ferromagnetically .
the dmi again leads to gap magnon bulk bands with counter - propagating gapless edge states and again with nonzero @xmath6 .
for the in - plane dmi @xmath168 shown in fig .
[ mo2 ] , the situation is different .
there is no magnon hall effect and @xmath6 vanishes in all regimes @xcite , but there is topological magnon insulator as we now explain .
indeed , we have degenerate magnon bulk bands between @xmath163 sectors at zero field , but the dmi has a profound effect since the spins are along the @xmath25-@xmath26 kagom plane . as shown in fig . [ mo2](a ) there is a pair of edge state modes for each spin sector , and they are related by time - reversal symmetry .
this is an evidence of topological magnon insulator .
however , @xmath6 vanishes as a consequence of time - reversal symmetry between the degenerate spin sectors .
in fact , this system is a magnonic counterpart to fermionic topological insulator with imaginary second - nearest - neighbour soc between electron spin up and down @xcite . for @xmath167
the bands cross at @xmath169 as shown above due to staggered flux configurations and the edge modes are not topologically protected as we have confirmed by computing the berry curvatures and the chern numbers . for this reason
@xmath6 again vanishes . at the saturation field
@xmath79 the in - plane
dmi disappears in the noninteracting limit and the system is topologically trivial with vanishing @xmath6 .
the key observation in the layer antiferromagnetic system is that although topologically protected edge states are present at zero magnetic magnetic field , @xmath6 is suppressed by antiferromagnetic coupling .
now , let us consider model ii which corresponds to the kagom material kfe@xmath1(oh)@xmath2(so@xmath3)@xmath4 @xcite .
as shown above this model differs significantly from model i due to the presence of spin scalar chirality which survives in the absence of dmi and magnetic ordering .
the associated magnon edge modes are depicted in fig .
[ mo3 ] . at zero field @xmath170
it is evident that there are no protected chiral edge modes .
this shows that the system is topologically trivial and @xmath6 vanishes for any values of dmi . in the presence of a magnetic field perpendicular to the kagom plane
there is an induced noncoplanar spin texture which provides spin scalar chirality @xcite .
figures [ mo3](b ) and ( c ) show that the system is topologically nontrivial in this regime with protected gapless edge states which yield nonzero chern number and finite @xmath6 even without the presence of dmi @xcite .
indeed , the presence of spin scalar chirality is the basis of chiral spin liquid physics , therefore it will not be surprising that the nontrivial topology of this system persists in the spin liquid phase of the frustrated kagom magnets .
model iii explicitly ignores the dmi and the easy - plane anisotropy stabilizes the @xmath5 magnetic ordering .
this system is analogous to model ii as shown in fig .
this is because the presence of the dmi does not have any topological effects in frustrated kagom lattice unlike insulating kagom ferromagnets .
the layer antiferromagnetic system and the frustrated system have similarities and differences . in both systems
@xmath6 vanishes at zero field which can be attributed to zero net magnetic moment . in other words ,
the degeneracy at zero field between @xmath163 sectors in layer antiferromagnetic system yields a zero net magnetic moment and for the coplanar / non - collinear system at zero field we have @xmath171 on each triangular plaquette which also yields a zero net magnetic moment .
however , the origin of finite @xmath6 is different in both systems . for the layer antiferromagnets ,
topological magnon bands is induced by the dmi , whereas in the frustrated system with coplanar / non - collinear ordering the concept of topological magnon bands originates from field - induced spin scalar chirality which is nonzero even in the absence of dmi and magnetic ordering @xmath172 .
it is natural to ask the importance of this investigation and whether such nontrivial topological effects can be experimentally realized in insulating antiferromagnets .
recently , topological magnon insulator has been realized in the kagom ferromagnet cu(1 - 3 , bdc ) @xcite .
this material is also the first kagom ferromagnet that shows magnon hall effect with finite @xmath6 @xcite .
a recent experiment has reported a finite @xmath6 in frustrated kagom volborthite cu@xmath1v@xmath4o@xmath11(oh)@xmath4 @xmath142h@xmath4o in the presence of an out - of - plane magnetic field @xmath173 @xcite .
this result is attributed to spin excitations in the spin liquid regime . as mentioned previously ,
the kagom volborthite is known to exhibit different magnetic - field - induced ordered phases for @xmath174 @xcite , and the frustrated kagom compound ca@xmath10cr@xmath11o@xmath12 @xcite also exhibits ferromagnetic ordered states for @xmath175 .
this suggests that the observed low temperature dependence of @xmath6 in kagom volborthite might not be due to spin excitations in the spin liquid regime , but magnon excitations in the field - induced ordered phases .
the iron jarosite kfe@xmath1(oh)@xmath2(so@xmath3)@xmath4 @xcite is an ideal kagom antiferromagnet with @xmath5 ground state and nonzero field - induced spin scalar chirality @xcite .
this is a perfect material to search for topologically nontrivial excitations with finite @xmath6 as described in the present study and ref .
@xcite . at the moment
, inelastic neutron scattering experiment has not figured out how to measure magnon edge state modes , which are consequences of the magnon bulk topology that gives rise to finite @xmath6 because it is a bulk sensitive method .
the magnon edge modes can be probed by edge sensitive methods such as light @xcite or electronic @xcite scattering method .
this is not an impossible task in principle , and we believe it will be measured in the near future .
research at perimeter institute is supported by the government of canada through industry canada and by the province of ontario through the ministry of research and innovation .
y. onose , t. ideue , h. katsura , y. shiomi , n. nagaosa , y. tokura , science * 329 * , 297 ( 2010 ) .
r. matsumoto and s. murakami , , 197202 ( 2011 ) ; , 184406 ( 2011 ) .
lifa zhang , jie ren , jian - sheng wang , and baowen li , phys .
b * 87 * , 144101 ( 2013 ) .
a. mook , j. henk , and i. mertig , , 024412 ( 2014 ) ; , 134409 ( 2014 ) .
max hirschberger , robin chisnell , young s. lee , n. p. ong , , 106603 ( 2015 ) .
h. lee , j. h. han , and p. a. lee , , 125413 ( 2015 ) .
s. a. owerre , j. phys . : condens .
matter 28 , 386001 ( 2016 ) .
s. a. owerre , j. appl . phys . * 120 * , 043903 ( 2016 ) . s. a. owerre , j. phys . : condens .
matter * 28 * , 436003 ( 2016 ) .
li , y .- d .
li , y .- b .
kim , l. balents , y. yu , and g. chen , nat . commun .
* 7 * , 12691 ( 2016 ) .
mook , a. , henk , j. and mertig , phys .
lett . * 117 * , 157204 ( 2016 ) .
ying su , x. s. wang , x. r. wang , arxiv:1609.01500 ( 2016 ) .
s. a. owerre , arxiv:1608.00545 ( 2016 ) . c. kane and t. c. lubensky , nat
10 * , 39 ( 2014 ) .
b. g. chen , n. upadhyaya , and v. vitelli , proc .
* 111 * , 13004 ( 2014 ) .
j. paulose , b. g. chen , and v. vitelli , nat .
phys . * 11 * , 153 ( 2015 ) .
j. paulose , a. s. meeussen , and v. vitelli , proc .
* 112 * , 7639 ( 2015 ) .
d. zeb rocklin , b. gin - ge chen , m. falk , v. vitelli , and t. c. lubensky , phys .
lett . * 116 * , 135503 ( 2016 ) .
olaf stenull , c. l. kane , and t. c. lubensky , phys .
* 117 * , 068001 ( 2016 ) .
i. dzyaloshinsky , j. phys .
solids * 4 * , 241 ( 1958 ) ; t. moriya , phys . rev . * 120 * , 91 ( 1960 ) .
guo and m. franz , phys .
b * 80 * , 113102 ( 2009 ) .
m. z. hasan and c. l. kane , rev .
82 * , 3045 ( 2010 ) .
qi and s .- c .
zhang , rev .
83 * , 1057 ( 2011 ) .
m. elhajal , b. canals , and c. lacroix , phys .
b * 66 * , 014422 ( 2002 ) .
o. cpas , c. m. fong , p. w. leung , and c. lhuillier phys . rev .
b * 78 * , 140405(r ) ( 2008 ) .
d. grohol , k. matan , j.h .
cho , s .- h .
lee , j.w .
lynn , d.g .
nocera , y.s .
lee , nature materials * 4 * , 323 ( 2005 ) .
k. matan , d. grohol , d. g. nocera , t. yildirim , a. b. harris , s. h. lee , s. e. nagler , and y. s. lee , phys .
* 96 * , 247201 ( 2006 ) .
y. taguchi , y. oohara , h. yoshizawa , n. nagaosa , y. tokura , science * 291 * , 2573 ( 2001 ) .
y. machida , s. nakatsuji , y. maeno , t. tayama , t. sakakibara , and s. onoda , phys .
rev . lett . * 98 * , 057203 ( 2007 ) .
y. machida , s. nakatsuji , s. onoda , t. tayama , and t. sakakibara , nature , * 463 * , 210 ( 2008 ) . c. srgers , g. fischer , p. winkel and h. v. lhneysen , nature commun .
* 5 * , 3400 ( 2014 ) .
jian zhou , qi - feng liang , hongming weng , y. b. chen , shu - hua yao , yan - feng chen , jinming dong , and guang - yu guo , phys .
116 , 256601 ( 2016 ) .
s. fujimoto , phys .
lett . * 103 * , 047203 ( 2009 ) . c. balz , b. lake , j. reuther , h. luetkens , r. schnemann , th .
herrmannsdrfer , y. singh , a. t. m. nazmul islam , e. m. wheeler , j. a. rodriguez - rivera , t. guidi , g. g. simeoni , c. baine , and h. ryll , arxiv:1606.06463 ; nature physics ( 2016 ) .
m. yoshida , m. takigawa , h. yoshida , y. okamoto , and z. hiroi , phys .
lett . * 103 * , 077207 ( 2009 ) .
m. yoshida , m. takigawa , s. kramer , s. mukhopadhyay , m. horvatic , c. berthier , h. yoshida , y. okamoto , z. hiroi , j. phys .
soc . jpn .
* 81 * , 024703 ( 2012 ) .
t. yildirim and a. b. harris , phys .
b * 73 * , 214446 ( 2006 ) .
a. l. chernyshev , phys .
b * 92 * , 094409 ( 2015 ) .
d. watanabe , k. sugii , m. shimozawa , y. suzuki , t. yajima , h. ishikawa , z. hiroi , t. shibauchi , y. matsuda , m. yamashita , proc .
usa * 113 * , 8653 ( 2016 ) . | topological magnon insulators in insulating kagom ferromagnets have been extensively studied in a series of papers .
it has been established that dzyaloshinskii - moriya interaction ( dmi ) is the key ingredient to observe a nontrivial topological magnon with edge modes . however ,
insulating antiferromagnets on the kagom lattice are frustrated systems considered as a playground for studying quantum spin liquid physics . in these systems the dmi can induce a coplanar but noncollinear magnetic orders with a @xmath0 propagating wavevector .
we show that topological magnon bands are absent in this coplanar spin texture in sharp contrast to collinear ferromagnets with dmi .
hence , geometrically frustrated kagom antiferromagnets can be deemed topologically trivial .
the presence of an out - of - plane magnetic field in these frustrated magnets induces noncoplanar spin textures exhibiting a nonzero spin scalar chirality .
we show that the field - induced spin chirality provides topological magnon bands in kagom antiferromagnets without the need of dmi and survives in the chiral spin liquid phase of frustrated magnets .
possible experimentally material includes iron jarosite kfe@xmath1(oh)@xmath2(so@xmath3)@xmath4 . |
the idea of an energy cascade through a hierarchy of scales ( richardson 1922 ; kolmogorov 1941a ; obukhov 1941 ) , the phenomenological theory of turbulent energy dissipation ( kolmogorov 1941a , b , 1942 ) , and the experimental and theoretical results related to the intermittency of turbulent velocity fields ( see frisch 1995 for references ) are the cornerstones of the present understanding of incompressible turbulence .
much less is known about compressible ( supersonic ) turbulence .
analytical approaches and laboratory experiments are still limited to low mach numbers , m@xmath3 , and the results of numerical simulations may depend crucially on the adopted boundary and initial conditions .
von weizscker ( 1951 ) was the first to point out that the interstellar medium ( ism ) is characterized by very high reynolds numbers and high mach numbers , and thus highly supersonic turbulence should be its typical state .
much work using the ism to study supersonic turbulence has since been attempted ( see , for example , dickman 1985 ; scalo 1987 ; falgarone & phillips 1990 ; elmegreen 1993 ; franco & carraminana 1999 ) . however , at the relatively large scales considered so far , @xmath4 cm , the ism may not be an adequate laboratory for studying supersonic turbulence in the traditional sense of the term `` turbulence . ''
the initially hypothesized turbulent energy cascade from the largest to the smallest scales of the ism , initiated by the differential rotation of the galaxy ( von weizscker 1951 ; fleck 1981 ) is an obvious idealization . even if the differential rotation works as the major energy source for the observed turbulence in the neutral gas at the perifery of galaxies ( sellwood , & balbus 1999 ) , it should be disrupted by the powerfull injection of energy from supernovae and stellar winds at intermediate scales in the major galactic disk ( spitzer 1978 ) .
ism turbulence at large and intermediate scales is also complicated by the effects of self - gravitation .
it has recently been recognized that the 1.35-cm wavelength h@xmath0o masers [ see reviews on masers in elitzur ( 1992 ) ; clegg and nedoluha ( 1993 ) ; migenes and reid ( 2002 ) ] may be promising tools for the study of `` kolmogorov - type '' supersonic turbulence .
vlbi studies of the proper motions of several bright h@xmath0o maser sources associated with newly - born stars have revealed expansion of the clusters of maser spots participation in gas outflows from these stars ( see anderson and genzel 1993 for a review ) .
this type of regular motion had been theoretically predicted before its discovery ( strelnitski & sunyaev 1973 ) . in some cases , including w49n , there are also indications of another regular component in the velocity field revealed by the masers rotation ( reid _ etal .
_ 1988 , gwinn _ etal . _ 1992 ) .
however , besides these regular components , vlbi measurements indicated the presence of a residual _ random _ component of motion .
typically , approximating the proper motion vectors by a simple model of expanding and rotating gas leaves a residual dispersion of @xmath5kms@xmath6 per axis , which is considerably larger than the errors of these observations ( reid _ etal .
_ 1988 ) .
this value corresponds to @xmath7kms@xmath6 for the total velocity vector and can be considered the characteristic turbulent velocity dispersion at the largest spatial scale covered by the maser cluster . walker ( 1984 ) , using the vlbi maps of the h@xmath0o source in w49n obtained by walker , matsakis & garcia - barreto ( 1982 ) , demonstrated that both the two - point velocity increments and the two - point spatial correlation function show power - law dependencies on maser spot separation .
this behavior is typical of a turbulent flow , although walker did not favor the turbulence interpretation .
gwinn ( 1994 ) carried out a similar statistical analysis using better vlbi results for w49n obtained by gwinn _ etal . _ ( 1992 ) .
he confirmed the power - law dependency of velocity dispersion and spatial density of masing spots on spatial scale and ascribed this behavior to turbulence .
this paper summarizes a series of our studies of h@xmath0o masers as tracers of supersonic turbulence in regions of star formation .
preliminary results were reported by strelnintski _ etal . _ ( 1998 ) ,
holder & strelnitski ( 1997 ) , gezari ( 1997 ) and gezari , reid , & strelnitski ( 1998 ) .
section 2 presents our vlbi results for the h@xmath0o maser source in sgr b2(m ) .
the geometrical properties of supersonic turbulence revealed by h@xmath0o masers in this and four other sources are presented in section 3 , and the statistical properties of the velocity field traced by the masers in section 4 . in section 5 , proceeding from the hypothesis that the h@xmath0o masers adequately probe the velocity field of turbulence , we discuss implications of our statistical results for the theory of supersonic turbulence . in section 6
we describe a new conceptual model of h@xmath0o masers in regions of star formation based on the surmised connection of maser pumping with the sites of ultimate dissipation of turbulent energy .
conclusions are summarized in section 7 .
the observations of sgr b2(m ) were conducted in 1986 on 23 january , 26 february , 27 march , and 26 april , as part of a campaign to measure proper motions of these h@xmath0o masers .
four telescopes spanning the u.s . were used : the haystack 37-m telescope in westford , ma , the nrao 43-m telescope in green bank , wv , one 25-m telescope of the vla near socorro , nm , and the ovro 40-m telescope in big pine , ca . the mk iii recording system was used with four 2-mhz bands covering the lsr velocity ranges of -40 to -14 , + 18 to + 43 , + 43 to + 69 , and + 68 to + 94 km s@xmath6 , assuming a rest frequency of 22235.08 mhz for the h@xmath0o @xmath8 transition .
the recorded data were correlated at the mk iii processor at haystack observatory in a mode which yielded 56 ( uniformly weighted ) spectral channels , each 35.71 khz or 0.48 km s@xmath6 wide .
the data were edited , calibrated , and imaged following the same general procedures as described in reid _ etal . _ ( 1988 ) for the source sgr b2(n ) .
the synthesized interferometer beam was approximately 2.4 by 0.4 mas fwhm , elongated in the north - south direction , owing to the low declination of sgr b2 .
compact maser spots with flux densities ranging between 135 and 0.4 jy were detected across a field of approximately 2 by 2 arcseconds .
the positions of the maser spots were obtained by fitting a circular gaussian brightness model for each spectral channel independently using the aips task imfit .
figure 1 shows a series of decreasing spatial scales for our vlbi map of sgr b2(m ) .
the scale changes by almost four orders of magnitude from @xmath9arcsec down to @xmath10mas which , at the assumed distance of 8 kpc ( reid 1993 ) , corresponds to a range of linear scales between , roughly , 10,000 and 1 a.u . as is typical of fractal dust - like structures , the distribution of masers looks qualitatively the same on all scales , with evident clustering . on all maps of figure 1 , except the last one , dot sizes are larger than the typical observed size of an individual maser spot .
dots on the last map show measured positions of spectral channels . since a spectral channel ( @xmath11kms@xmath6 ) is narrower than a typical spectral width of a single spot ( @xmath12kms@xmath6 ) , a dot on the last map typically represents the position and velocity of only a part of an individual spot .
we call the smallest groupings of heavily blended individual spots `` minimal clusters . ''
the dots on the map only approximate the extension of and the velocity dispersion in these minimal clusters .
inspection of the available data indicates that a typical size of a minimal cluster is a few a.u .
the velocity dispersion within minimal clusters varies from @xmath13 to 5 kms@xmath6 .
we used two methods to estimate the fractal dimension of the spatial distribution of h@xmath0o masers in sgrb2(m ) : the `` density - radius '' and the `` box - counting '' methods ( see e.g. crownover 1995 ; feder 1988 ) .
the density - radius measure is based on the generalization of the mass _ versus _ radius relation for objects of integer dimension , @xmath14 where @xmath1 is the dimension of the object .
equation ( 1 ) can be used as a general definition of the dimension of an object , including objects whose average density changes in a self - similar way with changing scale ( e.g. mandelbrot 1982 ) . for these objects ( fractals )
@xmath1 is non - integer .
average density , @xmath15 , within a given volume @xmath16 is @xmath17 .
therefore , @xmath18 where @xmath19 is the dimension of the supporting space , e.g. @xmath20 for a plane .
if the dimension of the object equals the dimension of the support , @xmath15 equals a constant . if not
, @xmath15 is a function of @xmath21 .
the steepness of this function depends on @xmath1 . a more practical procedure results from differentiating equation ( 1 ) , which gives @xmath22 or @xmath23 determining the dimension
is thus reduced to measuring the slope of @xmath24 _ versus _ log@xmath25 . in the case of a point set like ours ,
`` density '' means `` number density . '' in our numerical procedure , @xmath26 is calculated ( for a discrete set of @xmath21 values ) as the surface density of companions to a given maser hot spot at angular separation @xmath21 , averaged over all maser spots : @xmath27}}\;.\ ] ] the same procedure was used by walker ( 1984 ) and gwinn ( 1994 ) for demonstrating self - similar clustering of h@xmath0o masers in w49n .
however , these authors did not relate @xmath28 to the fractal dimension of the source .
larson ( 1995 ) uses a similar procedure to obtain a fractal dimension for a young stellar association in taurus .
we tested our numerical procedure by obtaining the density - radius fractal measure of a simple straight line and of the classical mathematical fractal , a sierpinski triangle .
the measured dimensions of 1.00 and 1.58 were in excellent agreement with their theoretical values @xmath29 and @xmath30 , respectively .
the box - counting measure associates a fractal object s dimension @xmath1 with the number @xmath31 of boxes of side length @xmath32 needed to cover the object ( crownover 1995 ) : @xmath33 the graph of log@xmath34 _ versus _
log@xmath35 is a straight line , having slope @xmath36 .
we used the following computational algorithm .
the square plane of minimal side length @xmath37 containing the whole object is divided into @xmath38 equal squares of side length @xmath39 . the number of these squares containing one or more points making up the object is determined and stored .
each non - empty square of side length @xmath39 is subdivided again into 4 squares of equal area ; the number of non - empty squares of side length @xmath40 is determined and stored , etc .
the procedure is repeated down to some minimal side length of sub - squares ; minimal side length is determined by the characteristic length of the smallest features of the object .
logarithm of the number of non - empty squares _ versus _ logarithm of their side length is plotted , and the slope of the straight line fitting the data points is measured ; this slope is equal to @xmath36 .
this numerical procedure was also tested with a straight line and a sierpinski triangle .
the measured values of dimension were again in agreement with the theoretical values .
we used equation ( 4 ) to determine the fractal dimension of the 2d projected ( @xmath20 ) h@xmath0o masers in the four observations of sgrb2(m ) .
the values of the fractal dimension for the four observations and for their average are indicated in the corresponding panels of figure 2 .
the average value is @xmath41 .
the box - counting result for the average of the four observations is shown in figure 3 .
the ensuing fractal dimension , @xmath42 , is noticeably lower than that obtained from the density - radius plot . for both measures ,
a single linear fit is a satisfactory first approximation ; the standard deviation of the residuals to the fit does not surpass @xmath43 for each epoch of observations and for the combined fit .
however , notable deviations from a single power - law approximation can be seen in figures 2 and 3 .
for example , a higher fractal dimension for the largest scales is evidenced by the steeping slope of the points in figure 3 .
some deviation from the linear dependence is seen in figure 2 , between log ( separation ) @xmath44 and @xmath45 .
it may indicate some depression of clustering at the scales around 0.003 arcsec ( @xmath46 cm for this source ) .
the origin of the systematic difference between the two applied fractal measures is unclear .
it may be rooted in technical particularities of the methods .
some practical problems in the application of the box - counting method are discussed in gouyet ( 1996 , section 1.4.4 ) .
this discussion indicates that derived fractal dimensions should be correct within a factor of two .
given this uncertainty and the range of the derived values of slopes in figures 2 and 3 , we estimate the fractal dimension of the observed cluster as @xmath47 . gwinn ( 1994 ) performed a statistical analysis of the vlbi maps of w49/h@xmath0o obtained by gwinn _ etal . _ ( 1992 ) .
he demonstrated a power - law dependence of the number density of neighbors on their separation .
however , he used the one - dimensional projection of the distance and did not interpret his results in terms of fractals . to make results for w49n comparable to those for sgrb2(m )
, we applied the density - radius fractal measure to the two - dimensional spatial distribution of the h@xmath0o maser spots in w49n , using the vlbi positions published by gwinn _ etal . _ ( 1992 ) .
we have applied this measure to three more sources two with the published vlbi results : w51(main ) and w51n ( genzel et al .
1981 ; schneps et al . 1981 ) and one source , w3(oh ) , for which we used our unpublished vlbi coordinates of the maser spots (
the corresponding map of the source was published alcolea et al .
1992 ) .
the results for all sources are shown in figure 4 .
a power - law dependence is a good approximation for all of them and it gives a low fractal dimension , @xmath48 , for all the observed sets of maser spots as projected on the sky .
given the small angular dimensions of maser clusters ( @xmath49arcsec ) , their projection on the sky is essentially an orthogonal projection .
therefore , the dimension @xmath50 of the real fractal , residing in 3-space , coincides with the dimension , @xmath51 , of its 2d projection if @xmath52 ( e.g. falconer 1990 ) .
we have demonstrated above that this condition is fulfilled for h@xmath0o masers .
thus , we conclude that the fractal dimension of h@xmath0o clusters in all five of the sources we have studied is low , @xmath53 .
this conclusion is new , although walker ( 1984 ) and gwinn ( 1994 ) obtained results that can be converted to estimates of fractal dimension .
walker ( 1984 ) obtained a high negative value of the power index ( @xmath54 ) in the power - law approximation of the density - radius dependence for w49n / h@xmath0o .
this corresponds to a fractal dimension @xmath55 .
gwinn ( 1994 ) reduced his two - point correlation analysis of h@xmath0o masers in w49n to the 1d projection of the observed map on the @xmath56 axis .
he obtained a power index , @xmath57 , for a large interval of scales .
although he did not connect this result with a fractal dimension , we note that for a 1d - projected fractal , @xmath58 , where @xmath59 is the fractal dimension of the 1d projection of the real fractal residing in 3-space .
thus , gwinn s result corresponds to @xmath60 . since @xmath61
, the same fractal dimension is ascribed to both the 2d - projection and the real fractal residing in 3-space .
thus , both the walker s ( 1984 ) and the gwinn s ( 1994 ) results support our conclusion that the fractal dimension of h@xmath0o clusters is @xmath48 . to better appreciate the fractal distribution of h@xmath0o masers , it is instructive to compare it with models of homogeneously distributed random points .
in our numerical model we created 90 points randomly and uniformly distributed in a thin spherical shell and then projected this distribution onto a plane .
a drastic difference between the model distribution and the observed maser distribution can be seen visually ( a lack of clustering in the model distribution ) and is confirmed by the measured spatial dimension of the model point sets . as anticipated , both box - counting and density - radius methods gave @xmath62 for the random , homogeneous cluster of model dots , to be compared with @xmath63 for the observed clusters of maser spots .
other types of masers should also be tested for possible fractal structure .
at least some of them do not seem to have such structure .
for example , the oh masers associated with regions of star formation do not show self - similar spatial distribution , rather they demonstrate strong clustering on one scale , @xmath64 cm ( reid _ etal . _ 1980 ) .
these masers form just outside an expanding ultra compact hii region and would not be expected to have a turbulent structure of the same kind as the h@xmath0o masers that are due to the shear between a stellar wind and surrounding gas ( see section 6 ) .
we investigated two statistical properties of the velocity field traced by h@xmath0o masers in the same five sources : ( 1 ) the low - order two - point velocity structure functions , and ( 2 ) the probability distribution for the deviations of the two - point velocity increment from its mean value at different spatial scales .
most statistical studies of the kinematics and structure of interstellar gas using masers as probes have so far been limited to one velocity component ( the line - of - sight ) and two coordinates on the celestial sphere . owing to the smallness of maser sources (
a whole source is only @xmath65arcsec across ) , the two spherical celestial coordinates are , with high precision , approximated by rectangular cartesian coordinates .
we assume , as all previous authors implicitly did , that if a power - law scaling relation takes place for velocity vectors in 3d space , the same relation , with the same exponent , holds for the dependence of the line - of - sight component of velocity on projected distances .
this is a reasonable assumption if the velocity field is isotropic .
it is analogous to the well - known use of the longitudinal velocity component in incompressible turbulence studies ( see e.g. frisch 1995 )
. statistical analysis of the velocity field probed by h@xmath0o masers has previously been performed for w49n on two independent sets of data ( walker 1984 , gwinn 1994 ) .
we discuss here only the low - velocity h@xmath0o maser spots ( @xmath66kms@xmath6 from the systemic velocity ) , which are more likely connected with the `` kolmogorov - type '' supersonic turbulence than the high - velocity spots ( section 6 ) . walker ( 1984 ) did not see an explicit dependence of two - point velocity increments on point separation for low - velocity features , shown in his figure 8 .
one can interpret this graph as a power law with the exponenent @xmath67 .
gwinn ( 1994 ) found @xmath68 for the dependence of the median velocity differences on the 1d projection of the maser pair separation .
one can conclude from these two studies that the value of @xmath69 for the two - point correlation function in w49n / h@xmath0o does not surpass 1/3 . for each of the five h@xmath0o sources ( section 3 ) we calculated the structure functions @xmath70 for low values of the order of the function , @xmath71 . in equation ( 7 ) the vectors
* r * and * l * determine the positions of the two points in the plane of the sky , @xmath72 designates the line - of - sight velocity and @xmath32 the 2d projection of the linear distance between the points .
our calculation procedure is as follows .
the whole range of maser spot angular separations ( up to four orders of magnitude , in both sources ) is divided into @xmath31 bins , with logarithmically increasing bin size and thus logarithmically increasing separation between bin - centers . using the vlbi relative position data
, the procedure selects all pairs within a given separation bin and calculates one of the functions ( eq.[7 ] ) .
this procedure is repeated for every separation bin and the resultant averages of the powers of velocity differences are plotted as a function of separation in a log - log graph
. a least - squares fit of a straight line to the points on the graph is then performed to obtain a power - law exponent , @xmath69 , and its uncertainty ( one standard deviation ) .
figure 5 gives the results for sgr b2(m ) , for @xmath73 .
the figure shows the results for the four epochs of observations ( section 2 ) , as well as their average .
the data are satisfactorily approximated by a power law .
we obtain @xmath74 , when the whole set of data is used .
when data for the largest scales ( log @xmath75 ) are excluded ( to avoid possible edge effects ) , the value of @xmath69 changes insignificantly : @xmath76 .
the second and third order structure functions have the exponents of @xmath77 and @xmath78 , respectively .
we note that for all three structure functions the power - law exponents are close to their classic kolmogorov values , which are 1/3 , 2/3 and 1.0 for @xmath792 and 3 , respectively .
the results for @xmath73 for other sources are presented in figure 6 .
the power - law approximation gives the values of @xmath69 close to 1/3 for all the sources except for w3(oh ) , where it is significantly lower ( @xmath80 ) . in this source
, however , the vlbi map reveals a strong regular component of motion ( strongly collimated bipolar outflow ) , which should significantly influence the results when the whole vlbi map is considered for the statistical analysis . to decrease the influence of the regular velocity component
, we obtained the first order structure function for only one of the two lobes of the bipolar outflow . in this case , the regular component of the relative velocities should be minimal , and one can anticipate that the bulk of the relative motions of the condensations will be due to turbulence . the result is shown in figure 7 ; the value of @xmath69 ( 0.30 ) is now much closer to the kolmogorov value . with all sources displaying the low - order structure functions close to kolmogorov s
, one might wonder how likely it is that this is simply coincidental .
can regular , non turbulent velocity fields , such as expansion and/or rotation , produce the observed power - law dependence of the velocity increments on spatial scales , with the power index close to 1/3 ? in order to answer this question
we applied the same statistical analysis to the results of numerical modeling that simulated regular motions of the maser spots only .
90 model dots were randomly and uniformly distributed in a thin spherical shell .
three types of regular motion were considered : ( 1 ) radial expansion ; ( 2 ) rotation around an axis perpendicular to the line of sight ; and ( 3 ) expansion plus rotation . in the last case we varied the ratio of the absolute values of the expansion velocity and the velocity of rotation on the equator .
the results of the ( line - of - sight velocity ) @xmath81 ( dot separation ) correlation analysis are shown in figure 8 .
the long straight line on the plots shows the slope 1/3 for reference .
the quickly growing dispersion of the data points at smaller scales , seen on all the plots , is due to the uniform , non - fractal distribution of the points a lack of clustering , and thus a poor statistics , at smaller scales .
this large dispersion makes a linear fit beyond about 1.5 orders of magnitude from the largest scale meaningless .
the slope of the fitting line , drawn in this limited interval , changes from about zero for the case of pure expansion ( @xmath82 , for the specific realization shown in figure 8) to approximately unity for the case of pure rotation ( @xmath83 in the example shown ) .
all intermediate values of the slope can be achieved by combining expansion and rotation in a due proportion ( the four lower panels ) . in particular , the `` kolmogorov '' value
1/3 is achieved when the ratio of the velocity of rotation on the equator to the velocity of expansion is @xmath84 .
it is quite improbable , however , that this _ ad hoc _ combination of kinematic parameters , plus the same orientation of the axis of rotation is realized in all the sources under study . in all the published models of the observed proper motions and radial velocities of h@xmath0o masers
the deduced model ratio of the expansion - to - rotation velocities is less than one . in w49n
these velocities are almost equal 17 and 16 kms@xmath6 , respectively ( we consider here only the low - velocity componenent of expansion ; see section 6 ) .
it is seen from figure 8 that such a low ratio of expansion - to - rotation should produce a flatter slope than 1/3 at large scales .
it is noteworthy that flattening of the slope of the two - point correlation function does indeed appear at larger scales for all the sources in figure 6 .
this reveals the contamination of the statistical properties of the turbulent component of motion by the regular component .
the role of the regular component(s ) of the velocity field relative to the turbulent component drops with the decreasing spatial scale , and it is remarkable how effectively the smaller scales `` compensate '' for the flattening at the larger scales in figure 6 and force the average slope to tend to its kolmogorov value .
an inherent manifestation of terrestrial turbulent flows is _ intermittency _ the spatial and temporal inhomogeneity of the velocity field .
intermittency results in enhanced , higher than gaussian , probability of large deviations of the two - point velocity increments from their average value at a given spatial scale .
deviations from a gaussian distribution have been observed in laboratory and atmospheric incompressible turbulent flows [ dutton and deaven ( 1969 ) , van atta and park ( 1972 ) ] .
falgarone and phillips ( 1990 ) and falgarone , phillips , and walker ( 1991 ) attributed the broad ( broader than gaussian ) wings of the emission line profiles observed in molecular clouds to an excess of large deviations from the average velocity difference in the cloud .
masers are more direct probes of the velocity field than thermal molecular lines observed in the cold clouds ( see section 6 ) .
typically , the available vlbi results provide coordinates and line - of - sight velocities for @xmath85 maser spots per observation .
thus , there are @xmath86 unique pairings for measuring the velocity - difference distribution . given this relatively large number
, we hoped that the statistics were sufficient to identify possible deviations from a gaussian distribution at various spatial scales .
figures 9 and 10 show examples of velocity - difference probability distributions obtained for particular spatial ranges in sgrb2(m ) .
the range for a given spatial scale was chosen to be equal to the scale .
unfortunately , some individual distributions were found to be not well - defined , centrally peaked distributions , which is evidently due to insufficient statistics . in order to produce a more statistically significant result
, we attempted to co - add the individual distributions .
the summation was done as follows .
the velocity difference ( @xmath56 axis ) and the number of pairs or counts ( @xmath87 axis ) of the individual distributions were independently normalized .
the @xmath87 axis was normalized by dividing all bin counts by the maximum bin count . in order to normalize the @xmath56 axis ,
we first considered how the absolute value of the maximum velocity difference of individual distributions scaled as a function of radial separation between maser pairs .
we found that this function was well approximated by a power - law .
the best fitting straight line gave the exponent @xmath88 .
given the previously established kolmogorov scaling law for the mean absolute velocity differences between maser points , this result could be anticipated if we assume that the dispersion of velocity differences at a given scale is proportional to the mean absolute velocity difference at that scale . using the resulting linear fit
, the dispersions of velocity differences for individual distributions were normalized by multiplying the @xmath56-values of a distribution by the ratio @xmath89 where @xmath90 is arbitrarily chosen to equal the maximum observed absolute value of the velocity difference occurring in the central spatial bin , and @xmath91 is the maximum absolute velocity difference value of an individual distribution .
once the @xmath56 and @xmath87 axes of the individual distributions were normalized , these distributions were co - added , producing a single histogram of the distribution averaged over spatial scales .
the conjoined histogram for sgrb2(m ) is shown in figure 10 .
it has a well - defined , centrally - symmetric shape .
however , fitting it with a single gaussian results in strong positive residuals in the wings of the distribution ( figure 10@xmath92 ) .
a two - gaussian fit , both gaussians being centered at zero , results in much smaller residuals ( figure 10@xmath93 ) . due to the method by which this histogram has been obtained , it contains only averaged ( over all the spatial scales ) information about the probabilities of deviations . comparing the conjoined histogram with the individual histograms shows that summation significantly improves the statistics .
this is indirect evidence that the velocity field at all or at most of the accessible spatial scales , has qualitatively similar statistical properties , including an excess of large deviations from gaussian distribution .
the two - gaussian fit of the conjoined distribution provides a quantitative measure of the excesses , averaged over the entire scale range .
the narrower gaussian approximately describes the central part of the distribution , and the broader one describes its wings .
the ratio `` narrow / broad '' of areas under these two gaussians measures the excess of large deviations .
> from figure 9@xmath93 , this ratio is @xmath94 , considerably less than unity .
thus , the super - gaussian wings are significant indeed .
figure 11 shows the results of the same analysis for four other h@xmath0o sources . in all of them , the wings of the distribution are much broader than those of the gaussian that fits the central part of the distribution .
this demonstrates that excess of large velocity - difference deviations is a common feature of the turbulent velocity fields probed by h@xmath0o masers .
in this section we discuss possible implications of our results for the theory of supersonic turbulence .
the statistical study of h@xmath0o masers described in the previous sections reveals three important results : ( 1 ) self - similarity ( fractality ) of the spatial distribution of the maser spots , ( 2 ) power - law character of the structure functions for the velocity field traced by the masers , with the power indices close to their classic kolmogorov values for incompressible turbulence , and ( 3 ) excess of large fluctuations of the two - point velocity increments .
as all these features are typical of the relatively well studied , incompressible turbulence , one can suspect that h@xmath0o masers in regions of star formation arise in a turbulent medium .
if this assumption is correct , h@xmath0o masers , due to their record brightness and small angular sizes , may become unique probes of astrophysical turbulence .
since the velocity increments at the largest scales in the h@xmath0o clusters are much greater than the probable speed of sound in neutral molecular gas , we deal here , by definition , with the poorly studied supersonic regime of turbulence .
comparison with the available theoretical knowledge about incompressible turbulence should therefore be done with high caution .
the major conclusion of our analysis in section 3 is that the spatial distribution of h@xmath0o masers is fractal and that the measured fractal dimension is low , @xmath63 .
if h@xmath0o masers trace the dissipation of supersonic turbulence , we should conclude that the fractal dimension of supersonic turbulence is considerably lower than that of incompressible turbulence , the latter being @xmath95 ( mandelbrot 1982 ) . if one accepts the hypothesis that mass fluctuations in star forming clouds are produced by supersonic turbulence ( larson 1995 ) , then the low fractal dimension ( @xmath96 ) of a cluster of young stars in taurus measured by larson can be considered as a corroboration of our conclusion .
another relevant fact may be the observed low fractal dimension of the large - scale distribution of galaxies ( @xmath97 ; mandelbrot 1982 ) , but the role of turbulence in shaping this structure is even less clear .
it is helpful to introduce the `` running '' filling factor , @xmath98 , of `` daughter '' turbulence elements within `` mother '' elements ( frisch 1995 ) .
the dimension of a fractal is expressed through @xmath98 by @xmath99 where @xmath19 is the dimension of the supporting space and @xmath100 is the scaling factor from the mother eddies to the daughter eddies .
constancy of @xmath98 from scale to scale guarantees a well - determined value of @xmath1 ( a linear plot in figurs 2 4 ) . from equation ( 9 )
@xmath101 which shows that the filling factor of active daughter eddies in mother eddies decreases when the fractal dimension of turbulence decreases .
the filling factor of active eddies is a direct measure of the degree of intermittency ( spottiness ) of turbulence .
we conclude that highly supersonic turbulence , as revealed by h@xmath0o masers and perhaps by the large - scale galaxy distribution and the distribution of stars in young clusters , is more intermittent than incompressible turbulence .
the very possibility of representing the observed spatial distribution of active turbulence elements by a single power law means that turbulence is intermittent on virtually all scales .
the all - scale intermittency is also corroborated by the fact that the strong excess of the large velocity - difference deviations is revealed by the statistics averaged over different spatial scales ( section 4.2 ) .
this seems to be an important conclusion , because , in the case of incompressible turbulence , with its relatively high fractal dimension , the existence of intermittency in the inertial subrange of the scales ( as opposed to the dissipation subrange ) has long been an open question ( frisch 1995 ) .
according to the classical work by kolmogorov ( 1941b , c , 1942 ) , incompressible turbulence is characterized by two limiting scales the outer scale @xmath37 , where energy is supplied to the turbulent flow , and the inner scale @xmath102 , where it is dissipated via molecular viscosity .
the `` inertial '' subrange of linear scales @xmath32 , where kinetic energy is neither injected into turbulence nor dissipated , but only transferred from larger to smaller scales , is between : @xmath103 . in the inertial subrange
, turbulence tends to become homogeneous and isotropic .
kolmogorov demonstrated that the inner scale of incompressible turbulence is of the order of @xmath104 where @xmath105 is the kinematic viscosity of the fluid and @xmath106 is the mean rate of energy dissipation per unit mass , given by the equation : @xmath107 where @xmath108 is the characteristic velocity difference at the outer scale @xmath37 .
expression ( 11 ) for the dissipation scale is readily obtained from dimensional considerations ; it is the only combination with the dimensions of length that can be constructed from the two parameters relevant to this mechanism of energy dissipation , @xmath105 and @xmath106 .
our attempt to derive a possible inner scale for highly supersonic turbulence is based on two assumptions .
first , proceeding from the common belief that the major mechanism of energy dissipation in supersonic turbulence is via shock waves , we assume that sonic speed , @xmath109 , rather than molecular viscosity , is the relevant parameter of the problem .
our second assumption may be more arguable .
we assume that the second relevant parameter of the problem is the same as for incompressible turbulence the mean rate of energy dissipation , @xmath110 .
thus , we assume that @xmath110 is an approximate constant of the energy cascade , equal to the rate , per unit mass , of supply of kinetic energy at the outer scale .
in other words , we assume no significant energy dissipation at intermediate scales . although some theoretical and observational arguments can be provided in favor of this hypothesis ( see below ) , we emphasize that , for the moment , it is only a hypothesis , whose consequences we would like to compare with observations . with these two assumptions
, we can derive the dissipation scale for supersonic turbulence , @xmath111 using standard dimensional analysis .
it is easy to show that only one combination can be formed by @xmath109 and @xmath110 with the dimensions of length , namely : @xmath112 substituting @xmath113 from equation ( 12 ) into equation ( 13 ) , we can give @xmath111 a more useful form : @xmath114 where m@xmath115 is the typical value of the mach number associated with the outer scale .
equation ( 14 ) shows that in highly supersonic turbulence ( m@xmath116 ) , the scale where shock waves begin to dissipate turbulent energy effectively is many orders of magnitude smaller than the outer scale @xmath37 .
an argument in favor of our assumption that highly supersonic turbulence does not dissipate much of its kinetic energy at larger scales is the observed kolmogorov , 1/3 , slope of the two - point velocity correlation function in h@xmath0o masers .
the 1/3 slope is a straightforward consequence of the conservation of energy during its cascade along the hierarchy of scales .
any energy dissipation in the inertial subrange would produce a steeper slope .
the specific kinetic energy associated with turbulent pulsations on a linear scale @xmath32 is @xmath117 , where @xmath118 is the r.m.s .
turbulent velocity on the scale @xmath32 .
this energy is passed to smaller scales in about one `` turn - over '' time , @xmath119 therefore , the rate of energy transfer is @xmath120 if @xmath110 is constant , equation ( 16 ) gives @xmath121 .
suppose now that some amount of energy is dissipated on each scale .
then @xmath110 is not a constant it decreases with decreasing @xmath32 . approximating this decrease by a power law ,
@xmath122 we have @xmath123 which demonstrates the steeping of the scaling law ( since @xmath124 ) .
why can energy dissipation in larger - scale shocks be hindered in highly supersonic turbulence ?
here is one possible answer .
although the potential component of the velocity field ( describing compression and expansion of the gas ) should play some part in a supersonic flow , this part may crucially depend on the unknown boundary and initial conditions of the flow ( e.g. falgarone _ etal . _ 1994 ) .
an extreme case is a purely vortical initial motion of gas at the larger scales . in this case , although the velocity increments for these scales ( the difference in velocities of two opposite peripheral points of an `` eddy '' ) may be highly supersonic , most converging flows within the eddy , which arise from fluctuations , will produce _ oblique _ shocks , whose normal velocity component , @xmath125 , will not exceed @xmath109 by much , and thus it will be small in comparison with the average vortical velocity increment : @xmath126 .
thus , even if a shock of large scale is formed , the time it needs to sweep the eddy and dissipate its kinetic energy is much longer than the turn - over time ( given in equation 15 ) , during which the eddy will disintegrate , passing its energy to smaller scales . only when @xmath118 drops down to @xmath127 , which happens at the inner scale , @xmath128 and @xmath118 become comparable , and massive dissipation of kinetic energy in random shocks becomes possible .
this gives a possible physical justification to equation ( 14 ) .
the important conclusion from the results presented in section 4.1 is that the exponents of the low - order structure functions for the highly supersonic turbulence are close to their classic kolmogorov values . in particular , the exponents of the second order structure function in all the investigated sources are close to 2/3 .
this value was predicted by kolmogorov for homogeneous incompressible turbulence at very high reynolds numbers ( kolmogorov 1941 a - c ) .
later , many authors , beginning with kolmogorov himself ( kolmogorov 1962 ) , attempted to introduce theoretical corrections to this exponent which would account for the experimentally detected intermittency of turbulence ( e.g. mandelbrot 1967 ; frisch , sulem & nelkin 1978 ; see also frisch 1995 ) .
for example , in the popular `` @xmath98-model '' of frisch , sulem , and nelkin , intermittency is assumed to have a fractal geometry , and the corrected exponent is given by @xmath129 where @xmath1 is the fractal dimension of the set on which intermittent turbulence concentrates . in the case of incompressible turbulence @xmath130 and , therefore , the codimension @xmath131 and the whole correction factor [ second term in equation ( 19 ) ] are small .
in fact , most of these intermittency corrections have historically been introduced as _
small _ parameters , to account for the small degree of observed intermittency in incompressible flow .
this has made it difficult to discriminate between different theoretical models , as well as to judge , in general , the plausibility of the approach treating intermittency as a `` disturbance '' to the classic kolmogorov theory .
highly supersonic turbulent flow should have specific features different from those of incompressible turbulence .
these two regimes of turbulence should , at least , differ in the ways they ultimately dissipate energy via shock waves and molecular viscosity , respectively . however , the power - law character of the observed spectrum of supersonic turbulence , over several decades in scale , is a strong indication that the energy cascade from larger to smaller scales is as intrinsic a property of supersonic turbulence as it is for incompressible turbulence .
moreover , the prominent fractal structure of the set on which supersonic turbulence dissipates its energy and the non - gaussian statistics of the two - point velocity increments , as revealed by h@xmath0o masers , are strong evidence for qualitative similarity of both regimes of turbulence also in the sense of an intermittent character of turbulent activity .
since the energy cascade and intermittency are the only two assumptions in the above mentioned models of incompressible turbulence , these models should likely also work in the case of supersonic turbulence .
the only application of incompressibility of turbulence in these models is the assumption that energy is not dissipated at the intermediate ( inertial subrange ) scales .
this assumption is not obvious for compressible , supersonic turbulence .
however , the observed slope of the two - point velocity correlation function , close to kolmogorov s , suggests that dissipation on intermediate scales is insignificant in the supersonic case too ( section 5.1 ) .
it is important to note , in this connection , that the h@xmath0o masers reveal a very low fractal dimension ( @xmath63 ) of the set on which supersonic turbulence ultimately dissipates in shock waves .
with such a low @xmath1 , the correction term from the beta - model [ equation ( 19 ) ] is large , and we anticipate that , due to intermittency , the slope of the second order structure function will be at least twice as steep as kolmogorov s classic value .
any dissipation of energy on intermediate scales the only possible difference from the incompressible case would steepen the structure function even more .
yet , the observed structure functions are close to kolmogorov s .
the observed pronounced intermittency of turbulence combined with the classic kolmogorov velocity structure functions can only be understood if we accept that intermittency is inherent in turbulence not a mere disturbance of its classic kolmogorov properties .
such an approach to incompressible turbulence is being developed by barenblatt and chorin ( 1997 ; hereafter b&c ) ) .
these authors claim , in particular , that both the tendency of the second order structure function to its classic dependence on @xmath32 and an increase of the degree of intermittency are natural asymptotic properties of turbulence when the reynolds number tends to infinity .
as first pointed out by von weizscker ( 1951 ) , the reynolds numbers of the interstellar gas are , in general , very high .
the reynolds numbers of the dense nuclei of the star - forming molecular clouds , where h@xmath0o masers reside , are especially high , due to the low viscosity of the dense gas .
the turbulent flows probed by h@xmath0o masers , have typical velocities @xmath132cm / s at a scale of @xmath133 cm . with the typical number density
@xmath134@xmath135 and temperature @xmath136k for a molecular cloud core , the kinematic viscosity is @xmath137 , and the typical value of the reynolds number , @xmath138 , is @xmath139 . for incompressible turbulence ,
the reynolds number is a measure of the width of the inertial subrange .
if the inertial subrange of supersonic turbulence probed by h@xmath0o had been limited by viscosity , it would have been very large . however , there is little doubt that , in the highly supersonic regime , the dissipation of energy starts at a much larger scale than the inner scale determined by viscosity . in section 5.2
, we postulated the existence in highly supersonic turbulence of an inner scale determined by energy dissipation in small - scale stochastic shocks and given by equation ( 14 ) .
although this inner scale is much larger than the kolmogorov dissipation scale ( @xmath140 cm ; @xmath141 cm ) , there is still a large inertial subrange in the highly supersonic regime
almost four decades in projected linear scale . one can try to reformulate the b&c theory and obtain an asymptotic law for the supersonic regime using the mach number rather than the reynolds number . as is customary in crude
phenomenological approaches , we shall ignore distinctions between structure functions involving different components of the velocity vectors .
specifically , we shall assume that the second order structure function @xmath142 involving the line - of - sight ( @xmath143 ) component of the difference velocity vector for two points reflects typical properties of all other structure functions .
following the reasoning of b&c , we can assume that in the inertial subrange of supersonic turbulence @xmath144 where @xmath113 is the mean rate of energy dissipation per unit mass .
it is assumed that in this inertial subrange the energy flux from larger to smaller scales is nearly constant and equal to @xmath113 . applying the standard dimensional analysis
, we find the scaling law for @xmath142 : @xmath145 where @xmath146 is a dimensionless function of its two dimensionless arguments .
we chose the mach number at the outer scale , m@xmath147 , and the running scale @xmath32 measured in the units of inner scale , @xmath111 , as the two dimensionless arguments .
in the case of incompressible turbulence the reynolds number , instead of the mach number , enters the parentheses in equation ( 20 ) . the dimensional analysis , by itself ,
does nt tell us anything about the properties of the function @xmath146 and its two arguments . in order to finalize the scaling law
, one must make a _ similarity assumption _ about the behavior of @xmath146 as its two arguments tend to infinity . as b&c point out , two different assumptions about this function mark the historical evolution of incompressible turbulence theory : \1
. complete similarity in both arguments : @xmath148const , when both dimensionless arguments tend to infinity ( kolmogorov 1941a ) ; \2 .
complete similarity in re , but incomplete similarity in @xmath149 , leads to a power - law dependence of @xmath146 on @xmath149 ( kolmogorov 1962 ) .
the first assumption leads to the classic kolmogorov @xmath150 .
the second assumption introduces an additive correction to the power index , which has been interpreted as a correction for intermittency .
b&c argue that the second assumption is internally contradictory and make the separate assumption of incomplete similarity in @xmath149 and _ no _ similarity in re .
this again generates an additive correction to the power index , but now the correction depends on the value of re . in the limit of re@xmath151
the correction tends to zero and one is left with the classic @xmath152 dependence .
the correction is substantial only when re is not large .
these similarity arguments can be repeated in the case of highly supersonic turbulence , with the formal substitution of m@xmath147 for re .
the b&c s theory makes no quantitative estimates of the magnitude of re ( or m@xmath147 , in our case ) necessary to reach the asymptotic behavior of @xmath153 .
we can suppose that in the case of turbulence probed by h@xmath0o masers , m@xmath147 is high enough for the correction to kolmogorov value of the power index to lie within the experimental errors .
thus , b&c s theory ( and its extension to supersonic turbulence ) rejects the notion that significant intermittency requires a significant correction to kolmogorov s 1941 law and thereby reconciles the observed low fractal dimension ( high intermittency ) of supersonic turbulence and its classic kolmogorov velocity structure functions .
one can speculate that a jet from a young star produces two basic flow regimes in the ambient gas : ( 1 ) a frontal , high - mach shock and ( 2 ) a high - vorticity flow due to the velocity shear at the side interface of the jet and the ambient gas ( e.g. masson & chernin 1993 ) .
we surmise that the low - velocity h@xmath0o masers are associated with the second regime .
the high - velocity h@xmath0o masers may be connected with the first regime , but one can anticipate that this connection , and turbulence produced by the frontal shock , would be more complicated than in the second regime . looking for the simplest cases of supersonic turbulence ,
we discuss only the second regime and the low - velocity masers in this investigation .
a power - law spatial - distribution correlation function in all of the h@xmath0o sources investigated here signifies self - similar clustering over almost four orders of magnitude in scale from @xmath154a.u . to @xmath65a.u .
the `` minimal clusters '' in this hierarchy [ `` features '' in gwinn s ( 1994 ) terminology ] , are actually spatial and spectral blends of elementary sources those observed through an element of spectral resolution . by the order of magnitude , both the elementary sources and the minimal clusters have a typical size of @xmath65a.u . , which is intriguingly close to the predicted dissipation scale of supersonic turbulence , given by equation ( 14 ) . in a typical h@xmath0o maser source , the low - velocity features are spread over a projected area of @xmath133 cm and occupy a velocity interval @xmath155km / s .
taking @xmath156 km / s ( appropriate for probable kinetic temperatures of several hundred k ) , we have @xmath157 and @xmath158 cm @xmath159a.u . unless this is a coincidence , the smallest clusters of h@xmath0o masers may be the sites of the ultimate dissipation of turbulent energy via stochastic shocks on the inner scale of supersonic turbulence .
this new conceptual approach to h@xmath0o masers may have several important consequences for understanding the very mechanism of masing in these sources .
it has often been argued that shock waves provide the best conditions for pumping h@xmath0o masers .
strelnitski and sunyaev ( 1973 ) interpreted the observed large dispersion of the h@xmath0o doppler velocities in w49n as due to supersonic gas outflow from a young host star and conjectured that the interaction of the outflow with the surrounding gas could produce shock waves necessary for pumping .
this hypothesis has been further developed by many authors , who have elaborated on details of the shock structure and collisional - radiative ( the first word standing for the source and the second for the sink of the quantum heat engine ) or collisional - collisional schemes of maser pumping behind a shock ( e.g. , shmeld , strelnitski & muzylev 1976 ; strelnitski 1984 ; kylafis & norman 1986 , 1987 ; hollenbach , mckee & chernoff 1987 ; elitzur , hollenbach & mckee 1989 ; kaufman & neufeld 1996 ) .
a common feature of all these models is that the pumping shock is a result of a direct collision of the outflow with a dense blob in the surrounding quiescent gas , or a direct collision of a dense blob in the outflow with the surrounding gas ( tarter & welch ( 1986 ) . in the present model , the pumping energy
is not imparted to the masing gas blobs directly by the stellar wind . instead
, the energy is channeled to the masers by a cascade from larger scales , which receive energy from the stellar wind or jets . in previous models ,
the shocks pumping h@xmath0o masers were assumed to have high velocities .
these were either high - speed ( @xmath160kms@xmath6 ) dissociative j - shocks ( elitzur , hollenbach & mckee 1989 ) , or slower ( @xmath161kms@xmath6 ) c - shocks propagating perpendicular to the magnetic field ( kaufman & neufeld 1996 ) .
two main goals were pursued in developing those models : ( 1 ) the achievement of the maximum possible abundance of h@xmath0o molecules via chemical reactions , and ( 2 ) the realization of sufficiently high kinetic temperatures for pumping the @xmath162 and other masing transitions .
it has been argued ( melnick _ etal . _ 1993 ; kaufman & neufeld 1996 ) that to fulfill both these requirements , temperatures @xmath163k are needed .
an important question is whether the slow shocks we advocate in this paper can provide such temperatures .
an analysis of the shock structure is beyond the scope of the present paper .
however , we note that a j - shock propagating _ along _ the magnetic field lines will have a post - shock temperature @xmath163k , if its velocity is @xmath164kms@xmath6 , which is a realistic velocity for an inner scale shock in our model .
if the new conception of h@xmath0o masers proposed here is correct , they may become an ideal tool for studying the properties of supersonic turbulence .
in contrast to the large - scale ism , supersonic turbulence probed by h@xmath0o masers has only one source of energy , supplied at the largest scale the interaction of the outflow from a young star with the surrounding gas .
furthermore , these flows are highly super - virial , so that gravitational effects are not important . most probably , these flows are also super - alfvenic . with the probable magnetic field strength @xmath165 g in the dense cores of molecular clouds , and @xmath166 g in the h@xmath0o maser clumps ( fiebig & gsten 1989 ) , and with the probable number densities of molecular hydrogen , @xmath167 and @xmath168@xmath135 respectively , the alfvenic velocity is @xmath169kms@xmath6 in both cases .
this is much less than the outer scale velocity ( @xmath170kms@xmath6 ) .
thus , turbulent velocities at most of the scales should be greater than the alfvenic velocity , which means that the magnetic field does not constrain turbulent pulsations .
masers are more effective than traditional ism probes of turbulence , such as thermal or fluorescent spectral lines , from the observational standpoint .
maser lines are bright and narrow , which allows the spatial and kinematic structure of the associated flow to be measured with high precision . because masing condensations are so small ( @xmath10 marcsec )
, every spectral feature detected by the interferometer gives a direct measure of line - of - sight velocity at a given _ point _ in the flow , projected onto the sky .
indeed , the requirement of the velocity coherence along the line of sight ( to produce lines with the observed widths @xmath10kms@xmath6 ) , together with the observed line - of - sight velocity gradients in the transverse direction of @xmath49kms@xmath6au@xmath6 , and under the assumption that velocity gradients in the region are more or less isotropic , limits the probable length of a maser hot spot along the line of sight to @xmath171au .
this is much less than the size of the whole active region ( @xmath172au ) and allows us to consider the h@xmath0o masers as point - like probes of the velocity field , virtually as effective as the direct probes used to study terrestrial turbulent flows .
doubts can arise on whether h@xmath0o masers are an adequate probe of the geometry of the turbulence dissipation .
for example , due to possible directivity of their radiation , some masing blobs may be unobserved .
can this distort the statistics we study ?
obviously , the non - observability of a fraction of places where turbulence dissipates results in underestimation of the space filling factor of dissipation
. however , this will not affect the deduced fractal dimension of the set on which dissipation takes place , if all the `` eddies , '' down to the smallest ones contain _ some _ amount of observable masers .
this follows from equation ( 4 ) : reduction of @xmath26 by any factor ( due to non - observability of a fraction of the masers ) will not change the value of the logarithmic derivative , and thus the value of @xmath1 .
if the directivity of the maser radiation is very high , it can affect observability of the smallest eddies , containing relatively small numbers of elementary masers . by an unlucky chance
, all the masers within such an eddy could be turned away from the observer .
the larger eddies , containing more elementary masers , should still be observable , although their contours will be delineated by reduced numbers of maser spots .
a lack of observability of the smaller eddies will result in a decrease of @xmath98 with the decreasing scale [ see equation ( 10 ) ] and in a corresponding change of the slope of the plots in figures 2 - 4
. using equations ( 9 ) , ( 4 ) , and ( 6 ) , it is easy to convince oneself that this would steepen the slope of the plot toward the smallest scales in figures 2 and 4 and flatten it toward the smallest scales in figure 3 .
these effects are either unseen or quite small in these figures .
we believe , therefore , that the majority of the scales , covering almost four orders of magnitude , are well represented by the h@xmath0o masers and that @xmath1 is determined adequately , regardless of possible omission of a fraction of the maser probes caused by radiation directivity .
\1 . vlbi maps of five h@xmath0o maser sources in regions of star formation
reveal fractal spatial distribution of the masing hot spots ; a power - law dependence of two - point velocity increments on spatial scale ; and the non - gaussian statistics of velocity increments ( a strong excess of large deviations from the mean value ) .
all these properties are known as typical of turbulence .
if the h@xmath0o masers trace turbulence indeed , our quantitative analysis shows that this highly supersonic turbulence is characterized by a much lower fractal dimension ( @xmath63 ) , and thus much stronger intermittency , than incompressible turbulence .
strong intermittency at virtually all the spatial scales is also confirmed by the excess of large velocity increments at all scales .
unexpectedly , the power indices of the low - order velocity structure functions for the putative supersonic turbulence are found to be close to the classic kolmogorov values for high - reynolds - number incompressible turbulence .
this is incompatible with the strong intermittency ( low fractal dimension ) in traditional approaches to turbulence , but may find its explanation in the framework of the new approch put forward by barenblatt and chorin .
supersonic turbulence with a high mach number at its greatest scale may possess an inner scale , at which the bulk of its energy is dissipated in low - mach - number stochastic shocks .
the predicted value of the inner scale is close , by the order of magnitude , to the observed sizes of the h@xmath0o hot spots .
we hypothesize that the h@xmath0o masers are generated in the random shocks at the inner scale of highly supersonic turbulence produced in the ambient gas by the intensive outflow from a newly - born star .
_ acknowledgements .
_ s.g . and b.h .
acknowledge with gratitude support of their participation in this study by the nsf ( research experiences for undergraduates ) grant ast 93 - 0039 and by the nantucket maria mitchell association .
alcolea , j. , menten , k.m , moran , j.m . , & reid , m.j .
1992 , in astrophysical masers , ed .
clegg & g.e .
nedoluha ( heidelberg : springer ) , p.225 anderson , n. , & genzel , r. 1993 , in `` astrophysical masers , '' eds .
clegg & g.e .
nedoluha ( berlin : springer ) , p. 97 barenblatt , g.i . and chorin , a. 1997 , comm
. pure appl .
, 50 , 381 clegg , a. and nedoluha , g. ( editors ) 1993 , astrophysical masers ( berlin : springer - verlag ) crownover , r.m .
1995 , introduction to fractals and chaos ( boston : johns and bartlett publishers ) dickman , r.l . 1985 , in `` protostars and planets ii , '' eds . d.c .
black & m.s .
matthews ( tucson : univ . of arizona press ) , p. 150
dutton , j.a .
& deaven , d.g .
1969 , radio sci . , 4 , 1341 elitzur , m. 1992 .
astronomical masers ( dordrecht : kluwer ) elitzur , m. , hollenbach , d. , and mckee , c. 1989 , apj , 346 , 983 elmegreen , b.g .
1993 , in `` protostars and planets iii , '' ed .
levy & j.i .
lunine ( tucson : univ . of arizona press ) falconer , k.j . 1990 , fractal geometry , mathematical foundations and applications ( chichester : john wiley & sons ) falgarone , e. and phillips , t.g . 1990 , apj , 359 , 344 falgarone , e. , phillips , t.g . , and walker , c.k .
1991 , apj , 378 , 186 falgarone , e. , lis , d.c . ,
phillips , t.g . , pouquet , a. , porter , d.h . , & woodward , p.r .
1994 , apj , 436 , 728 feder , j. 1988 , fractals ( new york and london : plenum press ) fiebig , d. & gsten , r. 1989 , a&a , 214 , 333 fleck , r.c . ,
jr . 1981 , apj ( lett ) , 246 , l151 franco , j. and carraminana , a. , eds . , 1999 ,
`` interstellar turbulence , '' cambridge university press , 1999 frisch , u. 1995 , _ turbulence _
( cambridge : university press ) frisch , u. , sulem , p.l . , and nelkin , m. 1978 , j. fluid mech .
87 , 719 genzel , r. , et al .
1981 , apj , 247 , 1039 gezari , s. 1997 , baas , 29 , # 5 , abstract # 20.01 gezari , s. , reid , m.j .
, & strelnitski , v. 1998 , baas , 30 , # 4 , abstract # 71.06 gouyet , j - f .
`` physics and fractal structures '' , springer , 1996 gwinn , c. 1994 , apj , 429 , 241 gwinn , c.r . ,
moran , j.m . , & reid , m.j .
1992 , apj , 393 , 149 holder , b. & strelnitski , v. 1997 , baas , 29 , # 5 , abstract # 121.08 hollenbach , d.j . ,
mckee , c.f . , & chernoff , d. 1987 , in `` star forming regions , '' ed .
m. peimbert & j. jugaku ( dordrecht : reidel ) , p. 334 kaufman , m.j . and neufeld , d.a .
1996 , apj , 456 , 250 kolmogorov , a.n .
1941 , dokl .
nauk sssr , 30 , 9 ; 31 , 99 , 538 kolmogorov , a.n .
1942 , izv .
nauk sssr , ser .
fiz . , 6 ( 1 - 2 ) , 56 kolmogorov , a.n .
1962 , j. fluid mech .
, 12 , 82 kylafis , n. & norman , c. 1986 , apj ( lett . ) , 300 , l73 kylafis , n. & norman , c. 1987 , apj , 323 , 346 larson , r.b .
1995 , m.n.r.a.s . , 272 , 213 mandelbrot , b.b .
1967_science _ , 156,636 mandelbrot , b.b .
1982 , `` the fractal geometry of nature , '' ( new york : freeman ) masson , c.r .
& chernin , l.m .
1993 , apj , 414 , 230 melnick , g.j . , menten , k.m . , phillips , t.g . , & hunter , t. 1993 , apj , 416 , l37 migenes , v , reid , m. eds .
2002 , `` cosmic masers : from protostars to blackholes , '' asp conf .
series , vol .
206 obukhov , a.m. 1941 , dokl .
nauk sssr , 32 , 22 reid , m.j .
1993 , ann .
, 31 , 345 reid , m.j . ,
hashick , a.d . ,
burke , b.f . ,
moran , j.m . ,
johnston , k.j .
, & swenson , jr , g.w .
1980 , apj , 239 , 89 reid , m. j. , schneps , m. h. , moran , j. m. , gwinn , c. r. , genzel , r. , downes , d. & rnnng , b. 1988 , apj , 330 , 809 richardson , l.f .
1922 , weather prediction by numerical processes ( cambridge : cambridge university press ) scalo , j.m .
1987 , in `` interstellar processes , '' eds .
hollenbach and h.a .
thronson , jr .
( reidel publishing co. ) , p. 349 schneps , m.h . , et al .
1981 , apj , 249 , 124 sellwood , j.a . & balbus , s.a .
1999 , apj , 511 , 660 shmeld , i.k . ,
strelnitski , v.s . , & muzylev , v.v .
1976 , soviet .
, 53 , 728 spitzer , l. 1978 , physical processes in the interstellar medium ( new york : wiley ) strelnitski , v.s .
1984 , m.n.r.a.s . , 207 , 339 strelnitski , v.s . , and sunyaev , r.a .
1973 , soviet .
astron . , 16 ,
579 strelnitski , v.s . ,
alexander , j. , moran , j.m .
, & reid , m.j .
1998 , iau colloquium 164 , radio emission from galactic and extragalactic compact sources , asp conference proceedings , ed .
zensus , g.b .
taylor & j.m .
wrobel , 144 , 369 .
tarter , j. & welch , w.j .
1986 , apj , 305 , 467 van atta , c.w . and park , j. 1971 , in _ statistical models and turbulence _ , lecture notes in physics , vol .
m. rosenblatt and c. van atta ( springer : berlin ) , p. 402 - 426
von weizscker , c.f .
1951 , apj , 114 , 165 walker , r.c .
1984 , apj , 280 , 618 walker , r.c . ,
matsakis , d.n . , & garcia - barreto .
1982 , apj , 255 , 128 fig . 1 .
multi - scale vlbi maps of h@xmath0o maser source in sgr b2(m ) as observed on january 23 , 1986 . on all maps except the last one , dot sizes are larger than maser spot sizes . on the last map ,
dots are smaller than observed maser spot sizes ( @xmath173au , which is perhaps affected by interstellar scattering ) .
the dots on the last map show measured positions of spectral channels whose radial velocities are indicated near the dots .
fig . 2 .
density - radius fractal measure for h@xmath0o maser source in sgr b2(m ) for the four epochs of observation ( @xmath92 @xmath1 ) and for their average ( @xmath174 ) .
the solid straight line shows linear fit to the data points ( empty circles ) .
two - point line - of - sight velocity correlation function for h@xmath0o masers in sgrb2(m ) .
the empty circles represent data , the straight solid line shows linear fit . _ a d _ the results for the four epochs of observation ; _ e _ the average for the four epochs .
two - point line - of - sight velocity correlation function for a planar projection of a randomly - filled spherical shell , for different ratios of expansion and rotation components .
asterisks represent data .
one of the two straight lines in each graph shows a linear fit ( possible only for the largest scales ) ; the other line indicates , for reference , a slope of 1/3 .
fig . 9 .
histograms of the number of pairs having a given value of the deviation of their relative velocity from its mean value .
@xmath92 : separation range @xmath175mas ; @xmath93 : separation range @xmath176mas ; @xmath177 : separation range @xmath178mas . fig
the probability of deviation of the two - point velocity difference from its mean value , averaged over scales for the h@xmath0o maser spots in sgrb2(m ) .
both abscissa and ordinate axes are normalized as described in text .
dots represent data .
solid lines show a one - gaussian fit ( @xmath92 ) and a two - gaussian fit ( @xmath93 ) to the data points .
same as in fig .
10 for h@xmath0o masers in w49n , w51n , w51main , and w3(oh ) ( `` cluster '' ) . solid line observed data , broken line one - gaussian fit to the central part of the observed distribution . | we use unpublished and published vlbi results to investigate the geometry and the statistical properties of the velocity field traced by h@xmath0o masers in five galactic regions of star formation sgr b2(m ) , w49n , w51(main ) , w51n , and w3(oh ) . in all sources
the angular distribution of the h@xmath0o hot spots demonstrates approximate self - similarity ( fractality ) over almost four orders of magnitude in scale , with the calculated fractal dimension @xmath1 between @xmath2 and 1.0 . in all sources ,
the lower order structure functions for the line - of - sight component of the velocity field are satisfactorily approximated by power laws , with the exponents near their classic kolmogorov values for the high - reynolds - number incompressible turbulence .
these two facts , as well as the observed significant excess of large deviations of the two - point velocity increments from their mean values , strongly suggest that the h@xmath0o masers in regions of star formation trace turbulence .
we propose a new conceptual model of these masers in which maser hot spots originate at the sites of ultimate dissipation of highly supersonic turbulence produced in the ambient gas by the intensive gas outflow from a newly - born star . due to the high brightness and small angular sizes of masing hot spots and the possibility of measuring their positions and velocities with high precision
, they become a unique probe of supersonic turbulence . _
subject headings : _ masers turbulence ism : jets and outflows |
eternal solutions are solutions of mean curvature flow defined on a whole time interval @xmath0 .
since mean curvature flow is quasi - linear parabolic equation , we can not solve the flow backward in time .
hence eternal solutions are very special and have interesting properties if they exist .
translating solutions are solutions of the mean curvature flow which have translation invariance under mean curvature flow , and they are typical examples of eternal solutions .
eternal solutions are type ii singularity models of mean curvature flow , that is , they naturally arise after the parabolic rescaling at type ii singularities . in codimension one
, r. hamilton @xcite showed that any complete convex eternal solution whose curvature attains its maximum at a point in space - time must be a translating solution .
however this fact is still not clear in higher codimension .
therefore , it is natural to consider not only translating solutions , but also eternal solutions to study type ii singularities . besides above facts , chen - li @xcite and m .-
wang @xcite independently showed that type i singularities do not appear under almost - calibrated ( see section 2 ) lagrangian mean curvature flow .
later neves @xcite generalize this result to the zero - maslov class case . in this case , therefore , we essentially need to investigate type ii singularities and eternal solutions ( including translating solutions ) as the singularity models .
our aim in this paper is to prove a non - existence theorem for eternal solutions of almost - calibrated lagrangian mean curvature flow . in this direction , we know some results by han - sun @xcite , neves - tian @xcite and sun @xcite .
however , all the known results mention only translating solutions . in @xcite , han and
sun showed a non - existence result of almost - calibrated lagrangian translating solitons with nonnegative sectional curvature .
we generalize their theorem @xcite into the class of lagrangian eternal solutions under a similar situation .
[ main ] there is no non - flat complete lagrangian eternal solution with nonnegative ricci curvature to the almost - calibrated lagrangian mean curvature flow in @xmath1 with @xmath2 for each time @xmath3 . in general ,
the blow - up limit of type ii singularity must be non - flat .
therefore , as a consequence of our result , we obtain the following . any eternal solution with nonnegative ricci curvature can not arise as blow - up limit of almost - calibrated lagrangian mean curvature flow in @xmath1 .
our result is a parabolic version of theorem by han and sun @xcite , and the proof is somewhat similar to @xcite .
however , to show the parabolic version of the curvature estimate , we adopt another technique by souplet - zhang @xcite and m. wang @xcite to show a parabolic version of curvature estimate .
thanks to the parabolic curvature estimate , we do not need to assume the translation symmetry which is needed in @xcite .
the author is supported by the grant - in - aid for jsps fellows . during the preparation of this paper
the author has stayed in max planck institute for mathematics in the sciences , leipzig .
the author is grateful to jrgen jost for his hospitality and his interest .
the author also would like to thank reiko miyaoka for her helpful comments .
in the following , our setting follows @xcite and @xcite .
let @xmath4 and @xmath5 , respectively , be the standard complex structure on @xmath6 and the standard symplectic form on @xmath1 .
we consider the closed complex valued @xmath7-form given by @xmath8 where @xmath9 , are complex coordinates of @xmath10 . a smooth @xmath7-dimensional submanifold @xmath11 in @xmath1 is said to be lagrangian if @xmath12 .
this implies that @xmath13 where @xmath14 denotes the volume form of @xmath15 and @xmath16 is some multivalued function called _ lagrangian angle _ or _
phase function_. when the lagrangian angle is single valued , the lagrangian is called _ zero - maslov _ class , and if @xmath17 then the lagrangian is said to be _ almost - calibrated_. we denote by @xmath18 the second fundamental form , and @xmath19 the mean curvature vector of the submanifold @xmath15 in @xmath1 .
now , by the gauss equation , we can write the scalar curvature @xmath20 of @xmath15 as follows : @xmath21 therefore , nonnegative ricci curvature of @xmath15 implies latexmath:[\ ] ] such a function can be given by a canonical way ( for example , see @xcite , @xcite , @xcite or @xcite ) .
we use a cut - off function supported on @xmath116 given by @xmath117 .
let @xmath118 . by using we can calculate @xmath119 note that @xmath116 is compact .
hence @xmath120 attains its maximum at some point @xmath121 in @xmath116 .
we may assume that such a point @xmath122 is not in the cut locus of @xmath71 ( see @xcite , @xcite or @xcite ) .
at the point @xmath123 , we have @xmath124 hence by , we obtain @xmath125 note that the following holds : @xmath126 recall that the following young s inequality : for any @xmath127 , and any @xmath128 , we have @xmath129 where @xmath130 . by using , young s inequality and the property of @xmath131 , we can estimate @xmath132 as follows : @xmath133 where @xmath128 is an arbitrary constant which is determined later , @xmath134 are constants depending only on @xmath135 .
similarly , we can calculate by using young s inequality and the property of @xmath131 , @xmath136 since the ricci curvature of @xmath15 is nonnegative , by the laplacian comparison theorem we have @xmath137 using and @xmath138 , we have @xmath139 hence we obtain for the second term of @xmath140 in the same way as above , @xmath141 where @xmath142 is a constant depending only on @xmath135 and @xmath7 .
( note that we may assume @xmath143 for the second inequality since @xmath144 for @xmath145 . ) as for the first term of @xmath140 , we have @xmath146 from , , and , we finally obtain @xmath147 since @xmath128 is arbitrary we can take a sufficiently small @xmath135 such that @xmath148 then we have @xmath149 since @xmath150 on @xmath151 , @xmath152 this completes the proof of theorem [ curvestimate ] .
in the last section , we mention arbitrary codimensional case without proof ( see @xcite , @xcite ) .
let us consider an isometric immersion @xmath153 , and its mean curvature flow : @xmath154 although mean curvature flow in higher codimension is interesting problem , in general , it becomes rather difficult to deal with since the behavior of the normal bundle is complicated .
few results are known in this field except for special cases ( for example @xcite ) .
let @xmath155 be a parallel @xmath7-form with @xmath156 .
then @xmath157 induces a function @xmath158 on the immersion @xmath159 by @xmath160 where @xmath161 is the induced volume form on @xmath162 . since @xmath156
, we must have @xmath163 the condition to be a graph then expressed by @xmath164 .
note that there holds @xmath165 is vanish everywhere . in @xcite ,
smoczyk , wang and xin considered mean curvature flow in arbitrary codimension with flat normal bundle , and showed that the flat normal condition is preserved under the mean curvature flow . in this case , we can compute evolution equations similarly to hypersurface cases .
let @xmath167 be an entire graphic eternal solution in arbitrary codimension with flat normal bundle .
suppose that there exist constants @xmath82 and @xmath168 so that @xmath169 then the eternal solution must be flat for each time slice @xmath170 .
smoczyk , k. , _ mean curvature flow in higher codimension -introduction and survey- _ , global differential geometry(ed . c. b@xmath171r , j. lohkamp , m. schwarz ) , springer proc .
, * 17 * , 231274 , springer , heidelberg , ( 2012 ) . | in this paper , we derive a mean curvature estimate for eternal solutions ( including translating solutions ) of almost - calibrated lagrangian mean curvature flow in complex euclidean space . as a consequence ,
we show a non - existence result for eternal solutions of almost - calibrated lagrangian mean curvature flow . |
following sustained interest from policy makers , recent years have seen a number of bio - economic models examining the effects of commercial fisheries on marine resources . even though over - fishing has been a fact since historical times @xcite
, the problem has gained a new quality due to the industrialization of fisheries @xcite .
in particular the latter has reduced fish biomass by 80% within 15 years of exploitation @xcite . in this context
, the impact of capital accumulation has been a major issue in fisheries economics over the last two decades @xcite . in these contributions ,
commercial fisheries is portrayed as a system in which a biological stock and a capital stock interact dynamically . as the capital stock is highly specialized and
can not readily be converted to other uses , investment decisions are characterized by irreversibility , which is assumed as a major cause of over - fishing . the previous literature has treated capital accumulation in various settings .
@xcite study the optimal management strategy for a renewable resource with irreversible investment , assuming that marginal investment costs are constant .
the latter assumption is abandoned by @xcite on the grounds that constant investment costs imply an immediate jump in the capital stock , which is then followed by a period of decline in both the capital and fish stock .
contrary , observed patterns of capital accumulation are characterized by an initial phase of continuous growth of fleet size .
assuming increasing marginal investment costs leads to better agreement with observations , but makes the model more complicated .
considerations of tractability therefore lead @xcite to assume that harvest productivity is independent of the size of the biological stock .
in contrast to these optimal - exploitation models , approaches which study capital accumulation in more realistic settings are rare .
an exception is @xcite who examines an open - access fishery with irreversible investment under both perfect and imperfect competition .
but the increased realism of these models comes at a cost in that marginal investment costs and harvesting productivity are kept constant in the analysis of out - of - equilibrium behaviour due to serious analytical difficulties . in general , the variety of modelling strategies pursued in the literature thus reflects the tension between realism and tractability , illustrating the need for new concepts in integrated modelling ( cf .
* ; * ? ? ?
* ) . in order to keep them accessible to analysis ,
most of the models mentioned above disregard at least one of these difficulties , e.g. those relating to investment costs , harvesting productivity , or industry structure . in many cases
the difficulties restrict analysis to equilibria or comparative dynamics in the neighbourhood of an equilibrium .
but since fisheries systems tend to stay far away from equilibrium , e.g. when catches reside above the maximum sustainable yield , there exists an urgent need for the analysis of the enfolding dynamics .
a common approach to this task is phase plane analysis .
although this technique is not impossible for systems with more than two ordinary differential equations ( odes ) , it becomes rather difficult ( cf .
another technique to tackle tractability problems might be to run an ensemble of numerical simulations or to use methods from nonlinear analysis ( e.g. bifurcation analysis or computing domains of attraction ) rather than solving analytical models .
these approaches are limited if precise parameters and exact functional specifications are not completely available .
in fact , as pointed out by @xcite , our understanding of bio - economic systems is characterized by low levels of knowledge .
the dynamics of fish stocks , the economic characteristics and strategies of the fishing industry are subject to a serious lack of information ( cf . * ? ? ?
* ; * ? ? ?
* ) . in this paper
we demonstrate a qualitative simulation technique which complements phase space analysis and numerical simulation in data - poor settings when nonlinear dynamics far from equilibrium are to be investigated .
the method , developed in the field of artifical intelligence ( c.f .
* ; * ? ? ?
* ; * ? ? ?
* ) , starts from the argument that imprecise understanding can be formalized and used for a variety of model - based tasks ( e.g. identification of general dynamic properties , scenario testing , hypothesis exploration , policy advice ) .
the approach allows all possible dynamic behaviours of the system to be characterized and classified on the basis of purely qualitative relationships ( i.e. in the absence of quantitative information ) .
it is used to an increasing extent in several fields ( * ? ? ?
* ; * ? ? ?
* economics and finance ) , ( * ? ? ?
* ; * ? ? ?
* epidemiology and genetics ) , ( * ? ? ?
* chemistry ) , ( * ? ? ?
* ; * ? ? ?
* ecology ) , ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* sustainability science ) , but seldom in bio - economics so far .
we use qualitative simulation to investigate the dynamics of capital accumulation in unregulated fisheries with nonlinear investment costs and stock - dependent harvesting productivity .
our model is able to trace the evolution of capital and fish stocks in good agreement with observed patterns .
a main result is that the model necessarily produces a period where the capital stock continues to rise even after the harvest has started to decline , i.e. the development of excess capacities is unavoidable . in general
the paper shows that qualitative models allow us to derive robust properties of bio - economic systems when we have only weak knowledge at hand .
we have organized this paper as follows : in section [ sec2 ] the basics of qualitative modelling are introduced .
section [ sec3 ] develops an analytical model of capital accumulation in fisheries . in section [ sec4 ]
we generalize this model to a qualitative one to characterize its global dynamic features . in section [ sec5 ]
we compare the results with some development paths recently observed in industrial fishery and draw general implications for fisheries management and the applicability of the qde method . a discussion and a summary conclude the paper .
qualitative differential equations ( qdes ) are a prominent methodology in qualitative modelling . for this goal
@xcite has developed the qsim runtime environment whose underlying concept is used for the analyses presented here ] . in the following
we describe the basic ideas of qdes and provide the necessary technical terms ( written in italics ) . for a more detailed introduction and a thorough overview of applications we refer to @xcite .
the input for a modelling task is a qualitative differential equation ( qde ) comprising the following parts : 1 . a set of continuously differentiable functions ( variables ) of time ; 2 . a quantity space for each variable , specified in terms of an ordered set of symbolic _ landmarks _ ; 3 . a set of _ constraints _ expressing the algebraic , differential or monotonic relationships between the variables .
a qde can be considered as an abstract description ( _ abstraction _ , cf .
fig.[fig : lab1 ] ) of a set of ordinary differential equations ( odes ) .
the abstraction is attained in a twofold manner : ( i ) variables take values from the set of symbolic _ landmarks _ or intervals between landmarks .
each landmark represents a real number , e.g. maximum sustainable yield , of which the exact quantitative value may be unknown . nevertheless , it is analytically distinguished whether the catch is above or below this threshold .
the landmark or the interval between landmarks where the value of a variable is at a given time , is called its _ qualitative magnitude_. ( ii ) monotonic relationships specified between variables , e.g. that the yield is monotonically decreasing with a decreasing stock , are expressed by _
constraints_. this is an abstraction in the sense that the constraint comprises an entire family of ( linear and nonlinear ) monotonic functions .
the only requirement is that these functions are smooth and that their derivatives have certain signs .
qualitative simulation achieves its result by performing a constraint satisfaction scheme ( for a general introduction see * ? ? ?
* ; * ? ? ?
* ) , where all combinations of qualitative magnitudes inconsistent with the constraints are filtered out . due to continuity of the variables values in time ,
the scheme additionally checks which admissible combinations of qualitative magnitudes can occur after other combinations , called _
successors_. the guaranteed coverage theorem ( an in - depth discussion of this theorem is beyond the scope of this paper , see * ? ? ?
* ) ensures that the algorithm predicts an abstract description of _ all _ solutions to any ode described by the given qde .
this implies , due to the indeterminacy of analysed systems , that a qde solution comprises not only one time development , but a set of trajectories .
in the following we illustrate these concepts considering a simple nonlinear model , which is not meant to solve real - world problems ( it will be extended in the succeeding sections ) . consider a natural resource stock @xmath0 with an associated recruitment function @xmath1 .
the industry chooses effort depending on @xmath0 such that there is a harvest function @xmath2 .
the time behaviour of the stock is expressed as @xmath3 it is assumed that @xmath4 , and that @xmath5 is strictly increasing in @xmath0 .
the recruitment function is of logistic type and attains a maximum @xmath6 ( maximum sustainable yield ) at @xmath7 .
furthermore , @xmath8 , where @xmath9 denotes the carrying capacity of the biological system . for @xmath10
the function @xmath11 is strictly increasing , but strictly decreasing if @xmath12 .
in the first abstraction step , for each variable the following sets of symbolic landmarks are defined ( including lower and upper bounds ) : @xmath13 for a given time @xmath14 and a ( continuous ) value for @xmath15 , @xmath16 and @xmath17 one can determine the qualitative magnitudes with respect to the defined landmarks , e.g. as an interval @xmath18 , if @xmath19 , or as a singular landmark , e.g. @xmath20 .
qualitative simulation explicitly tracks the direction of change of all variables in the model , which is called _ qualitative direction_. the arrow @xmath21 portrays that a variable is decreasing ( i.e. @xmath22 ) , @xmath23 increasing ( @xmath24 ) , and a circle @xmath25 that @xmath26 at a given time @xmath14 .
the qualitative magnitude together with the qualitative direction of a variable at a given time is called its _ qualitative value _ , e.g. written as .
the qualitative values of all variables of the model calculated for a certain time is called the _ qualitative state _ of the system . in the second abstraction step
the constraints of the nave model are formulated as relationships between the variables and their derivatives . to simplify the presentation
we only introduce the most important constraints of the model ( for the whole model we refer to the annex and to the download version ; cf .
footnote[fn1 ] ) . since @xmath27 with respect to @xmath28 is denoted as @xmath29 and the second derivative as @xmath30 . ] , @xmath31 and @xmath32 , using the the sign operator @xmath33 , @xmath34 we define @xmath35 and describe the relation between stock and recruitment by the monotonicity assumptions mentioned above by : @xmath36 the relation @xmath37 implies @xmath38 indicating that a recruitment below and a harvest above a sustainable level leads to a decreasing stock .
it is obvious , due to the indeterminacy of the relations , that many functions exist which comply with these constraints ( see , e.g. fig.[fig : lab2]a ) , but nonetheless a complete qualitative simulation is possible . )
([eq : rrule6 ] ) .
the uncertainty intervals regarding the landmarks are indicated by brackets . for @xmath39
all values are possible lying in the cross - hatched area .
( b ) an example for the qualitative behaviour of harvest @xmath16 ( dotted line ) .
such a behaviour is representative for a class of ( quantitative ) time developments as indicated by the dashed lines ( grey areas ) . ]
it starts with a qualitative state as initial value at time @xmath40 .
supposing that at the beginning the fishery is characterized by an almost unexploited biological stock and catches above @xmath6 , the qualitative state at @xmath40 is given by @xmath41 the qualitative magnitude of @xmath11 results from the qualitative magnitude of @xmath0 ( above @xmath7 ) by applying constraint([eq : rrule6 ] ) .
the qualitative direction of @xmath0 ( decreasing ) is a consequence of the qualitative magnitudes of @xmath5 and @xmath11 , i.e. rule([eq : dotxrule3 ] ) .
constraint([eq : rrule6 ] ) forces @xmath11 to increase .
finally , @xmath5 decreases due to constraint([eq : hrule1 ] ) .
a time evolution is now initiated by these starting conditions , which has to respect the direction of changes shown in eqs.([eq : lab3 ] ) and the defined constraints .
for the example three possible evolutions can be distinguished : 1 .
@xmath0 decreases below @xmath7 , where @xmath5 still resides above @xmath6 at this time . 2 .
@xmath5 decreases below @xmath6 , while @xmath0 stays above @xmath7 at this time .
@xmath0 decreases below @xmath7 and @xmath5 decreases below @xmath6 at the same time .
these ` events ' ( occurring at a time @xmath42 ) lead to new ( and different ) qualitative states .
qualitative simulation checks which of these states comply with all constraints of a defined qde .
those states ` surviving ' this test are the successors valid for the next time interval .
for case ( i ) , we obtain the following qualitative state as a successor : @xmath43 here @xmath44 indicates the time point until the calculated state is valid ( i.e. a new ` event ' will occur ) .
the value of @xmath0 is the direct outcome of case ( i ) , where constraint([eq : rrule4 ] ) determines the value of @xmath11 .
the magnitude of @xmath5 is unchanged ( since otherwise , we would be in case ( iii ) ) .
decreasing harvest is the result of constraint([eq : hrule1 ] ) . by similar arguments , case ( ii )
leads to the successor @xmath45 case ( iii ) is neglected here , because it is very unlikely that both harvest and recruitment drop below @xmath6 exactly at the same time . in the semantics of qualitative simulation
the exclusion of so - called marginal cases or other specific assumptions can be explicitly defined ( cf .
annex ) . in the next time steps ( valid for the intervals @xmath46 ) qualitative simulation takes into account all successors and performs the same consistency checks as introduced above .
the procedure is repeated until the system attains an equilibrium , no new successor is possible or until it enters a cycle .
a logical sequence of successors is called _ qualitative behaviour _ and represents a qualitative dynamical trajectory .
it should be emphasized here that a single behaviour represents a set of quantitative development paths ( including all solutions of an ode respecting the constraints , fig.[fig : lab2]b ) . since for some states
more than one successor might be possible , all states and their possible sequences have to be displayed as a state - transition graph ( stg ) .
figure [ fig : lab4 ] shows the stg for the example .
each vertex represents a consistent qualitative state .
if one state is a successor of another , they are connected by an edge ( arrow ) .
each qualitative behaviour can be traced as a path along edges through the graph .
however , for more complex bio - economic models the number of states typically increases rapidly . in these cases
a clustering algorithm can be applied to provide a further structuring of the graph . for the algorithm
, the modeller has to specify a set of relevant ( state ) variables of the model .
all states with identical qualitative values of the relevant variables are automatically joined to clusters . in the generalized state - transition graph ( gstg ) each cluster is represented by a single vertex .
its edges are inherited from the stg .
more formally , each vertex in the gstg is an equivalence class of vertices in the stg with respect to the qualitative value of the relevant variables .
there is an edge from a vertex @xmath47 ( a class of qualitative states ) to a vertex @xmath48 ( another class ) if there exists at least one edge in the stg where a qualitative state in @xmath48 is the successor of a qualitative state in the class @xmath47 .
this representation allows solutions of high - dimensional models to be displayed in a concise manner . :
harvest and @xmath0 : stock .
the position in the diagram refers to the qualitative magnitude , and the arrows between the boxes describe possible changes of qualitative states in time .
circles represent equilibrium states .
all states with a qualitative magnitude of @xmath49 are characterized by a decreasing stock and harvest . below the @xmath6 level @xmath0 and @xmath5 both decrease or increase .
one of the equilibria represents extinction ( @xmath50 ) , one a steady state below @xmath7 , another an equilibrium above this landmark .
] comparing the qde method with monte carlo simulations or markov chains , one has to emphasize that it is a deterministic approach .
in contrast to probabilistic methods , qualitative modelling determines possible transitions for the systems development in the logical sense .
this is quite different to techniques propagating changes on the basis of likelihoods , e.g. by transition matrices ( for an example of such approaches cf .
it is in general not possible to derive probabilities in the case of multiple successors of a single state , since these also depend on qualitative states visited before ( the markov property is violated ) .
furthermore , due to various uncertainties , it is often impossible to measure sufficient frequency records . a qde is a compact representation of a large family of odes , and the simulation result includes all their solutions .
the result is rather general in nature , since no parameters need to be specified .
therefore , the result of qualitative simulation is not exactly the same as of common phase - plane analysis .
there , multiple cases due to the number of intersections of the main isoclines ( where the derivatives of some state variables vanish ) have to be specified ( as in , e.g. * ? ? ?
* ) depending on the concrete parameterization , which is not known to the modeller in our case .
for the same reasons , different cases regarding stability of equilibria , bifurcations and other phenomena studied in nonlinear analysis are contained in one qualitative model . in the stg
one circle may represent multiple equilibria of different types .
if the modeller wants to exclude or distinguish some cases , additional equilibria must be made explicit by introducing new landmarks and linking them with appropriate constraints .
in this section we introduce a bio - economic model describing a situation where @xmath51 identical and profit - maximizing firms compete for an unregulated resource , i.e. a marine fish stock with the size @xmath0 . because the harvesting technology and the associated cost and profit functions have been formulated in several ways in the previous literature@xcite , we start by deriving a generic formulation based on standard production economics .
assume that any harvesting requires variable inputs ( labour , fuel , material ) , jointly referred to as effort @xmath52 , and fixed inputs @xmath53
( capital , e.g. ships and gear ) .
the productivity of these inputs depends on the biological stock @xmath0 .
thus , a production function @xmath54 can be defined , that determines the resulting harvest @xmath55 . for its partial derivatives
the properties @xmath56 , @xmath57 , @xmath58 are imposed .
the first two sets of inequalities describe the standard properties of positive , decreasing marginal productivity .
the third set of inequalities means that the marginal product of the variable input decreases with decreasing fish stock , but is raised by capital accumulation .
the derivative of harvest with respect to the fish stock also increases in capital , because certain attributes of capital enhance the accessibility of the fish stock ( improved fishing gear and technology , increased horsepower of boats , etc . ) . since
@xmath0 and @xmath53 are given when a firm undertakes efforts @xmath52 to obtain a chosen @xmath5 , the resulting variable costs @xmath47 depend on @xmath59 and @xmath53 .
the function @xmath60 can be determined by solving @xmath61 for @xmath52 and multiplying it with a renumeration rate @xmath48 which is assumed to be fixed .
due to the assumptions made for @xmath62 , the implicit function theorem implies that @xmath63 additionally , we assume that the hessian of @xmath47 is positive definite , which is no contradiction to the above inequalities .
this guarantees usual convexity properties as needed later on .
it should be noted that the marginal harvesting costs @xmath64 decrease in both the fish stock and the capital stock .
this assumption differs from previous approaches where capital only sets an upper limit for the harvest , such that an increase in capital equipment improves the productivity of the variable inputs , and capital accumulation may offset the negative effect of a declining fish stock on harvesting costs .
we now turn to the dynamics of the economic and biological stock .
the regeneration of the resource is given by a concave recruitment function @xmath1 , for which the assumptions made in section [ sec2 ] are valid .
we define @xmath65 as the equation of motion , where @xmath5 denotes the harvest of a firm under consideration and @xmath66 that of all the others .
each firm s capital stock is described by @xmath67 where @xmath68 represents ( irreversible ) investment and @xmath69 the depreciation rate which is assumed to be constant .
investment costs are expressed by a strictly convex increasing function @xmath70 .
the convexity reflects inelastic supply of highly specialized equipment and rising adjustment costs for higher investment .
the demand for fish is described by the downward sloping inverse demand function @xmath71 . in the following the decision of each firm on @xmath5 and @xmath72
has to be determined .
if each fishing company acts in an economically rational way , it chooses an investment and harvest plan that maximizes the discounted profit given by @xmath73 subject to eq.([eq : lab7 ] ) and eq.([eq : lab6 ] ) . here , @xmath74 denotes a constant discount rate and @xmath75 $ ] a planning interval . the optimization problem can be solved by a sufficiency theorem of mangasarian(@xcite ) . by introducing @xmath76 and @xmath77 as costate variables for @xmath0 and @xmath53 ,
the current - value lagrangian is given by @xmath78 the four slack variables appear due to the constraints @xmath79 . since @xmath80 is concave in @xmath81 , an interior solution to the first - order conditions @xmath82 and the costate conditions @xmath83 and @xmath84 ( where @xmath85 ) with @xmath86 and @xmath87 is an optimal path .
we assume that marginal investment costs become small for a low investment level , i.e. @xmath88 , avoiding that there is a negative solution of @xmath89 since @xmath90 .
it is not as easy to guarantee a non - negative solution of @xmath91 , which depends on the relation between prices and variable costs .
however , we do not investigate the optimal path in detail since it may be unrealistic because firms are most likely to ignore the effect of their harvesting decision on future stocks @xcite .
in other words , they disregard eq.([eq : lab7 ] ) in their optimization procedure , partly because of a lack of knowledge on the recruitment function , and partly because they consider their own influence on the fish stock to be negligible .
moreover , they tend to assume that other firms behave in the same way .
thus , we suppose that the shadow price for the biological stock is neglected by the individual firms .
we further assume that the number of firms is constant and investment takes place in the form of increasing fishing power or number of vessels per firm . from this perspective
it is consistent to set @xmath92 in the fourth summand of the current - value hamiltonian eq.([eq : lab9 ] ) . as a consequence
, the costate condition for @xmath76 is ignored and corner solutions of @xmath93 become irrelevant . the harvest decision is myopic in contrast to the investment decision . utilizing the ( constant ) inverse price elasticity of demand @xmath94 , one obtains the following equations : @xmath95 according to eq.([eq : lab10b ] ) the costate variable on capital equals the marginal cost of investment . if the latter were constant , as assumed in several previous models , eq.([eq : lab10d ] ) would therefore boil down to the usual condition that the user cost of capital , @xmath96 , should just be balanced by the induced reduction in variable costs , @xmath97 . in our model with increasing marginal investment costs we get a more complicated equation . by substituting @xmath77 from eq.([eq : lab10b ] ) and its time derivative in eq.([eq : lab10d ] ) one gets @xmath98 , where @xmath99 .
the investment programme is characterized by the condition that the user cost of capital should be balanced not only by reduced variable costs , but also by the change in the purchase cost of capital induced by a change in the level of investment . because we have assumed that all firms are characterized by the same technology and behave in the same way , one obtains @xmath100 .
consequently , the total amount of capital and investment is given by @xmath101 and @xmath102 , respectively .
our model can therefore be written in the following way : @xmath103 equation ( [ eq : ls3 ] ) has already been interpreted above and eq.([eq : ls4 ] ) represents the usual equality between marginal variable costs and marginal revenue .
it should be recalled that the marginal variable costs @xmath64 decrease in both the fish stock and the capital stock .
therefore , an increase in marginal costs due to a decreasing fish stock may trigger additional investment in an effort to keep marginal costs from rising excessively .
this is crucial for explaining why the capital stock may increase along with a decrease in the fish stock , a phenomenon which is at the heart of what is frequently referred to as over - capitalization .
we will show in the next section that this happens for every parameterization of the model .
to analyse whether over - capacities occur or not , and whether the fish stock recovers once it is in a critical state , we are interested in the dynamics of the ode model eq.([eq : ls1])([eq : ls2 ] ) far from equilibrium .
due to the difficulties with phase - plane analysis and numerical simulation as discussed before , we study the qde corresponding to the model in the next sections .
the abstraction procedure commences as outlined in section [ sec2 ] . at first ,
landmarks for all variables are chosen and constraints about the functions are defined . for the resource stock @xmath0 , recruitment @xmath11 and harvest @xmath5 the same landmarks as in section [ sec2 ]
are introduced .
the constraints relating @xmath0 to @xmath11 and @xmath104 to @xmath11 and @xmath5 are the same as given in ( [ eq : hrule1])([eq : dotxrule3 ] ) and ( [ eq : dotxruleall ] ) .
for capital and investment the landmarks @xmath105 and @xmath106 are defined .
further descriptions of the following constraints are provided in the annex .
eq.([eq : ls4 ] ) for the marginal variable costs @xmath64 can be solved for @xmath5 to yield a harvest supply function @xmath107 .
this function is increasing in both arguments , which can be shown from the assumptions made for the production function @xmath108 and the inverse demand function @xmath109 , which implies @xmath110 the constraints following from eq.([eq : pardrv ] ) for @xmath111 are given by ( [ eq : vk ] ) , those from eqs.([eq : ls3 ] ) and ( [ eq : ls2 ] ) by ( [ eq : irule])([eq : iirule ] ) ( cf .
annex ) . for technical reasons
@xmath111 is replaced by @xmath112 .
this implies that @xmath97 can be expressed by a function which is strictly monotonically increasing in @xmath5 and strictly monotonically decreasing in @xmath0 and in @xmath53 ( cf .
eq.([eq : pardrv ] ) ) .
further on , we assume that @xmath113 and @xmath69 are constant . since @xmath88 , it can be shown that @xmath72 has always the same sign and qualitative direction as the expression @xmath114 .
therefore , this term of eq.([eq : ls3 ] ) can be simplified to @xmath72 qualitatively .
for the same reason @xmath115 can be replaced by @xmath53 ( in eq.[eq : ls3 ] ) , and @xmath116 by @xmath5 ( in eq.[eq : ls1 ] ) .
thus , the abstraction of the model and the associated constraints as given by eqs.([eq : ls1])([eq : ls2 ] ) can be expressed in the following relational form , where the right hand side represents the typical notation developed by ( * ? ? ?
annex for further details ) : @xmath117 the constraint @xmath118 is the qualitative abstraction of quantitative addition .
the @xmath119 constraint represents a downward bent u - shaped function of logistic type ( cf .
the landmarks in the brackets are corresponding pairs / triples of argument and result values specifying correspondences between variables , e.g. in the first equation of ( [ math : lab3 ] ) : if @xmath120 and @xmath121 , then @xmath122 .
the qualitative simulation of the bio - economic model ( see footnote[fn1 ] ) provides 467 qualitative states . according to the methodology described in section [ sec2 ]
the potential systems developments are structured and reorganized as a gstg ( fig.[fig : lab5 ] ) with stock size , harvest and capital as relevant variables .
the gstg contains 19 vertices , where one , represents a catastrophic equilibrium ( where @xmath123 ) . for
each behaviour represented in the gstg , corresponding phase plots for the relevant variables are available .
now , the question about the real - world validity of these trajectories and the added value for the discussion of the problems in fisheries arises .
our argumentation follows three directions : * in order to validate the results obtained it can be checked whether case studies can be reconstructed by observed time behaviours . as an example we discuss the development of the blue whale fishery in the time period 1946 - 1980 ( fig.[fig : lab6 ] ) . *
the qualitative approach determines all ` dynamic patterns ' which are conceivable under the model settings .
the gstg enables us to determine general properties common to all time developments .
* the gstg has the capabilities to discuss potential management interventions , e.g. for scenario analysis or for detection of serious developments . in this sense the approach supplies knowledge for an ex - ante assessment . .
( b ) displays the situation in the blue whale hunting industry from 19461980 ( bwu @xmath124 blue whale units ; the dashed lines refer to model outputs generated by the @xcite model , and @xmath51 to the number of firms in this model ) . ]
the example of blue whale fishery ( fig.[fig : lab6]b ) shows that the qualitative bio - economic model fits real situations quite well .
figures [ fig : lab6]a and [ fig : lab6]c display the associated phase plots of the qualitative model , indicating a development path corresponding to transitions from vertex # 1 to # 12 ( fig.[fig : lab5 ] , dashed - dotted and white arrows ) . both the observed quantitative and the qualitative case are characterized by an initial expansion phase where the whale stock declines while the capital stock increases .
this is followed by a period in which the capital stock declines rapidly , while the whale stock is still declining .
finally , the capital decreases further , whereas the stock tardily recovers .
this is completely different to several models ( e.g. clark et al .
1979 , mckelvey 1986 , cf .
fig.[fig : lab6]b ) which , due to linearity assumptions , show an initial and a final jump of the capital stock .
since the gstg contains many paths , we may wonder if every fishery showing an intial expansion phase can be reconstructed by the model .
however , in the following we identify further structural properties of the graph the model can only explain real - world fisheries which reflect these patterns . during the expansion phase
the stock decreases while capital is still increasing .
in the course of further evolution the effect of declining stocks on harvesting costs becomes so important that it can not compensated any longer by investment .
then the capital stock begins to decrease , i.e. net investment @xmath125 becomes negative . at this stage
it is possible ( fig.[fig : lab6]a ) that the fish stock approaches zero ( a discussion of such branching points is provided in section [ sec5_1 ] . ) .
however , it is also possible that the decline of the fish stock is reversed .
in contrast to models like those of @xcite and @xcite , in this case fisheries do not need to attain an equilibrium with @xmath126 , but can exhibit a further boom - and - bust cycle through all quadrants .
it is not even safeguarded that the system converges to an equilibrium after repeated cycles .
we can also show that every fishery described by the model necessarily undergoes a phase of over - capitalization .
this is related to the fact that via the simulation irreversible transitions are identified , i.e. transitions between qualitative states which are possible only in one direction ( cf .
gstg , e.g. , from quadrant ii to quadrant iii ) .
the proof of this proposition is as follows : decreasing catches and increasing capital stock occur simultaneously , whenever the system approaches the critical vertices # 3 , # 6 , # 8 or # 18 .
if the system starts , for example , at vertex # 1 ( relatively undisturbed stock ) and does not directly shift to vertex # 18 , the only way to avoid vertex # 3 is to change from vertex # 2 to # 5 .
however , at # 5 the only way for a further development ( without collapse of the stock ) is via vertex # 6 .
thus , we always approach at least one of the critical vertices and in this sense over - capitalization is an unavoidable system property .
this property is rooted in the fact that @xmath127 and @xmath128 , i.e. that the harvest supply function increases in @xmath0 and @xmath53 ( via eq.([eq : ls4 ] ) ) .
as long as we observe increasing harvest although the fish stock is reduced , net investment must be positive to compensate losses from increasing marginal costs .
therefore , @xmath53 can not start to decrease before @xmath5 .
discussing the results and implications for policy actions , a variety of conclusions can be drawn .
the model is further validated by showing that mistakes arising from temporary management measures can be reconstructed by the model .
public decision makers may respond rather late to an emerging crisis and , in addition , with drastic , but temporary interventions .
such situations can be analysed with the model as follows : although the system does not follow the dynamics of an unregulated fishery during a period managed in this way , we can compare the qualitative state of the system before and after this time frame .
two such interventions are represented by the bold arrows in fig.[fig : lab5 ] starting at the dashed vertex # 8 and addressing the well - known problems in the cod fisheries of the north atlantic ocean ( grand bank and barents sea ) .
the situation in the grand bank region and in the barents sea exhibited similarities , but the results of political interventions in these two cases were different . whereas in the canadian case the worst consequences ensued , the norwegian authorities were able to prevent a severe disaster .
the question arises , how and why ?
the cod fiasco in the canadian waters at the beginning of the 1970s is an example of an maladjusted strategy leading to a severe economic and environmental catastrophe .
anticipating an ecological disaster , the exclusive economic zone ( eez ) was introduced in 1977 and foreign trawlers were banned from the grand banks .
the government implemented catch quotas which , however , were not always binding . in parallel
, the government responded with massive subsidies to the fishery .
the earliest situation when such a situation occurs in the gstg is at vertex # 6 where the harvest is above the sustainable level and the stock already below .
the only way to avoid more serious developments is to alter to vertex # 7 ( and from there to # 9 , # 11 , etc . ) , but this needs a reduction of the capital .
indeed the canadians did just the opposite , i.e. the awareness of increasing gains due to the ban on european trawlers led to investment in improved gear and to more efficient fish killing@xcite .
the situation at the end of the 1970s is represented by vertex # 8 ( fig.[fig : lab5 ] ) , where both stocks and catches are below a critical level and still decreasing while the capital continues to increase .
the reaction of the government in this situation was counterproductive ( cf .
bold black arrow in fig.[fig : lab5 ] ) .
it may be that the stocks slightly recovered ( possibly indicated by an increased catch , cf .
fig.[fig : lab7 ] ) , but they have not reached a sustainable level ( i.e. @xmath7 ) .
thus , when the total allowable catches ( tac ) were relieved , the system was not in a safe state , but approached a state represented in the model results by vertex # 13
. a second period of high catch levels occurred , which led to the critical state # 8 again ( along the path # 13#10#8 or via # 13#10#5#6#8 ) , finally approaching via # 9 .
this was the end of the story , because the cod stock has not recovered up to the present day . to prevent these short - termed cycles
, it would be better to shift the system to vertex # 15 .
this was the case in norway when the country faced a similar disaster in the barents sea . here
, individual transferable quotas helped to reduce the race for fish .
although there were a lot of bankruptcies in fisheries and many demonstrations , the politicians knew that there could be no giving in to protests for short - term political gain .
they set up subsidies in contrast to the canadians to remove ships from fisheries and to diversify the coastal economy@xcite .
finally , they banned fishing from spawning grounds .
this is equivalent to a path sequence via # 8#9#11#14#15 .
thus , norway now has a better basic position for a sustainable resource management .
however , a slackening of catch permits , as currently discussed , may lead to the outcome that the unregulated dynamics starts again at vertex # 15 .
we conclude from these two cases that short - term measures or non - binding catch quotas only suspend the intrinsic problem .
the gstg provides various clues on how management interventions can be implemented .
although some of them may not differ so much from earlier conclusions ( cf . those derived from the static model of
* ) , they are here based on a comprehensive system analytical background and a full dynamical perspective . for example , irreversible transitions indicate critical stages in the development of the fishery .
the fishery is in a ` high - risk situation ' when stock and harvest are decreasing , but capital is still increasing and if at least one irreversible transition occurred ( from # 3 to # 6 , or form # 4 to # 7 ) .
critical branchings are states characterized by multiple successors , where at least one of them is an irreversible transition . here , either a worsening or mitigation is possible ( e.g. # 1 , # 2 , # 4 , # 9 , or # 13 ) .
fishery management can use the information represented by the gstg and take account of the fisheries problems in two ways : 1 .
how can it be avoided that the fishery returns again to critical branching after the resource has recovered ? 2 . under
which conditions can critical branchings be prevented at an early stage , i.e. that the system does not leave quadrants i and ii ?
we can consider control measures for harvest , capital , or stock , but also - in a qualitative sense - the time when they have to be applied at the latest .
it would also be helpful to assign likelihoods to different transitions at these branchings .
although results of the latter kind are problematic in this context ( cf .
sect.[sec2 ] , last paragraph ) , the qualitative model provides a starting point for investigations in this direction .
we illustrate these issues by referring again to the example of the blue whale fishery ( dashed - dotted arrows , fig.[fig : lab5 ] ) .
assume that we start with a relatively undisturbed stock and an industry at a low level ( # 1 , quadrant i , @xmath129 ) .
the first irreversible transition occurs from vertex # 1 to # 2 ( quadrant ii ) .
already here the catch exceeds the sustainable level ( @xmath130 ) , while the capital stock is still increasing . to avoid such a situation ,
strong harvest limitations must be introduced , i.e. the harvest rate must decrease before @xmath6 is reached and the stock approaches @xmath7 , while over - capitalization might continue .
for such a transition to node # 18 the following scenario is conceivable : the harvest costs already strongly increase due to a reduced population although it is still above @xmath7 and/or the substitution of resource by capital stock has only minor effects . in reality
this is observable for less developed fisheries . for highly developed fisheries
the system evolves to vertex # 2 ( quadrant ii , @xmath131 , cf .
fig.[fig : lab6 ] ) representing fisheries having the potential to increase harvest and capital ( as , for instance , the blue whale fisheries between 1946 - 1960 ) .
a safe development under these conditions can only be guaranteed if a transition to vertex # 5 or # 6 is avoided . at vertex # 2
this is possible via # 3 , where we are still in quadrant ii , but the situation is economically worse than at vertex # 2 , because in addition to declining fish stocks , harvest is also decreasing .
sustainable limits can only be reached by massive harvest limitations ( evolution via # 4#17 ) . if these actions are not implemented , the transition to # 5 or # 6 is inescapable and fisheries necessarily enter to a situation in which the industry and the fish stock are likely to be ruined .
the further development depends on how early net investment decreases , i.e. in the phase when quadrant iii is reached , effort controls are more important than catch controls .
in contrast in quadrant iv , successful recovery is determined by the regeneration rate of the fish stock and the speed of harvest reduction .
since policy actions often have the tendency to come into play too late , over - capitalization and therefore over - exploitation is a permanent risk .
qualitative simulation shows that at these vertices the collapse of the fishery can be prevented if , due to a rapid capital reduction , the stock already recovers ( indicated by the white arrows , fig.[fig : lab5 ] ) . in general
the simulation results indicate that fisheries under the settings of the model are in a state of perpetual risk , because the high risk states ( # 5#9 ) are likely to occur repeatedly in every boom - and - bust cycle .
we conclude that if it is impossible to formulate a precise numerical model qualitative simulation has complementary advantages compared to other methods .
first , we can shift the perspective from equilibria to non - equilibrium dynamics and can unveil general dynamic patterns , i.e. intrinsic properties of unregulated fisheries which hold for every functional and parametric specification of the model and which can not be identified by comparative dynamics in a neighbourhood of the equilibrium .
examples are unavoidable overcapacities and the possibility of cyclical behaviour .
second , several stages of system development can be identified to enhance our knowledge of how and when management strategies should be introduced .
this allows alternative scenarios to be discussed .
third , critical branchings were identified .
these can be associated with regions in the phase space where the qualitative direction of state variables are in a configuration which admits irreversible problematic and positive changes . in our model ,
factors for the propensity of the system to recover include ( cf .
eqs.([eq : ls1])([eq : ls2 ] ) ) : ( i ) regeneration rate of the fish stock , ( ii ) depreciation of the capital stock , ( iii ) marginal investment costs and ( iv ) marginal variable costs with respect to capital and harvest .
here , qualitative reasoning comes to its limits , since no numerical estimates for critical parameters can be made .
yet we think that the method helps to identify decisive regions of the phase space as a starting point for the development of new hypotheses focusing specifically on them .
semi - quantitative techniques combining qualitative and quantitative methods show promise here ( cf . * ? ? ?
thus , qualitative reasoning is appropriate whenever we are dealing with imprecise knowledge . given the complexity of the systems in question it must be accepted that one may have to be content with `` soft prognoses '' only .
this paper addresses the global problem of industrial unregulated fisheries and the role of capital accumulation .
such systems are often intrinsically complex and the understanding of them is limited by low levels of knowledge with respect to both biological and economic properties . to keep models tractable ,
previous analytical approaches have relied on a variety of simplifying assumptions with respect to investment costs , harvesting costs or industry structure . additionally , they often concentrated on equilibrium analysis or on comparative dynamics near equilibrium .
we demonstrated qualitative modelling as a new method to approach uncertainty and tractability problems by applying it to a model which improves former results by relaxing assumptions .
the qualitative model describes the dynamical behaviour of a fishery without reference to quantitative values .
we have shown that this technique can improve our reasoning about global properties of the dynamics of the system .
the model features ( i ) increasing marginal investment costs and ( ii ) marginal harvesting costs that are decreasing in both the fish and the capital stock . due to the former
, the build - up of fishing capacities can be modelled more realistically . due to the latter ,
fish stock and capital stock are ( incomplete ) substitutes in the production of catch .
this proved to be the key factor in explaining why capital will keep rising while both the fish stock and the catch decline .
the qualitative simulation reveals that all fisheries described by the general model necessarily undergo a phase of over - capitalization .
it also shows other inherent and in the sense of political interventions
serious system properties , e.g. critical branchings and potentially recurring boom - and - bust cycles .
future work will run along several lines .
different policy measures can be assessed by incorporating them into the qualitative model and comparing the resulting graphs .
also a multi - species module , together with more detailed models of decision making in fisheries could enhance the results .
some other efforts are related to an integration of hard and soft knowledge in one model approach . in particular model approaches comprising a module for cooperative negotiations and
allows to test which management strategies are applicable in a fishery under additional settings , e.g. normative sustainability targets , are examined ( see , e.g. * ? ? ?
* ; * ? ? ?
summing up , we feel that the results and the technique open a promising road towards new insights in the dynamics and management of fishery systems . *
acknowledgements : * this work was supported by the german federal ministry for education and research ( bmbf ) under grant number 03f0205b ( diwa ) .
we wish to thank the anonymous reviewers for their helpful comments .
the annex completes the qualitative constraints derived from the bio - economic model ( cf .
eqs.([eq : ls1])([eq : ls2 ] ) ) and the assumptions made in section [ sec3 ] .
these constraints are translated to the specific model code necessary to run the qualitative simulation software . the relationship between stock @xmath0 and harvest @xmath5
is defined by the relations in eq.([eq : hrule1 ] ) and expressed by constraint ( c.1a ) in the subsequent table .
the recruitment is given by a u - shaped function @xmath1 , defined in eqs.([eq : rrule1])([eq : rrule6 ] ) and expressed by constraint ( c.1b ) .
to describe the change of fish stock @xmath37 , the complete model needs more than eq.([eq : dotxrule3 ] ) .
the derivative @xmath104 is increasing with @xmath11 and decreasing with @xmath5 .
in addition , if @xmath5 becomes zero , @xmath104 changes in the same direction as @xmath11 and if @xmath11 vanishes , @xmath104 changes in the opposite direction to @xmath5 . since @xmath132 , @xmath120 if @xmath122 and @xmath121 : @xmath133 these expressions are equivalent to those which are encoded in the qualitative constraint ( c.1c ) .
the constraints for the harvest supply function @xmath107 are already defined in eq.([eq : krule ] ) and expressed by constraint ( c.2a ) .
in addition , eq.([eq : pardrv ] ) states that the negative marginal costs @xmath97 increase in @xmath5 and decrease in @xmath0 and @xmath53 .
thus , @xmath134 which is expressed by the constraint ( c.2b ) , where @xmath111 is replaced by @xmath135 for technical reasons . taking into account eq.([eq : ls2 ] ) it is obvious that the change of capital stock is increasing with @xmath72 and decreasing with @xmath53 .
if @xmath53 vanishes , @xmath136 changes in the same direction as @xmath72 , if @xmath72 vanishes , @xmath136 changes in the opposite direction to @xmath53 , yielding @xmath137 the relationship between @xmath136 and @xmath72 is represented by ( c.3 ) . finally , referring to eq.([eq : ls3 ] ) , the change of investment has the form @xmath138 with strictly monotonic increasing functions @xmath62 and @xmath139 which vanish at zero .
thus , @xmath140 which is encoded in the constraint ( c.4 ) .
again we change @xmath111 to @xmath97 .
the constraints , expressed by a set of comprehensible and recurring keywords , are summarized in the following table : banks , r. , 1999 .
subsidising eu fleets : capacity reduction or capital subsidisation . in : hatcher , a. , robinson , k. ( eds . ) ,
overcapacity , overcapitalisation and subsidies in european fisheries .
cemare , portsmouth , pp .
200210 .
jackson , j. b. c. , kirby , m. x. , berger , w. h. , bjrndal , k. a. , botsford , l. w. , bourque , b. j. , bradbury , r. h. , cooke , r. , erlandson , j. , estes , j. a. , hughes , t. p. , kidwell , s. , lange , c. b. , lenihan , h. s. , pandolfi , j. m. , peterson , c. h. , steneck , r. s. , tegner , m. j. , warner , r. r. , 2001 . historical overfishing and the recent collapse of coastal ecosystems .
science 293 ( 5530 ) , 629638 .
kropp , j.p .
, zickfeld , k. , eisenack , k. , 2002 . assessment and management of critical events : the breakdown of marine fisheries and the north atlantic thermohaline circulation . in : bunde ,
a. , kropp , j.p . ,
schellnhuber , h. j. ( eds . ) , the science of disasters : climate disruptions , heart attacks , and market crashes .
springer verlag , berlin , pp .
192216 .
kropp , j. p. , eisenack , k. , scheffran , j. , 2004 .
sustainable marine resource management : lessons from viability analysis . in : path - wostl ,
c. , schmidt , s. , rizzoli , a. e. , jakeman , a. j. ( eds . ) , complexity and integrated resources management , trans .
2nd biennial meeting of the int .
software soc . ,
manno , switzerland , pp .
104109 .
mckelvey , r. , 1986 .
fur seal and blue whale : the bioeconomics of extinction . in : cohen , m. ( ed . ) , applications of control theory in ecology .
lecture notes in biomathematics .
springer verlag , berlin , pp .
5782 .
munro , g. r. , 1999 .
the economics of overcapitalisation and fishery resource management : a review . in : hatcher ,
a. , robinson , k. ( eds . ) , overcapacity , overcapitalisation and subsidies in european fisheries .
cemare , portsmouth , pp .
723 .
scheffran , j. , 2000 .
the dynamic interaction between economy and ecology - cooperation , stability and sustainability for a dynamic - game model of resource conflicts . mathematics and computers in simulation 53 , 371380 .
trelease , r. , henderson , r. , park , j. b. , 1999 . a qualitative process system for modeling nf-@xmath141b and ap-1 gene regulation in immune cell biology research .
artificial intelligence in medicine 17 ( 3 ) , 303321 . | capital accumulation has been a major issue in fisheries economics over the last two decades , whereby the interaction of the fish and capital stocks were of particular interest .
because bio - economic systems are intrinsically complex , previous efforts in this field have relied on a variety of simplifying assumptions .
the model presented here relaxes some of these simplifications .
problems of tractability are surmounted by using the methodology of qualitative differential equations ( qde ) .
the theory of qdes takes into account that scientific knowledge about particular fisheries is usually limited , and facilitates an analysis of the global dynamics of systems with more than two ordinary differential equations .
the model is able to trace the evolution of capital and fish stock in good agreement with observed patterns , and shows that over - capitalization is unavoidable in unregulated fisheries . |
large brightness changes of young stars were seen long before the class of objects was understood to be related to star formation , or even that stars formed @xcite .
the largest and most prominent physical changes are decades - long bursts of accretion at rates of @xmath5 m@xmath6 yr@xmath7 , called fuor objects ( after fu ori ; see reviews by * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , and months - long bursts of accretion at @xmath8 m@xmath6 yr@xmath7 @xcite , called exor outbursts ( after ex lup ; * ? ? ?
the fuor and exor classes of outbursts form our framework for interpreting large luminosity increases of young stars ( see review by * ? ? ?
the different timescales for the longer ( fuor ) and shorter ( exor ) outbursts suggest that they are different phenomena .
fuor outbursts are thought to be triggered by instabilities in the disk ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) at radii with rotational timescales of decades .
on the other hand , the shorter exor outbursts are thought to be triggered by instabilities in the magnetic connection between the star and disk @xcite .
these two phenomena require different disk structures , which is supported with spectroscopic evidence .
viscously heated , optically thick fuor disks produce low - gravity spectra similar to those of supergiant stars , including deep absorption in co overtone bands at 2.3 @xmath9 m ( e.g. , * ? ? ?
exors show a forest of optical emission lines , evidence of magnetospheric accretion , and a warm disk surface layer that produces strong co emission ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
since measuring the duration of an outburst usually requires impatient people to wait , these spectroscopic proxies are used to immediately discriminate between fuor and exor outbursts . in practice ,
outbursts of protostars are often forced into the exor / fuor classification scheme , even when the outburst does not fit well into either category .
many outbursts appear to be intermediate between the fuor and exor classes ( e.g. , * ? ? ?
* ) , despite the very different mechanisms and masses thought to be involved . in other cases , some characteristics of the outburst are inconsistent with this classification scheme @xcite .
either the classification system groups together diverse physics , or the same accretion burst physics of exors produces a wider range of observed phenomena than expected .
powerful new transient surveys are now leading to the discovery of 12 large outbursts of young stellar objects ( ysos ) each year ( e.g. , * ? ? ?
* ; * ? ? ?
a recent outburst of a candidate young star , asassn-15qi , was identified by the all - sky automated survey for supernovae ( asas - sn ) variability survey @xcite .
follow - up observations were obtained because the object was suspected to be young ( see discussion in * ? ? ?
* ) based on its projected spatial location near the h ii regions sh 2 - 148 and sh 2 - 149 and the small , low extinction molecular cloud tgu 676/dobashi 3359 @xcite .
this region is a few degrees southwest of the main cepheus star forming complexes , including ldn 1218 ( e.g. , * ? ? ?
* ; * ? ? ?
a distance of @xmath10 kpc is adopted from the parallax measured for [email protected] , which is visually located 1.5 deg from asassn-15qi and shares the same radial velocity @xcite .
initial optical spectroscopy of asassn-15qi showed p cygni profiles characteristic of strong winds , an indicator of accretion - ejection processes that occur on young stars @xcite .
however , subsequent low resolution near - infrared ( ir ) and high resolution optical spectra revealed a confusing picture of an outburst that does not neatly fit into either the fuor or exor category @xcite .
spatially extended emission in optical imaging @xcite confirmed that the object is likely young .
the source quickly faded from its peak @xcite . in this paper , we analyze optical and near - ir photometry and spectroscopy of the 2015 outburst of asassn-15qi . the outburst is especially remarkable because of ( 1 ) the dramatic 3.5 mag brightening in less than one day , and ( 2 ) the presence of a fast wind that either caused or was produced by the outburst .
the wind faded as the outburst decayed over 3 months . in 2 , we describe the wide array of observations used in this paper . in 3 , we analyze these data to arrive at some empirical conclusions . in 4 , we summarize the source properties , compare these properties to those of known outbursts where the physics is better understood , and describe some possible alternatives .
these descriptions and speculations are summarized in 5 .
the outburst of asassn-15qi ( 2mass j22560882 + 5831040 ) occurred on jd 2,457,298 ( 2 october 2015 ; utc dates are used herein ) .
this section describes the follow - up photometry and spectroscopy obtained for 3 months after the burst occurred .
archival photometry is listed in table [ tab : photarchive ] , while photometry obtained or measured as part of this paper is listed in tables [ tab : asassnphot][tab : nirphot ] .
the log of spectroscopic observations of asassn-15qi is listed in table [ tab : speclog ] . lcccc telescope / survey & epoch & band & mag & ref . + usno - a2 & 1953.83 & @xmath12 & 18.2 & 1 + usno - b1 & 1976.5 & @xmath12 & 17.81 & 2 + usno - b1 & 1976.5 & @xmath12 & 17.41 & 2 + usno - b1 & 1976.5 & @xmath13 & 14.62@xmath14 & 2 + usno - b1 & 1976.5 & @xmath13 & 14.76@xmath14 & 2 + usno - b1 & 1976.5 & @xmath15 & 14.22@xmath14 & 2 + gsc & 1989.669 & @xmath12 & 18.51 & 3 + gsc & 1989.669 & @xmath16 & 17.84 & 3 + gsc & 1989.669 & @xmath1 & 16.50 & 3 + gsc & 1989.669 & @xmath17 & 14.33 & 3 + 2mass & 2000.06.21.39 & @xmath18 & 13.70 & 4 + 2mass & 2000.06.21.39 & @xmath19 & 12.92 & 4 + 2mass & 2000.06.21.39 & @xmath20 & 12.65 & 4 + ptf & 2009.07.31.4 & @xmath21 & 17.06 & 5 + ptf & 2009.08.25.3 & @xmath21 & 17.79 & 5 + ptf & 2009.12.17.1 .
& @xmath21 & 18.01 & 5 + iphas & 2004.08.25 & @xmath22 & 16.66 & 6 + iphas & 2004.08.25 & @xmath23 & 15.59 & 6 + iphas & 2004.08.25 & h@xmath24 & 16.04 & 6 + tautenburg 1.34 m & 2004.09.09.90 & @xmath15 & 15.08 & 7 + leicester 0.5 m & 2010.10.09.96 & @xmath15 & @xmath25 & 8 + tautenburg 1.34 m & 2015.10.11.81 & @xmath12 & @xmath26 & 8 + tautenburg 1.34 m & 2015.10.11.81 & @xmath1 & @xmath27 & 8 + tautenburg 1.34 m & 2015.10.11.81 & @xmath28 & @xmath29 & 8 + tautenburg 1.34 m & 2015.10.11.81 & @xmath15 & @xmath30 & 8 + tautenburg 1.34 m & 2015.11.05.77 & @xmath12 & @xmath31 & 7 + tautenburg 1.34 m & 2015.11.05.77 & @xmath1 & @xmath32 & 7 + tautenburg 1.34 m & 2015.11.05.77 & @xmath13 & @xmath33 & 7 + tautenburg 1.34 m & 2015.11.05.77 & @xmath15 & @xmath34 & 7 + tautenburg 1.34 m & 2015.11.13.99 & @xmath12 & @xmath35 & 7 + tautenburg 1.34 m & 2015.11.13.99 & @xmath1 & @xmath36 & 7 + tautenburg 1.34 m & 2015.11.13.99 & @xmath28 & @xmath37 & 7 + tautenburg 1.34 m & 2015.11.13.99 & @xmath15 & @xmath38 & 7 + irtf & 2015.10.15.44 & @xmath18 & @xmath39 & 9 + irtf & 2015.10.15.44 & @xmath19 & @xmath40 & 9 + irtf & 2015.10.15.44 & @xmath20 & @xmath41 & 9
+ + + + + + + + + + + + the asas - sn is an all - sky @xmath1-band transient survey with a limiting magnitude of @xmath42 .
asas - sn photometry of asassn-15qi was obtained from two different telescopes on haleakala , hawaii with intervals of 13 days .
the outburst of asassn-15qi was first detected and reported in the asas - sn transient list @xcite .
asas - sn monitoring includes extensive photometry before and during the outburst .
the quiescent brightness of asassn-15qi is near the limiting magnitude of asas - sn . as a consequence ,
flux measurements prior to outburst required a different technique than the standard asas - sn photometry available online .
we first measured the flux in a reference image , and subsequently calculate the flux in difference images .
the fluxes are then extracted with a 2-pixel ( @xmath43 ) aperture radius .
small offsets may occur between the two telescopes .
the asas - sn photometry includes the central source and all nearby nebulosity because of the large pixels .
a selected subsample of asas - sn photometry is listed in table [ tab : asassnphot ] ; all photometry is available online .
ccc jd@xmath112,457,000 & @xmath1 ( mag ) & @xmath44 error + 275.91551 & 17.37 & 0.30 + 277.96895 & 16.88 & 0.17 + 288.89590 & 17.22 & 0.29 + 297.87394 & 17.37 & 0.34 + 297.88712 & 16.99 &
0.25 + 298.77584 & 13.53 & 0.02 + 298.78226 & 13.60 & 0.02 + 300.83682 & 13.84 & 0.02 + 304.95129 & 14.70 & 0.04 + 305.90914 & 14.73 & 0.04 + 309.81402 & 14.75 & 0.04 + 309.84820 & 14.81 & 0.04 + 314.80541 & 15.03 & 0.05 + 317.87615 & 15.29 & 0.07 + 319.69468 & 15.26 & 0.09 + 323.81185 & 15.36 & 0.09 + 324.84534 & 15.27 & 0.07 + 327.85558 & 15.44 & 0.07 + 333.76965 & 15.52 & 0.07 + 342.82343 & 15.73 & 0.09 + 344.73217 & 15.93 & 0.09 + 360.72276 & 15.90 & 0.08 + 361.79525 & 15.87 & 0.09 + 363.79191 & 15.93 & 0.09 + cccccccccc date & & & seeing + jd2,457,000 & @xmath12 & @xmath1 & @xmath22 & @xmath23 & @xmath45 & @xmath46 & @xmath47 & @xmath48 & ( @xmath49 ) + + 318.57 & 16.38 & 15.15 & 14.63 & 14.04 & 0.31 & 0.18 & 0.23 & 0.20 & 3.2 + 322.57 & 16.70 & 15.46 & 14.96 & 14.30 & 0.49 & 0.41 & 0.28 & 0.28 & 2.8 + 324.67 & 16.99 & 15.66 & 15.12 & 14.42 & 0.53 & 0.35 & 0.36 & 0.27 & 2.5 + 330.67 & 16.95 & 15.63 & 15.08 & 14.40 & 0.52 & 0.31 & 0.32 & 0.28 & 2.1 + 333.62 & 16.91 & 15.63 & 15.05 & 14.39 & 0.50 & 0.28 & 0.29 & 0.27 & 2.0 + 336.67 & 17.18 & 15.90 & 15.26 & 14.56 & 0.54 & 0.45 & 0.42 & 0.37 & 1.8 + 339.67 & 17.46 & 15.93 & 15.31 & 14.61 & 0.26 & 0.28 & 0.35 & 0.29 & 2.5 + 351.61 & 17.50 & 16.12 & 15.54 & 14.79 & 0.36 & 0.29 & 0.29 & 0.26 & 2.4 + 356.67 & 17.78 & 16.31 & 15.65 & 14.93 & 0.31 & 0.15 & 0.26 & 0.24 & 3.1 + 356.73 & 17.66 & 16.18 & 15.63 & 14.91 & 0.53 & 0.27 & 0.25 & 0.24 & 2.9 + 360.62 & 17.54 & 16.20 & 15.57 & 14.85 & 0.42 & 0.24 & 0.28 & 0.30 & 2.7 + 365.70 & 17.67 & 16.33 & 15.72 & 14.98 & 0.58 & 0.34 & 0.29 & 0.26 & 2.6 + 370.55 & 17.80 & 16.26 & 15.67 & 14.92 & 0.25 & 0.14 & 0.14 & 0.18 & 3.9 + 374.65 & 17.89 & 16.24 & 15.65 & 14.90 & 0.21 & 0.23 & 0.20 & 0.22 & 2.4 + 378.62 & 17.70 & 16.27 & 15.70 & 14.96 & ... & 0.23 & 0.20 & 0.23 & 2.3 + 379.60 & 17.82 & 16.27 & 15.76 & 15.02 & 0.27 & 0.30 & 0.20 & 0.21 & 2.2 + + 387.63 & & 16.18 & & & & 0.09 & & 4.5 + 390.66 & & 16.15 & & & & 0.10 & & 4.2 + 392.65 & & 16.17 & & & & 0.16 & & 3.5 + 397.63 & & 16.31 & & & & 0.13 & & 4.6 + 399.62 & & 16.26 & & & & 0.11 & & 4.6 + 401.61 & & 16.29 & & & & 0.05 & & 5.6 + 405.64 & & 16.35 & & & & 0.17 & & 3.8 + 408.59 & & 16.37 & & & & 0.15 & & 4.1 + 415.60 & & 16.45 & & & & 0.09 & & 5.1 + 424.62 & & 16.52 & & & & 0.09 & & 4.3 + 427.63 & & 16.55 & & & & 0.01 & & 5.5 + 554.90 & ... & 16.71 & ( 15.74 ) & ( 15.65 ) & .. & 0.07 & 0.10 & 0.06 & 3.5 + + 504.69 & 18.86 & 17.29 & 16.51 & 15.68 & 0.34 & 0.25 & 0.24 & 0.13 & 1.5 + 551.71 & 18.99 & 17.31 & 16.52 & 15.66 & 0.18 & 0.21 & 0.20 & 0.13 & 1.3 + + we obtained @xmath50 images with the las cumbres observatory global telescope network ( lcogt ; * ? ? ?
* ) 1-m telescope at mcdonald observatory between 2015 october 23 and december 23 and at the liverpool telescope on 2016 april 26 and 2016 june 11 .
additional optical photometry in approximately the @xmath1 band was obtained by joseph brimacombe in 2015 december 2016 january from new mexico skies , new mexico , usa using a 43-cm planewave cdk telescope on a software bisque pme ii mount with an sbig stl-6303 camera and astrodon johnson - cousins photometric filters .
results are presented in table [ tab : photometry ] .
the astrometry for these images was obtained using astrometry.net @xcite .
the pixel scales are @xmath51 for lcogt , @xmath52 for the liverpool telescope , and @xmath53 for the brimacombe photometry .
relative source fluxes were measured in apertures with a @xmath54 radius using the _ apphot _ task in iraf @xcite .
the absolute calibration was then calculated from the aavso photometric all - sky survey @xcite .
calibration uncertainties are typically 0.1 mag in @xmath12 , 0.06 mag in @xmath1 , 0.05 mag in @xmath22 , and 0.07 mag in @xmath23 .
random statistical errors introduced by photon noise and other relative calibration uncertainties are @xmath55 mag per band . however ,
systematic uncertainties of 0.10.5 mag exist and are dominated by methodological differences in how the nebulosity affects the flux measurement ( see examples for _ swift _ data in table [ tab : swift ] ) .
these uncertainties are minimal near the outburst peak but significant when the central source is weak . for the liverpool imaging , obtained 207 days after outburst peak , point - spread function ( psf ) photometry and photometry with a @xmath56 radius extraction region both yield brightnesses that are @xmath57 mag fainter in all bands .
cccccccccc date & post - burst & obsid & measurement@xmath58 & @xmath59 & @xmath60 & @xmath61 & @xmath62 & @xmath12 & @xmath1 + 12 oct 2015 & 10 days & 00034098001 & @xmath63 ( s ) & 314.7 & 1231.9 & 157.3 & 78.5 & 78.5 & 78.5 + & & & @xmath54 aperture ( mag ) & 19.12 & 20.70 & 17.97 & 16.85 & 16.24 & 14.94 + & & & gaussian fit ( mag ) & 19.67 & 20.73 & 18.41 & 16.95 & 16.25 & 15.12 + 27 dec 2015 & 87 days & 00034098002 & @xmath63 ( s ) & 1633.1 & & 817.2 & 407.8 & 451.3 & 407.8 + & & & @xmath54 aperture ( mag ) & & & 20.17 & 18.84 & 18.10 & 16.53 + & & & gaussian fit ( mag ) & @xmath64 & & 20.24 & 19.02 & 18.22 & 16.93 + + + .near - ir and mid - ir photometry [ cols="<,^,^,^,^,^",options="header " , ] ccccccccccc telescope & instrument & wavelength ( @xmath9 m ) & slit width ( @xmath49 ) & resolution & @xmath63 ( s ) & jd@xmath58 & post - burst@xmath58 & @xmath65 + kanata 1.5 m & howpol & 0.450.90 & 2.2 & 400 & 300 & 303.499 & 5.5 & 14.2 + bisei 1.01 m & ... & 0.400.80 & 2.0 & 1300 & 2400 & 304.014 & 6.0 & 14.3 + nasa / irtf & spex & 0.72.5 & @xmath66 & 1300 & 1440 & 310.940 & 12.9 & 14.8 + lick / shane & kast & 0.351.0 & 2 & 600 & 900 & 315.5 & 17.5 & 15.0 + mcdonald 2.7 m & igrins & 1.44 - 2.5 & 1 & 40000 & 4800 & 318.599 & 20.6 & 15.1 + keck - i & hires & 0.480.90 & 0.86 & 45000 & 1200 & 322.267 & 24.7 & 15.4 + mdm & osmos & 0.430.68 & 1.2 & 1800 & 1200 & 338.675 & 40.7 & 15.7 + keck - i & hires & 0.480.90 & 1.15 & 45000 & 1800 & 377.226 & 79.7 & 16.3 + keck - i & mosfire & 2.02.3 & 1.1 & 3000 & 400 & 386.245 & 88.7 & 16.3 + liverpool & sprat & 0.500.75 & @xmath67 & 350 & 500 & 392.395 & 94.4 & 16.3 + liverpool & sprat & 0.500.75 & @xmath67 & 350 & 1000 & 504.68 & 206.7 & 17.1 + lbt & mods & 0.350.9 & @xmath68 & @xmath69 & @xmath69 & 551.94 & 353.9 & 17.3 + + + we triggered the _
swift _ satellite to obtain images with the ultraviolet / optical telescope ( uvot ; * ? ? ?
* ) on 12 october 2015 ( i d 00034098001 ) and 27 december 2015 ( i d 00034098002 ) .
the uvot images have a field - of - view of @xmath70 with @xmath71 pixels and were obtained with filters covering 20006000 .
count rates were converted to fluxes and magnitudes using the zero points calculated by @xcite .
asassn-15qi was detected in all filters on 12 october and in the @xmath72 , @xmath62 , @xmath12 , and @xmath1 filters on 27 december 2015 .
the december observations were longer integrations , excluded the @xmath73 filter , and had a non - detection in the shortest wavelength filter , @xmath74 . in both epochs ,
the @xmath12 and @xmath1 emission is spatially extended beyond the psf measured from nearby stars ( see 3.1 ) .
other bands have too few counts to detect any spatial extent .
the counts are measured from aperture photometry with @xmath54 ( radius ) extraction regions and separately from gaussian fits to the object along the horizontal and vertical axes of the detector .
aperture photometry produces results that are consistent with those obtained from lcogt , but with larger fractional errors than gaussian fits for marginal detections .
psf photometry is also not applicable to spatially extended emission . in the gaussian fits ,
counts are summed over @xmath75 pixels in both directions on the detector , are fit with one - dimensional ( 1d ) gaussian profiles , and then have counts averaged between these two fits .
results from both methods are listed in table [ tab : swift ] .
we observed asassn-15qi in the @xmath76 bands with the notcam camera at the nordic optical telescope on 22 december 2015 ( table [ tab : nirphot ] ) .
the notcam data were reduced with an external iraf package notcam version 2.5@xmath77 .
asassn-15qi was also observed three times in the @xmath19 band with the liverpool telescope .
bright , isolated , single stars across the field were used to construct a characteristic psf for each mosaic image .
the psf photometry of all the sources was obtained using the _ allstar _ task in iraf .
the zero - point magnitude was calculated using 2mass photometry of isolated stars and was then applied to asassn-15qi .
the calibration accuracy is @xmath78 mag for the notcam images , @xmath79 mag for the liverpool telescope images obtained on 28 december 2015 and 12 june 2016 , and @xmath80 for the liverpool telescope image obtained on 4 january 2016 . _
spitzer_/irac imaging of this region was obtained in all four bands ( 3.6 , 4.5 , 5.8 , and 8.0 @xmath9 m ) on 29 december 2006 ( program i d 30734 , pi donald figer ) .
we downloaded the corrected basic calibrated data ( cbcd ) , processed the images with the automated pipeline , version s18.25.0 , and created mosaics with mopex ( version 18.0.5 , * ? ? ?
* ) at an image scale of @xmath68 per pixel .
asassn-15qi is located where the ends of two filaments meet . at 3.6 and 4.5 @xmath9 m
, the filaments do not contribute significantly to the total flux .
the object flux was measured with point - response function ( prf ) fitting using the tool astronomical point source extraction ( apex ) , developed by the _
spitzer _ science center .
however , at 5.8 and 8.0 @xmath9 m , the object is heavily contaminated by two filaments and also begins to blend with a nearby object . in these bands ,
both filaments are fit separately with gaussian profiles along the horizontal axis in the mosaic .
the amplitude , centroid , and width of both filaments are then calculated at the location of the source and are subtracted from the image .
the source flux is then measured using point - source photometry , constrained by the location of asassn-15qi at 3.6 and 4.5 @xmath9 m .
this approach yields a detected flux at 5.8 @xmath9 m and an upper limit for a non - detection at 8.0 @xmath9 m .
the final photometry is listed in table [ tab : nirphot ] . _ wise _ w1 and
w2 photometry of asassn-15qi is five times brighter than _
spitzer_/irac photometry at similar wavelengths .
the _ wise _ photometry of asassn-15qi is similar in three separate observations obtained over 500 days . since _
wise _ has a larger psf than _ spitzer _ , this emission is likely contaminated by diffuse emission from the filament .
the _ wise _ photometry is not included in our analysis .
we obtained james clerk maxwell telescope ( jcmt ) scuba2 sub - mm observations of asassn-15qi on 5 november 2015 , 34 days after the outburst peak , as part of program m15bi083 .
we also analysed archival 450 @xmath9 m and 850 @xmath9 m scuba-2 observations from project m11bgt01 obtained on 14 november 2011 and previously published by @xcite .
the data were reduced using the standard gould belt survey legacy release 1 reduction parameters @xcite .
the source was not detected in either observation , with @xmath81 upper limits at 850 @xmath9 m of 0.06 jy / beam in the archival data and 0.07 jy / beam in our data ( beam size of 14@xmath49 ) . at 450 @xmath9 m ( 8@xmath49 beam ) , the upper limits are 1.4 jy / beam and 3.4 jy / beam .
low - resolution optical spectra were obtained at the lick / shane 3 m telescope with the kast double spectrograph @xcite , at the mdm telescope with the osmos spectrograph , at the liverpool telescope with the sprat spectrograph , and with the multi - object double spectrograph at the large binocular telescope ( lbt ) we also include in our analysis the low - resolution optical spectra from the kanata 1.5 m telescope at higashi - hiroshima observatory and the 1.01 m telescope at bisei astronomical observatory @xcite and low - resolution near - ir spectra from irtf / spex @xcite .
the kast , osmos , sprat , and lbt spectra were obtained with the slit aligned with the parallactic angle @xcite , while the kanata and bisei spectra were obtained with the slit at a position angle of @xmath82 . the osmos , sprat , lbt , kanata , and bisei spectra were all flux calibrated .
keck / hires @xcite observations of asassn-15qi were obtained on 27 october 2015 and 21 december 2015 .
the october observations were obtained with the @xmath83 slit ( @xmath84 ) while the december observations utilized the @xmath85 slit ( @xmath86 ) , both chosen to match the seeing of @xmath87 and @xmath88 . the spectra span 48009200 with some gaps between orders .
spectra were reduced and calibrated with the @xmath89 pipeline written by t. barlow .
the region around the @xmath90 line was awkwardly corrected for telluric absorption with a cfht / espadons spectrum of 72 tau obtained on a different night .
keck / mosfire @xcite long slit @xmath20-band spectra ( 2.02.3 @xmath9 m ) of asassn-15qi were obtained on 30 december 2015 at 1.6 airmasses .
the observations consist of four ab patterns with a total integration time of 400 s. the @xmath91 slit was aligned at a position angle of @xmath92 .
the seeing of @xmath93 led to a spectral resolution of @xmath94 .
the data were reduced with custom - written programs in idl .
telluric absorption was corrected with a spectrum of the a0v star hip 24311 , which was observed at 1.4 airmasses .
the relative wavelength calibration from oh sky lines is accurate to 3 , although the absolute accuracy depends on object centering in the slit .
we obtained target - of - opportunity observations of asassn-15qi with the immersion grating infrared spectrograph ( igrins ) on the harlan j. smith telescope at mcdonald observatory on 23 october 2015 ( jd 2457318.598 ) .
igrins is a high - resolution near - ir echelle spectrograph providing @xmath84 spectra simultaneously from 1.482.48 @xmath9 m in 53 orders @xcite .
asassn-15qi was observed in two sets of @xmath95 minute exposures in abba nodding patterns with @xmath96 nods for a total exposure time of 80 minutes .
asassn-15qi was observed at @xmath97 airmasses , with an a0v standard star observed at a similar airmass between the two abba nod patterns .
the @xmath98 slit was fixed to a position angle of 45@xmath99 .
the seeing was @xmath100 in the @xmath19 band and @xmath101 in the @xmath20 band , as measured in the spatial distribution of the continuum emission in the cross - dispersion direction on the detector .
the spectral image was rectified using version 2.0 of the igrins pipeline@xmath102 .
each pixel on the detector covers @xmath103 .
the spectrum was extracted from a 9-pixel ( @xmath104 ) region centered on the star .
figure [ fig : lc ] shows the @xmath1-band lightcurve of asassn-15qi , with locations marked for each of our spectroscopic data points ( see also tables 13 ) .
the asas - sn monitoring , which began on 16 dec .
2014 , yields a quiescent ( pre - outburst ) average brightness of @xmath105 mag .
the scatter in the pre - outburst photometry@xmath106 is consistent with measurement uncertainties combined with a variability of @xmath107 mag ( @xmath44 standard deviation ) . a day before the outburst , two asas - sn telescopes measured @xmath108 mag . on 2 october 2015 , 23 hours after the previous observation
, asassn-15qi was found to have brightened to @xmath109 mag , an increase of a factor of @xmath110 in flux .
the brightness increase was at least @xmath0 mag and may have been larger because the asas - sn photometry includes all nebulosity within @xmath111 .
the inclusion of nebulosity would increase the pre - outburst brightness measurements much more than the brightness near the outburst peak .
indeed , the latest photometric data from the liverpool telescope , obtained 254 days past outburst ( 11 june 2016 ) , has @xmath112 mag when measured in a @xmath54 radius aperture and @xmath113 when measured with psf photometry .
the flux then decayed quickly .
fits to the flux lightcurve show an initial @xmath114-folding decay time from peak of @xmath115 days , based on few data points and assuming that the first post - outburst photometry is during this decay and not during the initial rise . after this initial rapid decay ,
asassn-15qi faded with an @xmath114-folding time of @xmath116 days .
the object then settled at @xmath117 mag for @xmath118 days before resuming its decay , with an @xmath114-folding time of @xmath119 days .
the flattening , delayed continuation of the decay , and the possible bump at 80100 days may all be related to the light travel time to the nebulosity around the object .
differences in the asas - sn and lcogt photometry between days 5070 are likely related to the inclusion of nebulosity in the large asas - sn pixels and extraction aperture .
the photometric measurements from observations with small pixels and better seeing should provide a more direct measure of the outbursting source itself ( see also photometry with different extraction methods listed in table [ tab : photometry ] . in the following subsections
we describe our analysis of this outburst using this lightcurve and the photometric and spectroscopic follow - up observations described in 2 .
these sections frequently describe the days past outburst peak , defined as jd 2,457,298 . the object is described as in quiescence during the pre - outburst phase .
figures [ fig : spitzerimage][fig : iphas ] show _
spitzer_/irac infrared and iphas optical images of the region around asassn-15qi .
bright h@xmath24 nebulosity , warm dust emission , and bright 850 @xmath9 m emission are all centered around the position of sh 2 - 148 , with a central location @xmath120 from asassn-15qi .
as projected on the sky , the object is located along a filament of excess dust emission , either at a bend in a single filament or at the projected location where the ends of two filaments meet .
a visual companion located @xmath121 ne is 2.8 mag fainter than the quiescent optical brightness of asassn-15qi , as measured in the iphas archive @xcite , and likely does not contribute to any of the observed variability . in optical images ,
some nebulosity is detected just south of the target , as described in 3.1.1 .
the cloud velocity of @xmath122 ( @xmath123 ) has been measured with sub - mm co emission from the nearby regions @xcite , and is adopted as the stellar radial velocity.@xmath124 this velocity is consistent with the velocity of emission and absorption lines from the nearby nebulosity ( 3.1.2 ) and with the velocity of photospheric lines ( 3.4 ) .
the cloud velocity is used instead of the stellar velocity because the detected photospheric lines are very broad , have low signal - to - noise ratios , and may suffer from velocity offsets introduced by the wind .
initial follow - up imaging did not show any nebulosity nine days after the outburst peak @xcite .
however , some extended emission was detected to the se of the outburst in slit acquisition images for the keck / hires spectra obtained 26 days after the outburst peak by @xcite .
the nebulosity also appears in our lcogt and _ swift _ imaging with a time dependence consistent with both the @xcite non - detection and the @xcite detection .
figure [ fig : nebemission ] shows @xmath1-band nebulosity in lcogt and _ swift _ images as a 1d spatial distribution , extracting counts from s to n along the direction of the nebulosity@xmath102 .
the first _ swift _
@xmath1-band image , obtained 10 days after outburst peak , includes little contribution from nebulosity , although this emission is detected by 87 days after peak .
lcogt images obtained 26 days after outburst show strong nebulosity , which then weakened in later epochs of lcogt imaging . by day 80 after the outburst peak , the nebulosity faded relative to the central object , and the central object itself had faded by 0.6 mag since day 26 .
this comparison is qualitatively consistent with results on other dates and bands .
the nebulosity is likely a reflection nebula or a variable region that brightened with the central source .
the delay between outburst and nebulosity brightening and the flattening in the lightcurve ( see figure [ fig : lc ] ) are likely caused by light travel time . given the estimated distance , light from the star would take 35 days to arrive at a projected distance of @xmath100 ( the approximate centroid location of the nebulosity projected on the sky ) .
a more complete analysis of the nebulosity brightness versus time is beyond the scope of this paper .
h@xmath3 emission is detected in all near - ir spectra . while the h@xmath3 1 - 0 s(1 ) line is most prominent , a rich forest of other h@xmath3 lines is also present in the high resolution igrins spectrum ( see appendix a for a line list and description of the fits ) .
the average h@xmath3 line centroid velocity of @xmath125 is consistent with the heliocentric radial velocity of the parent molecular cloud ( left panel of figure [ fig : h2vel ] ) .
emission is detected in transitions from the common 1 - 0 s(1 ) line at 2.1218 @xmath9 m to lines with high vibrational excitation , including 86 o(4 ) and 97 q(2 ) .
excitation to such high vibrational levels requires a strong ultraviolet ( uv ) field .
many of the detected emission lines , such as 86 o(4 ) , are prominent indicators of uv fluorescence , as predicted in the models of @xcite .
excitation to similarly high vibrational levels has previously been observed in the region around hd 37903 , a b1 star in orion @xcite , and is also found in spectra of planetary nebulae ( e.g. , * ? ? ?
h@xmath3 emission from lower vibrational levels is common in fuor objects ( e.g. , * ? ? ?
the h@xmath3 emission is spatially resolved in the cross - dispersion direction in both the igrins and mosfire spectra ( figure [ fig : h2vel ] ) . in the mosfire spectrum , which was obtained in better seeing ,
h@xmath3 emission is not detected from the star , has two peaks at @xmath126 off the star , and becomes undetectable at @xmath127 ( 0.1 pc ) s and @xmath128 ( 0.2 pc ) n. the br@xmath129 and co overtone emission is not spatially extended beyond the continuum emission .
the h@xmath3 lines are extended by @xmath100 beyond the continuum emission . in the igrins spectrum ,
the h@xmath3 1 - 0 s(1 ) centroid is located @xmath130 from the continuum emission in the sw direction along the slit , while other h@xmath3 lines are offset by @xmath131 from the continuum in the same direction .
the weaker , low excitation lines are also located @xmath132 redward of h@xmath3 1 - 0 s(1 ) and other low excitation lines .
both the velocity and spatial information suggest multiple components , one of which has strong uv excitation .
emission is only marginally detected in the h@xmath3 1 - 0 s(1 ) line in the spex spectrum of @xcite , which was obtained 13 days past outburst peak .
the equivalent width ( ew ) of @xmath133 is a factor of @xmath134 weaker than the ews of 1.2 and 0.81 in the subsequent igrins and mosfire spectra , respectively , despite little change in the @xmath20-band continuum flux .
comparisons between these spectra are challenging because of different seeing conditions and slit widths .
however , the spex image was obtained with seeing of @xmath131 and a @xmath131 slit width , so the results should be similar to that of the mosfire spectrum .
the h@xmath3 1 - 0 s(1 ) emission is likely variable , with a time delay similar to that seen in the brightness of the optical nebulosity .
emission is also detected in the forbidden [ ] @xmath135 line and [ ] @xmath1366716 , 6731 doublets .
the lines are located at @xmath137 .
the [ ] equivalent widths increased by a factor of @xmath138 from 26 days past the outburst peak to 80 days after the peak , when the object was near quiescence .
these equivalent widths are approximately consistent with a constant line flux against a weaker continuum .
the [ ] line is stronger during outburst than during the decay .
emission is not detected in either [ ] or [ ] forbidden lines .
a number of atomic lines were detected in the spex near - ir spectrum of @xcite .
these lines are generally too weak in the igrins spectrum for a robust line identification and analysis .
interstellar absorption lines are detected in the @xmath1365889 , 5996 and @xmath90 lines@xmath139 ( see figure [ fig : h2vel ] ) , tracing either the nearby nebulosity or the parent molecular cloud .
the lines are more optically thick and include more absorption components than the line .
the line has narrow absorption components centered at @xmath137 , likely local to the star , and at @xmath140 in a cloud closer to us .
diffuse interstellar bands ( dibs ) are also detected at several wavelengths , including 5780 , 5797 , 6177 , 6283 , and 6614 , with velocities roughly consistent with the parent molecular cloud .
the equivalent width of the 5780 dib is consistent with @xmath141 mag @xcite , with the caveat that their sample does not include lines - of - sight that are higher than @xmath142 mag .
the depth and velocity of the absorption components attributed to the interstellar medium remained constant in both lines and diffuse interstellar bands between our first hires observation , obtained 25 days after the outburst peak ( @xmath143 mag ) and our second hires observation , obtained 80 days after the oubtburst peak ( @xmath144 mag ) .
cccccc & pre - burst & & post - burst + band & quiescence & 0 d & 12.9 d & 81.5 d & 254 d + @xmath59 & & & 19.1 & + @xmath145 & & & 20.7 &
+ @xmath61 & & & 18.0 & 20.2 + @xmath62 & & & 16.8 & 18.8 + @xmath12 & 17.418.2 & & 16.0 & 17.7 & 19.0 + @xmath21 & 17.0618.01 & &
+ @xmath1 & 17.01 & 13.5 & 14.8 & 16.3 & 17.3 + @xmath22 & 16.66 & & 14.4 & 15.7 & 16.5 + @xmath23 & 15.59 & & 13.8 & 15.0 & 15.7 + @xmath15 & 15.08 & & & + @xmath18 & 13.704 & & 12.15 & 12.93 & 13.5@xmath14 + @xmath19 & 12.923 & & 11.62 & 12.30 & 12.76 + @xmath20 & 12.647 & & 11.29 & 11.85 & 12.2@xmath14 + @xmath146 & 12.36 & & &
+ @xmath147 & 12.30 & && + @xmath148 & 12.50 & && + @xmath149 & @xmath150 & &&
+ + + in this section , we describe the emission from the optical photosphere using spectra and the spectral energy distribution ( sed ) .
the section is divided into pre - outburst quiescence , the outburst and decay ( 0 - 150 days post - burst ) , and the final epoch ( about 250 days post - burst ) .
table [ tab : phot ] presents the seds in each epoch , while spectra and analysis are shown in figures [ fig : sed]-[fig : balmerabs ] .
the pre - outburst sed of asassn-15qi is constructed from the _ spitzer _ , 2mass , iphas , _ hst _ guide star catalogue , palomar transient factory , and pre - outburst asassn-15qi photometry .
the archival pre - outburst photometry at blue wavelengths varies within a range of @xmath151 mag , including @xmath21-band emission obtained over 4.5 months by the palomar transient factory . whether this level of variability is also seen at redder wavelengths is uncertain .
the quiescent sed from 0.64.5 @xmath9 m is reasonably fit with a synthetic spectrum of 6200 k @xcite and an extinction @xmath152 mag , using an extinction curve from cardelli et al .
1989 and total - to - selective extinction value of @xmath153 . the effective temperature and extinction are adopted from a measurement of the post - burst spectrum ( see 3.2.3 ) .
these values , the 2mass @xmath18-band magnitude , and the distance estimate of 3.24 kpc lead to a quiescent luminosity of 18 l@xmath6 for this component .
the luminosity is consistent with the expectation that asassn-15qi is significantly fainter than the nearby o9/b0 stars that power the compact regions sh 2 - 148 and sh 2 - 149 @xcite , but still bright enough and hot enough to excite the nearby molecular cloud material .
excess emission is detected in @xmath12 and @xmath1 during quiescence , pointing to the presence of hot emission in excess of a single temperature fit .
the quiescent sed may alternately be reproduced with hot ( @xmath154 k ) and cool ( @xmath155 k ) components . the hot component would need to be fainter than the cool component , either because of a higher extinction or because of a smaller surface area .
the hot component could have a smaller surface area if a hot star is surrounded by a warm gaseous disk , if the disk has a small hot spot , if the star has an accretion - induced hot spot , or in an evolved binary system consisting of a white dwarf with a dwarf or giant companion .
the quiescent sed provides a baseline for interpreting past photometry .
as discussed above , the pre - outburst asas - sn photometry indicates a stable @xmath1-band magnitude .
photometry in many previous epochs is also consistent with a quiescent sed , including epochs in 1953.83 and 1989.67 .
however , @xmath21 and @xmath12 photometry shows variability of @xmath151 mag . in 1976.5 , usno - b recorded photometry that was @xmath132 mag brighter than the quiescent sed , indicating a likely outburst .
the uv - optical - ir sed during the outburst is constructed at 13 and 81 days past the outburst peak .
these dates are selected to coincide with the dates when @xmath156 photometry was obtained .
the multi - band lcogt lightcurve ( figure [ fig : color ] ) is then used to estimate the @xmath50 magnitudes .
the uv photometry from _ swift _ was obtained contemporaneous to the @xmath156 photometry , and are listed with only minor adjustments to account for temporal changes .
the lcogt photometry demonstrates that the object was bluer during the outburst than in quiescence ( figure [ fig : color ] ) .
the @xmath157 and @xmath158 colors are roughly constant during the outburst decay and become redder when the decay flattens out .
the @xmath159 color becomes much redder as the outburst decays and flattened when the lightcurve flattened .
the few epochs with @xmath76 photometry have bluer near - ir colors during outburst than during quiescence .
the sed at 13 days after the outburst peak is reasonably well fit with a @xmath160 k blackbody of @xmath161 l@xmath6 that suffers from an extinction @xmath162 , as shown in figure [ fig : sed ] .
adjusting this luminosity to the peak photometry would lead to @xmath163 l@xmath6 .
this luminosity increases is larger than the brightness increase in the @xmath1-band because of the different bolometric corrections in quiescence and outburst .
alternately , if we assume that the temperature does not change , then the luminosity could be as low as 500 l@xmath6 at peak and @xmath164 l@xmath6 at 13 days post outburst although this scenario leads to a poor fit to the sed .
if the brightness change were interpreted as attributed to extinction , then the sed could be fit with @xmath165 k emission and @xmath166 , with a luminosity only three times higher than quiescence .
however , the extinction scenario is unlikely ( see 4.2.2 ) .
these scenarios all overestimate the @xmath23-band photometry during the outburst . the high resolution optical spectra obtained during outburst lead to photospheric temperatures that are mostly consistent with the 6200 k temperature measured during quiescence ( figure [ fig : speclo ] ) .
the and line complexes at 51005400 and the + blend at 6497 are the most identifiable photospheric features .
we fit these features with a photospheric spectrum produced by bt - settl models with solar metallicity @xcite .
the 6497 feature is well reproduced by a 6300 k photosphere with @xmath167 centered at the expected radial velocity ( @xmath168 in the heliocentric frame ) . the observed spectrum near 5200
is well reproduced by a low gravity 7000 k photosphere with a similar velocity broadening and centroid .
this 6000 - 7000 k photospheric temperature seen in both hires spectra obtained during outburst poses a challenge to the sed models , where the @xmath1-band emission is expected to be produced by the hotter ( @xmath169 k ) component .
morever , the single temperature fits fail to reproduce some of the sed .
most likely , the assumption of a single temperature is overly simplistic .
especially if the emission is produced in a gaseous disk with a temperature gradient , the emission may instead arise from a range of temperatures . as confirmation ,
the balmer continuum absorption in the low resolution lick shane / kast spectrum is consistent with a b5b9 star when compared to the @xcite spectral library which suggests temperatures of @xmath170 k , much hotter than the @xmath171 k obtained from the high resolution spectra of the photosphere . in two late epochs with photometry obtained at 207 and 254 days past outburst , the source brightness was constant . by these dates ,
the outburst had subsided and most signatures of the burst had disappeared . in this epoch ,
the sed had a similar shape but is fainter at blue wavelengths than the sed inferred from pre - outburst photometry .
the lbt optical spectrum obtained in this epoch is consistent with an f8 star ( figure [ fig : balmerabs ] ) , though some residual features of the wind that will be discussed in 3.4 .
the extinction of @xmath172 mag measured from the spectrum is similar to @xmath173 measured from the @xmath174 color .
the extinction @xmath172 is adopted throughout the paper .
this measurement may be affected by wavelength dependence in the scattered light within the extraction aperture .
the total luminosity of the object is @xmath175 l@xmath6 , as measured from fitting a template 6200 k photosphere to the observed spectrum .
the outburst sed and the balmer continuum absorption in the kast spectrum both demonstrate the presence of hot gas during the outburst .
the late - time spectrum no longer exhibits balmer continuum absorption ( see inset in figure [ fig : balmerabs ] ) .
the @xmath157 color is also redder than the pre - outburst and outburst photometry .
the hot ( @xmath176 k ) component that was present during outburst , and that may have contributed to short - wavelength emission before the outburst , is no longer visible in the system .
the mid - ir sed is consistent with a single temperature blackbody and no excess dust emission .
the sed fit shown in figure [ fig : sed ] does not fully recover the near - ir and mid - ir fluxes , but this small discrepancy could easily be recovered with a 2-temperature fit to the sed .
the presence of dust can not be inferred from this small difference .
adust disk could be masked by bright emission from a hot gaseous disk ; for fuor objects , the excess dust emission may be detected only longward of @xmath177 @xmath9 m @xcite . at longer wavelengths
, the 850 @xmath9 m flux upper limit corresponds to a maximum dust mass of @xmath178 m@xmath6 .
this limit is too high to provide any useful constraints on an envelope or disk .
bright emission is detected in the co overtone bands ( @xmath179 ) at 2.3 @xmath9 m ( figure [ fig : cocomp ] ) .
the presence of co overtone emission from young stars often indicates the presence of a warm disk .
we created synthetic co emission spectra assuming an optically thin slab in local thermodynamic equilibrium at a single temperature , with co energy levels and oscillator strengths were obtained from @xcite . the band shape and relative strengths
are reasonably well reproduced with a co excitation temperature of 4000 k and a broad line profile with a full width at half - maximum intensity ( fwhm ) of 300 centered at the stellar radial velocity .
individual lines blend together and are not resolved .
the co luminosity in the igrins spectrum leads to an estimate of @xmath180 warm co molecules .
the band shape and lack of resolved lines in the high resolution igrins spectrum requires the high velocity broadening
. the characteristics of the co overtone emission are consistent with expectations for emission from a hot gaseous disk .
the band equivalent width is similar in the igrins ( 21 days post - outburst ) and mosfire ( 89 days post - outburst ) epochs and lower in the early spex observation ( 13 days post - oubturst ) .
because of the decay of the continuum brightness , the total flux and therefore number of emitting molecules was roughly a factor of @xmath181 higher during the igrins observation than either the spex or mosfire observation .
the co was not spatially extended beyong the continuum in any observation .
the diversity of wind absorption lines in the optical spectrum of asassn-15qi during outburst is fascinating . in the following subsections ,
we discuss the wind absorption features in different diagnostics .
we then describe the disappearance of this wind .
the h lines in asassn-15qi all look remarkably different ( figure [ fig : hlines ] ) .
the balmer lines h@xmath24 and h@xmath182 both exhibit strong blueshifted absorption and redshifted emission . at velocities of @xmath183
to @xmath184 ( relative to the assumed radial velocity of @xmath168 ) , the lines show absorption that is filled in with some emission , likely from scattered light in the wind .
the wind scattering also leads to redshifted emission in h@xmath24 , strong out to 500 and detectable to 1000 .
the red wing of h@xmath182 emission is weak and only detected to @xmath185 .
the h@xmath182 wind absorption is deep between @xmath186 to @xmath187 and gradually becomes weaker until the terminal velocity around @xmath188 .
much of the wind absorption in h@xmath24 is contaminated by emission .
however , the h@xmath24 and h@xmath182 wind absorption depths appear to match@xmath189 between @xmath190 to @xmath188 , with a depth of @xmath191% at @xmath190 and a few percent at @xmath188 .
the wind absorption in these lines may be optically thick but covers only a fraction of the emission region .
alternately , the wind covering fraction may be high even at large velocities , but with a low optical depth .
high order paschen lines are detected with shallow blueshifted absorption to @xmath190 and a photospheric absorption component that extends to about @xmath192 . the lower order paschen lines show emission and weak absorption in the low resolution spectrum of @xcite .
in contrast , the brackett lines ( br@xmath129 and higher ) are seen in emission up to br 15 ( h 15 - 4 ) .
these emission line profiles are similar in shape to the inverse of the paschen absorption line profiles .
the lack of emission in higher order h lines suggests that collisions and some photoexcitation determine the hydrogen level populations .
@xcite detected a classic p cygni profile in the @xmath193 line 13 days after the outburst peak .
the line looks similar to the nearby paschen lines , with absorption seen out to about @xmath188 , strong emission from @xmath194 , and an emission wing that extends to about @xmath195 . the lower level of the @xmath193 line is meta - stable and may be dominated by scattering when the opacity in the lower level is sufficiently high ( e.g. , * ? ? ?
optical he lines are not detected in either emission or absorption .
emission in optical lines is commonly used as a diagnostic of magnetospheric accretion ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the lack of any emission in these lines indicates an absence of magnetopsheric accretion .
the lack of absorption in the optical he lines suggests temperatures cooler than @xmath169 k @xcite .
lcccccccc & days & h@xmath24 & h@xmath182 & & & & & + spectrum & post - peak & 6563 & 4861 & 8498 & 7774 & 5893 & 5018 & 4924 + kanata & 5.5 & @xmath196 & @xmath197 & @xmath198 & @xmath199 & @xmath200 & & + bisei & 6.0 & @xmath201 & @xmath202 & & @xmath203 & @xmath204 & @xmath205 & @xmath206 + spex & 12.9 & & & @xmath207 & @xmath208 & & & + kast & 17.5 & @xmath209 & @xmath210 & @xmath211 & @xmath212 & @xmath213 & @xmath214 & @xmath215 7 + hires & 24.7 & @xmath216 & @xmath217 & @xmath218 & & @xmath219 & @xmath220 & @xmath221 + osmos & 40.7 & @xmath222 & & & & @xmath223 & @xmath224 & @xmath224 + hires & 79.7 & @xmath225 & @xmath226 & @xmath227 & & @xmath228 & @xmath229 & @xmath230 + sprat & 94.4 & @xmath231 & & & & @xmath232 & & + sprat & 206.4 & @xmath233 & & & & @xmath234 & & + lbt & 255.9 & @xmath235 & @xmath236 & @xmath237 & @xmath238 & @xmath239 & &
+ + the ir triplet lines@xmath240 have strong emission , with blueshifted absorption that extends to about @xmath184 relative to the assumed radial velocity ( figure [ fig : hlines ] ) .
the blueshifted absorption is deepest in the @xmath241 line and shallowest in the @xmath242 line , following the order of oscillator strengths .
this ordering implies that the wind absorption is optically thin .
the emission is much stronger than the absorption .
the @xmath242 line has a centroid velocity of @xmath243 and a fwhm of 210 , although both values are likely affected by the wind absorption .
these observed properties are unlike most spectra of young stars .
when the line is seen with such large equivalent widths in other young stars , strong emission is also detected in lines and is also usually seen in a forest of lines .
( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the relative strength of the emission compared with the weak wind absorption also does not follow expectations for redshifted emission from wind scattering , as seen in the h lines and as expected for classic p cygni profiles . for young stars with such strong wind absorption
, the ir triplet should show stronger absorption features .
the upper levels of the infrared triplet lines are the same upper levels as the resonant h & k lines ( see description of energy levels in , e.g. , * ? ? ?
in a cool wind exposed to strong uv radiation , the h & k lines will absorb radiation .
radiative de - excitation then occurs through the ir triplet lines , producing strong emission
. the wind must be sufficiently cool and/or low density that most of the population is in the ground state , and the central source must be hot enough to provide enough photons for sufficient h & k absorption to occur to explain the ir triplet emission .
the lack of [ ] @xmath244 indicates electron densities @xmath245 @xmath246 @xcite .
figure [ fig : hlines ] shows profiles of the d lines and of the coadded @xmath247 lines .
not shown is the @xmath248 line , which has a profile very similar to @xmath249 .
the wind absorption in these lines reaches a maximum opacity at @xmath250 and gradually decreases to zero at @xmath251 . at @xmath250 , the flux in h@xmath182 and
is nearly 0 , indicating that the wind is optically thick and covers the entire emission component .
the wind therefore has an optical depth of @xmath151 in and several other lines at @xmath250 .
the 42 multiplet lines at 4923 , 5018 , and 5169 are deep and prominent even in the low resolution spectra .
some other lines of are weaker but detectable .
wind absorption is weakly detected in some lines and is not detected in lines .
the wind absorption is prominent in many absorption lines , with an optical depth that peaks at about @xmath252 , and some gas seen out to @xmath188 .
the few emission lines that are detected are also likely produced by absorption of continuum emission by the wind followed by radiative decay .
these features can all be understood as a fast spherical wind ( john kwan , private communication ) .
the decrease in the wind absorption to faster velocities likely indicates that the wind is more optically thin at higher velocities , although the covering fraction of the continuum emission surface may also be decreasing .
the deep absorption trough over a wide velocity range is likely explained by radial acceleration of the wind in our line of sight . in radiative transfer models ,
this acceleration and consequent deep trough is consistent with expectations for a spherical wind and inconsistent with narrow absorptions that are produced by disk winds @xcite .
the emission line strengths support the interpretation of the wind as spherical .
the line emission is likely produced by absorption and re - emission of photons by the wind into our line of sight .
most of the emission is seen at @xmath253 to @xmath254 , consistent with much of the wind having velocities that are perpendicular to our line of sight .
the strength of the emission indicates that the wind covers a large fraction of the solid angle centered at the continuum source .
the properties of the emission lines , together with the deep absorption troughs , both indicate that the wind is spherical .
the strong , fast wind is the most distinctive feature of the optical spectrum of the asassn-15qi outburst .
this strong wind was detected with deep absorption features in low resolution spectra and confirmed as a fast wind in a high resolution spectrum obtained 25 days after peak . at this phase ,
the equivalent width of the lines had already started to decline . in a low resolution spectrum obtained 40 days after peak , the lines were only weakly detected and the d lines were also weaker than seen in earlier spectra . by day 80 , our second hires spectrum showed that the wind had mostly disappeared ( figures [ fig : gonegirl][fig : fe2fade ] and table [ tab : eqws ] ) .
no wind is detectable in any lines at velocities faster than 500 . from @xmath186 to @xmath184 ,
the wind absorption disappeared entirely from the d lines , although weak absorption at these velocities is still present in the h@xmath182 and lines .
the wind absorption also disappeared .
the absorption in the high - order paschen lines is now symmetric about the stellar radial velocity and may be photospheric .
a low resolution spectrum obtained 94 days after the peak was nearly featureless except for balmer lines and confirmed that the wind continued to disappear .
the final spectrum obtained @xmath255 days after outburst no longer shows any wind absorption features .
the h@xmath24 line is a symmetric emission profile .
the emission lines also became gradually weaker during the decay , indicating that the wind is also absorbing fewer photons along other lines of sight .
the lines line fluxes decrease by a factor of 3.3 ( after adjusting the equivalent width by the continuum flux ) between the first two hires epochs ( 25 and 80 days after peak ) , and by another factor of 5 in the last epoch ( 254 days ) . in that final epoch ,
the @xmath241 is @xmath132 times weaker than the other lines - a very unusual ratio .
the three lines had similar fluxes in all previous epochs .
a central absorption component is present in the decay epoch , which may be related to either self - absorption , the photosphere , or lack of emission ( see figure [ fig : ca2 ] ) . by the final post - outburst spectrum ,
the lines are double - peaked . in this last epoch
, the irt emission corresponds to the location of absorption in the first epoch .
the double - peaked shape of the line profile may imply some directionality ( e.g. , collimation ) in the wind .
_ youth : _ asassn-15qi is likely young , as inferred from its projected location near molecular gas and regions , its radial velocity consistent with membership in this parent cloud , the projected association with filaments , and the direct association with nebulosity ( see 3.1 ) .
most of the discussion below assumes that asassn-15qi is young .
however , while the star must be physically located within the molecular cloud , this location could be a coincidence if asassn-15qi is an old object that has migrated into the molecular cloud .
such coincidences must be rare but may occur as transient surveys cover larger areas at higher cadence .
the location in a filament may even be caused by past outflow events that warm nearby dust .
_ near - instantaneous rise time : _ the rise time from quiescence to outburst peak is unresolved and @xmath256 hours .
the outburst initially decayed by 1 mag over 9 days and another 1.5 mag over 40 days .
the decay then stalled out for 30 days , perhaps because of the delayed interaction with nebular material , before resuming its decay and returning to the quiescent state within 200 days ( see 3 ) .
_ fast , cool wind : _ a fast , cool wind was associated with the outburst and disappeared with the outburst decay .
the wind during outburst reached 1000 , with a peak optical depth at 150 .
the wind is likely spherical and not a disk wind .
for a 2 m@xmath6 star , an escape velocity of 1000 corresponds to 0.75 r@xmath6 , which is smaller than the expected stellar radius ( but much larger than the radius of any compact object ) .
the peak optical depth in the wind absorption at @xmath252 corresponds to a launch radius of @xmath118 r@xmath6 ( see 3.4 - 3.5 ) .
_ hot photosphere during the outburst : _ the balmer continuum and optical colors during the outburst reveal the presence of a @xmath257 k photosphere associated with the outburst .
this hot gas disappeared as the outburst faded . a strong far - uv radiation field , which may be related to the outburst , is inferred through the presence of vibrationally excited h@xmath3 emission ( see 3.13.2 ) . _ underlying quiescent photosphere : _ the quiescent sed and the post - outburst spectrum are consistent with an f8 ( @xmath165 k ) star with @xmath142 mag .
this photospheric component was detected during the outburst ( see 3.2 ) .
_ quiescent and outburst luminosity : _ the quiescent luminosity is @xmath258 l@xmath6 , which would correspond to an object with @xmath259 m@xmath6 and @xmath165 k at 3 myr , based on the @xcite pre - main sequence evolutionary models . assuming no change in extinction , the peak outburst luminosity is @xmath260 l@xmath6 ( see 3.2 ) .
the total outburst energy that radiatively escapes is @xmath261 erg over 6 months , with 33% released over the first 10 days and 84% over the first 100 days . _ a gaseous inner disk but no warm dust : _ strong co overtone emission indicates the likely presence of a gaseous disk .
however , no dust excess is detected in the mid - ir .
the non - detection of sub - mm emission is not a significant constraint on the presence of either a disk or an envelope ( see 3.3 ) .
the similarity in wind absorption in h@xmath24 and h@xmath262 profiles could be explained with an absorption depth that corresponds to the fraction of emission intercepted by the wind .
_ recurring outburst : _ photometry from the usno survey indicates a likely outburst in 1976 ( see 3.2 ) .
the empirical characterization of the outburst of asassn-15qi is supposed to help to identify the responsible phenomenon . the fast rise time and the maximum wind velocity both point to an event at or very near the star .
the outburst coincided with a cool wind launched spherically ( or nearly so ) from the star and/or an inner disk .
in the following subsubsections , we discuss several different scenarios that might explain this outburst .
accretion is the default explanation for an increased luminosity from a young star . in this case
, a large accretion rate could lead to additional viscous heating in the inner disk .
the inner disk would then become hotter and brighter , leading to bluer spectrum .
if the observed luminosity is related to an accretion event , then the accretion rate at the outburst peak could be estimated as @xmath263 assuming that the luminosity released by viscous heating is directly related to the increase in accretion energy and that the accretion energy is efficiently converted into radiation .
the spectra lack any indication of magnetospheric accretion ; the emission features can all be explained by wind scattering@xmath264 . these magnetospheric accretion features , including many strong optical emission lines , would be expected in an exor ( e.g. , * ? ? ?
although strong co overtone emission is typical of exors and often used to classify outbursts ( e.g. , * ? ? ?
* ) , but given other discrepancies , this diagnostic is insufficient to diagnose the outburst of asassn-15qi .
the fast timescale could be explained by the lack of any magnetospheric cavity in the inner disk .
although the leading explanation for exor events explains a buildup of mass in the inner disk through a large magnetospheric cavity @xcite , exor events could instead be triggered by disk instabilities at larger radii ( see suggestion by * ? ? ?
* ) . however , the challenge for applying this interpretation to asassn-15qi is the lack of excess dust emission in the sed . in principle , in an fuor the near - ir dust emission could be masked by strong emission from a hot , optically thick disk , as seen in the sed models of @xcite .
the quiescent luminosity of @xmath258 l@xmath6 measured from the quiescent spectrum and sed would then be interpreted as disk emission .
this luminosity would be extreme for normal disks but fainter than the @xmath265 l@xmath6 emitted from fuor disks @xcite .
the b - star - like balmer jump seen in the asassn-15qi spectrum ( see fig . [
fig : speclo ] ) is present in other fuor outbursts @xcite .
the lack of magnetospheric accretion signatures is also consistent with expectations for fuors , since the accretion rates are high enough to crush stellar magnetic fields .
however , the 2015 outburst was an eruptive event with a near - instantaneous rise time and short decay , whereas fuor ( and exor ) outbursts are driven by changes in viscous disks on timescales of months - to - years . this challenge could be solved if the 2015 outburst occurred during a much longer fuor - like outburst . in principle
, the increased brightness could be caused by a decrease in the extinction .
such a rapid change in extinction would likely be related to geometry , as seen in some young stars with disks ( e.g. , * ? ? ?
* ) , rather than dust destruction
. however , the spatially extended nebulosity brightens with the outburst , which demonstrates that the outburst was seen in multiple directions and not just our line of sight .
an extinction scenario would require a dust shell that was cleared or destroyed on a @xmath266 d timescale , and then quickly reformed in our line of sight .
moreover , the interstellar and dib absorption did not change between the two hires epochs , obtained 25 and 80 days after the outburst peak , which implies that the extinction is unchanged .
therefore , the brightness increase is unlikely an extinction - clearing event .
young planetary systems are thought to evolve through dynamical interactions between the planets .
hot jupiters and super - earths that have been detected around more mature stars may arrive into their present positions by first being scattered into highly eccentric orbits . in some cases
, these highly eccentric orbits must take the planet into a grazing orbit through the stellar photosphere .
the acoustic waves driven by this interaction could drive a cool wind from the stellar photosphere .
the rocky core of the planet might be able to survive passage through the atmosphere , so that the interaction would recur on an orbital timescale .
the dynamical timescale for a grazing encounter would be a few hours , though the energetics may present a challenge .
a related alternative is a periodic mass transfer event between an unseen companion or `` companion - tesimal '' in an eccentric orbit around the primary star .
any multiplicity in this young star may still be unstable , with interactions that could lead to mass transfer and ejection .
the wind absorption and related emission are the most prominent spectral features of the outburst and disappeared as the outburst faded .
the wind likely plays an important role in this outburst .
a possible explanation may be a spherical ( or nearly spherical ) mass ejection event , perhaps driven by an material accelerated from the stellar surface or by reconnection at the star - disk interaction region .
an analogous scenario on the sun would be a coronal mass ejection , except that coronal winds are hot while the wind detected here is cool and likely has a much larger mass flux .
the fast wind velocity also suggests the a launch radius at or very near the star .
the wind reaches velocities of 1000 , much faster than the 200500 wind typically seen from herbig ae / be stars ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the one exception , z cma , has a wind that reaches @xmath267 , and is discussed below . in this scenario ,
the wind may form a photospheric surface , or be launched by the eruption that heated gas to @xmath160 k. the outburst strength versus time would be a function of expansion of the @xmath268 surface , radiative cooling , and the decrease of density as the ejected material expands . the asassn-15qi outburst could be related to structural instabilities that are caused by rapid rotation , similar to those seen on classical be stars . these main sequence ( or near main sequence )
stars rotate near their breakup velocities and as a result periodically release material and form excretion disks ( see review by * ? ? ?
the outburst of asassn-15qi could have similar physics in the pre - main sequence that classical be stars have at later evolutionary stages .
asassn-15qi may be a pre - main - sequence star that is still contracting , with a compact gaseous disk and no evidence for a dusty primordial disk .
if the disk is unable to break the spin - up of the protostar during contraction , the star may rotate at or close to breakup velocity . the rise and decay times of asassn-15qi are inconsistent with outburst timescales for classical be stars , but this discrepancy could be explained by differences in stellar mass and radius .
the projected rotational velocity ( @xmath269 ) of @xmath270 is about half of the break - up velocity of a @xmath271 m@xmath6 , @xmath271 @xmath272 star .
the rapid rise time of the asassn-15qi outburst corresponds reasonably well to that of novae .
the brightness increase is smaller than expected for a nova but could be consistent with that seen in a dwarf nova .
while most nova spectra are dominated by emission lines from ejecta , early - time novae show similar p cygni absorption profiles as those seen from asassn-15qi ( e.g. , * ? ? ?
* ; * ? ? ?
* ) . however
, these absorption spectra quickly turn into emission line spectra .
based on the spectra , asassn-15qi is not a nova or a cv ( bob williams and fred walter , private communication ) . despite different triggers ,
the similarity in absorption lines indicate similar excitation conditions in the winds of novae and asassn-15qi .
the 2015 outburst of asassn-15qi is not easily classified .
none of the above possibilities is compelling .
the default classifications of outbursts of young stars as exor or fuor accretion events also do not seem to apply to asassn-15qi .
the outburst rise time was @xmath266 day , compared to 1 month for ex lup and months - to - years for fuors ( e.g. , * ? ? ?
the total outburst duration was also @xmath273 orders of magnitude shorter than that of fuor outbursts and is also shorter than exor outbursts .
although these classifications may not apply , the star shares some characteristics with some previous yso outbursts .
figure [ fig : comparespecs ] compares the spectrum of asassn-15qi with that of the fuor candidate z cma , the outburst of v899 mon ( iras 06068 - 0641 ) , and fu ori .
the z cma and v899 mon outbursts have also been challenging to classify and have similar wind absorption features .
the comparisons in this section are not comprehensive .
other outbursts , such as v900 mon @xcite , have also been difficult to classify , have similar wind absorption features , and may be similar to asassn-15qi but are not discussed below . the spectrum of the yso outburst v899 mon shows wind absorption features with velocities up to @xmath274 in many of the same lines as asassn-15qi @xcite .
the initial discovery spectrum of the v899 mon outburst @xcite showed strong absorption in , which disappeared in later spectra @xcite .
the infrared triplet lines of v899 mon look similar to those of asassn-15qi , with redshifted emission and some weak p cygni absorption .
lines are weak or not present .
the primary spectroscopic difference is that the co overtone bands are seen in absorption in v899 mon and in emission in asassn-15qi , which may indicate differences in disk structure or in the surrounding environment .
the v899 mon outburst was not clearly identified as either an fuor or exor outburst .
the outburst timescale is the primary difference between v899 mon and asassn-15qi .
v899 mon has had many brightness increases and decreases @xcite .
the initial rise occurred over @xmath275 years , while subsequent fading and brightening episodes occurred over much shorter timescales ( weeks months ) .
another outburst with spectral similarities to asassn-15qi is z cma , a binary herbig ae / be system in which one of the two stars has been seen in outburst and the other is an emission line object .
the multiple outbursts of z cma were classified as fuor - like when the characteristic co absorption spectrum was detected and resolved from the co emission spectrum produced by the non - outbursting component @xcite .
the deep co absorption indicates the presence of a bright , viscously heated disk with the warm disk photosphere located beneath cooler material .
while the co absorption is one of the diagnostics for an fuor outburst , the outbursts themselves have only been @xmath132 mag in amplitude with durations of a year , both much smaller than the classical fuor objects .
wind absorption in lines is detected in the spectrum of z cma but is only rarely detected to other herbig ae / be stars .
the outflow from z cma has a maximum velocity of 750 , a factor of 2 faster than outflows from other haebe stars @xcite . with less data ,
the outburst of asassn-15qi would have been clearly and uncontroversially described as an exor .
the primary characteristics that differ from exors are the very short rise time , which is only measurable in surveys having a high cadence , the strong wind absorption seen especially in the early - time optical spectra , and the absence of spectroscopic signatures of magnetospheric accretion .
the near - ir spectra all show strong co emission .
the spectrum of @xcite also exhibits strong emission in many atomic lines . these spectral features along with an outburst duration of @xmath273 months are amongst the primary diagnostics for the exor classification .
the short outburst of asassn-15qi was discovered only because of the frequent , wide - field monitoring of the asas - sn project . the fast rise and
initial decay left only a @xmath276 day period when the outburst was 3 mag brighter than quiescence .
the frequency of these types of outbursts is highly uncertain because the detection of months - long bursts requires a high temporal cadence that are only now becoming available .
the proliferation of all - sky surveys will require strict criteria for efficient follow - up spectroscopy .
the rise times of fuor and exor events , ranging from a few weeks to 10 years , provide a criterion for selecting fuor and exor outbursts for follow - up observations . however , this choice may also bias surveys against eruptive outbursts of unknown origin , including those like asassn-15qi .
the quiescent and outburst seds indicate a quiescent luminosity of @xmath258 l@xmath6 , an outburst luminosity of 1000 l@xmath6 , and a total energy radiated of @xmath277 erg .
the quiescent sed and the spectroscopic features during outburst suggest a photospheric temperature of 60007000 k , although the blue / uv excess indicates that a hotter component must also be present .
no warm dust emission is detected , although a gaseous disk is likely detected in co overtone emission and in the shape of the wind absorption profiles .
the energy release accelerated matter spherically from the star to a maximum velocity of 1000 , as seen in wind emission and absorption features that accompanied the outburst and disappeared as the outburst faded .
archival photometry indicates a previous outburst of asassn-15qi in 1976 .
these properties can not be easily explained in the framework of fuor and exor outbursts , which are usually applied to large brightness increases of young stars .
however , other yso outbursts , including v899 mon and z cma , share some similar spectrosopic features as asassn-15qi .
either fuor or exor events sometimes produce unexpected physical characteristics , or some protostars have outbursts that are triggered by different instabilities .
if only sparse photometry and near - ir spectra were available , this outburst would have been classified as an exor .
the outburst timescale and fast wind velocity suggests that the trigger occurred very close to the star .
such an event may be related to mass transfer , interactions between a star and an eccentric planet , formation of an excretion disk , or by some magnetic reconnection and outflow event .
we caution readers that we are also unable to rule out with complete confidence that the outbursts of asassn-15qi are produced in the test or use of advanced alien weaponry .
we thank the anonymous referee for useful comments and a prompt report .
gjh thanks john kwan , bob williams , fred walter , dong lai , doug lin , and bo reipurth for interesting conversations about asassn-15qi .
we thank lcogt and its staff for their continued support of asas - sn .
gjh is supported by general grant 11473005 awarded by the national science foundation of china .
sd and pc are supported by `` the strategic priority research program the emergence of cosmological structures '' of the chinese academy of sciences ( grant no .
xdb09000000 ) and project 11573003 supported by nsfc .
this research uses data obtained through the telescope access program ( tap ) , which is also funded by grant no .
xdb09000000 from the chinese academy of sciences and by the special fund for astronomy from the ministry of finance .
bs is supported by nasa through hubble fellowship grant hf-51348.001 awarded by the space telescope science institute , which is operated by the association of universities for research in astronomy , inc .
, for nasa , under contract nas 5 - 26555 .
csk and kzs are supported by nsf grants ast-1515876 and ast-1515927 .
tw - sh is supported by the doe computational science graduate fellowship , grant number de - fg02 - 97er25308 .
support for jlp is in part provided by fondecyt through grant 1151445 and by the ministry of economy , development , and tourism s millennium science initiative through grant ic120009 , awarded to the millennium institute of astrophysics , mas .
development of asas - sn has been supported by nsf grant ast-0908816 and ccapp at the ohio state university .
asas - sn is supported by nsf grant ast-1515927 , the center for cosmology and astroparticle physics ( ccapp ) at osu , the mt .
cuba astronomical foundation , george skestos , and the robert martin ayers sciences fund .
avf s group at uc berkeley is grateful for financial assistance from nsf grant ast-1211916 , the tabasgo foundation , clark and sharon winslow , and the christopher r. redlich fund .
research at lick observatory is partially supported by a generous gift from google .
this research was made possible through the use of the aavso photometric all - sky survey ( apass ) funded by the robert martin ayers sciences fund , data provided by astrometry.net ( barron et al .
2008 ) , and filter curves from the visual observatory .
the liverpool telescope is operated on the island of la palma by liverpool john moores university in the spanish observatorio del roque de los muchachos of the instituto de astrofisica de canarias with financial support from the uk science and technology facilities council .
we thank the _ swift _ too team for responding quickly to our observation requests .
some data presented here were 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 astrofisica de canarias .
some data presented here were obtained at the w. m. keck observatory , which is operated as a scientific partnership among the california institute of technology , the university of california , and the national aeronautics and space administration .
some espadons data were downloaded from the cfht archive , which is supported by the canadian astronomy data centre .
the observatory was made possible by the generous financial support of the w. m. keck foundation .
this work used the immersion grating infrared spectrograph ( igrins ) that was developed under a collaboration between the university of texas at austin and the korea astronomy and space science institute ( kasi ) with the financial support of the us national science foundation under grant ast-1229522 , of the university of texas at austin , and of the korean gmt project of kasi .
the james clerk maxwell telescope is operated by the east asian observatory on behalf of the national astronomical observatory of japan , academia sinica institute of astronomy and astrophysics , the korea astronomy and space science institute , the national astronomical observatories of china and the chinese academy of sciences ( grant no .
xdb09000000 ) , with additional funding support from the science and technology facilities council of the united kingdom and participating universities in the united kingdom and canada .
the authors wish to recognize and acknowledge the very significant cultural role and reverence that the summit of maunakea has always had within the indigenous hawaiian community .
we are most fortunate to have the opportunity to conduct observations from this mountain .
table [ tab : h2 ] lists fits to 48 h@xmath3 lines detected in the igrins spectrum of asassn-15qi .
lines were identified based on a @xmath278 detection of flux in a narrow ( fwhm of 510 ) line located within @xmath276 of the expected line location .
lines were fit with gaussian profiles with flux uncertainties estimated from nearby spectral regions .
the ability to detect lines depends on the telluric correction at that wavelength , so additional lines between 1.42.5 @xmath9 m may be strong but undetected .
the strong h@xmath3 lines have a median centroid of @xmath279 and a median fwhm of 8 .
the weak lines are located at @xmath280 longward of strong lines , indicating a possible velocity offset between the warm h@xmath3 gas and the fluorescently excited h@xmath3 emission .
the standard deviation of 2 in the centroid velocities of strong lines is consistent with expectations for the calibration accuracy of igrins .
the equivalent widths are converted to fluxes from linear fits to the continuum flux , calculated from @xmath281 , @xmath282 , and @xmath283 mag ( inferred from the near - ir photometry obtained during the outburst ) .
lcccccc & @xmath284 & cent.@xmath14 & fwhm & ew & @xmath285(ew ) & flux + i d & ( @xmath9 m ) & & ( ) & ( ) & @xmath14 + 3 - 1 o(4 ) & 1.4677 & 2.0 & 9.7 & 0.238 & 0.052 & 0.740 + 5 - 3 q(1 ) & 1.4929 & 2.8 & 11.3 & 0.204 & 0.039 & 0.620 + 4 - 2 q(9 ) & 1.4989 & 3.2 & 7.2 & 0.039 & 0.015 & 0.118 + 4 - 2 o(3 ) & 1.5099 & 2.6 & 10.8 & 0.146 & 0.016 & 0.436 + 5 - 3 q(4 ) & 1.5158 & 3.1 & 5.8 & 0.046 & 0.009 & 0.136 + 3 - 1 o(5 ) & 1.5220 & 4.1 & 7.8 & 0.072 & 0.015 & 0.211 + 5 - 3 q(5 ) & 1.5286 & 3.5 & 9.4 & 0.040 & 0.015 & 0.117 + 6 - 4 s(0 ) & 1.5369 & -0.8 & 7.5 & 0.058 & 0.010 & 0.168 + 5 - 3 o(2 ) & 1.5607 & 4.0 & 8.7 & 0.116 & 0.012 & 0.331 + 7 - 5 s(3 ) & 1.5615 & 3.2 & 5.3 & 0.026 & 0.007 & 0.074 + 4 - 2 o(4 ) & 1.5635 & 3.0 & 9.1 & 0.090 & 0.012 & 0.254 + 7 - 5 s(2 ) & 1.5883 & 5.4 & 8.8 & 0.039 & 0.013 & 0.108 + 6 - 4 q(1 ) & 1.6015 & 4.6 & 9.7 & 0.122 & 0.011 & 0.333 + 6 - 4 q(2 ) & 1.6074 & 4.2 & 7.9 & 0.096 & 0.009 & 0.260 + 5 - 3 o(3 ) & 1.6135 & 0.2 & 11.2 & 0.193 & 0.015 & 0.519 + 7 - 5 s(1 ) & 1.6205 & 8.1 & 5.7 & 0.033 & 0.006 & 0.088 + 4 - 2
o(5 ) & 1.6223 & 5.7 & 9.2 & 0.064 & 0.011 & 0.172 + 6 - 4 q(4 ) & 1.6281 & 4.2 & 7.1 & 0.050 & 0.009 & 0.132 + 7 - 5 s(0 ) & 1.6585 & 8.4 & 6.5 & 0.040 & 0.011 & 0.103 + 5 - 3 o(4 ) & 1.6718 & 1.6 & 10.4 & 0.142 & 0.012 & 0.358 + 6 - 4 o(2 ) & 1.6750 & 5.6 & 10.1 & 0.128 & 0.012 & 0.323 + 7 - 5 q(1 ) & 1.7288 & 6.6 & 6.9 & 0.092 & 0.014 & 0.218 + 9 - 7 s(1 ) & 1.9430 & 4.1 & 6.2 & 0.062 & 0.012 & 0.109 + 7 - 5
o(4 ) & 1.9434 & 7.7 & 6.3 & 0.131 & 0.015 & 0.230 + 8 - 6 o(2 ) & 1.9708 & 9.6 & 6.8 & 0.170 & 0.013 & 0.284 + 7 - 5
o(5 ) & 2.0220 & 4.6 & 7.1 & 0.117 & 0.015 & 0.178 + 1 - 0 s(2 ) & 2.0338 & 3.1 & 7.9 & 0.694 & 0.014 & 1.035 + 8 - 6 o(3 ) & 2.0418 & 5.6 & 10.5 & 0.229 & 0.016 & 0.336 + 2 - 1 s(3 ) & 2.0735 & -1.9 & 8.5 & 0.327 & 0.014 & 0.451 + 9 - 7 q(2 ) & 2.0841 & 1.6 & 8.1 & 0.100 & 0.013 & 0.135 + 9 - 7 q(3 ) & 2.1007 & 3.5 & 6.5 & 0.058 & 0.008 & 0.075 + 8 - 6 o(4 ) & 2.1216 & 6.0 & 8.2 & 0.169 & 0.011 & 0.209 + 1 - 0 s(1 ) & 2.1218 & 2.7 & 8.3 & 1.234 & 0.025 & 1.531 + 3 - 2 s(4 ) & 2.1280 & 9.1 & 8.2 & 0.106 & 0.011 & 0.129 + 2 - 1 s(2 ) & 2.1542 & 1.2 & 8.1 & 0.457 & 0.013 & 0.524 + 9 - 7 o(2 ) & 2.1727 & 4.6 & 6.1 & 0.093 & 0.008 & 0.101 + 3 - 2 s(3 ) & 2.2014 & 2.8 & 8.8 & 0.229 & 0.012 & 0.232 + 1 - 0 s(0 ) & 2.2233 & 1.2 & 8.1 & 1.106 & 0.011 & 1.052 + 2 - 1 s(1 ) & 2.2477 & 1.2 & 7.9 & 0.772 & 0.020 & 0.681 + 9 - 7 o(3 ) & 2.2537 & 4.3 & 6.3 & 0.134 & 0.015 & 0.116 + 3 - 2 s(2 ) & 2.2870 & 4.2 & 9.6 & 0.333 & 0.016 & 0.256 + 9 - 7 o(4 ) & 2.3455 & 5.8 & 8.9 & 0.260 & 0.031 & 0.157 + 2 - 1 s(0 ) & 2.3556 & 0.8 & 9.7 & 0.611 & 0.060 & 0.350 + 1 - 0 q(1 ) & 2.4066 & 1.9 & 5.0 & 0.540 & 0.025 & 0.231 + 1 - 0 q(2 ) & 2.4134 & 2.5 & 7.7 & 1.228 & 0.021 & 0.502 + 1 - 0 q(3 ) & 2.4237 & 2.0 & 9.4 & 1.014 & 0.043 & 0.385 + 11 - 9 s(3 ) & 2.4374 & 8.4 & 8.0 & 0.567 & 0.035 & 0.193 + 1 - 0 q(4 ) & 2.4375 & 2.5 & 6.2 & 0.389 &
0.140 & 0.132 + + + + | outbursts on young stars are usually interpreted as accretion bursts caused by instabilities in the disk or the star - disk connection .
however , some protostellar outbursts may not fit into this framework . in this paper
, we analyze optical and near - infrared spectra and photometry to characterize the 2015 outburst of the probable young star asassn-15qi .
the @xmath0 mag brightening in the @xmath1 band was sudden , with an unresolved rise time of less than one day . the outburst decayed exponentially by 1 mag for 6 days and then gradually back to the pre - outburst level after 200 days .
the outburst is dominated by emission from @xmath2 k gas .
an explosive release of energy accelerated matter from the star in all directions , seen in a spectacular cool , spherical wind with a maximum velocity of 1000 .
the wind and hot gas both disappeared as the outburst faded and the source returned to its quiescent f - star spectrum .
nebulosity near the star brightened with a delay of 1020 days .
fluorescent excitation of h@xmath3 is detected in emission from vibrational levels as high as @xmath4 , also with a possible time delay in flux increase .
the mid - infrared spectral energy distribution does not indicate the presence of warm dust emission , although the optical photospheric absorption and co overtone emission could be related to a gaseous disk .
archival photometry reveals a prior outburst in 1976 .
although we speculate about possible causes for this outburst , none of the explanations are compelling . |
many soft materials consist of building blocks that are themselves `` soft '' , meaning that , under experimental conditions , they can interpenetrate or deform appreciably due to the forces acting between them .
linear polymers provide an example : @xcite at moderate free - energy cost , the centers of mass of two distinct polymers can be made to superimpose . these macromolecules are therefore fully interpenetrable .
other particles do not interpenetrate but are deformable , like polymer micelles : they deform elastically at high densities , thus exploring configurations that are forbidden to hard - particle systems .
systems of soft particles are interesting because crystallization in simple and colloidal systems is usually attributed to excluded - volume effects .
systems with soft - repulsive , short - range interactions are therefore expected to have a qualitatively different freezing behavior than those with hard particles .
@xcite indeed , one of the striking features of systems of particles with a bounded repulsive interaction is that they have a maximum melting temperature . above it ,
the solid phase is unstable .
this phenomenon can be understood as follows : consider a system with an interparticle potential that has its maximum energy @xmath0 at zero separation and decreases monotonically to zero within a distance @xmath1 . when the thermal energy @xmath2 , interparticle interactions become negligible and the system approaches ideal gas behavior . however , when @xmath3 particles can only overlap at a significant energetic cost . in the limit @xmath4 , the system will undergo a hard - sphere freezing transition .
in fact , some 30 years ago , stillinger already showed that the so - called gaussian core model ( gcm ) has a maximum melting point .
@xcite moreover , stillinger , and subsequently other authors , @xcite showed that the gcm also exhibits re - entrant melting upon compression , and a transition from face- to body - centered cubic crystal structures .
such a rich phase behavior is not limited to the repulsive gaussian potential , but has been shown to be common to many effective , soft potentials that were designed to reproduce the phase behavior and dynamics of microgels , @xcite star - polyelectrolyte solutions @xcite and star - polymer solutions .
@xcite however , not all soft - repulsive potentials give rise to re - entrant melting . the penetrable sphere model ( psm ) ,
@xcite a square - shoulder potential , is the simplest model that leads to freezing at any arbitrary temperature and to the appearance of cluster crystals ( multiple occupation of lattice sites ) .
multiple occupancy in the psm arises because , as soon as two particles interpenetrate , no further energetic cost exists for full overlapping . as a consequence , particles can sit on top of each other thus reducing the total number of overlaps and the interaction energy .
density functional theory and molecular simulations have confirmed the occurrence of cluster crystals for the generalized exponential model . @xcite it is expected that certain dendrimers could form such crystals , @xcite but thus far experiments are lacking ( see , however , ref . ) .
it is apparent from the discussion above that the shape of the pair potential determines when the re - entrant melting and clustering scenarios manifest in the phase diagram .
indeed , likos _ et al . _
@xcite established a simple criterion to distinguish between the two scenarios for bounded repulsive potentials : re - entrant melting happens for bounded potentials with a positive definite fourier transform .
otherwise , clustering and freezing are expected to occur at all temperatures .
although the type of scenario , either re - entrant melting or clustering , can be predicted , there is no method that guarantees an _ a priori _ prediction of all stable crystal structures given the bounded repulsive potential .
the usual method involves selecting a small number of candidate structures , for which the ground - state energy or finite - temperature free energies are computed . among the candidates ,
the stable structure at any particular density and temperature is the one with the lowest ( free ) energy .
this works well for typical pair potentials , for which one assumes that unusual crystals are unlikely to be stable structures .
however , soft potentials are known to induce multiple crystals , @xcite including the less typical non - cubic lattices , and exotic structures like the a15 and diamond @xcite ( for an extreme example , see ref . ) .
there exist powerful approaches to search for stable crystal structures , such as those based on metadynamics @xcite and genetic algorithms .
@xcite however , metadynamics can be very expensive , in particular when trying to escape deep , local free - energy minima . in a genetic algorithm , an initially random population of lattices
is modified using evolutionary rules according to a fitness function : high fitness corresponds to low ( free ) energy .
the structure with the highest fitness in the population after breeding a few generations is assumed to be the most stable lattice .
although genetic algorithms greatly alleviate the problem of searching for candidate crystal structures , they still suffer from the same numerical bottleneck of the usual method : the expensive computation of free energies .
for this reason , numerical searches for stable crystal structures are often based on energies , rather than free energies . to improve upon this approach
, one can approximate the free energy by that of a harmonic crystal or by using density - functional theory
. however , to improve accuracy , one has to turn to computer simulation methods that generate free energies that are `` exact '' to within statistical accuracy .
accurate free - energy calculations can be crucial in soft systems where , specially at high densities , crystals with very similar energies can compete for stability .
in this paper we focus on the phase behavior of hertzian spheres .
the hertz potential describes the change in elastic energy of two deformable objects when subjected to an axial compression .
@xcite the potential has the following form : @xmath5 where @xmath1 and @xmath0 set the length and energy scales and @xmath6 is the distance between the centers of the undeformed spheres . in the limit @xmath7
the hard - sphere model is recovered .
the hertz model was developed to describe the interaction between elastic spheres that are deformed only slightly .
clearly , for very large compressions or at high densities ( well above the overlap concentration ) the assumptions underlying the original hertz model no longer apply . here
we use the hertz potential as a simple representation of soft particles that is bounded ( i.e. it remains finite at @xmath8 ) , has a finite range ( viz .
@xmath1 ) and a positive , yet sharply decaying fourier transform . also , the hertz potential is computationally cheap and thus allows us to perform extensive simulations for the calculation of free energies and coexistence points .
we have not attempted to map the present model onto a specific soft - colloid or star - polymer system .
@xcite we just note that a judicious design of star polymers or dendrimers @xcite makes it possible to `` engineer '' the interaction between these particles to a considerable degree . using standard thermodynamic integration , we calculated free energies of the fluid phase and of candidate crystal structures . as candidates , we considered all bravais lattices and the hexagonal close - packed , diamond , and a15 structures .
our results show that the phase diagram of the hertz model exhibits multiple re - entrant melting and polymorphic transitions between a number of crystals .
in addition , we find that the re - entrant fluid shows unusual diffusivity curves , and speculate about the nature of this fluid at high density and low temperature . the general primitive cell and vectors @xmath9 , @xmath10 and @xmath11.,height=207 ] l c c c c c c bravais lattice ' '' '' & @xmath12 & @xmath13 & @xmath14 & @xmath15 & @xmath16 & @xmath17 + simple cubic ( sc ) ' '' '' & 1 & 1 & 1/2 & 1/2 & 1/2 & 1 + face - centered cubic ( fcc ) & @xmath18 & @xmath18 & 1/3 & 1/4 & 1/4 & 1/4 + body - centered cubic ( bcc ) & 1 & @xmath19 & 0.304 & 0.304 & 1/2 & 1/2 + hexagonal ( h ) & 1 & ( 1/2 ) & 1/2 & 1/2 & 1/3 & [ @xmath20 + simple tetragonal ( st ) & 1 & 1/2 & 1/2 & 1/2 & 1/2 & 1/2 + body - centered tetragonal ( bct ) & 1 & ( 3/4 ) & [ 0.268 ] & [ 0.268 ] & 1/2 & [ 1/4 ] + rhombohedral or trigonal ( r ) & 1 & 1 & ( 1/3 ) & [ 1/3 ] & [ 1/3 ] & [ @xmath18 ] + simple orthorhombic ( so ) & ( @xmath18 ) & ( @xmath19 ) & 1/2 & 1/2 & 1/2 & [ 0.612 ] + base - centered orthorhombic ( baco ) & ( 1 ) & ( @xmath18 ) & 1/2 & 1/2 & [ 1/4 ] & [ 1/2 ] + body - centered orthorhombic ( bco ) & ( @xmath18 ) & ( 4/5 ) & [ 0.285 ] & [ 0.285 ] & 1/2 & [ 0.265 ] + face - centered orthorhombic ( fco ) & ( 3/5 ) & ( @xmath19 ) & [ 0.340 ] & [ 0.304 ] & [ 0.186 ] & [ 0.235 ] + simple monoclinic ( sm ) & ( @xmath18 ) & ( @xmath19 ) & ( 1/3 ) & 1/2 & 1/2 & [ 0.530 ] + base - centered monoclinic ( bacm ) & 1 & ( @xmath18 ) & ( 1/3 ) & 1/3 & ( 1/4 ) & [ 0.420 ] + triclinic ( t ) & ( @xmath18 ) & ( @xmath19 ) & ( 1/3 ) & ( 0.4 ) & ( 2/3 ) & [ 0.306 ] + l c c c non - bravais lattice ' '' '' & reference bravais lattice & @xmath22 & @xmath23 + diamond ( d ) ' '' '' & fcc & 1 & @xmath24 + hexagonal close - packed ( hcp ) & simple hexagonal ( sh ) ( h with @xmath25 ) & 1 & @xmath26 + a15 & sc & 7 & @xmath27 + & ' '' '' & & @xmath28 +
all 14 bravais lattices can be generated from a general primitive unit cell @xcite defined by three independent vectors @xmath9 , @xmath10 and @xmath29 .
once these primitive vectors are defined for a particular crystal , the location of every particle in the crystal follows from @xmath30 where @xmath31 and @xmath32 , and @xmath33 is an arbitrary vector . for simplicity
, we can fix one of the particles at the origin of the coordinate system , and thus @xmath34 .
there are many ways of defining primitive vectors for each bravais lattice . here
we conveniently choose the general set of vectors depicted in fig .
[ fig : primitive_cell ] , where @xmath9 is parallel to the @xmath35 axis , and @xmath9 and @xmath10 lie on the @xmath36 plane of the cartesian coordinate system .
algebraically , @xmath37 where @xmath38 , @xmath39 , and @xmath40 are the lengths of @xmath9 , @xmath10 and @xmath11 , respectively , and @xmath41 , @xmath42 and @xmath43 denote the angles formed by each pair of primitive vectors , as indicated in fig .
[ fig : primitive_cell ] .
the angle @xmath44 is such as @xmath45 the volume @xmath17 of the primitive cell is @xmath46 and , therefore , @xmath47 from eq .
[ eq : volume ] we can see that the length @xmath38 sets the specific volume of the crystal . without loss of generality ,
we choose @xmath38 to be the longest among the three primitive vectors .
hence , according to eqns .
[ eq : position_vector]-[eq : alpha ] , the 14 bravais lattices can be described with five parameters : @xmath48 , @xmath49 , @xmath50 , @xmath51 and @xmath52 .
table [ table : bravais_lattices ] shows the specific values of the parameters for the 14 bravais lattices .
depending on the particular lattice , any of the five parameters can be free , fixed or dependent .
a free parameter , shown in round brackets in table [ table : bravais_lattices ] , can be changed at will to generate different specific structures belonging to the same lattice , provided that @xmath53 .
a dependent parameter , shown in square brackets , has values that depend on those assigned to the free parameters , according to the expressions in table [ table : lattice_dependent_parameters ] . by definition ,
some lattices are a subset of the tetragonal , trigonal , orthorhombic , monoclinic or triclinic , as indicated in table [ table : lattice_equivalents ] .
bravais lattices have 1 particle in their primitive unit cell .
the primitive unit cell of non - bravais crystals contains @xmath22 additional particles that are characterized by a set of translation or basis vectors @xmath54 , where @xmath55 runs from @xmath56 to @xmath22 . because of translational symmetry ,
one of the basis vectors can be chosen arbitrarily , and this is why in eq .
[ eq : position_vector ] we defined @xmath57 .
a non - bravais crystal is generated by translation of its reference bravais lattice using the set of basis vectors @xmath58 table [ table : non - bravais_lattices ] shows the reference lattice and basis vectors corresponding to the three non - bravais crystals considered in this work . according to the framework described in sections [ sec : bravais ] and [ sec : non - bravais ]
, one can generate any crystal structure as a periodic repetition of the primitive unit cell along its three primitive vectors .
equivalently , one can also construct a unit cell that consists of an arbitrary number of primitive cells , and replicate the unit cell along its lattice vectors to obtain the same periodic structure .
therefore , we can define a non - primitive unit cell that contains a sufficiently large number of particles , and use general periodic boundary conditions to simulate the bulk crystal .
the lattice vectors that define this unit cell are @xmath59 , @xmath60 and @xmath61 , where @xmath62 , @xmath63 and @xmath64 are the number of primitive cells contained in the unit cell in each of the directions of the lattice vectors .
the nearest image of a particle in a unit cell can be found using the reciprocal lattice vectors @xmath65 , @xmath66 and @xmath67 , which are defined such as @xmath68 . using the general primitive vectors of eqns .
[ eq : primitive_vectors1]-[eq : primitive_vectors3 ] , the corresponding reciprocal vectors can be written as @xmath69 following eq .
[ eq : position_vector ] , the location @xmath70 of a particle inside the unit cell with an image at position @xmath71 can be calculated as @xmath72 where @xmath73 and @xmath74 denotes the nearest integer .
note that , by definition , @xmath75 . l c bravais lattice ' '' '' & dependent parameters + bct , bco ' '' '' & @xmath76 $ ] + r ' '' '' & @xmath77 + fco & @xmath78 $ ] + fco & @xmath79 $ ] + baco , fco & @xmath80 $ ] + l r bravais lattice ' '' '' & parameters and equivalent lattice + st ' '' '' & @xmath81 sc + bct & @xmath82 bcc + r & @xmath83 sc + so & @xmath84 or @xmath25 or @xmath85 st + & @xmath86 sc + baco & @xmath87 sc + bco & @xmath88 bct + & @xmath84 and @xmath82 bcc + fco & @xmath87 fcc + sm & @xmath83 so + bacm & @xmath25 and @xmath89 r with @xmath90 + t & parameters of table [ table : bravais_lattices ] @xmath91 any of the lattices + l c c c c c c c c c c c phases 1 - 2 ' '' '' & @xmath0 & @xmath92 & @xmath93 & @xmath94 & @xmath95 & phases 1 - 2 & @xmath0 & @xmath92 & @xmath93 & @xmath94 & @xmath95 + f - fcc ' '' '' & 12000 & @xmath968.333@xmath97 & 12.34 & 1.000 & 1.104 & f - h & 280 & 3.571 & 497.4 & 3.644 & 3.709 + f - fcc & 1000 & 1.000 & 14.74 & 1.145 & 1.252 & f - h & 250 & 4.000 & 474.5 & 3.764 & 3.811 + f - fcc & 400 & 2.500 & 17.78 & 1.295 & 1.389 & f - h & 228 & 4.386 & 501.2 & 4.046 & 4.050 + f - fcc & 200 & 5.000 & 24.15 & 1.531 & 1.602 & h - f & 270 & 3.704 & 767.4 & 4.579 & 4.518 + f - bcc & 125 & 8.000 & 44.75 & 2.003 & 2.028 & h - f & 240 & 4.167 & 617.1 & 4.361 & 4.325 + f - bcc & 113.1 & 8.842 & 71.17 & 2.398 & 2.398 & h - sc & 12000 & @xmath968.333@xmath97 & 38290 & 4.725 & 4.822 + fcc - bcc & 12000 & @xmath968.333@xmath97 & 5912 & 2.214 & 2.279 & h - sc & 1000 & 1.000 & 3132 & 4.693 & 4.785 + fcc - bcc & 1000 & 1.000 & 494.6 & 2.216 & 2.275 & h - sc & 400 & 2.500 & 1189 & 4.597 & 4.676 + fcc - bcc & 400 & 2.500 & 194.3 & 2.206 & 2.252 & h - sc & 270 & 3.704 & 768.6 & 4.518 & 4.586 + fcc - bcc & 200 & 5.000 & 88.35 & 2.148 & 2.175 & sc - f & 270 & 3.704 & 802.4 & 4.674 & 4.677 + fcc - bcc & 150 & 6.667 & 58.23 & 2.071 & 2.086 & sc - f & 325 & 3.077 & 1215 & 5.184 & 5.227 + fcc - bcc & 130 & 7.692 & 41.28 & 1.959 & 1.966 & sc - f & 400 & 2.500 & 1647 & 5.419 & 5.478 + bcc - f & 280 & 3.571 & 491.9 & 3.540 & 3.625 & sc - bct & 12000 & @xmath968.333@xmath97 & 48510 & 5.379 &
5.506 + bcc - f & 200 & 5.000 & 310.1 & 3.355 & 3.427 & sc - bct & 1000 & 1.000 & 4091 & 5.402 & 5.526 + bcc - f & 140 & 7.143 & 164.6 & 2.997 & 3.045 & f - bct & 400 & 2.500 & 1665 & 5.508 & 5.558 + bcc - f & 125 & 8.000 & 123.9 & 2.808 & 2.840 & f - bct & 350 & 2.857 & 1595 & 5.760 & 5.799 + bcc - h & 12000 & @xmath968.333@xmath97 & 20470 & 3.453 & 3.686 & f - bct & 310 & 3.226 & 1574 & 6.076 & 6.105 + bcc - h & 1000 & 1.000 & 1724 & 3.477 & 3.701 & bct - f & 325 & 3.077 & 2069 & 6.750 & 6.796 + bcc - h & 400 & 2.500 & 699.5 & 3.516 & 3.699 & bct - f & 300 & 3.333 & 1683 & 6.381 & 6.384 +
in order to construct the phase diagram , we first performed a rough test of the stability of the fluid phase by running short nvt monte carlo simulations at 100 points equally distributed in the @xmath98 space , @xmath99 within the interval @xmath100 $ ] and @xmath101 within @xmath102 $ ] .
this allowed us to map approximately the solid region of the diagram .
secondly , we distributed 28 points along the isotherm at @xmath103 for @xmath101 within @xmath104 $ ] , and tested the mechanical stability of all bravais crystals and the three non - bravais structures considered by running a short nvt simulation for each specific structure . in particular , for lattices with free parameters , typically simulations at 5 different values for each free parameter were carried out .
this large set of simulations reduced significantly the number of crystal structures for which we subsequently computed free energies accurately : in particular , at @xmath105 we calculated free energies for the sc , bcc , fcc , r and h. at @xmath106 , for the sc , h , and r. at @xmath107 , for the st , bct , bco , h , and r. to determine the thermodynamically - stable crystals , we obtained free energies for the set of mechanically - stable candidate structures .
we employed standard thermodynamic integration using the einstein crystal as a reference state and a 20-point gauss - legendre numerical integration . to assess accuracy , particularly at high densities
, we carried out four integrations for each crystal structure , at values of @xmath108 equal to 50 , 100 , 150 and 200 , where @xmath41 is the einstein - crystal spring constant .
the calculated free energies are thus averages over four integrations .
our nvt simulations typically consisted of 500 - 700 particles , the actual number depending on the crystal structure . in particular , we used 500 spheres for the fcc , 686 for the bcc , 512 for the h and sc , and 648 for the bct .
occasionally , simulations with about 1000 particles were run at a few points distributed in the @xmath98 space , but we did not find any significant finite - size effects . the determination of coexistence points involved npt simulations to compute the local equation of state @xmath109 around each pair of guesses for the equilibrium densities .
we used 30 simulation runs for each guess and fitted a 10-degree polynomial function to the data .
the free energy of the solid phase at the estimated density was calculated via integration from the einstein crystal as explained in the paragraph above .
we computed the free energy of the fluid by integrating energies from 800-particle nvt simulations along a constant density path starting at the isotherm at @xmath110 .
this isotherm consisted of polynomial fits to the equation of state of the fluid for eight different equally - spaced intervals of @xmath101 within the range @xmath111 $ ] .
we obtained free energies along the isotherm by integrating the equation of state from the ideal - gas limit at zero density .
we generated as many simulation points as necessary to keep standard deviations of integrals below 0.01% .
accuracy was absolutely important also when comparing free energies of competing crystals , since differences were often below @xmath112 per particle .
once the local equations of state @xmath109 and chemical potentials @xmath113 at fixed @xmath114 were determined , we searched for the densities @xmath115 and @xmath116 that satisfy @xmath117 and @xmath118 .
we then checked that @xmath115 and @xmath116 fall within the interval for which @xmath109 and @xmath113 were fitted .
otherwise , we used @xmath115 and @xmath116 as new guesses for the coexistence densities , and repeated the procedure explained in the previous paragraph .
according to the criterion established by likos in ref .
, soft potentials that are purely repulsive and bounded give rise to re - entrant melting when its fourier transform exists and it does not have negative components .
if the fourier transform oscillates around zero , cluster solids appear instead .
the fourier transform of the hertz potential is definite positive and , indeed , our results confirm the re - entrant melting scenario , as fig . [
fig : phase_diagram ] shows .
table [ table : coexistence_points ] summarizes phase coexistence data .
the phase diagram of hertzian spheres shows that the fluid phase freezes upon compression to form an fcc crystal which , at higher densities , turns into a bcc structure .
however , the bcc packing is favored over the fcc at high temperatures because particles in the former have a higher vibrational entropy .
an inset in fig .
[ fig : phase_diagram ] shows in detail the area around the f - fcc - bcc triple point , which occurs at @xmath119 . the fcc phase does not have a maximum freezing point for the hertz potential , as opposed to the gcm .
@xcite the fluid re - enters at densities larger than the maximum freezing point at @xmath120 and @xmath121 . at lower temperatures and with increasing density , we find multiple re - entrant melting and four more crystals : hexagonal , simple cubic , body - centered - tetragonal and trigonal .
simple cubic crystals of one component are rare , because they are usually mechanically unstable .
free energies per particle of the hexagonal ( left ) and body - centered tetragonal ( right ) crystals , as a function of the respective aspect ratio @xmath13 of the primitive unit cell .
free - energy sets have been displaced arbitrarily by @xmath122 for each specific point in the phase diagram , as indicated in the legends .
errors are smaller than the size of the symbols .
the optimal @xmath13 , at which the free energy is at a minimum , is density dependent for the bct structure .
the inset shows that the density behavior of the optimal @xmath13 has inflection points such that aspect ratios close to that of the bct at the maximum melting point are favored .
lines are a guide to the eye . ] with regard to the h and bct structures , the geometry of their primitive cells is not unique , as explained in section [ sec : bravais ] . both crystals have one degree of freedom , namely the aspect ratio of the primitive cell @xmath13 ( see table [ table : bravais_lattices ] ) .
one expects that the _ optimal _ value of @xmath13 , at which the free energy reaches a minimum , is in general density dependent .
we show this in fig .
[ fig : free_energies ] , where we plotted free energies per particle as a function of the aspect ratio . for the h crystal ,
the minimum in free energy occurs at @xmath123 , independently of temperature . within the range of densities
at which the h structure is stable , we do not observe any significant density dependence either , as the free - energy curve is relatively flat around the minimum .
however , this is not the case for the bct . as shown in the inset of fig .
[ fig : free_energies ] , the maximum freezing point , occurring at @xmath124 , leaves its `` imprint '' on the density dependence of the optimal aspect ratio of the bct primitive cell , as there is a clustering of aspect ratios close to @xmath125 , which is the value at the maximum freezing point .
very recently , prestipino _ et al . _
@xcite calculated the ground - state structures for the hertz potential .
they found that the non - bravais ci16 and a5 lattices minimize the energy at densities within the interval @xmath126 $ ] .
we did not include the ci16 and a5 in the set of candidate structures , and hence we can not tell whether the ci16 and a5 remain stable at @xmath127
. however , we stress that entropic contributions are not negligible even at the lowest temperatures that we studied . for example , at @xmath106 , the trigonal crystal with @xmath128 yields the lowest energy of all the candidate structures considered in this work . yet ,
the simple cubic structure has the lowest free energy ; it is thus stabilized by entropy .
furthermore , as discussed in detail below , the phase diagram and both structural and dynamical anomalies seen in the fluid are interconnected .
for example , the maximum melting point of the bct lies at @xmath124 , which is approximately the density at which the radial distribution function and the diffusion coefficient reach local extrema .
but at @xmath129 a transition between the a5 and the bct happens to be close to that same density @xcite ( the a5 becomes unstable at @xmath130 ) , which makes an extremum occurring at the same location in the phase diagram unlikely .
the fact that there is such a strong correlation between the shape of the computed phase diagram and the dynamics of the fluid provides indirect support for the assumption that we have indeed identified the stable crystal phases at @xmath127 . in the next paragraphs we discuss a few interesting aspects of the fluid phase .
the re - entrance of the fluid at high densities and low temperatures ( e.g. at @xmath131 and @xmath132 ; see fig .
[ fig : phase_diagram ] ) appears to be different in nature from that of a hard - sphere fluid , which has a rugged potential - energy landscape and yet is a fluid because there is an extended and connected region in configuration space where the potential energy is strictly zero ( and therefore flat ) . for the hertz model at the aforementioned density ,
the typical inter - particle distance in the liquid phase is of the order of 0.5@xmath1 and the potential is at less than 20% of its maximum value ; also , at the aforementioned temperature , the relative radial displacement of two particles by as little as @xmath133 will change the potential energy of such a pair by an amount that is comparable to @xmath134 . hence , the potential - energy landscape is far from being flat at these conditions .
we suggest that the reason why hertzian spheres ( and other potentials showing re - entrant melting , like the gcm ) melt at high densities and low temperatures has to do with the fact that the fourier transform of the potential goes to zero rapidly .
indeed , st @xcite ( see also ref . ) has shown rigorously that integrable bounded potentials whose fourier transform is non - negative and vanishes above a wave number @xmath135 have an infinite number of continuously degenerate ground states above a well - defined threshold density .
this means that , for this class of pair potentials , the potential energy is completely flat along certain directions in configuration space .
although the hertz potential does not satisfy the st criteria ( its fourier transform does not vanish above any @xmath135 ) , its fourier transform does decay @xmath136 for wave vectors larger than @xmath137 ( see inset in fig . [
fig : phase_diagram ] ) .
we speculate that , as a result , certain collective motions in the dense phase cost very little potential energy .
it is the softness of the pair potential that causes melting into a rather peculiar liquid for which particle motions should be strongly correlated .
diffusion coefficient of the fluid phase obtained by molecular dynamics as a function of density at three temperature values .
breaks in the lines joining the simulation data indicate that points in between were identified as crystalline states .
the inset shows the radial distribution function @xmath138 for the isotherm at @xmath139 and the following values of @xmath101 : 0.5 , 1 , 1.5 , 2 , 2.5 , 3 , 4 , 5 , 6 , 7 , 8 , 10 , 12 .
the highest peak in each @xmath138 is marked with a black dot . in the set of black dots ,
the local maximum and minimum occur approximately at the densities at which the diffusion coefficient has a minimum ( @xmath101=2.40 ) and a maximum ( @xmath101=6.35 ) , respectively . ] in fig .
[ fig : diffusivity ] we show other aspects of the peculiarity of the dynamics of the fluid phase .
the diffusion coefficient @xmath140 , which decreases monotonically with increasing density in simple liquids , shows a clear non - monotonic behavior in the hertzian fluid .
moreover , the minimum in @xmath140 at @xmath141 coincides with the maximum freezing point at @xmath121 , and a maximum in the first peak of the @xmath138 ( see inset in fig . [ fig : diffusivity ] ) . also the maxima in @xmath140 for the three temperatures shown in fig .
[ fig : diffusivity ] are located at the same density as the maximum freezing point at @xmath124 and a minimum in the first peak of @xmath138 .
these multiple structural and dynamical anomalies , and their interrelation , has been seen before in bounded potentials , @xcite in liquids with water - like anomalies @xcite and colloids with short - ranged attractions .
@xcite this suggests that crystallization and fluid dynamics are inextricably connected , and that this connection is not independent of the underlying crystal structure .
for instance , the presence of the h and sc crystals does not imprint peaks in the diffusion coefficient as the bct does .
we think that this is related to the fact that single - particle mobility is affected by differences in packing efficiency : the intrinsic volume of the bct primitive cell is much smaller than that of the bcc , h and sc structures .
besides , even though particle motions in the low - temperature liquid phase are strongly correlated , the liquid does not appear to be particularly glassy .
in fact , within the range of densities studied , we observe spontaneous crystallization rather than structural arrest upon cooling even at temperatures below @xmath142 .
further study of the dynamics of the high - density , supercooled liquid would therefore be most interesting .
another interesting feature of the system at high densities and low temperatures is its freezing behavior .
we have used free - energy calculations to trace freezing curves up to a density of @xmath143 . even beyond that density
, there appears to be a succession of other freezing transitions .
although we have not traced the accurate phase boundaries completely for @xmath144 , we have calculated free energies at @xmath145 and @xmath146 , and found the stable structure to be the bct with @xmath147 . additionally , we have found that crystallization occurs at low temperatures and at densities up to @xmath148 . in other words : as far as we can tell , the low - temperature phase is always crystalline , but the structure of the stable crystalline phase changes as the density is increased .
there is also evidence @xcite at @xmath129 supporting this observation .
in fact , the ground - state behavior is interesting in the context of a result reported by torquato and stillinger : @xcite on the basis of duality relations between a soft potential and its fourier transform , these authors found a one - dimensional system , the overlap potential , that exhibits an infinite number of phase transitions between periodic ground states over the entire density range .
one can then wonder whether the hertz potential also shows an unbounded number of phase transitions between periodic structures .
interestingly , the hertz potential and the overlap potential in 1d and 3d , @xmath149 and @xmath150 for @xmath151 , studied in ref . , belong to the same class of non - negative , bounded functions , and both have oscillatory , decaying fourier transforms ( see ref . and inset in fig .
[ fig : phase_diagram ] ) . in summary , we have shown that the hertz potential gives rise to a phase diagram with multiple re - entrant melting transitions and a succession of bravais crystals spanning a wide range of densities .
in addition , the diffusion coefficient of the re - entrant fluid shows unusual non - monotonic curves with increasing density .
this rich behavior together with the simplicity of the potential s functional form makes the hertz model an attractive one for the study of kinetic phenomena in soft - core systems , like martensitic transformations and supercooled dynamics .
thanks gerhard kahl and bianca mladek for their help during a stay at tu wien . financial support from the eu contract mrtn - ct-2003 - 504712 and
the short - term scientific mission cost - stsm - p13 - 02014 are acknowledged .
a.c . acknowledges financial support from columbia university .
the work of the fom institute , where most of this work has been carried out , is part of the research program of fom and is made possible by financial support from the netherlands organization for scientific research ( nwo ) . | we report the phase diagram of interpenetrating hertzian spheres .
the hertz potential is purely repulsive , bounded at zero separation , and decreases monotonically as a power law with exponent 5/2 , vanishing at the overlapping threshold .
this simple functional describes the elastic interaction of weakly deformable bodies and , therefore , it is a reliable physical model of soft macromolecules , like star polymers and globular micelles . using thermodynamic integration and extensive monte carlo simulations
, we computed accurate free energies of the fluid phase and a large number of crystal structures .
for this , we defined a general primitive unit cell that allows for the simulation of any lattice .
we find multiple re - entrant melting and first - order transitions between crystals with cubic , trigonal , tetragonal and hexagonal symmetries . |
using eqs . ( [ rr ] ) and ( [ uv ] ) the relevant part of the contribution to fidelity comming from alice s registration @xmath95 and bob s cp map defined by angles @xmath73 and @xmath74 reads @xmath96 where @xmath97 and all terms are made positive by restricting the ranges of @xmath98 and @xmath74 to @xmath99 $ ] .
now @xmath100 is bounded from above by @xmath101 ( arithmetic vs. geometrical mean ) , and this in turn is smaller than @xmath102 ( convexity ) .
@xmath103 can be attained by applying degenerated cp map defined by angles @xmath104 , @xmath105 : @xmath106 .
hence for alice s povms ( [ povms ] ) and the singlet state shared by alice and bob the optimum bob s cp map is contained in the set of one - parametric cp maps .
99 d. bouwmeester , j. pan , k. mattle , m. eibl , h. weinfurter , and a. zeilinger , nature * 390 * , 575 ( 1997 ) ; d. boschi , s. branca , f. de martini , l. hardy , and s. popescu , phys . rev . lett . * 79 * , 2755 ( 1997 ) ; a. furusawa , j. srensen , s.l .
braunstein , c.a .
fuchs , h.j .
kimble , and e.s .
polzik , science * 282 * , 706 ( 1998 ) .
bennett , c. brassard , c. crpeau , r. jozsa , a. peres , and w.k .
wootters , phys .
lett . * 70 * , 1895 ( 1993 ) .
n. ltkenhaus , j. calsamiglia , and k .- a .
suominen , phys . rev . a * 59
* , 3295 ( 1999 ) .
c. trump , d. bru and m. lewenstein , phys .
a * 279 * , 1 ( 2001 ) .
r. horodecki , m. horodecki , p. horodecki , phys .
a * 222 * , 21 ( 1996 ) .
k. banaszek , phys . rev
* 86 * , 1366 ( 2001 ) .
k. kraus , ann .
phys . * 64 * , 311 ( 1971 ) .
no generality is lost by this assumption since mixed states are just convex combinations of pure states .
alice performes optimal state estimation and sends this to bob who then reconstructs the state on the basis of this information .
p. badziag , m. horodecki , p. horodecki , and r. horodecki , phys .
a * 62 * , 012311 ( 2000 ) .
ruskai , s. szarek , and e. werner , arxiv : quant - ph/0101003 . g. m. dariano and p. lo presti , arxiv : quant - ph/0101100 ; m. sacchi , phys .
a * 63 * , 054104 ( 2001 ) . | we study general teleportation scheme with an arbitrary state of the pair of particles ( @xmath0 and @xmath1 ) shared by alice and bob , and arbitrary measurements on the input particle 1 and one of the members ( 2 ) of the pair on alice s side .
we find an efficient iterative algorithm for identifying optimum local operations on bob s side .
in particular we find that simple unitary transformations on his side are not always optimal even if particles @xmath0 and @xmath1 are perfectly entangled .
we describe the most interesting protocols in the language of extremal completely - positive maps . of many potential applications of quantum information processing ,
the quantum teleportation is probably the most appealing example .
several experiments @xcite have been done since the first proposal in @xcite , confirming thus that teleportation a popular subject of sci - fi literature is indeed feasible ; at least if one deals with simple quantum objects .
it is well known that ideal teleportation requires a source of maximally entangled particles and a very delicate measurement on the sender s side .
this is not always easy to do with realistic experimental devices .
for example , it has been shown in @xcite that a never - failing bell - state measurement is impossible with linear elements and detectors only .
more severe limitations might arise if the teleported object gets more complicated .
however , even with restricted resources there is still the possibility to optimize the teleporting scheme .
a natural question is : what is the optimum local operation on particle @xmath1 ( output of teleportation ) for the given resources of the shared entangled pairs of particles ( @xmath0 and @xmath1 ) , and of the measurements performed jointly on particle @xmath0 and the input particle @xmath2 ?
several attempts have been done in this direction . in @xcite
the authors analyzed a teleportation with realistic linear elements .
they found that local unitary transformations on the side of receiver are not always optimum ; sometimes more general operations can improve the performance of the realistic protocol .
optimization of protocols with arbitrary shared entangled states has been pursued in @xcite and @xcite but the optimization was done over local unitary transformations only .
the main goal of this paper is to consider the most general case . for the given resources of the sender ( type of the measurement on particles @xmath2 and
@xmath0 ) and resources on the quantum channel shared by two parties ( the state of particles @xmath0 and @xmath1 ) we will optimize the teleportation protocol by finding the optimum local operation that should be applied by the receiver on particle @xmath1 .
in contrast to the naive picture suggesting that every interaction of a quantum system with environment leads to a `` loss of information '' , we here find that such an interaction on receiver s side might enhance the fidelity of teleportation even if the pair of particles constituting the quantum channel are perfectly entangled .
the paper has two parts . in the first one we will derive an iterative algorithm that solves our problem . in the second part we will try to describe the most interesting protocols involving two - level systems in the language of extremal completely - positive maps .
let us consider the teleportation of an unknown state @xmath3 of particle @xmath2 between two parties called alice and bob .
let us assume that before the teleportation starts they share two particles in an arbitrary state @xmath4 .
alice then performs a measurement on particles @xmath2 and @xmath0 and sends the outcome by a classical channel to bob . based on the classical communication he receives he performs a local transformation on particle @xmath1 .
the optimum local transformation is such that the final state of particle @xmath1 gets as close to the input state @xmath3 as possible on the average .
the measurement performed by alice will be described by a positive operator - valued measure ( povm ) @xmath5 , @xmath6 .
its elements generate probabilities of all possible outcomes of alice s measurement .
consider the situation where one such particular outcome , say @xmath7 has been registered with probability @xmath8 .
the state of the third particle conditioned on this result reads @xmath9 where @xmath10 now bob applies a local ( @xmath11-dependent ) transformation on this state .
the most general transformation is a trace preserving completely positive ( cp ) map of the form @xcite @xmath12 the state describing the ensemble of teleported particles is obtained by averaging ( [ cp ] ) over all possible outcomes , @xmath13 the optimality of the given set of local transformations will be judged by the fidelity of the corresponding teleported state . to simplify further considerations let us assume that the input state is pure @xcite . in that case
the fidelity can be defined as follows : @xmath14 , where @xmath15 is the input state in the hilbert space of particle @xmath1 . using eqs . ( [ state - cond ] ) and ( [ final ] )
the fidelity becomes @xmath16 where @xmath17 mean averaging over a priori distribution of input states .
the average fidelity @xmath18 will be our global measure of quality of the given teleportation protocol that is to be maximized over the set of all possible local operations applied to particle @xmath1 .
although the following optimization can be carried out for arbitrary dimensional hilbert spaces we will illustrate the idea on the simple example of spin-@xmath19 particles . the generalization to more dimensions is straightforward .
we start by decomposing the input density matrix in some basis of hermitian generators . in the case of spin @xmath19 the convenient choice is the basis of pauli spin matrices . after substituting the decomposition in eq .
( [ fidel ] ) and simple integration over the whole surface of poincare sphere ( we assume that all possible input states are equally - probable states , hence the isotropic a priori distribution ) , we get @xmath20 where @xmath21 where @xmath22 and where we have dropped now unnecessary subscript of particle @xmath1
. since cp maps corresponding to different registrations @xmath23 are independent , each term on the right - hand side of ( [ f2 ] ) can be maximized independently .
omitting therefore notation @xmath11 and taking the constraint @xmath24 into account , the expression to be maximized is @xmath25 variation of this expression with respect to @xmath26 gives the extremal equation in the form @xmath27 where @xmath28 is the ( hermitian ) lagrange operator .
it can be determined from eq .
( [ cp - extrem ] ) as follows : @xmath29 where we have introduced hermitian operators @xmath30 which provide another representation of the cp map @xmath31 .
eq .
( [ cp - extrem ] ) can be brought to the form suitable to iterations .
multiplying it by @xmath32 from the left and summing over @xmath33 we obtain @xmath34 a formula suitable to iterative solving is obtained by adding @xmath35 to the left - hand side of eq .
( [ cp - vec ] ) and rearranging , @xmath36 the iterative algorithm for finding optimum cp maps based on eqs .
( [ cp - lagrang ] ) and ( [ iter ] ) is the main formal result of the present article .
starting from some `` unbiased '' cp map , for example @xmath37 ( means that particle @xmath1 is always brought to the maximally mixed state ) , the equations can be successively iterated until the stationary point is attained . in this way
we get the operators @xmath38 corresponding to the optimum local transformation of particle @xmath1 .
notice that the average fidelity of teleportation bears very simple form when expressed in terms of @xmath39 , @xmath40 notice also that @xmath18 is a linear functional of @xmath38 .
this means that all its maxima lie on the boundary of the set of physically allowed operators @xmath38 that is determined by the constraint of complete positiveness of the corresponding local transformations .
so there is a clear connection between optimum teleportation protocols and extremal cp maps .
the topology of cp maps is a well studied field related to many problems in quantum information processing .
we will use some of the recently derived results on cp maps for the discussion of some cases of special interest in the present problem .
but before we come to this point let us first demonstrate the usefulness of our iterative optimizing algorithm on some interesting examples involving spin @xmath19 systems .
it is well known that ideal teleportation protocols ( @xmath41 ) utilize a source of maximally entangled pairs of particles and a bell state analysis on alice s side .
the latter is definitely more difficult to realize in a laboratory .
let us therefore analyze the realistic situation with an ( almost ) perfect source of shared particles but with an imperfect measurement .
the imperfect measurement will be drawn from this one - parametric family of povms , @xmath42 here @xmath43 and @xmath44 are two orthogonal states , for example states spin up and spin down in the @xmath45 direction and @xmath46 $ ] .
one boundary point @xmath47 corresponds to the perfect bell measurement .
alice gets no information about the incoming state but the teleportation with fidelity one is possible .
the other boundary point @xmath48 corresponds to a projective measurement on the first particle and no measurement on the second one . in this case
@xmath49 and @xmath50 , and we have only two distinct outcomes @xmath51 less information is extracted about the input state and an imperfect quantum teleportation is possible .
as we have mentioned above we will take particles @xmath0 and @xmath1 in a maximally entangled state , for instance let them be in the singlet state @xmath52 accordingly , the operators @xmath53 have the form , @xmath54 where @xmath55 the operators @xmath56 generated by the remaining three povm elements @xmath57 , @xmath58 and @xmath59 differ from ( [ rr ] ) only in signs of its elements and hence their contribution to the average fidelity ( [ f3 ] ) is the same
. fidelities ( [ f3 ] ) of the optimum bob s local transformations that were found by our iterative algorithm ( [ cp - lagrang ] ) and ( [ iter ] ) for the above alice s measurements are shown in fig .
[ fig - fidel ] ( solid line ) . as could have been anticipated , the fidelity continuously changes from the classical limit @xmath60 @xcite to the maximum value of @xmath41 .
it is interesting to note that the optimum cp maps for @xmath61 are actually local unitary operations .
however for @xmath62 unitary operations are not optimum , see fig .
[ fig - fidel ] ( dashed line ) .
this becomes trivial statement if @xmath63 . in that case
alice performs no measurement on particle @xmath0 and therefore the state of particle @xmath1 remains a complete mixture for any outcome @xmath11 she obtains .
so the fidelity of `` teleported '' state is @xmath19 if one is allowed to perform unitary transformations only .
in contrast , if one adopts more general transformations , bob can construct the input state with fidelity @xmath64 on the basis of the outcomes of alice s measurement ( which is here optimal state estimation ) thus attaining the classical limit .
the most interesting situations correspond to alice s measurements that allow non - classical teleportation only if followed by a _
non unitary _ operation on bob s particle . in our case
this happens for @xmath65 ( see later ) and the optimum cp map turns out to be a kind of decoherence process @xcite .
let us emphasize again that the above simple example of teleporting spin @xmath19 system has been chosen for the sake of simplicity only . if needed , our iterative algorithm could be straightforwardly generalized to larger hilbert spaces .
the generalization consists of replacing pauli spin matrices by the appropriate basis of hermitian operators in that space and replacing the integration over the surface of the poincare sphere by the integration over the surface of generalized @xmath66-dimensional sphere .
nevertheless , the relative simplicity of the set of cp maps operating on a @xmath0-dimensional space allows one to get further insight into the optimum teleportation protocols and obtain analytical results .
we will therefore stick to the teleportation of spin @xmath19 particles in the following .
first let us express the local operation acting on the bob s particle using operators @xmath38 , @xmath67 where @xmath68 and @xmath69 are defined as follows : @xmath70 , and @xmath71 is the bloch vector defining the state of the bob s particle before he applies the operation @xmath72 .
to each local operation on the bob s particle there correspond a simple transformation of the poincare sphere : the unit sphere is mapped onto an ellipsoid , the lengths of its axes being the eigenvalues of @xmath69 , which is translated by @xmath68 from the origin .
of course the ellipsoid has to lie within the unit poincare sphere ( positivity ) .
however not all such ellipsoids define _ completely _ positive maps which are the most general maps in quantum mechanics .
recently it has been shown @xcite how to parameterize the set of _ extremal _ cp maps comprising the boundary of the convex set of all cp maps .
this set contains all optimum bob s local transformation .
matrix @xmath69 can be always brought to the diagonal form by a local unitary transformation .
when in diagonal form , extremal cp maps can be parametrized by two angles @xmath73 and @xmath74 , @xmath75 with @xmath76 , @xmath77 .
now let us show how the above example can be elegantly solved using this trigonometric parameterization of extremal cp maps . due to the form of matrix @xmath78 eq .
( [ rr ] ) it can be shown that the optimum cp map is degenerated , @xmath79 , so there is only one free parameter left ( see appendix ) . substituting the map ( [ uv ] ) with @xmath79 together with the povm element ( [ rr ] ) into eq .
( [ f3 ] ) and maximizing the fidelity with respect to @xmath74 one easily finds the analytical expression for the optimum fidelity : @xmath80 ( \cos^2\theta+2\cos\theta+3)/6 , & \cos\theta\in [ \frac{1}{2 } , 1 ] \end{array } \right.\ ] ] in contrast to this , the fidelity of the optimum local unitary operation is given by the second of the two expressions for all angles @xmath81 and thus coincides with the optimum value if @xmath61 .
we note that the optimum unitary operation is the identity map @xmath82 ( bob leaves particle @xmath1 alone ) for all @xmath81 .
the analytical solution confirms the results obtained numerically with the help of our iterative algorithm shown in fig .
[ fig - fidel ] .
the most interesting alice s measurement ( [ povms ] ) is that for which optimum unitary operation just gives the classical limit @xmath83 , but enables non - classical teleportation using a non - unitary cp map .
this happens for @xmath84 , optimum fidelity being @xmath85 the enhancement of quantum teleportation by non - unitary process has recently been discussed by badziag and horodeckis family @xcite in a slightly different context .
they considered teleportation protocols with perfect bell analyzer but imperfect preparation of the shared pair of particles and found out that teleportation protocols can sometimes be enhanced by the interaction of the teleportation device with environment ( damping ) .
the most pronounced example yielded improvement corresponding to our @xmath84 case . in fact the choice of our povms ( [ povms ] ) which are bell states being subject to a kind of decohering process has been inspired by their result .
now we can use the language of extremal cp maps to explain why the result of badziag _ et al .
_ is so exceptional .
a general teleportation protocol is , in fact , a cp map composed of two different maps : the first one @xmath86 being the alice s transformation of the input state to the output state sent to bob ; the second being the local operation @xmath72 on the output state applied by bob .
the teleportation protocol @xmath87 becomes perfect if the two maps make up the identity map : @xmath88= \bar{\rho}_3,\quad \forall \bar{\rho}_1.\ ] ] obviously , the most interesting protocols are protocols where both parts @xmath86 and @xmath72 are _ extremal _ cp maps , because they contain optimal maps , see discussion after eq .
( [ f3 ] ) , and because all remaining protocols are just convex combinations of such `` extreme '' protocols . among the protocols consisting of two extremal maps one can find the standard teleportation , both maps here being unitary operations , but this is also the case of our example ( [ povms ] ) !
this is immediately seen using another equivalent representation of our @xmath86 @xcite : @xmath89 where @xmath90 , and @xmath91 is partial transposition with respect to system @xmath2 .
for povms from our example ( [ povms ] ) and shared singlets ( [ max - ent ] ) we have that the operator @xmath92 is at most rank @xmath0 operator and hence the map @xmath86 it generates is extremal . also , the associated operators @xmath56 , entering the fidelity expression ( [ f3 ] ) in a similar way as @xmath38 do , posses the form of ( [ uv ] ) . the trigonometric parameterization of optimum cp maps hints on a possible generalization of horodecki s considerations .
one could think of a situation where the first cp map @xmath86 would not be degenerated with respect to angles @xmath73 and @xmath74 [ unlike in ( [ rr ] ) ] .
such non - degenerated alice s povms would not however lead to substantially new physics since also in this case the optimum bob s operation would be realizable by a unitary operation on an enlarged space of the third qubit followed by tracing out the auxiliary degrees of freedom thus representing a damping channel @xcite .
the trigonometric parameterization ( [ uv ] ) also provides a nontrivial lower bound on the fidelity of teleportation .
let us assume that bob , for the lack of invention , decides just to repeat the first cp map which results from alice s measurement in the sense @xmath93 . in non - degenerated case
this simple protocol yields fidelity @xmath94 and angles @xmath73 and @xmath74 define both @xmath56 and @xmath38 .
this results shows another interesting , but not surprising , feature of extremal cp maps : if used by both parties they cover the whole range of non - classical fidelities . in conclusion
, we have derived an iterative algorithm for finding optimum cp maps for quantum teleportation .
it provides a convenient way of optimizing teleportation protocols on hilbert spaces of arbitrary dimensions .
we have applied our procedure to protocols with imperfect bell analysis and have identified situations where a unitary transformation on the third particle is not optimal and should be replaced by a more general completely - positive map . * acknowledgements * :
this work was supported by grant no .
ln00a015 of the czech ministry of education and by the program `` quantum measurement theory and quantum information '' of esi in vienna .
.
b. acknowledges the support by qipc program of the european union . |
the prevalence of computer driven trading has radically changed the market structure in several asset classes , most notably in equities and futures . this has attracted the interest of researchers driven by both practical and academic motivations .
early studies have followed an econometric approach in exploring the relationship between order flow and prices within electronic trading venues @xcite , whereas later studies focused on understanding and reducing the price impact caused by large orders @xcite .
more recently , several researchers have focused on the application of point processes @xcite , queueing theory @xcite and agent s utility functions @xcite to modeling electronic markets .
participants in electronic trading fall into several categories , with different characteristics and objectives .
although any formal classification is blurred by the fact that the same market participants often interact in several capacities at the same time , algorithmic traders can be roughly divided into the following categories : market makers , who provide the most liquidity in the markets and try to capture the bid - ask spread with minimum duration risk ; systematic traders and arbitrageurs , who typically try to profit from price dislocations and statistical relationships among the prices of different securities ; and agency brokers , who execute large trades on behalf of their customers .
irrespective of their objectives , all market participants in a public electronic venue contribute to price formation by adding and removing liquidity in a limit order book where quotes are published and orders are matched . here , we will analyze the dynamics of prices and trades from an agency broker s point of view , faced with the task of buying or selling a given number of shares within a set time horizon and at the best possible price .
the broker typically seeks to obtain liquidity from a variety of market venues , including public exchanges , dark pools and internal liquidity sources . in a public exchange , a broker has the choice to either remove liquidity from the far side ( that is , the ask side for a buy order or the bid side for a sell order ) or add liquidity to the near side of the order book .
either strategy has its advantages and trade - offs . when removing liquidity , one pays the full spread but is free to choose the timing of the trade .
when posting passively , one avoids paying the spread but gives up the timing option . in order to explore this trade - off
, we study the relationship between price dynamics and intensity of trade arrival . in the next sections
we will first compute empirically average price moves and waiting times conditional on the state of the order book .
we then attempt to capture the main features by introducing a stochastic model for diffusion in three dimensions .
we compute the probabilities of price movement and trade occurrence from the model , and calibrate them to recent historical market data . ] .price moves and waiting times until trade arrival .
we denote quote updates by tuples @xmath0 , with the quantities indicating respectively the time of the quote update , the new best bid , the new best ask , the number of shares at the best bid and those at the ask .
the change in just one of the prices or quantities is sufficient to trigger the publication of a quote update , i.e. a new row in the table . in the same table
we denote a trade execution by the tuple @xmath1 with quantities indicating the time , the price , the quantity and the side of the trade .
this sequence is used for computing price moves and waiting times until the arrival of the next trade , as described in the text[tab : taq - proc ] [ cols="^,^,^,^,^ " , ]
a common intuition among market practitioners is that the order sizes displayed at the top of the book reflect the general intention of the market .
when the number of shares available at the bid exceeds those at the ask , participants expect the next price movement to be upwards , and inversely , for the ask .
this is why , for example , a broker working an order on behalf of a client might be concerned about posting too much of it at the near side , thus showing his intentions to the rest of the market . in order to quantify and model this intuition ,
we calculate a few quantities related to the microstructure of the order book conditional on its bid - ask imbalance defined as @xmath2 where @xmath3 and @xmath4 are the bid and ask quantities posted at the top of the book .
positive ( negative ) imbalance indicates an order book that is heavier on the bid ( ask ) side . in figure
[ fig : cond_imbalance_emp-1 ] , we show the effect of the book imbalance on the average mid price change and on the waiting time until the next price change .
we calculate these quantities by considering the stopping time defined by the next change in either the best bid or the best ask .
the average price changes are therefore related to the conditional up - tick and down - tick probabilities , which have been previously considered in the literature @xcite . here
, we use the same probabilities to compute the average size of the price jump .
as expected , a high book imbalance is indicative of the general trading intention in the market and , on average , a good predictor of mid price movements .
the price change until the next tick is well approximated by a linear function of the imbalance and is typically well below the bid - ask spread , even for highly imbalanced order books .
in other words , although book imbalance might be used as a predictor for the next price movement , it does not by itself offer an opportunity for a straightforward statistical arbitrage .
the average waiting time until the next mid price change is also affected by the imbalance of the order book , with highly imbalanced books indicating a price move coming in a relatively short time .
this relationship is to be expected in a market where the typical broker posts part of his orders at the near side and both price variations and queue levels are mostly driven by the pressure of the order flow . in the context of microstructure studies , this is usually interpreted as the order flow having an impact on the limit book , but it can also be seen more generally as the natural supply and demand influence on the price of an asset .
another stopping time with economic significance is the one defined by the first arrival of a trade of a specified side .
this stopping time is important to a broker who is posting part of his order at the near side of the limit order book and has to wait for market orders originating from the opposite side to be matched with his resting order .
trades originating from the same side ( i.e. trades from buy market orders while the broker is posting at the bid ) may affect the share price and change the prevailing dynamics in the market , but they will not contribute to the broker s fill rate .
therefore , for an order posted at the bid ( ask ) side , the relevant stopping time is the time of first arrival of a sell ( buy ) trade . in table
[ tab : taq - proc ] we show a typical segment of the trades and quotes times series .
it consists of quote tuples @xmath0 , indicating the time of the quote update , the new best bid and ask prices and sizes , and trade tuples @xmath1 indicating the time , price , quantity and side of the trade .
note that although the side of the trade is not usually published by equities exchanges , it can be inferred by its price compared to the prevailing quotes .
we calculate average price variations and waiting times until the next buy or sell trade as follows : a ) for each quote @xmath0 we compute the prevailing book imbalance @xmath5 , as in eq .
( [ eq : imbaldef ] ) ; b ) given a quote , we identify the next buy and sell trade on the tape ; c ) we determine the prevailing mid price at the time of the trade ; d ) we compute the difference between the prevailing mid price and the mid price of the original quote , and assign it to the corresponding bucket based on the original quote imbalance and the side of the trade .
a similar procedure is used for computing the waiting time until the next buy or sell trade . in figure
[ fig : cond_imbalance_emp-2 ] we show average price moves and waiting times until the arrival of the next buy or sell trade , conditional on the book imbalance . as expected , the general trends of these quantities as a function of the book imbalance are similar to those obtained by using the next mid price move as a stopping time .
however average price movement from the observation time to the arrival of the next sell trade display a clear shift upwards compared to those derived until the arrival of the next buy trade .
this can be interpreted as the result of the information advantage of aggressive traders over traders posting at the near side .
it can also be interpreted more mechanically , as the impact of the intervening buy trades between the observation time and the next sell trade ( conversely , the impact of intervening sell trades if the stopping time is the next buy trade ) . as a result of this information ,
a trader who is posting his quote on the bid side of the limit order book will see on average an upwards price move by the time a sell trade matches his quote .
conversely , a trader posting on the ask side of a book displaying the same book imbalance will experience a price movement with a downward bias . finally , in figure [ fig : cond_prob_empir-1 ] we compute the empirical probability of the next market event as a function of the current book imbalance .
the events of interest are a favourable or unfavourable price move or the occurrence of a matching trade from the point of view of a broker pegging part of his order on the near side , i.e. the broker keeps updating his quote to stay at the prevailing best bid or ask price . in the next sections
, we will explain the empirical observations summarized in figures [ fig : cond_imbalance_emp-1 ] and [ fig : cond_prob_empir-1 ] by introducing a three - dimensional stochastic model for the joint evolution of the bid / ask queues and the near side trade arrival process . before developing the full model we will first review a simpler , two - dimensional model for the bid and ask queues only , in order to introduce some of the analytical techniques that are used in the full three - dimensional model .
our starting point is the two - dimensional diffusion model for the top of the book @xcite.we model the number of shares @xmath3 and @xmath4 posted at the top of the order book for the bid and ask with the following stochastic processes : @xmath6 where @xmath7 are brownian motions . because the number of shares available at the top of the book is always a positive quantity , when either of the two processes cross the positive axis we reset its value by drawing from two positive distributions @xmath8 and @xmath9 .
one can think of these distributions as modeling the next price levels in the book ; once the first level is depleted , they serve as a new starting point for the top queue . following this interpretation
, we will also assume that every time one of the two queues is depleted the price also moves in the direction of the depleted queue .
if the ask queue is depleted the price moves up ; if the bid is depleted the price moves down .
in other words , we assume that as one side of the book is depleted under the pressure of incoming aggressive trades and cancellations of existing limit orders , the other side follows it to keep the bid - ask spread fixed .
given the queue model of equation ( [ eq : b ] ) , it is natural to ask whether an up - tick or a down - tick is the most likely future price move .
we can calculate such probabilities by identifying the @xmath10 and the @xmath11 axes of the plane with the bid and the ask quote sizes respectively and writing the general equation for the evolution of the hitting probability : @xmath12 where @xmath13 is the correlation between the diffusion processes governing the depletion and replenishment of the bid and ask queues .
this correlation is typically negative in a normal market .
the choice of boundary conditions selects the event with corresponding probability @xmath14 .
for example , for an up - tick price movement , the boundary conditions are @xmath15 moreover , since the probability @xmath14 is for event occurrence up to the first stopping time , the pde in eq .
( [ eq : pde2dxy ] ) is time independent .
in fact , the solution is the @xmath16 limit of the fixed @xmath17 horizon problem .
as shown in @xcite among others , it is possible to solve the above pde by introducing two changes of variables .
the first transformation removes the correlation between the queue processes : @xmath18 yielding the equation @xmath19 the second transformation casts the modified problem in polar coordinates : @xmath20 where @xmath21 .
then the equation for the hitting probability becomes @xmath22 and the boundary conditions for an up - tick price movement become @xmath23 the solution in polar coordinates is straightforward , @xmath24 , which in the original set of coordinates has the form @xmath25 this is the probability of upward movement of a markovian order book in the diffusive limit .
in order to capture the joint dynamics of the bid and ask queues and trade arrival , we introduce another stochastic process to model the arrival of trades on the near side of the book ( the bid side for a broker executing a buy order , or the ask side for a sell order ) : @xmath26 the process @xmath27 does not correspond to a market observable .
however , in analogy to our model for the bid and ask queues , we assume that a new trade hits the near side whenever @xmath27 crosses the origin . within this extended three dimensional model ,
the queue processes @xmath3 and @xmath4 are now driven exclusively by the addition and cancellation of limit orders in the book until the arrival of a trade , while the timing of near side trade arrivals is governed by the @xmath27 process .
more importantly , we can model the relationship between book imbalance , order flow and price variations by introducing correlations between the stochastic processes @xmath28 , @xmath29 , and @xmath30 .
the introduction of an unobserved diffusion process @xmath30 for modeling trade arrival leads to a non - markovian model . as in the two - dimensional case
, we assume that once a queue gets depleted or a trade arrives , the process @xmath31 restarts at a value @xmath32 which characterizes the trade arrival time distribution . this model parameter is to be determined by calibration . during the dynamic optimization of an execution schedule
, one needs to consider the relative likelihood of favourable and adverse price moves and of trade arrival at the near side of the book . in terms of the model described above
, these events correspond to the three dimensional stochastic process @xmath33 crossing the positive orthant . in the following ,
we calculate the probability of these market events conditional on the state of the queue and the value of the trade process @xmath27 . as in the previous section
we identify the @xmath10 and @xmath11 coordinates with the processes governing the bid and ask queue sizes , and the @xmath34 coordinate with the level of the process @xmath27 .
the equation for the hitting probability becomes : @xmath35 where @xmath13 is the correlation between the processes governing the bid and ask queues , and @xmath36 and @xmath37 are the correlation of those processes with that governing the arrival of the trades at the near side .
the different events are identified by the boundary conditions of @xmath14 , which are set to one on the plane corresponding to the market event considered , and zero on all other boundaries .
for example , the probability of a near side trade before any price move corresponds to the boundary conditions @xmath38 as in the two dimensional case , it is possible to remove the correlations between @xmath28 , @xmath29 , and @xmath27 via the coordinate change @xmath39}{\sqrt{1-\rho^2_{xy } } \sqrt{1 - \rho_{xy}^2 - \rho_{xz}^2 - \rho_{yz}^2 + 2\rho_{xy}\rho_{xz}\rho_{yz } } } , \end{aligned } \right.\ ] ] yielding the equation @xmath40 then we cast the problem into a convenient set of curvilinear coordinates via the transformation @xmath41 yielding the modified problem @xmath42 and the boundary condition for a near side trade becomes @xmath43 the new integration domain after the introduction of spherical coordinates is shown in figure [ fig : integration - domain ] . the problem ( [ eq : phitheta ] ) can be further simplified by introducing one extra transformation , @xmath44 which changes the integration domain into the semi - infinite strip , @xmath45 @xmath46 , and the diffusion equation ( [ eq : phitheta ] ) into the form @xmath47 in this domain the near side trade boundary conditions become @xmath48 the solution to problem ( [ eq : phizeta ] ) that satisfies the first two boundary conditions ( [ eq : phizeta_trade_bc ] ) can be expressed as a generalized fourier series @xcite : @xmath49 with @xmath50 .
the expansion coefficients @xmath51 can be determined by enforcing the third boundary condition in ( [ eq : phizeta_trade_bc ] ) . to compute the coefficients , we introduce the integrals @xmath52 @xmath53
then the third boundary condition in ( [ eq : phitheta_trade_bc ] ) becomes the matrix equation @xmath54 and the coefficients @xmath51 can be computed by matrix inversion as @xmath55 . ] ) in the @xmath56 space , for @xmath57 and boundary conditions as in ( [ eq : phitheta_trade_bc ] ) .[fig : sol_profile ] ] when the boundary @xmath58 is approximately linear , the integrals @xmath59 and @xmath60 can be computed analytically .
figure [ fig : sol_profile ] shows the solution profile for the case where the near side trade arrival process is uncorrelated from the bid and ask queue sizes , i.e. @xmath61 . in summary we have a semi - analytic method of computing the probability @xmath62 or equivalently @xmath63 for given initial values @xmath64 and corresponding correlations .
we proceed by calibrating the model jointly to the average mid - price moves shown in figure [ fig : cond_imbalance_emp-2 ] and to the empirical probabilities of price moves and trade arrivals summarized in figure [ fig : cond_prob_empir-1 ] .
it was shown in the previous section that the model probabilities are determined by the book imbalance ( relative size of the bid and ask queues ) , the initial position of the trade arrival process @xmath32 and the three elements of the correlation matrix @xmath65 .
the book imbalance is a market observable , and the remaining four parameters are to be estimated by the calibration process .
model probabilities are computed semi - analytically as in the previous section .
expectations of price changes up to the stopping time defined by the arrival of a trade can not be easily computed with analytic methods and monte carlo simulations were used instead .
we calibrate the model against empirical probabilities of price movements for equity vod.l for all trading days in the first quarter of 2012 .
the results are shown in figures [ fig : cond_imbalance_emp-2-cal ] and [ fig : passive - fill - probabilities ] .
calibrations against data from other liquid stocks displayed qualitatively similar results .
we note that the model reproduces the gap between the average mid - price moves until the arrival of a buy or sell trade .
the gap is controlled by the correlation matrix @xmath66 and in particular by the two correlations between the trade arrival process and the bid and ask queue processes .
this shift in average price moves conditioned on the side of the trade can be interpreted as the expected slippage of a passive fill .
when executing an order , a broker will typically post a fraction of the total quantity at the near side in an attempt to save part of the spread . however , when measuring the differences in the average price achieved by aggressive and passive fills , we observe that passive fills rarely achieve their expected savings of one full spread . while part of this effect is generally attributed to adverse selection ( that is , the ability of other market participants to take advantage of the timing given up by the broker when posting at the near side ) , the slippage of orders posted at the near side is also due to the interaction between queue depletion and trade flow , as our model predicts . in the calibration described here ,
this effect alone is responsible for the loss of about 60% of the theoretical spread captured by a passive fill .
figure [ fig : cond_imbalance_emp-2-cal ] also compares the empirical and model derived average arrival times of a buy or sell trade as a function of book imbalance .
the model captures the gross features of the empirical shapes for moderate book imbalance values , but is not as accurate in reproducing the steep decrease of the arrival times at extreme book imbalance values .
finally , we note that the model is able to reproduce the empirical shapes of the event probabilities .
as shown in figure [ fig : passive - fill - probabilities ] , the probability of an unfavorable price movement increases as the book gets heavier towards the near side , reaching almost 90% in cases of high imbalance . knowing this , a broker can decide to keep the order posted on the near side for moderate imbalance values and cross the spread in a highly imbalanced book . deriving such optimal spread crossing policies is left for future work . ,
@xmath67 and @xmath68 @xmath69 [ fig : cond_imbalance_emp-2-cal ] ] ]
we analyzed the microstructure of trade arrival and its relationship to the state of a limit order book .
we observed empirically that the arrival time of trades at the near side and the dynamics of the mid - price until the arrival of a trade of a given side depend strongly on the order book imbalance .
we introduced a stochastic model with correlation between the processes for the order queues at the top of the book and a process representing the arrival of the trades at the near side of the book .
we were able to compute probabilities of price movement and trade arrival in a semi - analytical form .
this allowed us to perform an efficient calibration of the model to empirical probabilities .
since the model captures the dependence of the trade arrival probabilities on the order imbalance , it can be used to construct optimal order execution strategies .
we would like to thank ioana savescu for many insightful discussions and comments .
bouchaud , j. d. farmer , and f. lillo .
how markets slowly digest changes in supply and demand . in t.
hens and k schenk - hoppe , editors , _ handbook of financial markets : dynamics and evolution_. elsevier : academic press , 2008 . | we examine the dynamics of the bid and ask queues of a limit order book and their relationship with the intensity of trade arrivals . in particular , we study the probability of price movements and trade arrivals as a function of the quote imbalance at the top of the limit order book . we propose a stochastic model in an attempt to capture the joint dynamics of the top of the book queues and the trading process , and describe a semi - analytic approach to calculate the relative probability of market events .
we calibrate the model using historical market data and discuss the quality of fit and practical applications of the results . |
white dwarfs are evolved , degenerate stars which have exhausted their nuclear fuel and are supported by electron degeneracy pressure .
they are destined to cool to ever lower temperatures over billions of years as the end product of most stars in the galaxy , main sequence stars with @xmath5 .
next to low mass stars , they are the most common stellar objects in the milky way , providing a window into the history of stellar evolution and star formation in the galaxy .
ultracool white dwarfs ( @xmath6 k ) have been of much interest to researchers in recent years for several reasons .
first , it remains an unanswered question as to whether cool white dwarfs are a significant component of galactic halo dark matter @xcite .
second , the ages and compositions of cool white dwarfs have obvious implications for the age and evolution of the disk , bulge and halo components of the galaxy .
finally , the distribution and origins of atmospheric and core compositions of the coolest white dwarfs is still not well understood . in this paper
the discovery of gd392b , a common proper motion companion to the helium white dwarf gd392 ( wd2058 + 342 , @xmath7 j2000 ) , is discussed .
the spectral energy distribution ( sed ) of gd392b strongly suggests flux deficiency redward of @xmath8 m , a @xmath9 k , and possibly a helium component in its atmosphere as well @xcite .
tentative stellar parameters are derived for gd392b based on the data and different assumptions about the mass of the primary .
possible implications based on the model fit parameters and similarities to known ultracool white dwarfs are then discussed .
near infrared data on the gd392 system were obtained at lick observatory in june & august 2003 .
@xmath10 images were acquired using the gemini camera @xcite on the 3 meter shane telescope .
total observation time was 7.5 minutes at each filter , consisting of 15 second exposures in a repeated five point dither pattern .
multiple infrared photometric standard stars from @xcite & @xcite were observed twice during the night .
the conditions in august were very good with clear skies and average seeing of @xmath11 .
in june the weather was good but the average seeing was @xmath12 .
a few @xmath13 band images were taken at mauna kea in july 2003 using the near infrared camera ( nirc ) @xcite on the 10 meter keck i telescope .
images consisted of 10 second exposures for the primary and 50 seconds for the secondary .
observing conditions were poor with considerable cloud cover and seeing of @xmath14 .
optical data were obtained at lick observatory in july & september 2003 .
@xmath1 images were acquired with the ccd camera on the 1 meter nickel telescope .
total exposure time was 5 minutes at each filter and photometric standard stars from @xcite were observed immediately prior to the gd392 system .
conditions in august were excellent with clear skies and seeing of @xmath15 . in july ,
the sky was clear but the average seeing was @xmath16 images were reduced using standard programs in the iraf environment .
optical images were filtered clean of bad pixels and cosmic rays in the area of interest , then flat fielded with a normalized dome flat .
near infrared images were sky subtracted , flat fielded , registered and averaged into a single image at each wavelength .
straightforward aperture photometry was used , including airmass / extinction corrections , for @xmath17 measurements of the primary . using a circular aperture centered on the target star and an annulus on the surrounding sky
, both the flux and signal to noise ratio ( snr ) was calculated for a range of apertures from one to four full widths .
the target flux was measured at or near the aperture size which produced the largest snr . in this way
the flux of all targets was measured , including standard stars . a large aperture ( @xmath18 )
was used for all calibrators , and flux measurements for science targets were corrected to this standard aperture to account for the smaller percentage of the total flux contained within apertures smaller than the standard . extracting all photometry for gd392b
was complicated by the presence of a nearby background star ( figure [ fig1 ] ) . in a single subtraction @xmath19 band image ,
the separation between gd392b and background star was measured to be @xmath20 at a pa of @xmath21 . since even small aperture flux measurements of gd392b would be contaminated by light from the background star , the iraf package daophot was employed to fit the psf of both stars simultaneously and extract their magnitudes . since daophot is sensitive to the separation and brightness ratio of two sources with overlapping psfs , the optical and infrared data were broken into two sets ; primary data for which the full widths at half maximum in the reduced images were less than @xmath12 , and secondary data for which they were greater than or equal to @xmath12 .
this is an especially important consideration in the infrared where the brightness ratio of gd392b and background star is far from unity .
all photometric data are listed in table [ tbl-1 ] and the adopted values used for the analysis are listed in table [ tbl-2 ] .
the secondary data were not included in the analysis due to the uncertainty in psf fitting for sources separated by less than one full width at half maximum the data are consistent with the adopted values to within @xmath22 and are included for completeness .
conversions between near infrared filter sets were not performed for three reasons : 1 ) the corrections are typically very small @xcite ; 2 ) the colors of gd392b are drastically different from the colors of standard stars used to derive the transformations ; 3 ) the near infrared photometry errors in table [ tbl-1 ] are likely to be several times larger than any corrections . due to the poor weather conditions , there was only a single nirc frame acquired with gd392b on the chip , three frames containing gd392a , and several sky frames . since the common proper motion system is separated by more than the @xmath23 field of view , the binary pair could not be observed simultaneously .
the reduced images of the primary and secondary contained three field stars in common .
aperture photometry was performed on these stars twice ; once with gd392a as a calibrator , and again using the three field stars as calibrators for gd392b and the background star .
these intermediate calibrator stars had a standard deviation of only a few percent in their measured relative fluxes between the two reduced images . gd392b and
the background star were spatially resolved from each other in this observation .
the snr in the primary data set was greater than 30 for both stars in the reduced images at all wavelengths with the exception of the following . in the gemini images ,
the snr was calculated to be 8.6 at @xmath24 , 3.7 at @xmath25 for gd392b ( from the extracted daophot data ) . in the nickel data ,
the snr at @xmath26 was 11.4 for gd392b and 15.1 for the background star .
in the single subtraction , flat fielded nirc @xmath13 frame , the snr for gd392b was 5.8 .
a @xmath27 m spectrum was obtained using the low resolution imaging spectrograph ( lris ) @xcite in october 2003 at mauna kea on keck i by c. steidel , d. erb & n. reddy . the setup consisted of the 300/5000 grism on the blue side and the 400/8500 grating on the red side , separated at 6650 by a dichroic beamsplitter .
the spectra of both stars were spatially resolved from each other in the observation .
the spectral images were cleaned of bad pixels and cosmic rays , then flat fielded using an internal halogen lamp .
a spectrum of the sky was extracted at two positions for each wavelength region , averaged , and then subtracted from the spectrum of each star .
standard programs in iraf were used to extract the spectra of both gd392b and the background star .
the lamp used to flat field the spectral images is itself spectrally nonuniform and has a strong rise from @xmath28 .
hence , the instrument and chip response were removed by flat fielding but the remaining shape was a convolution of the stellar spectra and the `` flat '' lamp .
the background star appears to be a k dwarf whose exact spectral type cannnot be established from photometry because its reddening is unknown and the low resolution spectrum precludes the line measurements necessary for a determination .
spectra extracted without flat fielding confirm that the flux of both gd392b and k star rises toward the dichroic cutoff near 6600 ( a fact corroborated by the photometry ) . attempting to reconstruct the true shape of the continuua , the reduced spectra were multiplied by a blackbody with a temperature similar to that of the halogen flat lamp .
this information was obtained from the manufacturer .
the resultant spectra are shown in figures [ fig2 ] and [ fig3 ]
a wide field , infrared proper motion survey for low mass stellar and substellar companions to nearby white dwarfs is being completed @xcite .
because the observations are being conducted primarily at @xmath19 band ( @xmath29 m ) , the survey is particularly sensitive to cool objects such as late m , l and early t type dwarfs .
a near infrared search is also sensitive to cool white dwarfs .
this survey s sensitivity is slightly less for cool degenerates which suffer collision induced absorption ( cia ) opacity at this wavelength relative to those which do not . in general , common proper motion companions brighter than @xmath30 mag can be detected . to measure proper motions and detect companions , geomap is used .
this is a standard program in the iraf environment , which generates a transformation between two sets of coordinates corresponding to sources in the same field at two different epochs . in this way
proper motion stars can be identified and their motions measured against the near zero motion of background stars and galaxies , which provide a measure of the standard error . the common proper motion companion to gd392 ( wd2058 + 342 , greenstein 1984 ; greenstein & liebert 1990 , wesemael et al .
1993 ) was detected in a routine examination of the digitized poss i & ii plates for confirming identity , proper motion and coordinates of the primary for a finder chart prior to an observing run .
a simple blinking of poss i and poss ii frames reveals the comoving companion at a separation of @xmath31 and a pa of @xmath32 ( figure [ fig1 ] ) .
measurement of proper motion between poss epochs reveals that both gd392a and b have the same proper motion over a 41 year baseline , namely @xmath33 @xmath34 at @xmath35 .
furthermore , astrometric analysis of 2003 infrared images confirms that the separation and pa between primary and secondary have remained constant since 1951 .
the usno b1.0 catalog @xcite gives a proper motion of @xmath36 @xmath34 at @xmath37 for gd392 and no value is given for the companion .
@xcite published a detected proper motion of @xmath38 @xmath34 at @xmath39 for the primary .
both of these measurements are consistent with the value presented in this paper .
figure [ fig4 ] shows the unchanging position angle of the pair over several epochs . at a nominal distance of 50 pc ( see 4.2 ) , the proper motion of the pair gives @xmath40 km @xmath41 and thus there is no reason to suspect this system is not a member of the local disk population .
despite some relatively low snr photometry on gd392b in the near infrared , there are seven independent measurements at these longer wavelengths , three of these at @xmath42 m ( table [ tbl-1 ] ) . in stark contrast with the similar optical magnitudes and colors ,
gd392b appears roughly 1.0 magnitude fainter than the nearby background k star at @xmath43 m , 2.0 magnitudes fainter at @xmath44 m , and 2.5 magnitudes fainter at @xmath42 m in all the measurements .
the optical and infrared colors of gd392b point strongly towards an ultracool white dwarf with strong cia longward of @xmath8 m .
no other stellar object could be so red in @xmath45 and yet blue in @xmath46 @xcite .
the optical spectrum confirms that gd392b is a featureless dc star ( figure [ fig2 ] ) .
hence gd392b must be a very cool white dwarf .
plotting gd392b on a @xmath45 vs. @xmath47 color - color diagram , one sees that it is located beyond the turnoff for log @xmath48 pure hydrogen atmosphere white dwarfs , which corresponds to @xmath2 k ( figure [ fig5 ] ) .
there are only three other ultracool white dwarfs in the literature for which infrared photometry is available , wd0346 + 246 , lhs3250 and sdss1337 + 00 @xcite .
gd392b s optical colors are strikingly similar to those of wd0346 + 246 @xcite .
although they have almost identical optical colors , gd392b appears to have more flux in the near infrared than does wd0346 + 246 . from recent model atmosphere analyses of the three aforementioned stars ,
it is possible that gd392b is warmer , has different gravity , and/or contains a different ratio of hydrogen to helium in its atmosphere than does wd0346 + 246 @xcite .
the best way to estimate the temperature of gd392b is to fit models to the @xmath17 photometry . to evaluate its sed
, the magnitudes were converted into isophotal or average fluxes , following the method of @xcite .
different fits to the data were tried using the pure hydrogen and mixed h / he atmosphere model grids of p. bergeron ( 2003 , private communication ) .
as in the case of wd0346 + 246 @xcite , pure hydrogen and mixed atmosphere models fail to reproduce the sed of gd392b in detail .
and like @xcite , a low gravity solution appears to provide a decent fit to the data .
in fact , every solution at log @xmath49 seems to provide optical colors that are too blue ( too much cia ) , or infrared colors that are too red ( too little cia ) .
as shown in figure [ fig6 ] , a decent fit is acheived at log @xmath50 , @xmath51 k , and @xmath52 .
also shown in figure [ fig6 ] is a blackbody of the same temperature scaled to the peak flux of gd392b relative to the blackbody , the sed of gd392b appears suppressed at wavelengths longer than 8000 and enhanced at shorter wavelengths .
because gd392b is a companion to a previously studied white dwarf , it should be possible to deduce the distance to this system and use it to constrain stellar radius @xmath53 and @xmath54 .
one must keep in mind that the model predicted temperature of gd392b will depend on its mass and atmospheric composition and thus can not be known exactly at present . in
4.2 a range of possibilities is explored . using basic relations between luminosity and flux
, one can calculate a radius for gd392b if its distance and @xmath55 are known . to obtain a distance , the @xmath17 photometry of the primary ( table [ tbl-1 ] ) , gd392 ,
was fitted with pure helium model grids ( p. bergeron 2002 , private communication ) and its measured sed was integrated directly .
the model fit provides a @xmath55 , and integrating the sed of gd392 yields a total flux .
a radius must be specified in order to calculate a distance to the primary .
the model fit to the sed of gd392 does not provide a radius determination because the colors of a db star are largely unaffected by gravity .
hence one is forced to make various assumptions for log @xmath56 .
table [ tbl-3 ] lists stellar parameters for gd392 with various gravities , and thus distances . in the literature ,
gd392 has been noted as a db5 @xcite , dba?4 @xcite and finally a db5 again while explicitly stating `` revised from @xcite '' @xcite .
the optical and infrared photometric data presented here are more consistent with a db4 than a db5 .
good agreement between model and measured colors is acheived at @xmath57 k ( regardless of log @xmath56 ) , which actually corresponds to db4.5 .
this is the temperature used in all calculations for the parameters of gd392 .
for each distance estimate to the gd392 system , there is an analytic constraint on @xmath58 for gd392b .
it is also critical that the model predicted parameters , such as @xmath54 , optical and infrared colors corresponding to each combination of @xmath53 and @xmath55 , fit the photometric data on gd392b .
if gd392 is an average mass white dwarf ( log @xmath48 ) , a solution is found almost identical to the preliminary fit in figure [ fig6 ] .
that is , at 57.8 pc gd392b must have a temperature near 3500 k and a very large radius either an unresolved binary or a low mass white dwarf .
this model fit is consistent with the measured colors ( excepting @xmath59 , see 4.3 ) and matches the absolute magnitude of the secondary at this distance ; in fact , only a low gravity solution appears to provide such agreement between measured and predicted fluxes . in figure
[ fig7 ] , four fits are shown satisfying the @xmath54 constraint at 57.8 pc for a range of gravities .
it is clear that only the log @xmath50 fit comes near to matching the photometry .
although these fits have all been scaled to match the flux at @xmath60 m , the obvious discrepancies of the higher gravity solutions occur at any scaling ; they predict too little flux in the optical or too much flux in the near infrared .
in addition , the log @xmath61 fits do not satisfy the @xmath58 constraint at this distance .
this scenario is instructive for two reasons .
one is that there is no independent reason to suspect that gd392 is anything but an average mass white dwarf . in fact , the spectroscopic mass distribution of db stars peaks at @xmath62 , with a very small dispersion of @xmath63 @xcite .
the other reason is the failure of the higher gravity fits to match the data persists for any reasonable distance estimate . assuming the primary is a higher than average mass white dwarf ( log @xmath64 ) results in serious discrepancies between model predicted and measured colors for gd392b .
this disagreement appears to be insensitive to the type of atmosphere in the models . shown in figure [ fig8 ]
is a model fit for gd392b at 40.4 pc .
higher gravity solutions at this distance or closer are worse and are not shown for the sake of brevity .
problems arise as well if one assumes the primary is a lower than average mass white dwarf ( log @xmath65 ) .
this leads to extreme low gravity solutions for gd392b and the models used here do not include gravities below log @xmath50 .
table [ tbl-4 ] lists all resulting fits for gd392b .
while the possibility exists that gd392b is an unresolved binary or helium core white dwarf , this is uncertain until the distance to the system is known with sufficient precision and a good model fit to its sed is achieved .
it seems very unlikely that gd392b could have a such an unusually low mass as @xmath66 , but it can not be ruled out either .
it is conceivable that a good model fit with a much higher @xmath55 ( hence predicting a smaller radius and a larger mass ) could show good agreement with the measured colors of gd392b .
all models used here do not give such agreement .
a parallax measurement of the primary is currently in progress ( h. harris 2003 , private communication ) .
all attempts to fit the sed of gd392b with mixed h / he atmosphere models having @xmath67 are problematic .
in fact , for all temperatures and gravities , the same problem mentioned previously occurs ; colors too blue in the optical and/or too red in the infrared to fit the measured photometry , given the absolute magnitude constraint .
the only possible exception being the pure hydrogen models at log @xmath50 . shown in figure [ fig9 ] are several model fits with varying ratios of h / he for the lowest gravity case
. only the pure hydrogen and the @xmath52 models provide any kind of decent fit to the data .
however , the pure hydrogen fit has one minor fault when compared to the mixed atmosphere fit it predicts an @xmath54 corresponding to 57.8 pc but does not precisely fit the @xmath58 constraint .
the @xmath52 , @xmath68 k solution comes very close to hitting all the data points except @xmath26 and satisfies all constraints surprisingly well at 57.8 pc .
there currently exists a discrepancy between model predicted and measured @xmath26 band fluxes this wavelength is excluded from fits to the energy distributions of very cool hydrogen atmosphere white dwarfs @xcite .
it is noteworthy that gd392b appears to be the first ultracool white dwarf for which a pure hydrogen atmosphere is not inconsistent with the data .
at present there remains a general failure by model predictions to fit the photometric observations of ultracool white dwarfs in detail @xcite a fact exemplified by gd392b .
another case in point is the sed of wd0346 + 246 , where both @xcite and @xcite found that a mixed atmosphere model with an extremely low value of @xmath69 provides a good fit to the data .
but because of the unlikelihood and volatility of such a solution , @xcite introduced a pseudocontinuum opacity characterized by an ad hoc damping function .
although a good fit with @xmath70 was achieved when the opacity was added to the models , it is unclear whether this pseudocontinuum source plays a significant physical role in the atmospheres of ultracool white dwarfs .
therefore , it is not yet possible to determine the atmospheric parameters of the coolest degenerates with confidence .
it should be mentioned that with the possible exception of the ultracool white dwarfs in figure [ fig5 ] , all suspected or confirmed low mass white dwarfs have hydrogen rich atmospheres @xcite .
hence a model fit to the photometry of gd392b with low mass and a mixed h / he atmosphere is inconsistent with these findings .
gd392b is not the first or even the second ultracool white dwarf for which an apparent large radius is possible or likely .
lhs3250 , wd0346 + 246 and sdss1337 + 00 _ all appear to be potentially overluminous _ @xcite .
in fact , the best fits acheived by @xcite to wd0346 + 246 and f351 - 50 were with log @xmath71 , although for pure hydrogen atmospheres .
these solutions are problematic because they predict temperatures below 3000 k and such extreme low gravities . although unlikely , the present possibility that these ultracool degenerate stars have anomalously low mass can not be ruled out . white dwarfs with @xmath72
have relatively large radii and are understood to be the product of close binary evolution ; their post main sequence lifetimes cut short before helium burning begins @xcite . on the other hand , the potential overluminosity of these stars can be explained equally well by binary membership with a white dwarf companion of comparable luminosity .
if this is correct , the gd392 system would then be the second known triple degenerate . in either case , binarity is a strong possibility for most if not all four objects in figure [ fig5 ] , if the models are correct .
is it possible that the discovery of these stars represents a luminosity detection bias ?
this would imply all known ultracool degenerates were detected because they are relatively bright compared to normal to high gravity single white dwarfs with comparable temperatures . in fact , the faintest white dwarf known , eso439 - 26 @xcite , has an extremely high mass ( log @xmath73 ) , a temperature around 4500 k , and @xmath74 just above the detection limit of very large sky surveys like the serc / eso survey in the southern hemisphere and the palomar survey ( poss ii ) in the north @xcite .
what if eso439 - 26 had @xmath51 k instead of @xmath75 k ?
would it have still been detected ?
almost certainly not , since models predict it would have been too faint at @xmath76 and @xmath77 .
a similar calculation was performed for gd392b , assuming it is a 3500 k mixed h / he atmosphere degenerate with log @xmath50 at 57.8 pc .
would it still have been detected were it a higher mass white dwarf with log @xmath78 ?
likely not at @xmath79 and @xmath80 which is just beyond the poss ii red plate limit and right at the blue plate limit .
it probably would not have been detected in the infrared at @xmath81 .
similar conclusions can be drawn for wd0346 + 246 at @xmath82 , but possibly not lhs 3250 at @xmath83 .
if we similarly push back the magnitude of sdss1337 + 00 to @xmath84 it would still be detectable in the sloan survey but not on the poss ii plates .
the claim that all known ultracool white dwarfs are overluminous is not being made here .
it is simply being stated that _ if _ the majority of them are indeed brighter than expected due to low gravity or binarity , then all surveys with the exception of sloan have been biased against detecting average to higher mass single white dwarfs at the very bottom of the disk cooling sequence . in fact
, one should expect that the oldest white dwarfs those which spent very little time on the main sequence would have come from fairly massive progentiors .
furthermore , the initial mass to final mass relation for white dwarfs indicates that higher mass main sequence stars tend to form higher mass degenerates @xcite .
this implies that the oldest and least luminous white dwarfs should have high masses , small radii , and be up to 2.4 magnitudes fainter than a low mass white dwarf with the same temperature .
hence there is a possibility that the lowest luminosity disk white dwarfs have yet to be detected .
of course the other possibility ( and , according to occam s razor , the most likely ) is simply a failure of the current models to explain the photometric observations .
also , in this case , the data do not constrain the mass of gd392 and therefore we do not know distance to the system . despite the fact that the model predictions fail to reproduce the colors of gd392b in all but the log @xmath50 case , it is certainly possible that it has a higher mass .
this can not be ruled out and future improvements in models may result in the disappearance of the overluminosity issue .
spectroscopy and photometry confirm a cool dc white dwarf common proper motion companion to gd392 .
it is the first ultracool white dwarf ( @xmath6 k ) to be discovered as a companion to another star .
although the primary is a well studied db white dwarf , gd392 does not have a mass estimate , nor any kind of reliable parallax .
in fact , its helium lines are very weak @xcite and therefore it is not possible to spectroscopically determine its mass .
neither is it possible to photometrically determine its mass because the optical and infrared colors of relatively hot helium atmosphere white dwarfs are insensitive to gravity .
therefore the distance to the gd392 system is not well constrained and the parameters we derive for gd392b must come from a range of assumed values for the mass of the primary . despite this difficulty , it appears that all model fits fail to reproduce the sed of gd392b with the exception of a solution with log @xmath50 , @xmath68 k , and @xmath85 ( or similar values with pure hydrogen ) .
it is the only set of parameters for which the model predicted fluxes fall within @xmath86 of all but one photometric data point and which satisfies all constraints .
however , the model fit alone is not sufficient to conclude that gd392b is overluminous .
it is possible that current models simply can not reproduce the observed spectral energy distribution of gd392b with a normal to higher surface gravity , but perhaps ongoing and future improvements will .
based on the data and the models used here , along with the published analyses of similar objects , gd392b may be an unresolved binary with a companion of comparable luminosity or is itself a low mass white dwarf .
albeit unlikely , this could also be the case for other known ultracool degenerates .
if correct , this may hold interesting consequences for the population of stars at the faint end of the local disk white dwarf luminosity function .
some of the data presented herein were obtained at the keck observatory , which is operated as a scientific partnership among the california institute of technology , the university of california and the national aeronautics and space administration ( nasa ) .
the author would like to extend his sincere gratitude to c. steidel , d. erb and n. reddy for obtaining the optical spectrum of gd392b and to e. e. becklin & i. song for making the nirc observations .
acknowledgement goes to e. e. becklin & b. zuckerman for careful readings of the manuscript and to b. hansen for many helpful conversations .
bergeron kindly provided his models for use in this work .
thanks to the referee n. hambly for very helpful suggestions and comments .
a special thanks goes to c. mccarthy for donating his time and computing skills much to the author s benefit . the digitized version of the poss i & ii plates were provided as a service by the space telescope science insititute .
this research has been supported in part by grants from nasa to ucla .
@xmath26 & 0.44 & 15.71 & @xmath87 & @xmath88 + & & 15.79 & @xmath89 & @xmath90 + @xmath91 & 0.55 & 15.67 & @xmath92 & @xmath93 + & & 15.69 & @xmath94 & @xmath95 + @xmath53 & 0.64 & 15.61 & @xmath96 & @xmath97 + & & 15.63 & @xmath98 & @xmath99 + @xmath100 & 0.80 & 15.62 & @xmath101 & @xmath102 + & & 15.69 & @xmath103 & @xmath104 + @xmath19 & 1.25 & 15.73 & @xmath105 & @xmath106 + & & 15.77 & @xmath107 & @xmath108 + @xmath24 & 1.65 & 15.79 & @xmath109 & @xmath110 + & & 15.81 & @xmath111 & @xmath112 + @xmath25 & 2.12 & 15.86 & @xmath113 & @xmath114 + & & 15.88 & @xmath115 & @xmath116 + @xmath13 & 2.21 & & @xmath117 & @xmath118 + @xmath26 & 0.44 & @xmath87 & @xmath119 + @xmath91 & 0.55 & @xmath92 & @xmath120 + @xmath53 & 0.64 & @xmath96 & @xmath121 + @xmath100 & 0.80 & @xmath101 & @xmath122 + @xmath19 & 1.22 & @xmath105 & @xmath123 + @xmath24 & 1.63 & @xmath109 & @xmath124 + @xmath13 & 2.19 & @xmath125 & @xmath126 + @xmath127 ( pc ) & 57.8 & 40.4 & 25.9 + @xmath128 & 0.59 & 0.90 & 1.19 + @xmath129 & 0.0127 & 0.0089 & 0.0057 + @xmath54 & 11.87 & 12.65 & 13.61 + @xmath130 & 11.20 & 11.98 & 12.93 + log @xmath131 & -2.58 & -2.89 & -3.27 + cooling age ( gyr ) & 0.45 & 0.91 & 1.78 + @xmath55 ( k ) & 3509 & 3639 & 3875 + log @xmath56 & 7.00 & 7.95 & 8.72 + @xmath128 & 0.153 & 0.576 & 1.066 + @xmath129 & 0.0205 & 0.0133 & 0.0075 + @xmath54 & 15.69 & 16.47 & 17.43 + @xmath130 & 15.35 & 16.14 & 17.27 + log @xmath131 & -4.24 & -4.56 & -4.94 + cooling age ( gyr ) & 2.92 & 9.92 & 8.62 + | the discovery of a low luminosity common proper motion companion to the white dwarf gd392 at a wide separation of @xmath0 is reported .
@xmath1 photometry suggests a low temperature ( @xmath2 k ) while @xmath3 data strongly indicate suppressed flux at all near infrared wavelengths .
thus , gd392b is one of the few white dwarfs to show significant collision induced absorption due to the presence of photospheric @xmath4 and the first ultracool white dwarf detected as a companion to another star .
models fail to explain gd392b as a normal mass white dwarf .
if correct , the cool companion may be explained as a low mass white dwarf or unresolved double degenerate .
the similarities of gd392b to known ultracool degenerates are discussed , including some possible implications for the faint end of the white dwarf luminosity function . |
heavy - to - light decays are very important processes in flavor physics .
the successful running of @xmath0 factories babar and belle has enabled the semileptonic @xmath0 decays like @xmath4 to work out the information about the least known cabibbo - kobayashi - maskawa ( ckm ) matrix element @xmath5 , for which @xmath6 form factors play the crucial roles .
the same form factors enter the nonleptonic modes such as @xmath7 , @xmath8 , which are interesting because of the factorization as well as the angles of ckm unitarity triangle .
heavy - to - light decay is also responsible for @xmath9 transitions @xmath10 or @xmath11 , which are good probes for new physics beyond the standard model(sm ) .
however , the theoretical understanding of heavy - to - light transition is quite poor and still controversial .
the transition matrix elements are usually parametrized by the form factors .
they are nonperturbative objects which can not be calculated from first principles , and all the essential features of the decay processes are encoded on them . at large recoil limit where the final state meson energy is sufficiently large ,
these form factors are not independent .
conventionally , there are 3(7 ) decay form factors for @xmath0 to @xmath12 decays , where @xmath12 stands for the pseudoscalar ( vector ) mesons .
the number of independent form factors reduces to 1(2 ) for @xmath13 at large recoil limit @xcite .
this is due to the so called spin - symmetry relations .
it is a common lore that there are two kinematically distinctive contributions to the form factors .
one is the soft overlap ( or feynman mechanism ) where one of the partons in the daughter meson carries almost all the momentum .
the other is the
hard scattering where none of the partons in the daughter meson is in the end - point region of momentum configuration by exchanging hard gluons .
but there is no general agreement on which of the two is the dominant one .
one of the main issue related is the end - point singularity .
contributions of hard gluon exchange with the spectator quarks are described by the convolution integrals involving meson distribution amplitudes ( da ) and some kernel . at the heavy quark and large recoil limit
, the kernel behaves like @xmath14 where @xmath15 is some momentum fraction of the partons while the das of the final state meson do as @xmath16 in their asymptotic forms .
the resulting convolution integral diverges at the end point @xmath17 , and this is called the end - point singularity . the end - point singularity has been dealt in many different ways in different theoretical frameworks . in perturbative qcd ( pqcd )
@xcite , the end - point singularity is absent due to the sudakov suppression near the end - point region .
the point is that when one of the partons carries all the momentum , the remaining part of the daughter meson can not shield the color charge of the fast parton , so many gluons should be emitted @xcite .
thus the soft overlap contribution is very suppressed in this picture , and the hard scattering is the dominant contribution .
but the sudakov suppression is not generally accepted in the literature . in ref .
@xcite , it is argued that the sudakov suppression is not severe at the heavy quark scale @xmath18 gev ( for more recent discussions , see @xcite ) . on the other hand , in ref .
@xcite , the problem of end - point singularity is evaded by absorbing the singular terms into the so called soft form factor . here
the heavy - light @xmath19 ( @xmath20 is a light meson ) form factors @xmath21 are compactly written as @xmath22 where @xmath23 are the soft form factors for @xmath20 of energy @xmath24 , and @xmath25 are the hard vertex renormalization ; @xmath26 are hard kernels which are convoluted ( @xmath27 ) with the das @xmath28 and @xmath29 , and @xmath30 is the momentum transfer .
it was shown that the soft form factors @xmath31 satisfy the spin - symmetry relations .
the second term of convolution arises from the hard spectator interactions .
terms involving end - point singularities are already absorbed into @xmath31 , so the remaining convolutions are end - point finite .
they are shown to break the spin - symmetry relations .
the soft form factors @xmath31 are totally nonperturbative objects ; they can not be calculated from first principles , and they contain the end - point singularities .
the meson das @xmath32 are also nonperturbative , but they can be treated systematically and are well understood .
other quantities @xmath25 and @xmath26 are perturbatively calculable .
in fact , one of the great merit of eq .
( [ soft ] ) is the perturbative calculability of the hard kernel , @xmath26 .
this is the main philosophy of the qcd factorization @xcite . from the numerical analysis , the authors of ref .
@xcite found that the symmetry breaking convolutions contribute about 10% ; they conclude that heavy - light form factors are largely from the soft form factor .
the advent of the soft - collinear effective theory ( scet ) @xcite shed new lights on the heavy - to - light decays . in this framework ,
the @xmath19 form factors are described as @xcite @xmath33 where @xmath34 and @xmath35 are the hard functions and @xmath36 is the jet function , and @xmath37 with @xmath38 being the meson decay constants . here
@xmath39 is the scet version of the soft form factor . just as in @xcite ,
the end - point singular terms are absorbed into @xmath40 , and as a whole satisfy the spin - symmetry relations .
from the fact that the same form factors enter the nonleptonic two - body @xmath0 decays , ref .
@xcite showed that the two contributions of eq .
( [ scet ] ) are comparable in size .
this point differs from the qcd factorization ( qcdf ) analysis @xcite , where the hard scattering contribution is very small .
although the brief summary above shows impressive achievements in heavy - to - light transitions , there are still some ambiguities and confusions .
first of all , the quantity @xmath40 in eq .
( [ scet ] ) is not
soft in the sense that the involved quarks are not in the asymmetric momentum configuration ; it is defined by the _ collinear
_ quarks . in this context , the scet description of eq .
( [ scet ] ) is much closer to the pqcd prediction where the contributions from the asymmetric momentum configuration ( i.e. , soft overlap ) are highly suppressed .
it is very helpful to see how the soft overlap is identified in the light - cone sum rule ( lcsr ) .
the lcsr has improved their predictions for the @xmath19 form factors @xcite . at tree level after the borel transformation , the weak form factor is proportional to @xcite @xmath41 where @xmath42 is the process - dependent amplitude . here
@xmath43 where @xmath44 is the continuum threshold is a lower limit of the convolution integral .
it scales as @xmath45 ; thus only the highly asymmetric momentum configuration is relevant .
this is nothing but the exact meaning of the soft overlap .
hard spectator interactions as well as the vertex corrections appear at @xmath46 . numerically eq .
( [ lcsr ] ) is dominant compared to the @xmath46 contributions .
recently , a new insight into the heavy - light transition was proposed @xcite .
it was worked out that there is an upper limit on the momentum of the outgoing quark from the weak vertex , less than the maximum recoil energy of @xmath47 . when the heavy quark is changed into the light quark with very high energy via weak interaction
, it suddenly moves through the brown muck consisting of the light degrees of freedom .
the situation is very similar to the case when an electron goes through a dense medium , where the cherenkov radiation should occur .
much more similar processes have been studied in the heavy - ion collisions recently .
here an energetic parton enters through a dense hadronic medium , and possible cherenkov gluon radiation has been considered extensively @xcite .
the energy loss due to the cherenkov radiation can be easily calculated : @xmath48~,\ ] ] where @xmath49 is the index of refraction .
the nonperturbative nature is encoded in @xmath49 .
the amount of energy loss for heavy - ion collisions varies around @xmath50 gev / fm up to the model .
but it is generally accepted that the energy loss due to the cherenkov radiation is smaller than the normal radiation by multiple scattering . roughly speaking , the cherenkov energy loss is about @xcite @xmath51 , where @xmath52 is the gluon energy
. the total energy loss might be @xmath53 where @xmath54 is the flight length of the energetic parton during the formation of @xmath55 .
more precise estimation requires the detailed structure of the index of refraction @xmath49 .
but this naive power counting is enough to give an important message for the soft overlap . if we take into account the cherenkov energy loss , the convolution integral of eq .
( [ lcsr ] ) will be changed into @xmath56 since @xmath57 , the integration domain shrinks severely .
consequently the soft overlap is highly suppressed .
it will be a good phenomenological trade to introduce the cutoff @xmath58 for the nonperturbative @xmath49 .
it was shown in @xcite that for a natural scale of @xmath59 one gets a compatible result for @xmath60 form factors with other approaches . in this paper
, we extend this approach to several @xmath3 decays .
the extension is quite straightforward . in the next section ,
the relevant decay form factors are defined and the end - point singularities are identified . in sec .
iii , various form factors are evaluated with the momentum fraction cutoffs , which originate from the cherenkov energy loss .
section iv contains discussions and conclusions .
the form factors of @xmath61 decays into @xmath62 are defined by @xcite @xmath63 + f_0(q^2)~\frac{m_b^2-m_p^2}{q^2}q^\mu~,\\ % % % \langle p(p)|\qbar\sigma^{\mu\nu}q_\nu b|\bbar(p_b)\rangle & = & \frac{if_t(q^2)}{m_b+m_p}\big[q^2(p_b^\mu+p^\mu)-(m_b^2-m_p^2)q^\mu\big]~,\\ % % % \langle v(p,\epsilon^*)|\qbar\gamma^\mu b|\bbar(p_b)\rangle & = & \frac{2iv(q^2)}{m_b+m_v}\epsilon^{\mu\nu\rho\sigma } \epsilon^*_\nu p_\rho p^b_\sigma~,\\ % % % \langle v(p,\epsilon^*)|\qbar\gamma^\mu\gamma_5 b|\bbar(p_b)\rangle & = & 2m_v a_0(q^2)\frac{\epsilon^*\cdot q}{q^2}q^\mu + ( m_b+m_v)a_1(q^2)\left[\epsilon^{*\mu}-\frac{\epsilon^*\cdot q}{q^2}q^\mu\right ] \nn\\ & & -a_2(q^2)\frac{\epsilon^*\cdot q}{m_b+m_v}\left[p_b^\mu+p^\mu- \frac{m_b^2-m_v^2}{q^2}q^\mu\right]~,\\ % % % \langle v(p,\epsilon^*)|\qbar\sigma^{\mu\nu}q_\nu b|\bbar(p_b)\rangle & = & 2t_1(q^2)\epsilon^{\mu\nu\rho\sigma}\epsilon^*_\nu p^b_\rho p_\sigma~,\\ % % % \langle v(p,\epsilon^*)|\qbar\sigma^{\mu\nu}\gamma_5 q_\nu b|\bbar(p_b)\rangle & = & -it_2(q^2)\big[(m_b^2-m_v^2)\epsilon^{*\mu}-(\epsilon^*\cdot q)(p_b^\mu+p^\mu ) \big]\nn\\ & & -it_3(q^2)(\epsilon^*\cdot q)\left[q^\mu-\frac{q^2}{m_b^2-m_v^2}(p_b^\mu+p^\mu ) \right]~,\end{aligned}\ ] ] where @xmath64 the end - point singularity appears when one considers the gluon exchange diagrams with the spectator quarks ( fig . [ diagram ] ) .
gluon exchanges with the spectator quarks.,title="fig : " ] + their amplitude is proportional to the scattering kernel @xcite @xmath65_{ij}~,\ ] ] which will be convoluted with the meson das to give the full amplitude . at the heavy quark and large recoil limit , the kernel reduces to @xmath66_{ij}~ , \label{kernel}\ ] ] where @xmath67 , @xmath68 , @xmath69 , @xmath70 , and @xmath71 with @xmath72 being a light - like vector .
the end - point singularity occurs when @xmath73 or @xmath74 unless the meson das fall fast enough to compensate the factors of @xmath75 or @xmath76 in the denominator . actually , the asymptotic form of das of the light mesons are @xmath77 . as for @xmath0 das ,
the specific forms depend on models , but they can also suffer from the end - point singularity with @xmath78 factors . in this work ,
we just concentrate on the problem of @xmath73 for simplicity , because it is conceptually more important . from this observation , it is commonly argued that heavy - light transition is dominated by the soft
physics where @xmath73 . and the troublesome divergent terms are absorbed into the nonperturbative soft form factor @xcite .
dominance of the soft - overlap contribution is then supported by the phenomenological fitting within this parametrization .
note that near the end point where @xmath73 , @xmath79 and @xmath80 ; the parton momenta of the outgoing light meson are highly asymmetric .
however , one can not conclude that the soft physics dominates from unphysical divergences .
even worse is that we can not extrapolate the nonperturbative physics from the perturbative analysis .
when @xmath81 is very close to 0 , the offshellness of the exchanged gluon momentum @xmath82 is vanishing , so we are entering the nonperturbative regime . in this case the neglected terms of order @xmath83 become significant , and thus eq .
( [ kernel ] ) is not reliable any longer . at this stage
, it will be very interesting to see how the soft form factor is treated in scet . in scet
the soft form factor is defined by a series of operators containing _ collinear _ quarks only @xcite .
in other words , energetic final - state meson is described solely by the collinear quarks ; soft+collinear combination is not allowed .
this is also justified by the large rapidity gap in @xcite .
that s the reason why the defining operators have the interaction lagrangians @xmath84 or messenger modes which convert the soft spectator quark ( @xmath30 ) into collinear one ( @xmath85 ) through the exchange of collinear gluons or into the soft - collinear quark . in this sense , there is no soft - overlap contributions in scet _ a priori .
_ this makes a sharp contrast with the works of @xcite where the soft - overlap contribution plays an important role .
also in many literatures the end - point singular terms are absorbed into the soft form factor as in eq .
( [ soft ] ) @xcite .
thus the soft form factor @xmath31 contains both soft overlap and hard spectator interactions .
this parametrization is not bad in a viewpoint of the spin symmetry because all the terms in the soft form factor satisfy the spin - symmetry relations .
but it may cause some confusions ; for example , it is not adequate to directly compare @xmath31 with @xmath39 since the soft overlap is included in the former while not in the latter from the construction .
as pointed out in @xcite , it might be unfruitful to extract the hard scattering effects from @xmath31 to leave it purely nonperturbative .
however , it is at least conceptually important to separate the soft overlap from the hard scattering when which of the two is dominant matters .
one more profound matter in scet regarding the end - point singularity is double counting .
the phase space region where @xmath73 corresponds to the zero - bin of the collinear momentum , which must be subtracted to avoid double counting @xcite .
the zero - bin subtraction is equivalent to introducing the ir cutoff regulator . as argued in sec .
i , there is a new possibility for the energy cutoff due to the cherenkov gluon radiation in the full theory .
the origin of this cutoff is of course quite different from that of zero - bin subtraction ; the latter is necessary for an effective theory where the various momentum scales are involved .
it is related to the self - consistency of the effective theory .
but there is a situation where only collinear modes are considered , as in producing energetic light mesons from heavy - light decays . in this case
the zero - bin subtraction corresponds to establishing a cutoff below which one should not enter .
the cherenkov energy loss explains why this must be the case ; the energetic parton can not reach far near end point .
the suppression of soft overlap can be easily seen in eqs .
( [ lcsr ] ) , ( [ cutoff ] ) . in numerical calculations , however , the value of @xmath86 is not so close to 1 .
typically , @xmath87 for @xmath60 .
its deviation from unity is much larger than the usual @xmath88 for @xmath89 mev .
it is very difficult and ambiguous to determine what portion of momentum should be transferred to insure the soft overlap configuration , or to make soft quark collinear .
but it is quite true that eq .
( [ lcsr ] ) contains more than soft overlap with @xmath87 in numerics .
furthermore , the approximation in @xcite @xmath90 tends to increase the numerical value compared to the original integral ( @xmath91 ) .
this is because @xmath92 is not sufficiently small .
in short , the soft overlap contribution is overestimated in lcsr .
the pure soft overlap contribution comes from the much narrower range of momentum fraction , which would be shrunken again by the cherenkov radiation .
in what follows , we _ assume _ that the soft overlap is negligible because of the cutoff by the cherenkov energy loss .
this approach is on the same line as pqcd or scet where the soft overlap is ignored .
the @xmath3 form factors are then given by the hard gluon exchange processes .
there is now one nonperturbative parameter @xmath93 which regulates the divergent convolution as a cutoff .
explicitly @xcite , @xmath94\phi^b_+(k^+)\right\}~,\\ % % % f_0&=&\left(\frac{2e}{m_b}\right)f_+ + \left(\frac{\alpha_sc_f}{4\pi}\right)\left(\frac{m_b-2e}{2e}\right ) \delta f_p~,\\ % % % f_t&=&\left(\frac{m_b+m_p}{m_b}\right)f_+ -\left(\frac{\alpha_sc_f}{4\pi}\right)\left(\frac{m_b+m_p}{2e}\right ) \delta f_p~,\end{aligned}\ ] ] for @xmath95 , where @xmath96 , @xmath97 , and @xmath98 are the higher twist das . here , @xmath99 where @xmath100 for @xmath101 , we have @xmath102\phi_+^b(k^+)\bigg\}~ , \\ % % % a_0&= & \left(\frac{\alpha_sc_f}{4\pi}\right)\left(\frac{\pi^2f_bm_b}{n_ce^2}\right ) \int_{u_c}^{\ubar_c}du\int_0^\infty dk^+\bigg\ { \frac{f_v\phi_\parallel(u)\phi_+^b(k^+)}{\ubar k^+ } + \frac{1+\ubar}{\ubar^2 k^+}f_v\phi_\parallel(u)\phi_-^b(k^+)\nn\\ & & + \frac{f_\perp m_v}{2e}\left [ -\frac{2e}{\ubar(k^+)^2}h'^{(s)}_\parallel(u)\phi_+^b(k^+ ) + \frac{1}{\ubar^2k^+}\left(h^{(t)}_\parallel(u)-\frac{h'^{(s)}_\parallel}{2}\right ) \phi_+^b(k^+)\right]\bigg\}~,\\ % % % a_1&=&\left[\frac{2em_b}{(m_b+m_v)^2}\right]v~,\\ % % % a_2&= & v-\left[\frac{m_v(m_b+m_v)}{em_b}\right]a_0 + \left(\frac{\alpha_sc_f}{4\pi}\right)\frac{m_v(m_b+m_v)}{em_b } \frac{m_b^2 - 2m_be}{4e^2}\delta f_\parallel~,\\ % %
% t_1&= & \left(\frac{m_b}{m_b+m_v}\right)v + \left(\frac{\alpha_sc_f}{4\pi}\right)\frac{m_b}{4e}\delta f_\perp~,\\ % % % t_2&= & \left(\frac{2e}{m_b+m_v}\right)v + \left(\frac{\alpha_sc_f}{4\pi}\right)\frac{1}{2}\delta f_\perp~\\ % % % t_3&= & \left(\frac{m_b}{m_b+m_v}\right)v-\left(\frac{m_v}{e}\right)a_0 + \left(\frac{\alpha_sc_f}{4\pi}\right ) \left(\frac{m_b}{4e}\delta f_\perp+\frac{m_vm_b^2}{4e^3}\delta f_\parallel\right)~.\end{aligned}\ ] ] as in the @xmath95 case , we define @xmath103 where @xmath104 the twist-3 das @xmath105 and @xmath106 can be written in terms of the twist-2 das by wandzura - wilczek type relations @xcite @xmath107+\cdots~,\nn\\ % % % g_\perp^{(a)}(u)&= & 2\left[\ubar\int_0^u\frac{\phi_\parallel(v)}{\vbar}~dv + u\int_u^1\frac{\phi_\parallel(v)}{v}~dv\right]+\cdots~,\nn\\ % % % h_\parallel^{(t)}(u)&= & ( 2u-1)\left[\int_0^u\frac{\phi_\perp(v)}{\vbar}~dv -\int_u^1\frac{\phi_\perp(v)}{v}~dv\right]+\cdots~,\nn\\ % % % h_\parallel^{(s)}(u)&= & 2\left[\ubar\int_0^u\frac{\phi_\perp(v)}{\vbar}~dv + u\int_u^1\frac{\phi_\perp(v)}{v}~dv\right]+\cdots~.\nn\\\end{aligned}\ ] ] note that the integration domain is changed into @xmath108 to avoid the soft overlap region .
the integral over @xmath109 will cause another divergence at @xmath110 .
we simply introduce an ir cutoff @xmath111 for @xmath0 meson sector . to get the numerical estimate
, we use the asymptotic form of @xmath112 , @xmath113 , and @xmath114 where @xmath115 is a model parameter @xcite .
the results are summarized in tables [ b2p ] , [ b2v ] for some @xmath93s .
.[b2p ] @xmath95 form factors for various @xmath93 .
@xmath116 is taken at @xmath117 gev . [
cols="^,^,^,^,^",options="header " , ] in these tables , @xmath118 is given as a natural representative scale in heavy - light decays .
note that the larger @xmath93 , the smaller form factors .
results from lcsr is also listed for comparison .
the value of @xmath119 is chosen to coincide the value of @xmath120 ( @xmath121 ) with that from lcsr . comparing the columns of @xmath119 and lcsr , @xmath21s for @xmath95 are generally in good agreement except @xmath122 .
as for @xmath101 , we can find the following tendency ; @xmath123 , @xmath124 , @xmath125 , @xmath126 , @xmath127 .
present analysis is done with the asymptotic forms of light meson das .
typically , nonasymptotic effects contribute by a few tens of percent , so inclusion of them will be required to refine the results .
one of the main source of uncertainty is the @xmath0 meson da , @xmath128 .
usually the problematic divergent terms involving @xmath129 are also absorbed into the soft form factors .
the remaining part contains the inverse moment of @xmath130 ; @xmath131 with eq .
( [ phib ] ) , one simply has @xmath132 .
we fix @xmath133 gev @xcite .
the divergent moment is parametrized as @xmath134 where @xmath135 acts as an ir regulator . in this parametrization with eq .
( [ phib ] ) , the singular inverse - square moment over @xmath130 is @xmath136 it might be that the bad thing is not the singularity itself ; the singular _ behavior _ would be even worse in the sense that the form factors are very sensitive to the cutoff or regulator . at least , we need to reduce the hadronic uncertainty on , say , @xmath137 . but current analysis shows that for a reasonable set of cutoffs the form factors are compatible with the literature .
the point is that there must be a fundamental cutoff in heavy - light decay due to the cherenkov energy loss .
the necessary condition for the cherenkov radiation is @xmath138>1 $ ] , where @xmath49 is the index of refraction .
analogous to the photon case , @xmath139 is proportional to the forward scattering amplitudes @xmath140 . at low energies , @xmath141>0 $ ]
if @xmath142 for the breit - wigner resonance @xmath143 where @xmath144 is the resonant energy and @xmath145 is the decay width @xcite . since the light mesons are possible intermediate resonances of the brown muck , the necessary condition can be easily satisfied also in @xmath3 transitions . in exclusive decays , the cherenkov gluons will eventually couple and transfer the energy to the light degrees of freedom to make the final state meson . on the other hand , in inclusive decays , they can appear as conelike jet events .
if the jets form a specific angle which can not be explained by conventional arguments , then they will be a strong candidate for the cherenkov gluons . at a glance
, the suppression of soft overlap by cherenkov energy loss looks like sudakov effects .
but the cherenkov radiation is purely medium effects .
the sudakov factor plays a role of weight for meson das which falls very rapidly ( though controversial ) near the end - point region .
however , this is basically the extrapolation into the nonperturbative regime from the perturbative calculations .
the spirit of this work is that we should not _ imagine _ the nonperturbative physics in the line of perturbative approaches .
the index of refraction @xmath49 is a robust nonperturbative object which forbids the energetic parton from the weak vertex to go far into the end - point region . in conclusion
, we propose a fundamental cutoff for the heavy - to - light transitions due to possible cherenkov gluon radiation .
it naturally avoids the end - point singularity and double counting problems . in this picture
the soft overlap contribution is highly suppressed ; heavy - to - light decay is dominated by the hard scattering processes .
though the numerical values of the weak form factor are very sensitive to the choice of the cutoff , for a natural scale of @xmath93 one gets a compatible result with other approaches .
the existence of cherenkov gluons can be directly checked from the conelike jets in inclusive decays .
99 j. charles , a. le yaouanc , l. oliver , o. pene and j. c. raynal , phys .
d * 60 * , 014001 ( 1999 ) [ arxiv : hep - ph/9812358 ] .
h. n. li and h. l. yu , phys .
lett . * 74 * , 4388 ( 1995 ) [ arxiv : hep - ph/9409313 ] ; phys .
d * 53 * , 2480 ( 1996 ) [ arxiv : hep - ph/9411308 ] .
y. y. keum , h. n. li and a. i. sanda , phys .
b * 504 * , 6 ( 2001 ) [ arxiv : hep - ph/0004004 ] ; phys .
d * 63 * , 054008 ( 2001 ) [ arxiv : hep - ph/0004173 ] . y. y. keum and h. n. li , phys .
d * 63 * , 074006 ( 2001 ) [ arxiv : hep - ph/0006001 ] .
t. kurimoto , h. n. li and a. i. sanda , phys .
d * 65 * , 014007 ( 2002 ) [ arxiv : hep - ph/0105003 ] .
h. n. li , phys .
d * 66 * , 094010 ( 2002 ) [ arxiv : hep - ph/0102013 ] .
s. descotes - genon and c. t. sachrajda , nucl .
b * 625 * , 239 ( 2002 ) [ arxiv : hep - ph/0109260 ] .
b. o. lange and m. neubert , nucl .
b * 690 * , 249 ( 2004 ) [ erratum - ibid .
b * 723 * , 201 ( 2005 ) ] [ arxiv : hep - ph/0311345 ] .
r. j. hill , t. becher , s. j. lee and m. neubert , jhep * 0407 * , 081 ( 2004 ) [ arxiv : hep - ph/0404217 ] .
z. t. wei and m. z. yang , nucl .
b * 642 * , 263 ( 2002 ) [ arxiv : hep - ph/0202018 ] ; c. d. lu and m. z. yang , eur .
j. c * 28 * , 515 ( 2003 ) [ arxiv : hep - ph/0212373 ] .
m. beneke and t. feldmann , nucl .
b * 592 * , 3 ( 2001 ) [ arxiv : hep - ph/0008255 ] .
m. beneke , g. buchalla , m. neubert and c. t. sachrajda , phys .
lett . * 83 * , 1914 ( 1999 ) [ arxiv : hep - ph/9905312 ] ; nucl .
b * 591 * , 313 ( 2000 ) [ arxiv : hep - ph/0006124 ] .
c. w. bauer , s. fleming and m. e. luke , [ arxiv : hep - ph/0005275 ] ; c. w. bauer , s. fleming , d. pirjol and i. w. stewart , phys .
d * 63 * , 114020 ( 2001 ) [ arxiv : hep - ph/0011336 ] ; c. w. bauer and i. w. stewart , phys .
b * 516 * , 134 ( 2001 ) [ arxiv : hep - ph/0107001 ] ; c. w. bauer , d. pirjol and i. w. stewart , phys .
d * 65 * , 054022 ( 2002 ) [ arxiv : hep - ph/0109045 ] . c. w. bauer , d. pirjol and i. w. stewart , phys .
d * 67 * , 071502 ( 2003 ) [ arxiv : hep - ph/0211069 ] .
a. v. manohar and i. w. stewart , arxiv : hep - ph/0605001 . c. w. bauer , d. pirjol , i. z. rothstein and i. w. stewart , phys
. rev .
d * 70 * , 054015 ( 2004 ) [ arxiv : hep - ph/0401188 ] .
m. beneke , g. buchalla , m. neubert and c. t. sachrajda , phys .
d * 72 * , 098501 ( 2005 ) [ arxiv : hep - ph/0411171 ] .
p. ball and r. zwicky , jhep * 0110 * , 019 ( 2001 ) [ arxiv : hep - ph/0110115 ] ; phys .
d * 71 * , 014015 ( 2005 ) [ arxiv : hep - ph/0406232 ] .
p. ball and r. zwicky , phys .
d * 71 * , 014029 ( 2005 ) [ arxiv : hep - ph/0412079 ] .
p. ball and v. m. braun , phys .
d * 55 * , 5561 ( 1997 ) [ arxiv : hep - ph/9701238 ] .
j. p. lee , phys .
d * 75 * , 031501 ( 2007 ) [ arxiv : hep - ph/0612083 ]
. v. koch , a. majumder and x. n. wang , phys .
lett . * 96 * , 172302 ( 2006 ) [ arxiv : nucl - th/0507063 ] ; a. majumder and x. n. wang , [ arxiv : nucl - th/0507062 ] .
i. m. dremin , [ arxiv : hep - ph/0507167 ] ; p. ball , arxiv : hep - ph/0308249 . m. beneke and t. feldmann , nucl .
b * 685 * , 249 ( 2004 ) [ arxiv : hep - ph/0311335 ] .
p. ball , v. m. braun , y. koike and k. tanaka , nucl .
b * 529 * , 323 ( 1998 ) [ arxiv : hep - ph/9802299 ] .
a. g. grozin and m. neubert , phys .
d * 55 * , 272 ( 1997 ) [ arxiv : hep - ph/9607366 ] . | decay form factors of @xmath0 to light pseudoscalar ( @xmath1 ) and vector ( @xmath2 ) mesons are analyzed with the momentum cutoff near end point .
the cutoff is caused by possible cherenkov gluon radiation when an energetic light parton from the weak vertex travels through the brown muck of light degrees of freedom .
the end - point singularities and the double countings are naturally absent .
the soft - overlap contributions where the partonic momenta configuration is highly asymmetric are very suppressed in this framework .
a simple calculation gives a plausible results for @xmath3 form factors . |
in recent years , the unconventional superconductivity found in iron - based superconductors(ibscs ) has attracted much attention in condensed matter community . due to the proximity of superconductivity and the collinear spin density wave(sdw ) phase
, it is generally believed that there exists an intrinsic link between the magnetic fluctuations and the superconductivity .
thus , the understanding of the normal state magnetic fluctuations will be helpful to identify the mechanism of superconductivity .
experimentally , it was found that the sdw transition is either preemptive by or coincident with a tetragonal - to - orthorhombic(to ) structural transition at @xmath1@xcite which signals the @xmath2 symmetry breaking above the sdw transition temperarture @xmath3 .
such a @xmath2 symmetry breaking phase is called the nematic phase in literatures@xcite .
further studies show that the nematicity persists to @xmath4 above @xmath1@xcite .
so far , there is clear evidence that nematicity comes from the electronic rather than the lattice degree of freedom@xcite .
however , it is difficult to identify the fingerprint of nematicity due to the coupling of various electronic degrees of freedom .
experimentally , it was found that the presence of nematicity is accompanied by anisotropic magnetic correlations which indicates its magnetic origination@xcite .
meanwhile , it was also found that the characteristic temperature @xmath4 of nematicity coincides with that of the orbital order@xcite , which indicates that nematicity is orbitally driven .
so , the origin of nematicity is still debated , both the magnetic and orbital fluctuations are proposed to be responsible for the emergence of nematicity@xcite . thus , the study of the origin of nematicity is highly desired as it may shed light on the mechanism of unconventional superconductivity in ibscs .
one of the effective ways to distinguish between different scenarios of nematicity is to compare the theoretical results of each scenario to the experimental observations . in this regard
, we notice that the electronic structure can be mapped effectively through the analysis of the quasiparticle scattering interference ( qpi ) patterns@xcite and the spatial modulation of the electronic states resulting from qpi can be probed directly by the stm experiments . in fact , the recent stm experiments@xcite for nafeas have been performed to investigate the electronic structure in the nematic phase in ibscs .
the stm experiments@xcite reveal that the qpi patterns exhibit highly anisotropic dimer structure deep in the sdw phase .
it was further found that the anisotropic features persist to high temperatures above @xmath1 , but the anisotropy weakens gradually with the increase of temperature .
therefore , the qpi may offer a playground for the test of the origin of nematicity . in this paper
, we study theoretically the quasiparticle interference induced by impurity in the nematic phase based on the magnetic and orbital scenarios of nematicity , respectively .
deep in the collinear @xmath5 sdw state , the low energy qpi patterns exhibit a dimer structure with its orientation along the ferromagnetic direction of the sdw state .
this is due to the fully opening of the sdw gap along the @xmath6 direction .
when the energy is increased to be slightly below the fermi level , it is composed of two sets of dimers along the same direction . in this case , the qpi reflects the topology of the distorted fermi surface and the two sets of dimers come from the inter - hole - pocket and intra - hole - pocket scatterings .
it is further shown that the above features of the qpi patterns remain in the magnetically driven nematic phase , which is modelled by the fluctuating short - range antiferromagnetic order . in the orbital scenario of nematicity
, it is found that the scatterings of quasiparticles between dirac cones dominate the qpi process . due to the inequivalent energy positions of the two dirac cones resulting from the @xmath2 symmetry breaking induced by the orbital order , the dimer structure in the qpi patterns undergoes a @xmath0 rotation with increasing energy .
the transition is irrespective of the momentum dependence of the orbital order , so long as the energy splitting between the @xmath7 and @xmath8 bands is fixed near the dirac cones .
we propose that these results can be used to distinguish the origin of the nematicity in various iron - based compounds . as a preliminary comparison to available experiments
, we note that the obtained anisotropic features of the scattering patterns and the orientation of the dimer in the magnetic scenario are qualitatively consistent with the experimental observations on nafeas@xcite .
the hamiltonian we use to carry out the calculations can be divided into three parts : the tight binding part , the impurity part and that modelling the nematicity .
we adopt the five - orbital tight binding hamiltonian of ref@xcite which reproduces the lda energy bands .
the tight binding hamiltonian reads @xmath9 . where @xmath10 and @xmath11 are the orbital and spin indices , respectively .
the tight binding hopping parameters for @xmath12 are given in ref@xcite and the energy unit ev will be used throughout the paper . in this paper
, we focus on the qpi phenomenon induced by impurity potential scattering . the impurity hamiltonian can be written as @xmath13 .
for simplicity , we adopt a @xmath14-function type scattering potential which resides on the given lattice site @xmath15 .
the scattering matrix @xmath16 is orbital diagonal and momentum independent . in this paper ,
the average electron occupation number is fixed to be @xmath17 per unit cell which corresponds to the undoped parent compound . in order to model the sdw and the magnetically driven nematic phase , we include the coulomb interaction hamiltonian .
so , the full hamiltonian reads @xmath18 , and @xmath19 is given by , @xmath20 where @xmath21 and @xmath22 are the orbital indices .
@xmath23 are the coefficients of the intraorbital interaction , interorbital interaction , hund - coupling , and pair hopping terms , respectively . @xmath24 and @xmath25 are assumed as required by the spatial rotational symmetry .
without loss of generality , @xmath26 is assumed .
@xmath19 is treated at the mean field level , and is decoupled into the orbital diagonal channel in the following way@xcite , @xmath27 where @xmath28 and @xmath29 are the electron number and spin operators at site @xmath30 with orbital @xmath21 , respectively .
the values of @xmath31 and the magnetic moment @xmath32 which is defined via @xmath33 for the @xmath5 sdw state ( @xmath34 the number of lattice sites , @xmath35 the @xmath36-coordinate of site @xmath30 ) are obtained through the self - consistent calculations . in this paper , the intraorbital interaction @xmath37 is adopted , which is slightly above the critical value @xmath38 for the appearance of sdw .
the results for the magnetic moments are @xmath39 , @xmath40 , @xmath41 , @xmath42 , and @xmath43 . in eq.(2 ) , we have ignored the orbital off - diagonal magnetic terms .
in fact , we have checked numerically the effects of these off - diagonal terms .
we find that the nonzero orbital off - diagonal sdw moments are @xmath44 , where @xmath45 .
these sdw moments are smaller compared to their diagonal parts , @xmath46 , @xmath47 , @xmath48 , @xmath49 , and @xmath50 .
furthermore , we have checked that the off - diagonal sdw orders have no qualitative influences on the behaviors of the qpi patterns in the sdw and spin driven nematic state . here , we focus mainly on the qpi within the orbital diagonal sdw approximation in the following discussions . within the @xmath5 ansatz of the sdw order
, we introduce the annihilation operator with ten components as @xmath51 , where @xmath52 is the sdw wave vector . in this way
, the green s function reads @xmath53 in the sdw state , where @xmath54 is the time - ordering operator .
, respectively .
the blue square marks the location of the dirac node in the momentum space .
( c ) the intensity map of the sdw distorted fermi surface . @xmath55 and @xmath56 are two dirac pockets .
( d ) the intensity map of the single - particle spectral function along the high symmetry directions .
the intensity is in logarithmic scale in this panel.,width=307 ] experimentally , the nematicity in fe - pnictides and fe - chalcogenides has been shown to develop at a temperature above the sdw transition@xcite .
therefore , the magnetically driven nematicity where the long - range sdw order is absent but the c@xmath57 symmetry is broken by magnetic fluctuations has been proposed@xcite .
this phase is modelled by the short - range antiferromagnetic correlation , which is approximated by the lee - rice - anderson formula@xcite . in this way
, the green s function is written as @xmath58 , where @xmath59 and @xmath60 are the matrix representation of the mean - field hamiltonian @xmath61 without the sdw terms .
@xmath62 is a lorentzian which represents the @xmath63 modulated magnetic correlation with @xmath64 the correlation length and @xmath63 the momenta derivation from @xmath5 .
the orbital diagonal @xmath65 is the order parameter matrix of the fluctuating magnetic order .
the elements of @xmath65 read as @xmath66 and @xmath67@xmath68 .
where @xmath32 is the magnetic moment of orbital @xmath21 .
following the previous study@xcite , the mean - field sdw order parameters obtained at @xmath69 are taken to be @xmath32 . in the orbital scenario ,
the orbital orders are used to model the orbitally driven nematicity as used before@xcite .
generally , the hamiltonian for the orbital orders can be written as @xmath70 which breaks the @xmath2 symmetry , where @xmath71 are the order parameters .
the value of @xmath71 is determined by the arpes data .
thus , the full hamiltonian reads @xmath72 for the orbital scenario of nematicity in the presence of the impurity scattering .
both the momentum dependent and the momentum independent orbital orders will be considered in the following . with the hamiltonians shown above , we can now construct the corresponding green s functions .
for the orbital scenario of nematicity , the green s function can be defined in the orbital basis as @xmath73 which can be obtained directly from @xmath74 .
the quasiparticle interference occurs between the ingoing and scattered outgoing electrons by impurity .
the resulting spatial modulation of the electronic states can be visualized directly by the stm experiments@xcite . after a fourier transformation
, one can get its manifestation in the momentum space , i.e. , the density of states in the momentum space @xmath75 .
thus , the features of the qpi patterns can be qualitatively understood by the analysis of the joint density of states of the initial and final states .
theoretically , @xmath75 is expressed as @xmath76 , where the @xmath54 matrix reads @xmath77 . in each scenario ,
the corresponding green s function @xmath78 is used to calculate @xmath75 .
a born limited scattering potential of @xmath79 is adopted . in this case , the qpi patterns are not disturbed by the impurity resonance states , thus they are directly related to the underlying bands structure .
let us start with the discussion of the distortion in the energy bands due to the collinear @xmath5 sdw order with its moment antiferromagnetically aligning along the @xmath36 direction but ferromagnetically along the @xmath80 direction . in fig .
[ f.1 ] , we show the normal state fermi surface ( fs ) and the resulted fs after the introduction of the sdw order . clearly , the normal state hole fs around @xmath81 develops into three distinct hole pockets @xmath82 , @xmath55 and @xmath56 in the presence of the sdw order which induces the hybridization between the hole pockets around @xmath82 and the electron pocket around @xmath83 . as revealed by previous study@xcite , the two small hole pockets @xmath55 and @xmath56 exhibit dirac cone like dispersions .
these dirac cone dispersions persist even when the off - diagonal sdw orders are taken into account .
the sdw distorted fss calculated here are in qualitative agreement with the arpes observations@xcite . at the same time
, the similar distortion of the fermi pocket at @xmath84 occurs , while the fermi pocket around @xmath85 is affected less , because the @xmath5 sdw order is considered here .
quite similar sdw distorted fermi surfaces have been obtained by the previous calculations @xcite based on the five orbital model . meanwhile ,
our results for the single - particle spectrum are fully consistent with the previous study@xcite , as shown in fig [ f.1](d ) . ,
@xmath86 , @xmath87 and @xmath88 , respectively .
the intensity maps of the corresponding quasiparticle spectral functions are plotted in the right side panels.,width=307 ] further analysis shows that the highly anisotropic fs induced by the sdw is related directly to the orbital characters of the normal state energy bands . as presented in fig .
[ f.1](a ) , the normal state fs is dominated by the @xmath7 , @xmath8 and @xmath89 orbitals .
the overlap between the @xmath82 and @xmath84 pockets are dominated by the @xmath8 orbital , as a result , their hybridization is maximum .
consequently , the coupling between the inner @xmath82 pocket and the @xmath84 pocket gives rise to a large sdw gap @xmath90 along @xmath91 axis . on the other hand ,
the sdw gap @xmath92 along the @xmath93 axis is smaller .
this attributes to the fact that the overlap of the @xmath7 or @xmath89 band between the @xmath82 pocket and the @xmath84 pocket is weak .
however , we note that the sdw gap opens along the full @xmath6 axis , while only at individual @xmath94 point along the @xmath95 axis . in addition , the dirac cone dispersions located at @xmath93 axis are directly related to the orbital character of the relative bands@xcite .
thus , the orbital degree of freedom plays an important role in the bands reconstruction in the sdw state .
now , we study the energy evolution of the qpi patterns in the sdw state and the results are presented in fig .
[ f.2 ] . in fig .
[ f.2](a ) , we show the qpi pattern for @xmath96 .
its main feature is obviously anisotropic with only @xmath97 symmetry and exhibits a dimer - like structure orienting along the @xmath98 direction .
we also show the intensity map of the corresponding quasiparticle spectral functions(qsfs ) at the same energy in fig .
[ f.2](b ) .
one can see that the electronic states along the @xmath93 axis are fully gapped , which is the consequence of the opening of the sdw gap @xmath99 as can be seen from the single - particle spectrum presented in fig . [
we notice that the sdw gap does not exist along the whole @xmath91 axis , although it opens at some @xmath95 points[fig .
[ f.1](d ) ] .
thus , the qsfs exhibit clear weights along the @xmath91 axis , and the qpi pattern is dominated by the scattering processes along the @xmath91 direction .
it should be noted that at the center region around the @xmath82 point the qpi pattern exhibits like a short dimer .
however , it is not the substantial feature of the scattering pattern . when the energy is slightly away from @xmath96 , the short dimer diminishes but the vertical dimer structure remains
actually , this dimer - like structure of the qpi pattern persists in the energy window of @xmath100 , which coincides basically with the energy region of the sdw gap in the single - particle spectrum as shown in fig . [ f.1](d ) . with the increase of energy , the pattern develops gradually into a broad peak around the @xmath82 point and it has no obvious anisotropy as shown in fig . [ f.2](c ) for @xmath101
moreover , the intensity of the qpi pattern is nearly three times larger than that with @xmath96 .
this is the specific case in that this energy happens to be near the top of the hole band around @xmath102 ( not shown here ) , where the density of states is large and consequently the qsfs show a noticeable intensity around @xmath102 as shown in fig .
[ f.2](d ) . above this energy region ,
one approaches gradually to the fermi energy , so the qsfs will copy the main features of the fs . in this case ,
the qpi pattern is dominated by two sets of vertical dimers along the @xmath103 direction as indicated by the @xmath104 and @xmath105 arrows in fig .
[ f.2](e ) for @xmath106 , which arises from the inter - hole - pocket and intra - hole - pocket scatterings , respectively [ see fig .
[ f.2](f ) ] . what we distinguish
the two sets of vertical dimers comes from the features exhibited in the qsf shown in fig .
[ f.2](f ) . from this figure
, one can see that the contour of the qsf is elongated along the @xmath103 direction , so that the portions of the contours connected by @xmath107 become more or less flat , and thus the scatterings across the vertical nearly flat portions are dominant and give rise to the vertical dimers .
actually , this kind of dimer structure of the qpi patterns persists in the energy regime of @xmath108 . in the positive energy regime ,
the dimer structure of the qpi pattern diminishes gradually . the typical qpi pattern is shown in fig . [ f.2](g ) for @xmath109 , which is dominated by a peak around the @xmath82 point
this is consistent with the qsfs analysis , where its anisotropy weakens significantly in the positive energy regime [ fig .
[ f.2](h ) ] .
it is mainly because the energy band above the fermi level is less distorted by the sdw formation , as can be seen from a comparison between fig .
[ f.1](b ) and ( d ) .
it was carefully checked that the above mentioned features of the qpi patterns persist for weak impurity scattering regardless of the sign of @xmath110 .
theoretically , the features of the qpi patterns in the sdw state have been investigated by previous studies@xcite .
similar to the results presented in fig .
[ f.2](e ) and ( f ) , the highly anisotropic qpi patterns obtained by previous studies@xcite are directly related to the topology of the sdw distorted constant energy contours . here
, we emphasize another cause leading to the dimer structure in qpi pattern , that is the highly anisotropic sdw gaps in the momentum space which is directly related to the orbital characters of the folded bands as addressed above . for @xmath111 and @xmath112 , respectively .
panels ( b ) , ( d ) are the qpi patterns at @xmath106 for @xmath111 and @xmath112 , respectively.,width=307 ] in the magnetically driven nematic phase , the long - range sdw order gives way to the short - range antiferromagnetic correlation which is modelled by the lee - rice - anderson formula@xcite as introduced above . in this case , the magnetic correlation length @xmath64 is the relevant parameter . therefore , let us study the evolution of the qpi patterns with the reduction of the magnetic correlation length @xmath64 .
the qpi patterns for @xmath96 are shown in fig .
[ f.3](a ) and ( c ) for @xmath111 and @xmath112 , respectively .
compared to that with the sdw long - range order [ fig .
[ f.2](a ) ] , the main feature still exhibits the anisotropy with the noticeable intensity of the pattern along the @xmath103 direction .
it is the consequence of the gap opening due to the short - range magnetic correlation similar to that in the sdw scenario discussed above .
the variation is that two new peaks emerge at the two ends of the dimer - like structure . from fig .
[ f.3](a ) and ( c ) , one can also see that this kind of qpi structure is less affected by the change of @xmath64 from @xmath113 to @xmath114 .
we also notice that it preserves in the same energy window of @xmath100 to the case of the sdw order .
it is interesting to notice that similar qpi patterns have been experimentally observed in the ferropnictide @xmath115 compounds@xcite , although the observed dimers are shorter .
the results for @xmath106 are presented in fig .
[ f.3](b ) and ( d ) .
it shows that the @xmath104 and @xmath105 dimers presented in the sdw state merge together to give rise to a loop around the @xmath82 point when the correlation length is @xmath111 .
the mergence becomes strongly with the further decreasing of @xmath64 , and eventually a nearly flat - top broad peak around the @xmath82 point will appear , as has already been seen from fig .
[ f.3](d ) for @xmath112 .
thus , it can be found that the anisotropy of the qpi patterns weakens significantly with the reduction of @xmath64 . with a finite correlation length @xmath64 ,
the sdw gives way to the damped spin excitations .
the damping becomes stronger with the decrease of the correlation length @xmath64 . as a result
, the quasiparticle peak is broadened correspondingly due to its coupling to these damped spin excitations as indicated by the lee - rice - anderson formula@xcite used above , so do the main features of the qpi around the @xmath82 point . in the case of @xmath96 , the dimer structure along the @xmath103 direction in the sdw phase
is mainly caused by the opening of the sdw gap along the @xmath6 direction of the single - particle spectrum which is affected less by the quasiparticle damping .
thus , the anisotropic feature of the qpi pattern for @xmath96 persists even when @xmath64 is significantly reduced . without and with the off - diagonal sdw orders , respectively .
the intensity maps of the corresponding qsfs are plotted in the right panels.,width=307 ] it was checked that the above mentioned features of the qpi patterns are qualitatively unchanged when the interacting parameters vary within a realistic region suggested in ref.@xcite .
we have also considered the effects of the band structure on qpi patterns by using the lda energy bands of lifeas@xcite .
it was found that the qpi patterns also exhibit the prominent dimer structure when energy is in the sdw gap region .
the dimers of the qpi patterns persist when the energy is around the fermi level , and they weakens gradually with the reduction of the magnetic correlation length in the magnetic nematic phase .
it should be noticed that there exist significant differences between the energy bands of lifeas and those of laofeas@xcite .
thus , the qpi features established above is robust .
furthermore , it was checked that the main features of the qpi patterns remain when the off - diagonal sdw orders are considered . here , we do not present detail comparisons between the qpi patterns with and without the off - diagonal sdw orders in all cases . instead , we give a typical example to show the behaviors of the scattering patterns when the off - diagonal sdw orders are taken into account .
we show the qpi patterns for @xmath116 in fig .
[ f.8](a ) and ( c ) in the sdw state without and with the off - diagonal sdw orders , respectively . as shown in fig .
[ f.8](a ) , the qpi pattern for @xmath116 exhibits a dimer - like structure orienting along the @xmath98 axis which is similar to that presented in fig . [ f.8](c ) with the off - diagonal sdw orders , although some differences around the @xmath82 point exists .
this stems from the effect of the highly anisotropic sdw gaps in momentum space , which is less subjected to the inclusion of the off - diagonal sdw orders as shown by the qsfs in fig .
[ f.8](b ) and ( d ) .
we also note that there is an approximate @xmath117 increase in magnitudes of the converged diagonal sdw orders after including the off - diagonal terms as presented in sec.iia .
as a consequence , the qpi pattern shown in fig .
[ f.8](c ) for @xmath116 has nearly the same structure as that shown in fig .
[ f.2](a ) for a slightly different energy of @xmath96 .
experimentally , it was found that the degeneracy between the @xmath7 and @xmath8 bands is lifted at a temperature coincident with the onset of nematicity@xcite . theoretically , it was proposed@xcite that the nematicity may originate from the orbital fluctuations .
however , it is difficult to distinguish the orbital driven nematicity from its spin counterpart due to their mutual coupling@xcite . in this section , we study the qpi patterns induced by the potential scattering in the orbital scenario of nematicity .
we note that it was recently revealed by arpes experiments that the orbital order is strongly momentum dependent in fese@xcite . with the help of the symmetry analysis
, we will first obtain the general form of the orbital order up to the nearest - neighbor sites .
in the present model , the point group of the fe plane is @xmath118 , where @xmath119 with @xmath120 a @xmath0 rotation along the @xmath121 axis followed by a fe plane mirror reflection , @xmath122 and @xmath123 are @xmath124 rotations along the @xmath36 and @xmath80 axis of the fe - fe bond , respectively . @xmath125 and @xmath126 are two mirrors determined by the nearest - neighbor fe and as atoms , respectively .
it was found experimentally that the dirac cone dispersion exists in the sdw state@xcite . as pointed out by previous study@xcite ,
the emergence of the dirac cones is a consequence of the @xmath122 and @xmath123 symmetries . in this way
, we expect that the @xmath122 and @xmath123 symmetries are preserved in the sdw state . as proposed theoretically@xcite
, the sdw state is expected to occur as a result of the spontaneous symmetry breaking of the nematic phase .
thus , the @xmath122 and @xmath123 symmetries are preserved in the nematic phase .
this is further supported by recent arpes observations@xcite . in this way , the point group of the nematic phase is @xmath127 . generally , the hamiltonian for the orbital order can be written as @xmath70 , where @xmath128 and @xmath10 are the lattice sites and orbital indices , respectively .
considering the point group symmetry and that the @xmath129 orbitals dominate the low energy electronic bands , we find that the @xmath130 with only the on - site orbital order reads @xmath131 .
other terms , such as @xmath132 , are forbidden since they break the @xmath122 and @xmath123 symmetries .
the @xmath130 with the orbital orders up to the nearest - neighbor sites are written as @xmath133 .
we can of course construct the orbital orders up to the next - nearest - neighbor sites in a similar way , however , this introduces more parameters which complicate the problem . in general , the magnitude of the order will decrease with the increase of the lattice distance .
so , we will focus on the orbital orders up to the nearest - neighbor sites which are believed to have captured the main physics .
we use the above obtained hamiltonian for the orbital orders to fit the splitting energy between the @xmath7 and @xmath8 bands observed in experiments@xcite .
experimentally , the splitting at @xmath82 point is about @xmath134 , and that between the @xmath8 band at @xmath84 point and the @xmath7 band at @xmath85 point is increased to be @xmath88 .
the parameters are @xmath135 , which can reproduce the experimental data well .
the resulted energy bands are shown in fig .
[ koob](a ) .
the outer hole band along the @xmath136 direction is mainly of the @xmath8 orbital character , while that along the @xmath137 direction is of @xmath7 character . and
@xmath137 directions are mainly composed of the @xmath8 and @xmath7 orbitals , respectively .
the two black squares mark the location of the two dirac nodes which reside on the @xmath93 and @xmath91 axes .
( b ) linear dispersion for the dirac cone located at @xmath93 axis .
this panel is plotted along the @xmath138 line .
the dirac node locates at @xmath139.,width=288 ] let us first discuss the effects on the energy band due to the orbital orders . because the outer hole band and the electron band along the @xmath6 axis are even and odd parities under the @xmath122 operation . as a result ,
no gap opens at the intersection point @xmath139 between the outer hole band and the electron band .
it leads to a dirac cone dispersion around @xmath5 .
equivalently , there exists another dirac cone band around @xmath140 .
the locations of the two dirac points are marked by the black squares in fig .
[ koob](a ) .
these two dirac cones are inequivalent in energy due to the @xmath2 symmetry breaking induced by the orbital orders .
the node energies are @xmath141 and @xmath142 for the dirac cones located at the @xmath93 and @xmath91 axes , respectively . in fig .
[ koob](b ) , we show the dispersion of the low energy dirac cone at @xmath139 along the @xmath138 direction .
it is interesting to notice that the similar cone like dispersion has been observed by the recent arpes measurements on thin fese films@xcite , although the observed dirac cone is just below the fermi level . on the other hand ,
compared to the complicated energy bands reconstruction induced by the sdw order as shown in fig .
[ f.1](c ) , the distorted energy bands due to the orbital orders are dominated by two hole bands around the @xmath82 point and two electron bands around the @xmath5 and @xmath140 points . now , we turn to the study of the energy evolution of the qpi patterns in the orbital scenario of nematicity .
the results are obtained in a wide energy window from @xmath143 to @xmath144 , and the anisotropy of the qpi pattern is rather weak when @xmath145 suggesting that it is irrelevant to the orbital orders above this energy .
thus , we show the typical patterns in fig .
[ kqpi ] ( a ) , ( c ) , and ( e ) for @xmath146 , @xmath147 and @xmath148 , respectively . for @xmath146 ,
the scattering pattern is highly anisotropic with its main intensity along the @xmath149 axis , leading to a horizontal dimer .
when the energy is increased to be near @xmath150 , the qpi pattern turns to be a vertical dimer .
this vertical dimer remains up to a higher energy @xmath151 which is slightly above @xmath152 , although additional structures develop around the dimer . with the further increasing of energy , the anisotropy of the qpi patterns weakens significantly .
especially , the scattering patterns exhibit little anisotropy in the positive energy region . ,
@xmath147 and @xmath148 , respectively .
right panels are the intensity map of the corresponding quasiparticle spectral functions.,width=307 ] the transition of the qpi patterns from the horizontal to vertical structure is associated with the inequivalent energy positions of the two dirac cones due to the @xmath2 symmetry breaking induced by the orbital orders . as shown in fig .
[ kqpi](b ) , ( d ) and ( f ) , the spectral function shows a large intensity around the dirac cones .
it attributes to the small velocity of the outer @xmath8 and @xmath7 bands around the two dirac cones .
so , the qpi process is dominated by the dirac cone to dirac cone scatterings of quasiparticles . for @xmath146 ,
the nearby dirac cones situate at the @xmath6 axis .
but , for @xmath150 and -0.1 , they situate at the @xmath95 axis .
so , the dirac cone to dirac cone scatterings lead to the transition of the qpi patterns which is related to the presence of the orbital orders .
though the orbital order is momentum dependent in fese@xcite , this may vary in different kinds of ibscs . as a comparison
, we will study the energy evolution of the qpi patterns with a momentum independent orbital order in this section .
the parameters @xmath153 give rise to a momentum independent orbital order as that used in previous studies@xcite . as shown in fig .
[ f.5](a ) , the resulted energy splitting between the @xmath7 and @xmath8 bands are @xmath88 at @xmath82 . and
it has the same splitting energy between the @xmath7 band around @xmath85 and the @xmath8 band around @xmath84 .
the magnitude of this splitting energy is comparable to that obtained by the arpes experiments@xcite . and
@xmath137 directions are mainly of the @xmath8 and @xmath7 characters , respectively . panels ( b ) , ( c ) and ( d ) are the qpi patterns for @xmath146 , @xmath147 and @xmath148 , respectively.,width=307 ] to make a comparison to the results for the momentum dependent orbital order , we present the qpi patterns at @xmath146 , @xmath147 , and @xmath148 in fig .
[ f.5](b)-(d ) .
quite similarly , the main feature of the qpi patterns exhibits a transition from the horizontal dimer at a low energy @xmath146 to the vertical dimer at @xmath150 , and the vertical dimer persists up to a high energy @xmath151 .
it should be noticed that the horizontal dimers with the highest intensity in fig .
[ f.5](c ) are not the substantial feature of the qpi pattern due to the fact that they diminish when the energy is away from @xmath150 .
it is carefully checked that the main features of the qpi patterns keep unchanged when the orbital orders evolve smoothly from the strongly momentum dependent case to the momentum independent case , if the splitting energy between the @xmath7 band at @xmath85 and the @xmath8 band at @xmath84 is fixed to be @xmath88 .
thus , the above established transition of the qpi patterns is a robust phenomenon , regardless of the momentum dependence of the orbital orders .
the difference in the energy splitting of the @xmath7 and @xmath8 bands between the momentum dependent and independent orbital orders is that there is a constant energy splitting in the latter case .
thus , the two dirac cones that situate near the @xmath5 and @xmath140 points are not affected by the difference of the momentum dependence of the orbital order as shown clearly in fig .
[ koob](a ) and fig .
[ f.5](a ) .
thus , the main feature of the qpi patterns is unchanged because it mainly arises from the dirac cones to dirac cones scatterings . before concluding sec.iiib
, it is worth mentioning that the transition of the qpi patterns is related to the @xmath2 symmetry breaking induced by the orbital order and the parities of relative bands which are both involved with the symmetry of the orbital nematic phase .
thus it is expected that the above established transition remains unchanged when the interactions between electrons are taken into account .
actually , there exists numerical evidence@xcite that the symmetries of relative bands do not change when the correlation effects are considered .
we also note that the established transition of qpi patterns occurs at energies much below the fermi level .
considering that the bandwidth is significantly reduced when the coulomb interactions are taken into account@xcite , we expect that the energy difference between the transition and the fermi level will be reduced and it will facilitate the experimental measurements . on the other hand , it was experimentally found on the fese films that the dirac cones reside slightly below the fermi level@xcite .
this indicates that the transition of qpi patterns can be observed at energies close to the fermi level in some iron pnictides , at least in the fese films .
in conclusion , we have studied theoretically the quasiparticle scattering interference patterns in the nematic phase of the iron - based superconductors based on the magnetic and orbital scenarios , respectively .
deep in the sdw state , the qpi patterns exhibit a dimer structure in a wide energy region and develop a bi - dimer structure when the energy is increased to be near the fermi level . it is also shown that the dimer structure of the qpi patterns still exists when the sdw state is replaced by the state with fluctuating magnetic order , though the bi - dimer structure is smeared due to the mergence of the two sets of dimers when the short - range correlation length is significantly decreased . thus , we identify that the qpi dimers present in the sdw and the magnetic nematic phase orient along the ferromagnetic direction of the sdw order as long as the correlation length is not significantly decreased .
our results based on the magnetic scenario of nematicity are qualitatively consistent with the stm observations on nafeas@xcite . in the orbital scenario of nematicity , the qpi patterns are dominated by a dimer structure along the @xmath154 or @xmath103 axes in a wide energy region .
a @xmath0 rotation of the dimer structure occurs when the energy increases from the lower dirac node to the higher one . above the two dirac nodes ,
the anisotropy of the qpi patterns is significantly weakened .
furthermore , it is found that this transition is insensitive to the momentum dependence of the orbital order .
theoretically , the suggestions for distinguishing the sdw order from the orbital driven nematicity by using stm technique have been proposed@xcite .
et al._@xcite focus on the contributions of the band extrema ( local maxima or minima in the bands ) to the local density of states ( ldos ) and qpi .
they found that the features related to the differences between two band extrema which track the orbital splitting or the sdw gap can be used to distinguish between the orbital splitting and the sdw order .
the features are easy to be detected in ldos , but are obscured in qpi by contributions from orbitals other than @xmath7 and @xmath8 .
et al._@xcite focus on the qpi at an energy near the fermi level , so the anisotropic qpi patterns are closely related to the topology of the sdw or orbital order distorted constant energy contours near the fermi level . here
, we uncover the significant contributions to qpi from the dirac cone to dirac cone scatterings of quasiparticles in the orbital scenario . for the magnetic scenario ,
we elaborate the robust dimer structure in qpi patterns resulting from the highly anisotropic sdw gaps in the momentum space which are directly related to the orbital characters of the folded bands .
we further show that due to the inequivalent energies of the two dirac nodes , the dimer structure in the orbital scenario undergoes a @xmath0 rotation with the increase of energy , which contrasts clearly with the case of the magnetic scenario .
these features of the qpi patterns have not been addressed in previous studies@xcite .
in addition , compared to ref .
@xcite in which the two - orbital model is used , we carry out the calculations by using the more realistic five - orbital model . from the results presented in this paper ,
we conclude that the qpi patterns exhibit dimer structure in a wide energy region in both the magnetic and the orbital scenarios of nematicity .
the dimer tends to orient along the ferromagnetic direction in the magnetic case .
however , it undergoes a @xmath0 rotation with the increase of energy in the orbital case .
thus , our results established in this paper for the qpi patterns may be used to probe the origin of nematicity in various iron - based superconductors .
this work was supported by the national natural science foundation of china ( 11190023 and 11374138 ) , and national key projects for research @xmath155 development of china ( grant no .
2016yfa0300401 ) .
s. avci , o. chmaissem , d. y. chung , s. rosenkranz , e. a. goremychkin , j. p. castellan , i. s. todorov , j. a. schlueter , h. claus , a. daoud - aladine , d. d. khalyavin , m. g. kanatzidis , and r. osborn , phys .
b * 85 * , 184507 ( 2012 ) .
t. shimojima , t. sonobe , w. malaeb , k. shinada , a. chainani , s. shin , t. yoshida , s. ideta , a. fujimori , h. kumigashira , k. ono , y. nakashima , h. anzai , m. arita , a. ino , h. namatame , m. taniguchi , m. nakajima , s. uchida , y. tomioka , t. ito , k. kihou , c. h. lee , a. iyo , h. eisaki , k. ohgushi , s. kasahara , t. terashima , h. ikeda , t. shibauchi , y. matsuda , and k. ishizaka , phys .
b * 89 * , 045101 ( 2014 ) . m. yi , d. h. lu , j. h. chu , j. g. analytis , a. p. sorini , a. f. kemper , b. moritz , s. k. mo , r. g. moore , m. hashimoto , w. s. lee , z. hussain , t. p. devereaux , i. r. fisher , and z. x. shen , proc .
sci , usa * 108 * , 6878 ( 2011 ) . | in this paper , we investigate the quasiparticle scattering interference(qpi ) in the nematic phase of iron pnictides , based on the magnetic and orbital scenarios of nematicity , respectively . in the spin density wave(sdw )
state , the qpi pattern exhibits a dimer structure in the energy region of the sdw gap , with its orientation along the ferromagnetic direction of the sdw order . when the energy is increased to be near the fermi level , it exhibits two sets of dimers along the same direction .
the dimer structure of the qpi patterns persists in the magnetically driven nematic phase , although the two dimers tend to merge together with energies closing to the fermi level .
while in the orbital scenario , the qpi patterns exhibit a dimer structure in a wide energy region .
it undergoes a @xmath0 rotation with the increasing of energy , which is associated with the inequivalent energies of the two dirac nodes induced by the orbital order .
these distinct features may be used to probe or distinguish two kinds of scenarios of the nematicity . |
in the past decades , the lattice boltzmann ( lb ) method , as a numerical approach originated from the lattice gas automata or developed from the simplified kinetic models , has gained great success in the simulation of the complex flows @xcite , and has also been extended to solve some partial differential equations , including the diffusion equation @xcite , wave equation @xcite , burgers equation @xcite , poisson equation @xcite , isotropic and anisotropic convection - diffusion equations ( cdes ) @xcite .
however , most of these available works related to cdes mainly focused on the linear cde for its important role in the study of the heat and mass transfer @xcite .
recently , through constructing a proper equilibrium distribution function , shi and guo @xcite proposed a single - relaxation - time ( srt ) lb model or the bhatnagar - gross - krook ( bgk ) model for nonlinear convection - diffusion equations ( ncdes ) where a nonlinear convection and/or diffusion term is included .
although the model has a second - order convergence rate in space @xcite , and can also be considered as an extension to some available works , it still has some limitations .
the first is that when the bgk model is used to solve the ncdes where the diffusion coefficient is very small , the relaxation parameter would be close to its limit value , thus the model may suffer from the stability or accuracy problem @xcite .
secondly , the bgk model is usually limited to solve the nonlinear isotropic cdes since it does not have sufficient relaxation parameters to describe the anisotropic diffusion process , while the multiple - relaxation - time ( mrt ) lb model with more relaxation parameters seems more suitable and more reasonable in solving ncde with anisotropic diffusion process . and finally , it is also well known that in the bgk model , only a single relaxation process is used to characterize the collision effects , which means all modes relax to their equilibria with the same rate , while from a physical point of view , these rates corresponding to different modes should be different from each other during the collision process @xcite . to overcome these deficiencies inherent in the bgk model , in this work we will present a general mrt lb model for the nonlinear anisotropic cde ( nacde ) , and show that , through the chapman - enskog analysis , the nacde with a source term can be recovered correctly from the proposed model . besides , it is also found that , similar to our previous bgk model @xcite , the present mrt model also has a second - order convergence rate in space , while it could be more stable and more accurate than the bgk model through tuning the relaxation parameters properly .
the rest of the paper is organized as follows . in section 2 , a general mrt model for nacde with a source term
is first presented , and then some special cases and distinct characteristics of the present model are also discussed . in section 3
, the accuracy and convergence rate of the mrt model are tested through some classic cdes , followed by a comparison between the present mrt model and the bgk model , and finally , some conclusions are given in section 4 .
the @xmath0-dimensional ( @xmath0-d ) nonlinear anisotropic convection - diffusion equation with a source term can be written as @xmath1+r(\mathbf{x } , t),\ ] ] where @xmath2 is a scalar variable and is a function of time and space , @xmath3 is the gradient operator , @xmath4 is the diffusion tensor . @xmath5 and @xmath6 are two differential tensor functions with respect to @xmath2 , @xmath7 is the source term . it should be noted that eq .
( 1 ) can be viewed as a general form of some important partial differential equations @xcite , such as the ( anisotropic ) diffusion equation , burgers equation , burgers - fisher equation , buckley - leverett equation , nonlinear heat conduction equation , ( anisotropic ) convection - diffusion equation , and so on .
generally , the models in the lb method can be classified into three kinds based on collision operator , i.e. , the lattice bgk model @xcite , the two relaxation - time model @xcite , and the mrt model @xcite . in this work
, we will focus on the mrt model for its superiority on the stability and accuracy both in the study of fluid flows @xcite and solving cdes @xcite .
the mrt model with d@xmath0q@xmath8 lattice ( @xmath8 is the number of discrete directions ) @xcite for the nacde [ eq . ( [ eq1 ] ) ] is considered here , and the evolution equation of the model can be written as @xmath9 \nonumber \\ & + & ( \delta t+\frac{\delta t^{2}}{2}\bar{d}_{k})r_{k}(\mathbf{x } , t),\end{aligned}\ ] ] where @xmath10 is time step , @xmath11 with @xmath12 and @xmath13 being two parameters to be determined in the following part .
@xmath14 is the distribution function associated with the discrete velocity @xmath15 at position @xmath16 and time @xmath17 , @xmath18 is the equilibrium distribution function , and is defined as @xcite @xmath19,\ ] ] where @xmath20 is the unit matrix , @xmath21 is a positive parameter related to the diffusion tensor @xmath22 [ see eq .
( [ eq26 ] ) ] , @xmath15 and @xmath23 are discrete velocity and weight coefficient , and in different lattice models , they can be defined as [ eq4 ]
@xmath24 @xmath25 [ eq5 ] @xmath26 @xmath27 [ eq6 ] @xmath28 @xmath29 where @xmath30 with @xmath31 representing lattice spacing , and usually it is not equal to 1 , @xmath32 . to recover the correct nacde from present mrt model , the second - order differential tensor @xmath33 in eq . ( [ eq3 ] )
should satisfy @xmath34 @xmath35 in eq .
( [ eq2 ] ) is a transformation matrix , and can be used to project the distribution function @xmath36 and equilibrium distribution function @xmath37 in the discrete velocity space onto macroscopic variables in the moment space through following relations @xcite , @xmath38 where @xmath39 , @xmath40 .
@xmath41 is the relaxation matrix , @xmath42 is the discrete source term , and is defined by @xmath43 where @xmath44 is a differential tensor to be determined below .
in addition , to derive correct nacde , i.e. , eq .
( [ eq1 ] ) , the following conditions also need to be satisfied , [ eq10 ] @xmath45 @xmath46 in this part , we will present the chapman - enskog analysis on how to derive the nacde from the present mrt model . in the chapman - enskog analysis , the distribution function ,
the time and space derivatives , and the source term can be expressed as @xcite [ eq11 ] @xmath47
@xmath48 where @xmath49 is small parameter . substituting eqs .
( [ eq11 ] ) into eq .
( [ eq2 ] ) and using the taylor expansion , we can obtain zero , first and second - order equations in @xmath49 , [ eq12 ] @xmath50 @xmath51 @xmath52 where @xmath53 , @xmath54 . if we rewrite eqs
. ( [ eq12 ] ) in a vector form , and multiply @xmath35 on both sides of them , we can obtain the corresponding equations in moment space , [ eq13 ] @xmath55 @xmath56 @xmath57 where @xmath58 , @xmath59 , @xmath60 , @xmath61 and @xmath62 with @xmath63
. if we take the d2q9 lattice model as an example , the transportation matrix @xmath35 can be written as @xmath64 where @xmath65 is a diagonal matrix and @xmath66 is given by @xcite @xmath67 consequently , one can also obtain @xmath68 from eq .
( [ eq8 ] ) , @xmath69 where @xmath70 , @xmath71 . for this lattice model ,
the relaxation matrix @xmath41 can be defined as @xmath72 where the diagonal element @xmath73 is the relaxation parameter corresponding to @xmath74th moment @xmath75 , and the off - diagonal components ( @xmath76 and @xmath77 ) correspond to the rotation of the principal axis of anisotropic diffusion @xcite . besides , based on eq .
( [ eq14 ] ) , one can easily obtain the following equation , @xmath78 and simultaneously , the evolution eq .
( [ eq2 ] ) can be rewritten as @xmath79 \nonumber \\ & + & ( \delta t+\frac{\delta t^{2}}{2}\bar{d}_{k})r_{k}(\mathbf{x } , t).\end{aligned}\ ] ] based on eq .
( [ eq13]b ) , we can rewrite the first - order equations in @xmath49 , but here we present the first , fourth , and sixth ones since only these three equations are useful in the following process of obtaining nacde , [ eq20 ] @xmath80
@xmath81 @xmath82 if we introduce a matrix @xmath83 and a vector @xmath84 , which are defined as @xmath85 then eqs . ( [ eq20]b ) and ( [ eq20]c ) can be rewritten in a vector form , @xmath86 similarly , we can also use eq .
( [ eq13]c ) to derive the second - order equations in @xmath49 , but only the first one corresponding to the conservative variable @xmath2 is presented , @xmath87 = \frac{\delta t}{2}(\theta-1)\partial_{t_1}r^{(1 ) } + \frac{\delta t}{2}[(\gamma-1)\nabla_{1}\cdot(\tilde{\mathbf{b}}r^{(1)})].\ ] ] substituting eq .
( [ eq22 ] ) into eq .
( [ eq23 ] ) , we have @xmath88\nonumber \\ & = & \frac{\delta t}{2}(\theta-1)\partial_{t_1}r^{(1 ) } + \frac{\delta t}{2}[(\gamma-1)\nabla_{1}\cdot(\tilde{\mathbf{b}}r^{(1)})],\end{aligned}\ ] ] with the aid of eqs .
( [ eq7 ] ) and ( [ eq20]a ) , we can rewrite eq .
( [ eq24 ] ) as @xmath89+\frac{\delta t}{2}(\theta-1)\partial_{t_1}r^{(1)}\nonumber \\ & + & \delta t\nabla_{1}\cdot\{[(\frac{\gamma}{2}\mathbf{i}-\mathbf{a}^{-1})\tilde{\mathbf{b}}-(\frac{1}{2}\mathbf{i}-\mathbf{a}^{-1})\mathbf{b}^{'}]r^{(1)}\},\end{aligned}\ ] ] where the diffusion tensor @xmath22 is given by @xmath90 from above equation , it is clear that , for a fixed diffusion tensor , the parameter @xmath21 can be used to adjust the relaxation parameters . through combining the results at @xmath91 and @xmath92 scales , i.e. , eqs .
( [ eq20]a ) and ( [ eq25 ] ) , we can recover the following nacde , @xmath93+r \nonumber \\
& + & \frac{\delta t}{2}(\theta-1)\partial_{t}r + \delta t\nabla\cdot\{[(\frac{\gamma}{2}\mathbf{i}-\mathbf{a}^{-1})\tilde{\mathbf{b}}-(\frac{1}{2}\mathbf{i}-\mathbf{a}^{-1})\mathbf{b}^{'}]r\}.\end{aligned}\ ] ] to derive correct nacde , the parameter @xmath12 and the tensor @xmath44 should be set as @xmath94 we noted that although above analysis is only carried out for the two - dimensional mrt model with d2q9 lattice , it can be extended to three - dimensional model without any substantial difficulties .
now we give some remarks on the present model .
* remark i : * we first want to present some discussion on the diffusion tensor @xmath22 . actually , if the diffusion tensor @xmath22 is taken by @xmath95 with @xmath96 being a constant or variable , the nacde [ eq . ( [ eq1 ] ) ] would be reduced to the ncde considered in the previous work @xcite , and can still be solved in the framework of the bgk model . however
, if @xmath22 is a diagonal matrix or full matrix where the element @xmath97 is a function of space , the mrt model rather than bgk model should be adopted .
we would also like to point out that for the special case where @xmath22 is a diagonal matrix , @xmath98 the relaxation matrix @xmath41 and the matrix @xmath83 are also diagonal matrices ( @xmath99 ) , then the tensor @xmath100 and the relation between the nonzero elements of @xmath22 and relaxation parameters can be written in simple forms , [ eq30 ] @xmath101 @xmath102 * remark ii : * based on the choice of the parameter @xmath13 , two _ special _ schemes of the present model can be obtained .
* scheme a * ( @xmath103 ) : @xmath104 .
for this special case , a time derivative is included in second term of the right hand side of the evolution equation [ see eq .
( [ eq2 ] ) ] , and thus we need to use the finite - difference method to compute the time derivative term @xmath105 . here for simplicity an explicit finite - difference scheme , i.e. , @xmath106/\delta t$ ] , is adopted .
although a little larger memory cost would be caused for this scheme , the collision process can still be conducted locally . * scheme b * ( @xmath107 ) : @xmath108 . under the present choice of the parameter @xmath13 , @xmath109 ,
both the time derivative and the space derivative are contained in the evolution equation .
although we can still use explicit finite - difference schemes to calculate time and the space derivatives @xmath110 , which is similar to the approach used in * scheme a * , the collision process can not be performed locally . to solve the problem ,
an implicit finite - difference scheme can be applied to compute @xmath111 , @xmath112 which not only can result in that the collision process can be implemented locally , but also cause the implementation of present model to be explicit . substituting eq .
( [ eq31 ] ) into the evolution equation , we can obtain @xmath113\nonumber \\
& + & \frac{\delta t}{2}[r_{k}(\mathbf{x}+\mathbf{c}_{k}\delta t , t+\delta t)+r_{k}(\mathbf{x } , t)].\end{aligned}\ ] ] to avoid the implicitness , a new variable is introduced @xcite , @xmath114 then one can rewrite evolution equation as @xmath115\nonumber \\ & + & \delta t[\mathbf{m}^{-1}(\mathbf{i}-\frac{\mathbf{s}}{2})\mathbf{m}]_{kj}r_{j}(\mathbf{x } , t).\end{aligned}\ ] ] which is the same as the evolution appeared in ref .
@xcite . based on the eqs .
( [ eq10]a ) and ( [ eq33 ] ) , the variable @xmath2 in * scheme b * can be calculated by @xmath116 here it should be noted that if the source term @xmath117 is a function of @xmath2 , usually one needs to use some other methods to solve the algebraic equation ( [ eq35 ] ) .
* remark iii : * we noted that there is still a limitation in the applications of above mrt model since it may be difficult or impossible to derive the function @xmath33 analytically [ see eq .
( [ eq7 ] ) ] .
following the idea presented in the work @xcite , however , one can solve the problem through adding a new source term @xmath118 in the evolution equation , i.e. , @xmath119 \nonumber \\ & + & ( \delta t+\frac{\delta t^{2}}{2}\bar{d}_{k})r_{k}(\mathbf{x},\;t ) + \delta t g_{k}(\mathbf{x},\;t ) , \end{aligned}\ ] ] where @xmath118 is defined as @xmath120 and meanwhile , the equilibrium distribution function @xmath18 can be simplified as @xmath121,\ ] ] which can be derived from eq .
( [ eq3 ] ) through setting @xmath122 . through the chapman - enskog analysis
, one can also find that eq .
( [ eq1 ] ) can be recovered correctly from eq .
( [ eq36 ] ) .
in addition , we would also like to point out that if @xmath123 , the d@xmath0q@xmath124 and d@xmath0q@xmath125 lattice models can also be used .
* remark vi : * although there are many lb models for cdes @xcite , most of them are limited to the linear cdes with isotropic diffusion @xcite , and what is more , some of them can not give correct cde @xcite .
actually in the past decade , some lb models for anisotropic cdes have also been developed @xcite , but usually they can only be used to solve anisotropic cdes where the convection term @xmath126 or diffusion term @xmath127 is a linear function of @xmath2 @xcite .
we also note that the lb model proposed by ginzburg @xcite can be used to solve the cdes with nonlinear convection and diffusion terms , but some additional assumptions have been adopted to recover the correct cde , as pointed out in ref . @xcite .
recently , shi and guo proposed a new bgk model for ncde @xcite , but the model is usually used to solve the isotropic ncde , and can not be directly applied to solve the nacde . from above discussion , however , it is clear that the present mrt model can be viewed as a general lb model for the nacde .
to test the accuracy and stability of present mrt model for nacdes , some classic examples , including the isotropic convection - diffusion equation with a constant velocity , burgers - fisher equation , buckley - leverett equation , and anisotropic convection - diffusion equations , will be considered in this section . in our simulations , the following global relative error ( @xmath128 ) defined by eq .
( [ eq39 ] ) is used to test the accuracy of the present mrt model , @xmath129 where the subscripts @xmath130 and @xmath0 denote the analytical and numerical solutions .
the distribution function @xmath36 is initialized by its equilibrium distribution function @xmath37 , i.e. , @xmath131 unless otherwise stated , the parameter @xmath21 appeared in the equilibrium distribution function is set to be 1.0 , the * scheme b * is adopted since the computation of the time derivative in * scheme a * can be avoided , and meanwhile , the non - equilibrium extrapolation scheme @xcite is adopted since it can be used to treat different boundary conditions and also has a second - order convergence rate in space .
in addition , it should be noted that , besides the relaxation parameters ( @xmath132 , @xmath76 , @xmath133 and @xmath77 ) related to diffusion tensor , the other relaxation parameters are simply taken as @xcite @xmath134 we first consider a simple two - dimensional isotropic cde with a constant velocity , which can be expressed as @xmath135 where @xmath136 and @xmath137 are constants , and are set to be 0.1 , @xmath96 is the diffusion coefficient .
@xmath117 is the source term , and is given by @xmath138\{\sin[\pi(x+y)]+\pi(u_{x}+u_{y})\cos[\pi(x+y)]\}.\ ] ] under the periodic boundary conditions adopted on the domain @xmath139\times[0 , 2]$ ] and the following initial condition , @xmath140,\ ] ] we can derive the analytical solution of the problem , @xmath141\sin[\pi(x+y)].\ ] ] when the present mrt model is used to study this problem , the function @xmath126 , @xmath33 , @xmath127 and the diffusion tensor @xmath22 should be given by @xmath142 with @xmath143 , @xmath144 , @xmath123 and @xmath95
. we now performed some simulations under different time and different pclet numbers ( pe=@xmath145 , @xmath146 is characteristic length ) , and presented the results in fig . 1 where the lattice size is @xmath147 .
as seen from the figure , the numerical results are in good agreement with analytical solutions .
profiles of scalar variable @xmath2 at different time and pclet numbers [ ( a ) : pe=100 , ( b ) : pe=1000 ; solid lines : analytical solutions , symbols : numerical results].,title="fig:",width=288 ] profiles of scalar variable @xmath2 at different time and pclet numbers [ ( a ) : pe=100 , ( b ) : pe=1000 ; solid lines : analytical solutions , symbols : numerical results].,title="fig:",width=288 ] besides , this problem is also applied to test the convergence rate of the present model since the boundary effect can be eliminated by the periodic boundary conditions adopted . to this end , we conducted a number of simulations and computed the @xmath128s under different lattice resolutions .
as shown in fig .
2 where lattice spacing is varied from @xmath148 to @xmath149 and numerical simulations are suspended at time @xmath150 , it is clear that present mrt model has a second - order convergence rate in space .
the global relative errors at different lattice sizes ( @xmath151 , @xmath152 , @xmath153 , @xmath154 and @xmath149 ) , the slope of the inserted line is 2.0 , which indicates the present mrt model has a second - order convergence rate in space.,width=288 ] the two - dimensional burgers - fisher equation , as a special case of the nacdes , can be written as @xcite @xmath155 where @xmath156 is the source term , @xmath157 , @xmath130 , @xmath158 and diffusion coefficient @xmath96 are constants . compared to the first problem considered above
, the present problem is more complicated since it is nonperiodic and nonlinear , but we can still obtain its analytical solution under the proper initial and boundary conditions , @xmath159\}^{\frac{1}{\delta}},\ ] ] where @xmath160 and @xmath161 are two parameters , and are defined by @xmath162 compared with the nacde defined by eq .
( [ eq1 ] ) , the function @xmath5 should be given by @xmath163 based on eq .
( [ eq7 ] ) , one can further derive the tensor @xmath33 in eq .
( [ eq3 ] ) , @xmath164 the diffusion tensor @xmath22 and tensor @xmath127 can be simply determined as @xmath165 we now consider how to use present mrt model to solve the burgers - fisher equation . if the * scheme b * is adopted to solve eq .
( [ eq46 ] ) , one needs to use some other methods to solve nonlinear equation ( [ eq33 ] ) since the source term @xmath117 is a nonlinear function of @xmath2 . to avoid such process , the * scheme a * is applied to solve the burgers - fisher equation .
we carried out some simulations in the computational domain @xmath166\times[-1 , 2]$ ] , and presented numerical results and corresponding analytical solutions under different time and different modified pclet numbers ( pe=@xmath167 , @xmath168 is the characteristic length ) in fig.3 where the parameters @xmath130 , @xmath158 , and lattice size are fixed to be 4.0 , 1.0 and @xmath169 . as seen from the figure
, the numerical results agree well with corresponding analytical solutions .
profiles of the scalar variable @xmath2 at different time and modified pclet numbers [ ( a ) : pe=100 , ( b ) : pe=1000 ; solid lines : analytical solutions , symbols : numerical results].,title="fig:",width=288 ] profiles of the scalar variable @xmath2 at different time and modified pclet numbers [ ( a ) : pe=100 , ( b ) : pe=1000 ; solid lines : analytical solutions , symbols : numerical results].,title="fig:",width=288 ] we note that this problem is nonperiodic and the boundary effect can not be excluded , thus it can be used to test the convergence rate of the mrt model coupling with the non - equilibrium extrapolation scheme . to this end
, we also carried out some simulations , and presented the @xmath128s in fig .
4 where the lattice spacing is ranged from @xmath170 to @xmath153 and simulations are suspended at time @xmath171 . as shown in this figure , the present mrt model coupling with non - equilibrium extrapolation scheme also has a second - order convergence rate in space . besides
, it is also found that the @xmath172 at pe=600 is less than those at pe=120 , this may be because the relaxation parameters ( @xmath132 and @xmath133 ) corresponding to the case of pe=600 are more close to 1 , which usually give more accurate results . the global relative errors at different lattice sizes ( @xmath173 , @xmath174 , @xmath175 , @xmath176 and @xmath153 ) , the slope of the inserted line is 2.0 , indicating that the present mrt model has a second - order convergence rate in space.,width=288 ] we also consider the two - dimensional buckley - leverett equation @xcite @xmath177 with the following initial condition , @xmath178 where @xmath96 is the diffusion coefficient .
@xmath179 and @xmath180 are a function of @xmath2 , and are defined as @xmath181.\ ] ] we note that , similar to the burgers - fisher equation considered previously , the buckley - leverett equation is also a special nacde , but there is no analytical solution to this problem .
when the present mrt model is used to solve the problem , the evolution equation [ eq . ( [ eq2 ] ) ] can be written more simply since there is no source term included in the the buckley - leverett equation , @xmath182,\ ] ] then from eqs .
( [ eq1 ] ) and ( [ eq7 ] ) , one can further determine the functions @xmath126 , @xmath33 , @xmath127 and diffusion tensor @xmath22 , @xmath183^{\top},\ \
\mathbf{c}=\left ( { \begin{array}{*{20}{c } } c_{xx } & c_{xy } \\ c_{yx } & c_{yy } \\
\end{array } } \right ) , \ \
\mathbf{d}=\phi\mathbf{i},\ \ \mathbf{k}=\kappa\mathbf{i},\ ] ] where the elements of the matrix @xmath33 are given as [ eq57 ] @xmath184 @xmath185 @xmath186\nonumber \\ & + & \frac{1}{16a}(-89 + 258\phi)+\frac{1}{16a^2}(13 + 14\phi)+\frac{1}{24a^3}(1 - 42\phi).\end{aligned}\ ] ] in our simulations , the computational domain of the problem and the lattice size are set to be @xmath187\times[-1.5 , 1.5]$ ] and @xmath169 .
we performed simulations at two different diffusion coefficients ( @xmath188 and @xmath189 ) or equivalently two different pclet numbers ( pe=@xmath190=30 and 300 , @xmath168 is the characteristic length , @xmath191 is the characteristic velocity ) , and presented the results at time @xmath192 in figs . 5 and 6 .
as expected , when the pe is smaller or the diffusion coefficient is larger , the role of the diffusion becomes dominant , and thus the distribution of scalar variable @xmath2 is more smooth [ see fig .
6(a ) ] . besides ,
to show more details on the distributions of scalar variable @xmath2 at different time , we also conducted some simulations , and presented the results of @xmath2 along the vertical centreline in fig .
as seen from the figure , with the increase of time , the distribution of the scalar variable with a small pe becomes more smooth .
contours of the scalar variable @xmath2 at the time @xmath192 and different pclet numbers [ ( a ) : pe=30 , ( b ) : pe=300].,title="fig:",width=288 ] contours of the scalar variable @xmath2 at the time @xmath192 and different pclet numbers [ ( a ) : pe=30 , ( b ) : pe=300].,title="fig:",width=288 ] distributions of the scalar variable @xmath2 at the time @xmath192 and different pclet numbers [ ( a ) : pe=30 , ( b ) : pe=300].,title="fig:",width=288 ] distributions of the scalar variable @xmath2 at the time @xmath192 and different pclet numbers [ ( a ) : pe=30 , ( b ) : pe=300].,title="fig:",width=288 ] profiles of the scalar variable @xmath2 at different time and pclet numbers [ ( a ) : pe=30 , ( b ) : pe=300].,title="fig:",width=288 ] profiles of the scalar variable @xmath2 at different time and pclet numbers [ ( a ) : pe=30 , ( b ) : pe=300].,title="fig:",width=288 ] we note that although the problem has no analytical solution , the present results [ see figs .
5(b ) and 6(b ) where @xmath193 agree well with those reported in some previous studies @xcite , which can also be used to conclude that the present mrt model is also accurate in solving the buckley - leverett equation .
we now consider the problem of the gaussian hill with constant velocity and diffusion tensor , which is also a classic benchmark example and has also been used to validate lb models for anisotropic cdes @xcite .
the cde for this problem can be written as @xmath194 where @xmath195 is a constant velocity , @xmath22 is the constant diffusion tensor , and can be defined as @xmath196 under the proper initial and boundary conditions , one can also derive the analytical solution of the problem , @xmath197}{2}\},\ ] ] where @xmath198 , @xmath199 , @xmath200 is inverse matrix of @xmath201 , @xmath202 is the absolute value of the determinant of @xmath201 . actually , there are two approaches that can be adopted to study the gaussian hill problem .
the first is that we directly use the mrt model to solve eq .
( [ eq58 ] ) with an anisotropic form , and set the functions @xmath126 , @xmath33 , and @xmath127 as @xmath203 while in the second approach , we first need to write eq .
( [ eq58 ] ) in an isotropic form , @xmath204.\ ] ] which is then solved by the mrt model .
in addition , it should be noted that eq .
( [ eq62 ] ) can also be solved by the previous bgk model @xcite .
based on eq .
( [ eq62 ] ) , one can also find that the functions @xmath126 and @xmath33 should be the same as those appeared in eq .
( [ eq61 ] ) , but the tensor @xmath127 should be given by @xmath205 with @xmath96 being a positive constant .
similar to some previous works @xcite , we also considered the gaussian hill problem in a bounded domain @xmath206\times[-1 , 1]$ ] , and adopted the periodic boundary condition on all boundaries . in our simulations , @xmath207 which is small enough to ensure that the periodic boundary condition adopted is reasonable and accurate at a finite time @xmath208 , @xmath209 , @xmath210 , and
the lattice size is @xmath211 . to test the capacity of the present mrt model in solving the anisotropic cdes , the following three types of diffusion tensor
are considered , @xmath212\times10^{-3},\ ] ] which are usually denoted as isotropic , diagonally anisotropic and fully anisotropic diffusion problems .
we conducted some simulations by using above two approaches , and presented numerical results in figs . 8 , 9 and 10 where @xmath213 is used in the second approach .
as seen from these figures , the numerical results qualitatively agree with analytical solutions . to quantitatively measure the deviations between numerical results and analytical solutions , the @xmath128s of isotropic , diagonally anisotropic and fully anisotropic diffusion problems
are computed , and the values of them are @xmath214 , @xmath215 and @xmath216 for the first approach , while they are @xmath214 , @xmath217 and @xmath218 for the second approach , which illustrate that the present mrt model is accurate in studying these problems . besides , the convergence rate of the mrt model for anisotropic cdes is also investigated , and the results are shown in fig .
11 where the lattice size is varied from @xmath147 to @xmath219 . from this figure
, one can find that , similar to some available mrt models for anisotropic diffusion problems @xcite , the present mrt model also has a second - order convergence rate in space .
distributions of the scalar variable @xmath2 at the time @xmath220 [ isotropic diffusion problem : numerical solution ( a ) , analytical solution ( b)].,title="fig:",width=288 ] distributions of the scalar variable @xmath2 at the time @xmath220 [ isotropic diffusion problem : numerical solution ( a ) , analytical solution ( b)].,title="fig:",width=288 ] distributions of the scalar variable @xmath2 at the time @xmath220 [ diagonally anisotropic diffusion problem : first approach ( a ) , second approach ( b ) , analytical solution ( c)].,title="fig:",width=288 ] distributions of the scalar variable @xmath2 at the time @xmath220 [ diagonally anisotropic diffusion problem : first approach ( a ) , second approach ( b ) , analytical solution ( c)].,title="fig:",width=288 ] distributions of the scalar variable @xmath2 at the time @xmath220 [ diagonally anisotropic diffusion problem : first approach ( a ) , second approach ( b ) , analytical solution ( c)].,title="fig:",width=288 ] distributions of the scalar variable @xmath2 at the time @xmath220 [ fully anisotropic diffusion problem : first approach ( a ) , second approach ( b ) , analytical solution ( c)].,title="fig:",width=288 ] distributions of the scalar variable @xmath2 at the time @xmath220 [ fully anisotropic diffusion problem : first approach ( a ) , second approach ( b ) , analytical solution ( c)].,title="fig:",width=288 ] distributions of the scalar variable @xmath2 at the time @xmath220 [ fully anisotropic diffusion problem : first approach ( a ) , second approach ( b ) , analytical solution ( c)].,title="fig:",width=288 ] the global relative errors at different lattice sizes ( @xmath221 , @xmath175 , @xmath148 , @xmath152 , @xmath153 and @xmath154 , @xmath146 is the characteristic length ) , the slope of the inserted line is 2.0 , which indicates that the present mrt model has a second - order convergence rate in space.,width=288 ] in this part , we will consider the following anisotropic cde with a constant velocity @xmath195 and a variable diffusion tensor @xmath22 , @xmath222 where @xmath117 is the source term .
we note that the problem is more complicated since the diffusion tensor @xmath22 is a function of space @xmath16 . for this reason , we can not write eq .
( [ eq64 ] ) in an isotropic form , and thus the problem can not be solved directly by the previous bgk model @xcite . in this work ,
the diffusion tensor @xmath22 is simply given by a diagonal matrix , @xmath223 where @xmath96 is a constant , and is fixed to be @xmath224 .
the source term @xmath117 is defined as @xmath225\{\sin(2\pi x)\sin(2\pi y)+4\kappa\pi^{2}\cos(4\pi x)\sin^{2}(2\pi y)\nonumber\\ & + & 2\pi[u_{x}\cos(2\pi x)\sin(2\pi y)+u_{y}\sin(2\pi x)\cos(2\pi y)]\}.\end{aligned}\ ] ] under the periodic boundary conditions on the physical region @xmath226\times[0 , 1]$ ] and the following initial condition , @xmath227 one can derive the exact solution of the problem , @xmath228\sin(2\pi x)\sin(2\pi y).\ ] ] based on eq .
( [ eq64 ] ) , one can determine the functions @xmath126 , @xmath33 and @xmath127 , which are the same as those appeared in eq .
( [ eq61 ] ) .
we now performed some simulations with a fixed lattice size @xmath211 , and presented the results at time @xmath150 and different pclet numbers ( pe=@xmath145 , @xmath229 is the characteristic length , @xmath230 is the characteristic velocity ) in figs . 12 and 13 where pe=100 and 1000 .
as seen from these figures , the numerical results are very close to the analytical solutions . to quantitatively measure the deviations between numerical results and corresponding analytical solutions
, we also computed the @xmath128s of these two cases , and found that the values of them are @xmath231 and @xmath232 , which are small enough and can be used to demonstrate that the present mrt model is accurate in the study of the anisotropic cde with a variable diffusion tensor .
distributions of the scalar variable @xmath2 at the time @xmath233 and pe=100 [ ( a ) : numerical solution , ( b ) : analytical solution].,title="fig:",width=288 ] distributions of the scalar variable @xmath2 at the time @xmath233 and pe=100 [ ( a ) : numerical solution , ( b ) : analytical solution].,title="fig:",width=288 ] distributions of the scalar variable @xmath2 at the time @xmath233 and pe=1000 [ ( a ) : numerical solution , ( b ) : analytical solution].,title="fig:",width=288 ] distributions of the scalar variable @xmath2 at the time @xmath233 and pe=1000 [ ( a ) : numerical solution , ( b ) : analytical solution].,title="fig:",width=288 ] to show the convergence rate of the present mrt for such complicated problem , we also carried out a number of simulations under different lattice resolutions , and presented the results in fig .
14 where the lattice size is varied from @xmath147 to @xmath219 .
as shown in this figure , the present mrt model also has a second - order convergence rate for this special problem . the global relative errors at different lattice sizes ( @xmath221 , @xmath175 , @xmath148 , @xmath152 , @xmath153 and @xmath154 ) , the slope of the inserted line is 2.0 , indicating that the present mrt model has a second - order convergence rate in space.,width=288 ]
as reported in some available works @xcite , through tuning the relaxation parameters properly , the mrt model could be more accurate and more stable than the bgk model . to show the superiority of the mrt model over the bgk model , a comparison between two models is also conducted .
we first performed a comparison of accuracy between the bgk model and mrt model through adopting a simple problem defined in a physical region @xmath234\times[0 , l]$ ] , which can be described by the following cde and boundary conditions , [ eq69 ] @xmath235 @xmath236 where @xmath96 is a constant diffusion coefficient , @xmath195 is a constant velocity with @xmath237 , @xmath238 and @xmath239 are two constants , @xmath240 is the source term with @xmath241 . under an assumption that the problem is steady and unidirectional , i.e. , @xmath2 is only a function of @xmath242 , we can derive analytical solution of the problem , @xmath243 the reason for choosing this problem is that , following a similar procedure reported in ref .
@xcite , one can readily derive the analytical solutions of the bgk model and mrt model ( * scheme b * with @xmath244 ) with adopting the anti - bounce back boundary condition @xcite , @xmath245 where @xmath246 with @xmath247 representing the grid number used in @xmath242 direction , @xmath248 is _ numerical _ slip caused by the model adopted , and can be given by [ eq72 ] @xmath249,\ ] ] @xmath250.\ ] ] based on eq .
( [ eq72]a ) , one can find that although the bgk model has a second - order convergence rate for this simple problem , the _ numerical _ slip of the scalar variable @xmath2 can not be eliminated unless @xmath251 . while for the mrt model , we can make the solution of mrt model [ eq .
( [ eq72]b ) ] consistent with that of the physical problem [ eq .
( [ eq70 ] ) ] through setting @xmath248 to be zero , which means that the relaxation parameters @xmath252 and @xmath132 should satisfy the following relation , @xmath253 from above theoretical analysis , it is clear that the mrt model can be more accurate than the bgk model through tuning the free relaxation parameter @xmath252 .
in addition , similar to the procedure used in above examples , the functions @xmath126 , @xmath33 and @xmath127 , the diffusion tensor @xmath22 , and the source term @xmath254 used in our model can also be determined , @xmath255 to validate above analysis and confirm our statement , we also performed some simulations with different lattice resolutions and relaxation parameters ( @xmath132 ) , and presented the results in fig .
15 where @xmath229 , @xmath256 , @xmath257 , @xmath258 , and the diffusion coefficient @xmath96 is set to be 0.1 . as seen from fig .
15 , the numerical results obtained by the mrt model are in good agreement with the analytical solution [ eq .
[ eq70 ] ] even with a coarse grid ( e.g. , @xmath259 ) , while the results given by the bgk model deviate from the analytical solution unless the relaxation parameter @xmath260 is fixed to be @xmath261 , which is consistent with above theoretical analysis . comparisons between the bgk model and mrt model under different lattice sizes and relaxation parameters [ ( a ) : @xmath262 , ( b ) : @xmath263 , ( c ) : @xmath261 , ( d ) : @xmath264.,title="fig:",width=288 ] comparisons between the bgk model and mrt model under different lattice sizes and relaxation parameters [ ( a ) : @xmath262 , ( b ) : @xmath263 , ( c ) : @xmath261 , ( d ) : @xmath264.,title="fig:",width=288 ] comparisons between the bgk model and mrt model under different lattice sizes and relaxation parameters [ ( a ) : @xmath262 , ( b ) : @xmath263 , ( c ) : @xmath261 , ( d ) : @xmath264.,title="fig:",width=288 ] comparisons between the bgk model and mrt model under different lattice sizes and relaxation parameters [ ( a ) : @xmath262 , ( b ) : @xmath263 , ( c ) : @xmath261 , ( d ) : @xmath264.,title="fig:",width=288 ] then a comparison of stability between the bgk model and mrt model is also carried out through using the gaussian hill problem , which has been investigated previously . here we only take the fully anisotropic diffusion problem as an example , while to simply perform a comparison between the bgk and mrt models , the second approach presented in section _ 3.4 _ [ the anisotropic cde is written in an isotropic form , see eq .
( [ eq62 ] ) ] is adopted . in the following simulations , the diffusion tensor @xmath22 is taken as @xmath265 from which one can further determine the function @xmath205 in the second approach [ see eq .
( [ eq61 ] ) ] .
@xmath96 is a constant , and is varied to test the stability of the bgk model and mrt model .
the lattice speed @xmath266 is equal to 1.0 , the velocity @xmath267 is fixed to be @xmath268 . for the relaxation parameters appeared in the mrt model , @xmath132 and @xmath133
can be determined from @xmath96 , and the others are fixed through the following equation , @xmath269 where @xmath270 can be varied in a proper range to ensure that the mrt model is stable , but for simplicity , only a special case @xmath271 is considered .
the other parameters used in simulations are the same as those appeared in section _ 3.4 _ except for time @xmath272 .
we first conducted simulations with @xmath213 , and presented the results in fig.16 .
as shown in this figure , both numerical results obtained by the bgk model and mrt model are close to the analytical solution . however , when @xmath96 is decreased to @xmath273 , the bgk model is unstable , while the mrt model can give a stable solution ( see fig .
17 ) , and the @xmath128 is about @xmath274 . a comparison between the bgk model and mrt model for fully anisotropic diffusion problem ( @xmath213 ) [ ( a ) : bgk model , ( b ) : mrt model , ( c ) : analytical solution].,title="fig:",width=288 ] a comparison between the bgk model and mrt model for fully anisotropic diffusion problem ( @xmath213 ) [ ( a ) : bgk model , ( b ) : mrt model , ( c ) : analytical solution].,title="fig:",width=288 ] a comparison between the bgk model and mrt model for fully anisotropic diffusion problem ( @xmath213 ) [ ( a ) : bgk model , ( b ) : mrt model , ( c ) : analytical solution].,title="fig:",width=288 ] a comparison between the bgk model and mrt model for fully anisotropic diffusion problem ( @xmath275 ) [ ( a ) : mrt model , ( b ) : analytical solution].,title="fig:",width=288 ] a comparison between the bgk model and mrt model for fully anisotropic diffusion problem ( @xmath275 ) [ ( a ) : mrt model , ( b ) : analytical solution].,title="fig:",width=288 ] from above discussion , it is clear that , through tuning the relaxation parameters properly , the mrt could be more accurate and more stable than the bgk model .
in this work , a general multiple - relaxation - time lattice boltzmann model for nonlinear anisotropic convection - diffusion equations is proposed , and is then tested by some classic nacdes , including the simple linear cde , nonlinear burgers - fisher equation , nonlinear buckley - leverett equation and some anisotropic cdes .
the numerical results show that the present mrt model is efficient and accurate in solving the nacdes , and also has a second - order convergence rate in space .
besides , we also conducted a comparison between the bgk model and mrt model , and found that the present mrt model could be more accurate and more stable than bgk model through tuning the relaxation parameters properly . and
finally , we would also like to point out that , based on the superiority of the mrt model and the role of the nacdes in describing the physical phenomena caused by the convection and diffusion processes , the present work may promote the mrt model in the study of heat and mass transfer @xcite , multiphase flows and crystal growth based on phase - field models @xcite .
the authors acknowledge support from the national natural science foundation of china ( grants no .
11272132 and no .
51125024 ) .
j. perko and r. a. patel , single - relaxation - time lattice boltzmann scheme for advection - diffusion problems with large diffusion - coefficient heterogentities and high - advection transport , phys .
e 89 ( 2014 ) 053309 .
i. ginzburg , truncation errors , exact and heuristic stability analysis of two - relaxation - times lattice boltzmann schemes for anisotropic advection - diffusion equation , commun .
phys . 11 ( 2012 ) 1439 - 1502 .
b. servan - camas , f. t .- c .
tsai , lattice boltzmann method with two relaxation times for advection - diffusion equation : third order analysis and stability analysis , adv .
water resour . 31
( 2008 ) 1113 - 1126 .
i. ginzburg , f. verhaeghe , and d. dhumires , two - relaxation - time lattice boltzmann scheme : about parametrization , velocity , pressure and mixed boundary conditions , commun .
phys . 3 ( 2008 ) 427 - 478 .
d. dhumires , generalized lattice - boltzmann equations , in : b.d .
shizgal , d.p .
weave ( eds . ) , rarefied gas dynamics : theory and simulations , in : prog .
159 , aiaa , washington , dc , 1992 , pp .
450 - 458 .
k. h. karlsen , k. brusdal , h. k. dahle , s. evje , and k .- a .
lie , the corrected operator splitting approach applied to a nonlinear advection - diffusion problem , comput .
methods appl .
engrg . 167
( 1998 ) 239 - 260 .
t. zhang , b. shi , z. guo , z. chai , and j. lu , general bounce - back scheme for concentration boundary condition in the lattice - boltzmann method , phys .
e 85 ( 2012 ) 016701 .
w. j. boettinger , j. a. warren , c. beckermann , and a. karma , phase - field simualtion of solidification , annu .
res . 32 ( 2002 )
163 - 194 . | in this paper , based on the previous work [ b. shi , z. guo , lattice boltzmann model for nonlinear convection - diffusion equations , phys .
rev .
e 79 ( 2009 ) 016701 ] , we develop a general multiple - relaxation - time ( mrt ) lattice boltzmann model for nonlinear anisotropic convection - diffusion equation ( nacde ) , and show that the nacde can be recovered correctly from the present model through the chapman - enskog analysis .
we then test the mrt model through some classic cdes , and find that the numerical results are in good agreement with analytical solutions or some available results .
besides , the numerical results also show that similar to the single - relaxation - time ( srt ) lattice boltzmann model or so - called bgk model , the present mrt model also has a second - order convergence rate in space .
finally , we also perform a comparative study on the accuracy and stability of the mrt model and bgk model by using two examples . in terms of the accuracy ,
both the theoretical analysis and numerical results show that a _ numerical _ slip on the boundary would be caused in the bgk model , and can not be eliminated unless the relaxation parameter is fixed to be a special value , while the _ numerical _ slip in the mrt model can be overcome once the relaxation parameters satisfy some constrains .
the results in terms of stability also demonstrate that the mrt model could be more stable than the bgk model through tuning the free relaxation parameters .
multiple - relaxation - time lattice boltzmann model , nonlinear anisotropic convection - diffusion equations , chapman - enskog analysis |
it is well known that selective resummations can play an important role in quantum field theory .
a prominent example is the hard thermal loop theory , which includes screening effects that regulate infra - red divergences .
the convergence of such perturbation theories can typically be improved using an @xmath0-particle irreducible ( @xmath0pi ) effective action .
in addition , @xmath0pi approximation schemes can be used to study far from equilibrium systems .
although the @xmath0pi effective action is consistent with the global symmetries of the theory , symmetry identities for @xmath0-point functions may not be satisfied during intermediate steps of the calculation @xcite . for scalar theories , goldstone s theorem may not be satisfied , and for gauge theories , ward identities can be violated . to address this problem we look at the resummed @xmath0pi effective action , which is defined with respect to the self consistent solutions of the @xmath0-point functions .
this type of strategy was originally proposed by baym and kadanoff @xcite .
the resummed effective action respects all symmetry properties of the theory , and the @xmath0-point functions which are obtained by functional differentiation satisfy goldstone s theorem , or the ward identities .
transport coefficients can be calculated from the kubo formulae .
conductivity and shear viscosity are obtained from the zero frequency limit of the 2-point function .
this limit produces pinch singularities , which come from pairs of retarded and advanced propagators which carry the same momenta .
when integrating a term of the form @xmath1 , the integration contour is ` pinched ' between poles on each side of the real axis , and the integral contains a divergence called a ` pinch singularity . '
these divergences are regulated by using resummed propagators which account for the finite width of thermal excitations .
this procedure introduces extra factors of the coupling in the denominators which change the power counting . as a consequence , in even the leading order ( lo )
calculation , there is an infinite set of graphs which contain products of pinching pairs that need to be resummed .
this is accomplished by solving an integral equation whose kernel is the square of the matrix elements that correspond to the 2 @xmath2 2 scattering and production processes . in gauge theories , in addition to pinch singularities , the presence of collinear singularities makes 2 @xmath2 3 scatterings as important as 2 @xmath2 2 scatterings .
these collinear terms are resummed by another integral equation .
the complete lo result is obtained by solving the two coupled integral equations .
the integral equations that resum pinch and collinear singularities can be obtained from the equations of motion ( eom s ) of the 3-loop 3pi effective action @xcite .
gauge invariance is automatically satisfied , and all lo terms appear without the need for any kind of power counting arguments .
little progress has been made on the calculation of transport coefficients beyond lo ( see however @xcite ) .
since power counting is notoriously difficult , and becomes increasingly complicated at higher orders , @xmath0pi effective theories could provide a useful method to organise the calculation of transport coefficients at next - to - leading order ( nlo ) . in spite of the complexity of the @xmath0pi effective action , a systematic expansion can be done in a self - consistent way @xcite . using the 4-loop 4pi effective action for a scalar theory with cubic and quartic interactions , with non - vanishing field expectation value
, one can derive the integral equations that would produce the nlo order contribution to the viscosity @xcite .
we remark that the numerical solution of these equations is more difficult than anything that has been accomplished so far .
in addition , renormalizability has not been demonstrated beyond the level of the 2pi effective action @xcite .
we use a compactified notation in which a single numerical subscript represents all space - time co - ordinates .
for example : the field expectation value is written @xmath3 , the variational propagator is written @xmath4 , the bare 4-point vertex is written @xmath5 , etc .
we also use an einstein convention in which a repeated index implies an integration over space - time variables . using this notation
we write the classical action : @xmath6=\frac{1}{2}\phi_1\big[i\,(d^{0}_{12})^{-1}\big]\phi_2-\frac{i}{\;3 ! } u_{123}^{oo}\phi_1\phi_2\phi_3-\frac{i}{\;4!}v_{1234}^0\phi_1\phi_2\phi_3\phi_4\,.\end{aligned}\ ] ] the 4pi effective action has the form : @xmath7 = s_{cl}[\phi]+ \frac{i}{2 } { \rm tr } \,{\rm ln}d^{-1}_{12}\\ & & + \frac{i}{2 } { \rm tr}\left[(d^0_{12}(\phi))^{-1}\left(d_{21}-d^0_{21}(\phi)\right)\right]-i\phi^0[\phi , d , u , v]-i\phi^{\rm int}[d , u , v ] \ , .
\nonumber\end{aligned}\ ] ] the terms @xmath8 $ ] and @xmath9 $ ] contain all contributions to the effective action which have two or more loops .
the first piece @xmath8 $ ] includes all terms that contain bare vertices .
the diagrams up to 4-loops are shown in fig .
[ phi ] .
4-loop 4pi effective action.,width=566 ] the re - summed propagator and self energy are defined as : @xmath16=i\big[(d^{0}_{12}(\phi))^{-1}- \pi^{\rm ext}_{12}\big]\,.\end{aligned}\ ] ] we can derive an expression for the re - summed 2-point function by taking derivatives of the resummed effective action and using the chain rule .
the result has the same form as the schwinger - dyson ( sd ) equation and depends on the 3- and 4-point vertices defined in ( [ vert - defns ] ) .
we also define a 5-point vertex that we will need below .
@xmath17 the vertices in equation ( [ vert - defns ] ) satisfy integral equations obtained by taking functional derivatives with respect to the field expectation value of the appropriate eom ( [ eom1 ] ) .
these integral equations are shown in fig .
[ ompsint ] .
one of the kernels is shown in fig .
the integral equations for the vertices in ( [ vert - defns ] ) depend on the vertices @xmath18 and @xmath19 , which in turn satisfy their own integral equations , that are obtained directly from ( [ eom1 ] ) and shown in fig .
[ uint ] . in all figures
we combine diagrams that correspond to permutations of external legs .
the full nlo contribution to the shear viscosity can be obtained from the kubo formula where the 2-point function is given by the sd equation with vertices obtained by solving the set of coupled integral equations shown in figs .
[ ompsint ] and [ uint ] . the integral equations for the vertices @xmath20 , @xmath21 and @xmath22 defined in equation ( [ vert - defns ] ) .
the kernels are drawn as grey boxes.,width=604 ] one of the kernels in fig .
[ ompsint].,width=453 ] integral equations for the variational vertices @xmath18 and @xmath19 from the 4-loop 4pi effective action.,width=604 ]
it is interesting to consider how the calculation would proceed at higher orders .
the @xmath0pi effective action is defined as the @xmath0th legendre transform of the generating functional obtained by coupling to @xmath0 source terms .
a direct calculation beyond the 4-loop 4pi level is extremely difficult @xcite , but it is possible to calculate the effective action without taking any legendre transforms @xcite .
the key to the method is the introduction of a set of fictitious bare vertices : to obtain the @xmath0-loop @xmath0pi effective action we include in the lagrangian the vertices @xmath23 for @xmath24 .
the inclusion of the non - renormalizable interactions ( @xmath25 ) is an organizational trick , and these vertices are set to zero at the end of the calculation . from this point on , we consider @xmath26 for simplicity , and use the notation @xmath27 and @xmath28 to represent bare and proper vertices with @xmath29 legs
. using these fictitious vertices , we can show that the eom s and sd equations are equivalent to the order at which the truncated theory respects the symmetries of the original theory.pi effective action . ]
this result allows us to construct the @xmath0-loop @xmath0pi effective action directly from the sd equations .
the method has been used to reproduce , with comparatively little effort , the known results for the @xmath0-loop @xmath0pi effective action with @xmath30 and @xmath31 . the technique has also been used to calculate the 6-loop 6pi effective action which is , realistically speaking , impossibly tedious to obtain using legendre transforms
. item 1 : : if we convert a set of skeleton diagram for the vertex @xmath28 into a series of perturbative diagrams using ( [ eom2 ] ) , the leading loop order of the new set of diagrams is greater than or equal to the leading loop order of the original set .
item 2 : : if we include fictitious vertices @xmath38 for @xmath39 , we can convert skeleton diagrams to perturbative diagrams using ( [ eom2 ] ) , or perturbative diagrams to skeleton diagrams using ( [ subbereom ] ) , and the leading loop order of the new set of diagrams is equal to the leading loop order of the original set .
we illustrate these statements with an example .
we use @xmath40 to indicate the loop order of the perturbative expansion . consider the skeleton diagram shown in part @xmath41 of fig .
[ ptexample ] , which is of order @xmath42 .
we can expand this diagram as a series of perturbative diagrams using equations of the form ( [ eom2 ] ) which are shown for this example in part @xmath43 of the figure . also have to be expanded to obtain a perturbative diagram .
this will produce extra loops that correspond to self energy corrections . in this paper we do not introduce notation to distinguish skeleton and perturbative propagators in diagrams . ] the lo term is shown in part @xmath44 , and is of order @xmath45 .
thus we have @xmath46 .
now consider the result if we set @xmath47 , which means we remove the first diagram on the right side of part @xmath48 . in this case
the lo term is shown in part @xmath49 and is of order @xmath50 .
thus we see that if the fictitious vertex @xmath51 is set to zero we have @xmath52 .
diagrams used to explain items 1 and 2 in section [ newapproachsection].,width=415 ] we consider truncating the @xmath0pi effective action at @xmath53-loop order .- loops , the @xmath0pi effective action is the same as the @xmath54pi effective action for @xmath55 @xcite . ] the functional derivative of an @xmath53-loop graph with respect to the variational vertex @xmath28 opens @xmath56 loops .
this means that an arbitrary @xmath53-loop graph in the effective action which contains the vertex @xmath28 produces a term with @xmath57 $ ] loops in the skeleton expansions of the eom for the vertex @xmath28 , where we define : @xmath58 : = m - j+1\,.\end{aligned}\ ] ] now we consider the schwinger - dyson equations , which form an infinite hierarchy of coupled non - linear integral equations .
they have the form : @xmath59\,.\end{aligned}\ ] ] although the structure of the sd equations is very different from the eom s , it is possible to show @xcite : item 3 : : the perturbative expansions of the @xmath0-loop @xmath0pi eom s and the sd equations truncated by setting @xmath60 for @xmath61 both match the perturbative expansion obtained from the 1pi effective action , and therefore each other , to order @xmath62 $ ] .
this result is perhaps not surprising , since both formalisms should contain the correct perturbative physics up to the order at which the calculation is done .
item 3 tells us that both the variational and sd vertex functions respect crossing symmetry to order @xmath62 $ ] , which we refer to as the `` truncation order . '' we can formally write equation ( [ sdgeneral ] ) as : @xmath63 + i_j[v_l^0,v_k^{sd}]\,,\\[2 mm ] & & i_j[v_l^0,v_k^{sd}]:= { \rm fcn}_j^{sd}[v_l^0,v_k^{sd}]-{\rm fcn}^{\prime}_j[v_l^0,v_k^{sd}]\,.\nonumber\end{aligned}\ ] ] comparing ( [ eom2 ] ) and ( [ newsd ] ) and using item 3 , it is clear that @xmath34 $ ] and @xmath64 $ ] must match each other in the perturbative expansion to order @xmath65 $ ] .
therefore we can rewrite @xmath64 $ ] as : @xmath66={\rm fcn}_j[v_k^{sd}]~+ ~{\rm extra}\,,\end{aligned}\ ] ] where the extra term is of order @xmath65 + 1 $ ] .
using item 2 _ which is only true in the presence of fictitious bare vertices _ , the extra term can be rewritten as a series of skeleton diagrams of order @xmath67 + 1 $ ] .
thus we have shown that : item 4 : : _ in the presence of fictitious bare vertices _ the sd equations can be rearranged to have the same form as the @xmath0pi eom s , plus additional terms of order @xmath68 + 1 $ ] in the skeleton expansion , which is beyond the truncation order .
it is straightforward to use the result in item 4 to calculate the @xmath0-loop @xmath0pi effective action from the sd equations .
first use ( [ subbereom ] ) to remove bare vertices in @xmath69 to order @xmath70 $ ] for @xmath71 .
then join the legs of the resulting equation for @xmath69 , set @xmath72 for @xmath73 , and multiply each diagram by the symmetry factor : @xmath74=s(1/v_j)(1/j!)\,,\end{aligned}\ ] ] where @xmath75 is the numerical factor in front of the diagram in the @xmath69 , and @xmath76 is the number of times the vertex @xmath28 appears in the diagram .
it is clear that joining the legs in diagrams in the eom s will produce all graphs in the effective action at a given order .
the trick is to obtain the correct symmetry factor .
consider starting from a known result for the effective action @xmath77 , taking derivatives of each graph with respect to each variational vertex , and trying to reconstruct the effective action by joining the legs in each of these eom s .
a given graph in the effective action will produce contributions to the eom s of each vertex it contains . in order to produce the correct symmetry factor when joining legs , we must drop the corresponding contribution in all but one eom , which we can take to be the eom for the largest vertex present .
this is accomplished by imposing the condition @xmath72 for @xmath73 .
if the largest vertex appears in a given diagram in the effective action more than once , the graph that is produced by joining the legs will have a symmetry factor that is too large by a factor equal to the number of times the vertex appears .
the correct symmetry factor is produced by including the factor @xmath78 in equation ( [ cals ] ) .
following the simple procedure described above , one can reproduce the known results for the 4-loop 4pi and 5-loop 5pi effective actions .
the method has also been used to calculate the 6-loop 6pi effective action @xcite .
the lo calculation of transport coefficients involves infinite resummations of diagrams that can be identified by power counting arguments .
the integral equations that resum these diagrams are produced directly from the equations of motion of the 3-loop 3pi effective action .
this result leads to the conclusion that @xmath0pi effective theories provide a natural framework for the calculation of transport coefficients beyond leading order .
the calculation of next - to - leading order shear viscosity in scalar theory can be formulated using the 4-loop 4pi effective action .
higher order effective actions can be calculated in a efficient way by working with the schwinger - dyson equations . 99 a. arrizabalaga and j. smit , phys .
rev . * d66 * , 065014 ( 2002 ) .
carrington , g. kunstatter and h. zaraket , eur .
j. * c42 * , 253 ( 2005 ) .
g. baym and l. kadanoff , phys .
124 , 287 ( 1961 ) ; g. baym , phys . rev .
127 , 1391 ( 1962 ) .
carrington and e. kovalchuk , phys . rev .
* d76 * , 045019 ( 2007 ) .
carrington and e. kovalchuk , phys . rev .
* d77 * , 025015 ( 2008 ) .
carrington and e. kovalchuk , phys .
d 80 * , 085013 ( 2009 ) .
moore , phys . rev . *
d76 * , 107702 ( 2007 ) .
s. caron - huot and g.d .
moore , phys .
100 * , 052301 ( 2008 ) .
s. caron - huot and g.d .
moore , jhep * 0802 * , 081 ( 2008 ) .
j. berges , phys . rev . *
d70 * , 105010 ( 2004 ) .
carrington and e. kovalchuk , phys .
rev . * d81 * , 065017 ( 2010 ) .
j. berges , s. borsanyi , u. reinosa and j. serreau , annals phys . *
320 * , 344 ( 2005 ) .
u. reinosa and j. serreau , jhep * 0607 * , 028 ( 2006 ) .
carrington and yun guo , phys . rev . * d83 * , 016006 ( 2011 ) .
carrington and yun guo - _ arxiv:1010.2978_. | transport coefficients can be obtained from 2-point correlators using the kubo formulae .
it has been shown that the full leading order result for electrical conductivity and ( qcd ) shear viscosity is contained in the re - summed 2-point function that is obtained from the 3-loop 3pi effective action .
the theory produces all leading order contributions without the necessity for power counting , and in this sense it provides a natural framework for the calculation and suggests that one can calculate the next - to - leading contribution to transport coefficients from the 4-loop 4pi effective action .
the integral equations have been derived for shear viscosity for a scalar theory with cubic and quartic interactions , with a non - vanishing field expectation value .
we review these results , and explain how the calculation could be done at higher orders . |
traditionally , flux - limited selection has been used to chart galaxy evolution .
the measurement of quasar absorption lines allows an independent approach to studying the history of galaxies .
the highest hydrogen column density absorbers seen in the spectra of background qsos the damped lyman-@xmath0 systems ( dlas ) contain most of the neutral gas in the universe at @xmath1 ( ref . 1 ) , although their exact relation to galaxies is not well established .
the global history of star formation can be inferred from the evolution in the co - moving density of neutral gas ( predominantly in dlas ) as it is consumed in star formation ( fig .
1a & refs .
the average star formation rate in each dla depends then on their space density .
one school of thought has @xmath7 dlas being thick gaseous disks , the progenitors of massive spirals ( e.g. , ref . 4 ) .
alternatively , dlas could be more numerous gas - rich dwarfs , potentially sub - galactic building blocks ( e.g. , fig .
1b and refs . 5 , 6&7 ) . hence , to determine the nature of the dlas requires a measurement of their star formation rate
this is most directly done through the hydrogen recombination lines ( e.g. , ref .
8) , which should have luminosity proportional to the photo - ionizing flux produced by the most massive and shortest lived ob stars ( and therefore trace the near - instantaneous star formation rate ) .
however , the largely unsuccessful searches for [email protected] nm emission do not provide useful limits as this line is resonantly scattered and hence is selectively suppressed by dust absorption , which is difficult or impossible to quantify . moving to the balmer lines [email protected] nm and [email protected] nm greatly reduces the problem of dust extinction , enabling the true star formation rates to be more accurately measured ( e.g. , ref .
9 ) . for interesting redshifts where large samples of dlas exist ( @xmath1 )
, these lines are redshifted into the near - infrared pass - bands .
( -300,143)*(a ) * ( -130,142)*(b ) *
typically , we do not know where the star forming regions in the galaxy associated with the dla actually are all we know is that the sight - line to the background qso passes through a large column of neutral hydrogen .
the difficulties of point spread function subtraction of the qso mean that broad - band imaging is unlikely to reveal the foreground galaxy in emission at small impact parameters ( separations from the qso sight line ) . however , spectroscopy is more effective for line emission searches because it is more sensitive and any emission line from the foreground dla should be readily apparent against the continuum of the background qso .
recent near - infrared @xmath9-band long slit spectroscopy ( ref .
10 ) with cgs4 on the ukirt 3.8 m telescope failed to detect h@xmath0 from star formation in 8 dlas at @xmath10 . however , the long - slit approach may be ineffective , because of the high probability of any star forming region in the dla galaxy falling outside the narrow slit .
if dlas really are @xmath11 galaxies , then they could extend over tens of kpc ( several arcsec ) . clearly , a larger solid angle needs to be surveyed .
this will be achieved with the new generation of integral field units such as the cambridge infrared panoramic survey spectrograph ( cirpass , ref .
11 ) , which can cover @xmath12 with its 499 fibres .
although the gaseous disks may be large , the star - forming h ii regions are likely to be compact ( as in local spirals
e.g. , fig .
the small size of star forming regions is consistent with observations of high - redshift galaxies in the rest - uv ( fig .
2b & ref .
12 ) , and will be well - matched to the 0.25arcsec cirpass fibres ( @xmath131kpc across ) . a 3-hour integration on one of the 8 m gemini telescopes should achieve a 10@xmath14 sensitivity between the sky lines of @xmath15 at @xmath16 m ( @xmath5-band ) for a compact source , conservatively assuming that 70% of the line emission from an h ii region will fall within 4 fibres with seeing effects . for a cosmology with @xmath17 , @xmath18 and @xmath19 ,
then the star formation rates reached are @xmath20 for h@xmath0 at @xmath21 and @xmath22 for h@xmath8 at @xmath23 ( see ref .
such star formation rates are well below @xmath6 for the high-@xmath2 lyman - break population ( e.g , refs .
9&13 ) and are comparable locally to the luminous giant h ii complexes in m101 .
it appears that instruments such as cirpass on gemini will have both the sensitivity and the survey area to measure star formation rates in @xmath1 dlas .
coupled with studies of the evolution of neutral hydrogen ( ref . 2 ) and metal enrichment ( ref .
14 ) , these observations will probe the nature of damped lyman-@xmath0 systems and address their relation to galaxies . | we assess the feasibility of detecting star formation in damped lyman-@xmath0 systems ( dlas ) at @xmath1 through near - infrared spectroscopy using the forthcoming integral field units on 8m - class telescopes .
although their relation to galaxies is not well established , high-@xmath2 dlas contain most of the neutral gas in the universe , and this reservoir is depleted with time presumably through star formation .
line emission should be an indicator of star formation activity , but searches based on lyman-@xmath0 are unreliable because of the selective extinction of this resonant uv line . using more robust lines such as h@xmath0 forces a move to the near - infrared at @xmath1 . for line emission searches , spectroscopy is more sensitive than imaging , but previous long - slit spectroscopic searches have been hampered by the likelihood that any star forming region in the dla galaxy disk would fall outside the narrow slit .
the new integral field units such as cirpass on gemini will cover sufficient solid angles to intercept these , even in the extreme case of large galactic disks at high redshift . on an 8m - class telescope
, star formation rates of @xmath3 will be reached at @xmath4 with h@xmath0 in the @xmath5-band .
such star formation rates are well below @xmath6 for the high-@xmath2 lyman - break population , and are comparable locally to the luminous giant h ii complexes in m101 .
it appears that instruments such as cirpass on gemini will have both the sensitivity and the survey area to measure star formation rates in @xmath1 dlas .
these observations will probe the nature of damped lyman-@xmath0 systems and address their relation to galaxies . |
in a theoretical analysis of impurity signatures , principally important is the fact that the equilibrium between the ( nearest neighbors ) lattice atoms , and the implanted impurity atom represents the structural balance between local geometry and chemical bonding , which depends strongly on the interatomic distances . in order to explore the dynamical nature of substituent effects on small
length scales one can employ an approach based on a detailed atomistic view : it can be straightforwardly implemented within ab - initio numerical simulations , which are able to calculate the relative ionic displacements and the charge density redistribution accompanying the effect of the substitution .
the results of calculations will then allow us both to capture the final binding geometry around the impurity site and to answer the fundamental question as to what extent the impurity atom is capable of adapting itself to the local environment and the existing chemical bonds . in this paper
, we apply such integrative approach to investigate the effects of substitutions for several @xmath2 transition - metal impurities such as divalent @xmath3 , @xmath4 and tetravalent @xmath5 in magnesium diboride @xmath0 .
we consider @xmath0 containing substitutions of @xmath6 as a doped model system for which we perform a series of the relevant numerical simulations .
there are two reasons that make this system attractive for detailed study .
first , while different properties of magnesium diboride doped by a variety of impurity atoms have been extensively studied in the past decade , the investigation of impurities in this material is still a topical subject , raising a number of actual issues that are worthy of special attention ( for details see @xcite , and the references cited therein ) .
the other reason is that knowledge on how the arrangement of a single impurity ion influences on a local structural order may have useful predictive value in the rational design of magnesium diboride solid solutions .
one of those issues , which is especially important to all aspects of local geometry and stability , is the influence of substitutions on a principal atomic configuration and the bonding properties of the cation environment .
however , in many cases , information on the bonding situation around a substitutional impurity turns out to be limited , and there remain several questions about binding of metal elements incorporated into the place of the host @xmath7 cation . understanding the latter is very essential because the response of the system on doping is not clear _ a priori_. for example , it was experimentally established that the suppression of the superconducting properties proceeds much slower in the @xmath8- , @xmath9- and @xmath10-doped systems , as compared to compounds that contain other dopant ions of the same formal valences ( details can be found in @xcite ) . in table 1 , using the data available from the literature , we have summarized some examples related to this observation .
@rcccccccc @xmath11 & @xmath12 & @xmath13 & @xmath14 & @xmath15 & @xmath16 & @xmath17 & @xmath18 & @xmath19 + @xmath20 @xmath21 & @xmath22@xmath23 & @xmath24@xmath25 & @xmath26@xmath27 & @xmath24@xmath28 & @xmath29@xmath30 & @xmath31@xmath32 & @xmath33@xmath32 & @xmath34@xmath35 + & @xmath26@xmath36 & @xmath37@xmath38 & @xmath39@xmath40 & @xmath41@xmath28 & @xmath42@xmath32 & @xmath43@xmath32 & @xmath44@xmath32 & @xmath45@xmath35 + & @xmath46@xmath36 & & @xmath47@xmath40 & @xmath48@xmath28 & @xmath49@xmath32 & & & @xmath50@xmath35 + + @xmath23@xcite ; @xmath36@xcite ; @xmath25@xcite ; @xmath38@xcite ; @xmath27@xcite ; @xmath40@xcite ; @xmath28@xcite ; @xmath30@xcite ; @xmath32@xcite ; @xmath35@xcite from the rate of decrease of the superconducting temperature @xmath51 with chemical substitution of the original @xmath52 with metal elements in @xmath0 one can expect that not all the impurity properties are matched well with those of @xmath6 .
thus , the given example of table 1 shows that conceptually @xmath8- , @xmath9- and @xmath10-containing solutions stand among other doped systems as a particular example . this feature prompted us to examine in more detail the role of these substitutional ions and ion - specific effects in the context of impurity - induced changes in local structure , dynamics and chemical bonding .
therefore , the aim of the present paper is by employing ab initio calculations to study variations of certain lattice and charge characteristics , such as lattice distortions , interatomic forces , charge densities and bonding features , caused by compositional variations in @xmath0 . by comparative analysis of features of the impurity atom with those related to the host one we pay attention in our work to what is happening within an ionic environment in the elementary cell when an intensity of interionic and electronic correlations is affected by the replacements of a magnesium cation .
according to the literature data ( e.g. , @xcite ) , a crystalline structure of the bulk @xmath0 consists of alternating honeycomb layers of @xmath1 interleaved with closed packed layers of @xmath6 .
each @xmath6 atom is located at the center of a hexagon formed by boron ; by donating two valence electrons it becomes a strongly charged ion with a charge very close to @xmath53 .
consensus on the nature of the superconductivity in @xmath0 is based on a generic multiband model which employs the intra- and interband pair - scattering mechanisms .
the presence of the interband channel connecting the @xmath54 and the @xmath55 electronic bands @xcite gives rise to a unique property of @xmath0 - two - gap superconductivity , which is characterized by two gaps of different sizes that are simultaneously closed at the same temperature @xmath51 of the superconducting transition .
the first - principles modeling of the effects of @xmath56 ion substitutions in the @xmath57 compositions , where @xmath58 stands for a transition metal cation such as @xmath8 , @xmath9 or @xmath10 , was performed using the vienna ab - initio simulation package ( vasp ) @xcite in conjunction with a plane - wave - basis code and the projector - augmented wave ( paw ) method @xcite .
exchange - correlation effects were described through the perdew - burke - ernzerhof ( pbe ) framework @xcite . for geometry optimization ,
one of the host atoms was replaced by one @xmath58 element in a @xmath59-atoms supercell ( @xmath60 ) .
such arrangement shown in figure 1 corresponds to @xmath61 in terms of the formal stoichiometry of the doped system .
the lattice distortion caused by the substitution has the following features : ( 1 ) it is totally symmetric on the cation site , and ( 2 ) it involves displacements of @xmath62 nearest - neighbor boron atoms from regular positions . for a central boron atom , these displacements
lead in turn to local symmetry breaking .
full relaxation of atomic positions was made with periodic boundary conditions and theoretical ( structurally relaxed ) bulk lattice parameters .
the convergence criteria for total energy and residual forces calculations were set to be within @xmath63 @xmath64 and @xmath63 ev / , respectively .
a cut - off energy of @xmath65 ev and a monkhorst - pack @xmath66-point grid @xcite of @xmath67 has been employed .
this computational scheme was successfully tested by a comparison of the calculated values of the lattice constants , @xmath68 ; @xmath69 ; @xmath70 , with the experimental data @xcite , @xmath71 ; @xmath72 ; @xmath73 ( the deviations are @xmath74 and @xmath75 , respectively ) .
changes in chemical bonding upon incorporation of a dopant atom into a @xmath52-site was studied in terms of charge - density distributions . in order to characterize variations of the partial charges at the magnesium position upon substitution and to assign a specific meaning of ionic charges in the analysis , values of the bader effective charges @xcite have been evaluated .
their calculations were carried out by using software @xcite , which utilizes a grid - based bader analysis algorithm @xcite .
the contribution of local many - body polarization effects was analyzed in terms of the born effective charges , values of which have been obtained by using the special extended method adopted at vasp .
in tables 2 to 5 , we present the main results which characterize the influence of the @xmath12 , @xmath13 and @xmath14 substitutional impurities on displacements of the boron atoms , the electronic charge redistribution around a cation , and the local properties of chemical bonding in @xmath76 .
these results and several consequences of magnesium replacement are discussed below .
the main conclusions are summarized at the end of the paper .
@lccccc & & & & + host / imp . &
b out & b in & along @xmath77-axis & in @xmath78-@xmath79 plane + cation & ( ) & ( ) & ( ev / ) & ( ev / ) + @xmath6 & @xmath80 & @xmath80 & @xmath80 & @xmath80 + @xmath81 & @xmath82 & @xmath83 & @xmath84 & @xmath85 + @xmath86 & @xmath87 & @xmath88 & @xmath89 & @xmath90 + @xmath91 & @xmath92 & @xmath93 & @xmath94 & @xmath95 + + @xmath23denoted : b out / b in boron out - of / in - plane displacement as shown in figure 1 ; the minus sign indicates inward directions .
we pay attention to the data of table 2 as they reveal certain trends .
first , no significant changes are seen local lattice distortions , and
the induced changes in transverse ionic strengths caused by cationic substitution appear to be notably small .
particularly , this observation relates to cases with @xmath86 and @xmath8 .
this implies that the relevant bond lengths are preserved , and the boron layer remains planar .
further , it is also seen from table 2 that longitudinal ( in - plane ) stresses are more strongly affected by the impurities , though the induced changes remain relatively small .
the overall comparison of the data of table 1 with the simulation results of table 2 reveals a certain correlation between the variation in the magnitude of the @xmath51 suppression rate ( @xmath96 ) and the doping induced changes in the boron sublattice .
this can be emphasized by observing the following difference : while in the majority of the doped solutions , the presence of several percents of divalent impurity metal atoms provides already a considerable suppression of @xmath51 , there are substitutions such as @xmath8 , @xmath9 , or even tetravalent @xmath10 low concentrations of which slightly influence both the superconducting order and the specific positions of the boron atoms .
the reason for this discrepancy seems to be related primarily to the strong @xmath97 @xmath98-@xmath98 linkage dominating in the boron layer @xcite which , as we have shown above , tends to keep the existing structural stability upon a change in the host cation .
considering @xmath52 as a reference cation , another interesting result is that one may say according to @xcite that the incorporation of just @xmath8 , @xmath9 or @xmath10 into the @xmath52-site has little effect on actual interband pairing channels .
correspondingly , in these cases , the protection function of the covalent boron linkage can not be markedly influenced ( at low doping levels ) what may provide a much weaker dependence of the superconducting temperature @xmath51 on impurity content . in order to get further insight into why the substitutional @xmath8 , @xmath9 and @xmath10 cause only minor changes in a planar geometry of the covalent boron layer
, we undertook an investigation of how local properties of chemical bonding are affected by the overall electronic structure and elemental characteristics of each of the considered impurity ions . for characterization of impurity - induced changes in chemical bonding via the differences of the electron states between the host and foreign cations , we performed an extensive comparative analysis of the relevant charge partitioning schemes . figure 2 shows charge density maps calculated for @xmath0 and all three doped cases ; the maps are projected onto the @xmath99 plane and contain the central @xmath6 or @xmath100 cations , and surrounding @xmath1 atoms
. figure 3 is instructive in another way : for each cation considered , it shows the line charge density distributions along the @xmath52-@xmath98 and @xmath58-@xmath98 directions , respectively .
computed values of the bader effective charges ( @xmath101 ) presented in table 3 reflect the topology features of the underlying electron - density distribution emerging as a direct consequence of the relevant arrangements .
@lcccc & @xmath102 & @xmath12 & @xmath13 & @xmath14 + electronegativity difference & & & & + values with respect to boron & @xmath103 & @xmath104 & @xmath105 & @xmath106 + @xmath107 & @xmath108 & @xmath109 & @xmath110 & @xmath111@xmath23 + + @xmath23a comparison shows that this value appears to be equal to that of @xmath112 obtained from the self - consistent lapw calculations for the bulk zirconium diboride @xmath113 @xcite .
comparison of figures 2 , 3 and the data of table 3 for the reference case of the @xmath56 ion reproduces the well - established result @xcite that @xmath52 is an ionically - bonded cation ( with an admix of some weak covalency ) in a stable divalent state .
analysis of cases where @xmath8 or @xmath9 substitutes for @xmath52 revealed three significant differences .
first , as depicted in table 3 , the bader effective charges are quite different from the formal chemical charge @xmath114 .
two other differences are very clearly delineated on figures 2 and 3 ; common to both ions they are : a depletion of the electron charge along the metal - boron bonding direction , and a large degree of electron charge accumulation of fairly good spherical form around the nuclei .
correspondingly , the physical picture of the outer electronic structure of @xmath8 and @xmath9 incorporated into magnesium diboride can be drawn as follows : ( i ) the @xmath115 states appear to be completely filled ( what provides increased electron density at the substituent , causes the observed spherical charge distribution around it , and induces the smaller effective charge transfer , in comparison with magnesium ) , ( ii ) zinc and copper cations acquire fractional quantities of the @xmath116 electron populations , about @xmath117 and @xmath118 , respectively , ( iii ) the partially filled @xmath116 states constitute the resulting outermost orbitals , and ( iv ) the @xmath119-electron covalency effects are extremely low because the filled @xmath115 levels of the substituent are shielded by its @xmath116 states . remarkably , smallness of the effective charges relative to the nominal chemical valence @xmath120 is an indication of the fact that the substitutional cations have retracted the certain amount of the delocalized electrons that participate in metallic bonds .
the reason why just the electrons shared by the boron atoms are excepted is that the significant changes in an equilibrium planar geometry of the covalent boron layer are absent .
furthermore , of note is another observation from table 3 that the local loss of charge exchange activity is directly connected with the drop in the electronegativity difference with respect to boron in the series of elements @xmath52 , @xmath8 , @xmath9 . it may be concluded thus that both substitutions give rise to a negative tendency which manifests itself ( i ) in increasing the electron density around outer orbitals of the @xmath8 or @xmath9 cations , and ( ii ) in decreasing the positive charge at the cation center up to the values of @xmath121 and @xmath122 , respectively .
this leads to partial neutralization of the overall charge on the substituents , makes the initial charge transfer channels much weaker and therefore causes the significant reduction of the polar contribution into the cationic nature of the substitutional ion .
as a result , the initial , predominantly ionic character of the local environment is degraded .
the variances of the bader effective charges from the nominal @xmath114 may therefore be regarded as a degree of @xmath8 or @xmath9 `` underbonding '' that should have a certain effect on physical properties of @xmath0 doped by these elements .
for example , since the interband channel is unaltered ( because of minor changes in a planar geometry of the covalent boron layer ) , the experimentally observed smallness of suppression of the superconducting temperature is perfectly understandable via influence of the above - mentioned retraction of `` previously expelled electrons '' on the density of electronic states at the fermi level .
this result correlates well with the conclusion made in @xcite for a series of @xmath123-@xmath124 codoped @xmath0 systems that the superconductivity is much more affected by the relevant lattice distortion induced by impurity ion substitution than by the band filling .
the trend towards electron density accumulation and a weakening of charge transfer channels breaks down for the case of the substitutional @xmath91 in the structure of @xmath0 .
although @xmath91 possesses almost the same electronegativity difference as @xmath6 , the calculated value of the bader effective charge , @xmath125 , is considerably different from the formal valence @xmath126 . from chemical point of view
, this implies that the zirconium cation in the @xmath0 lattice tends towards a lower oxidation state configurations of low - valent @xmath127 in complex sheet and cluster structures @xcite . ] .
in contrast to zinc and copper ions , @xmath91 attempts to expand definite amounts of charge density from itself toward an interstitial area .
analysis of figures 2 and 3 strongly suggests the moderate enhancement of the positive charge on the cation which takes place due to delocalization of its @xmath128 electrons caused by partial donation of @xmath129 additional electrons .
such a delivery of a partial charge ( along the metal non - metal charge - transfer configuration ) is well correlated with a chemical readiness of the zirconium cation to give up the valence @xmath130 electrons in order to move back to the predominant oxidation state of @xmath131 and to activate the covalent bonding with neighboring atoms .
consequently , as compared with the parent compound @xmath0 , an ionization increase @xmath132 @xmath133 @xmath134 on a substitutional cation site and the presence of appreciable covalency in the same region may give rise to a tendency towards dielectrization of the system .
this observation is confirmed by first - principles studies of the bulk compound @xmath113 @xcite , which show the presence the strong hybridization between the @xmath91 @xmath128 and the @xmath1 @xmath135 states as well as drastic fall of the density of electronic states at the fermi level .
thus , we can conclude that electronic charge distribution at the @xmath52 position in the bulk magnesium diboride is very sensitive to the cation replacement .
firstly , the substitution for an impurity ion of the same nominal charge @xmath120 leads to a substantial charge redistribution around the cation .
secondly , the change in the chemical bonding character depends on the type of metal ion and its electronegativity versus boron , so that among substitutional ions considered two different trends can be distinguished .
all this underlines in the light of the bonding situation that ion substitution not only always affects the existing chemical bonds but also that the resulting charge rearrangement is exhibited quite differently for the cations considered in the present work . moreover , in the context of superconductivity , these features can seriously limit the targeting efficiency of a number of potential substituents or even make them unfavorable towards the enhancement of superconducting properties . to complete the results obtained for impurity - induced changes in the cation charge states ,
it should be interesting to study how these changes proceed under pressure .
for this purpose , the @xmath10- and @xmath8-doped case were considered .
we employed the same set of calculations as used throughout this work and fully relaxed a doped system .
the resulting pictures of the charge density maps are shown in figure 4 .
the comparison of the charge distributions with those shown in figure 2 reveals no significant changes between them . the most remarkable difference is a slight increase of the bader effective charge of the @xmath10 cation from @xmath136 at zero pressure to @xmath137 under pressure of @xmath138 gpa ( what evidently indicates the above - mentioned oxidation tendency to the tetravalent state ) .
this implies that the charge state of @xmath10 remains relatively stable with compression until at least @xmath138 gpa .
the @xmath8 substitutional cation also demonstrates the similar tendency to maintain its charge state under pressure .
for example , the calculated change in the bader effective charge under compression of @xmath139 gpa , @xmath140 , shows that due to contraction of the filled @xmath115 levels , valence electrons do not favor extra oxidation . as a result ,
the overall charge configuration of @xmath8 in magnesium diboride is essentially unaffected by pressure .
another possible microscopic reason of lattice distortions stems from the fact well - known in the theory of local phase transitions that distortion of local geometry around an impurity may be caused by sufficient strength of the purely dynamical part of the charge transfer ( e.g. , @xcite ) .
generally speaking , significance of dynamical charge transfer processes generated by vibronic mixing of the proper electronic states is one of most important characteristic features of polar crystals @xcite . on the atomic level , this feature is associated with many - body polarization effects that contribute to ionic dipole polarization , and it reflects a mixed ionic(polar)-covalent character of the relevant chemical bonds @xcite . since the latter can be interpreted in terms of the born dynamic effective charges with respect to some reference ( nominal ) ionic value ( e.g. , @xcite ) , the calculations of the principal values of the born dynamic effective charge tensor @xmath141 can be employed as a test of how electronic effects of the many - body nature translate the polar contribution into ionic chemical bonding .
it is also of note that although the idea of calculation of the born dynamic effective charges for a conducting material may seem senseless at first glance , it can be , due to logic behind the ionic bonds and charge transfer , reasonably useful for a mixed bonded solid , indicating how specific pieces of a composite picture of chemical bonding may be distinguished . in table 4 , we have presented the converged values of the born effective charge for the magnesium cation in @xmath0 ( due to metallic properties of @xmath0 , we focused on dynamic charges only for cations and not for boron atoms ) .
it may be assumed that the numerical convergence to stable values of table 4 is guaranteed by the nearly ionic character of the @xmath52 sublattice which , in the spirit of layered superconductivity approach @xcite , can be figuratively imagined as a stack of the intrinsic insulating layers .
calculations have also predicted principal values of the electronic dielectric constant for the bulk magnesium diboride : @xmath142 , @xmath143 .
the theoretical value @xmath144 , evaluated as an average @xmath145 , is quite comparable with the experimental estimate of @xmath146 @xcite thus proving the consistency of the present analysis .
@*7cccc @xmath147 & @xmath148 & @xmath149 & @xmath150 + @xmath151 & @xmath151 & @xmath152 & @xmath153 + the tensorial nature of the born effective charge provides subtle information concerning differences in directions and strengths of dynamical processes related to the ionic sites and caused by the polarization effects .
as seen by comparing the values of table 4 , there exists a certain anisotropy , as is indicated by the observation that in the directions perpendicular ( the @xmath154-plane ) and parallel to the @xmath77-axis the components of the dynamic charge differ markedly from the formal value of @xmath155 .
this suggests that , although on the average @xmath52 remains quite well ionized in the full @xmath156-particle ensemble ( @xmath157 ) , ( i ) the attached effective potential , which gives rise to polarization force fields acting in a local environment around magnesium , appears non - spherical , and ( ii ) due to deviation from the nominal ionic charge some contribution of the pure dynamic covalent nature in the magnesium
boron bonding should be assumed .
the latter is entirely consistent with previous studies of dynamical covalency effects in @xmath0 @xcite . in order to analyze how the change in the host cation influences the attached effective potential
, we carried out the relevant calculations of the born effective charges for the substitutional impurities we are considering here .
in contrast to the previous case of magnesium , the computational method failed to give uniform results consistent with hexagonal symmetry . in the physical sense , the lack of numerical consistency is not discouraging , and should be considered as an additional confirmation of the drastic changes of a charge state which arise around the host cation position due to its replacement . in other words ,
impact of the substitutions with such unavoidable outcomes as falling of ionic character of chemical bonding , and the expansion of a metallic environment beyond the range of the boron layers @xcite make it impossible to get from the method of calculations the target values of dynamic charges for the foreign cations . to obtain some approximate estimates of the born effective charges ,
we have constructed a smooth procedure based on averaging the obtained results over the supercell , so that all the cations employed in the calculations are taken into account .
this can be done since overall charge balance is maintained .
due to electroneutrality of the simulation supercell , one can then expect that summation eliminates off - diagonal contributions , suppresses discontinuities , random perturbations and other artificial symmetry effects and thus will return the weighted - average components of the born effective charge tensor in the original diagonal form . in table 5
, we have presented the numerical values of the dynamic charges for @xmath81 , @xmath86 and @xmath91 substitutional cations as averages over the supercell .
@*7lcccc & @xmath158 & @xmath159 & @xmath160 & @xmath161 + @xmath81 & @xmath162 & @xmath162 & @xmath163 & @xmath164 + @xmath86 & @xmath165 & @xmath166 & @xmath167 & @xmath168 + @xmath91 & @xmath169 & @xmath170 & @xmath171 & @xmath172 + it is seen that the remaining uncertainties on the averaged values of charges are really negligible , so that this type of approximation is quite reasonable . nevertheless , one additional aspect which should be discussed here is the sensitivity of the retrieval accuracy within such a calculation procedure .
first of all , we note that all off - diagonal contributions are exactly eliminated . on the other hand ,
the required identity of the diagonal terms ( @xmath173 ) is restored at different accuracy levels : for zinc and copper , we get a more accurate retrieval than for zirconium .
we believe that this result is not surprising because some discrepancy in the case of zirconium suggests that due to the expansion of metallic bonds into the cationic environment , the effect of charge fluctuations is not completely eliminated after summation over the supercell .
the results listed in table 5 allow us to characterize the specific features of chemical bonds in terms of the born effective charges as follows .
( i ) a dynamic component in resulting chemical bonding becomes smaller upon cation substitution ( as compared with that of @xmath52 ) , and ( ii ) this partial loss in initial dynamic covalency is caused by the relevant charge redistribution around the cation position . at the same time , as seen from table 5 , a rate of an anisotropy in the dynamics of valence charge distribution remains relatively unchanged in the cases of zinc and zirconium .
moreover , of all three components just the @xmath174 component of the born effective charge , which is particularly interesting for dynamical covalency effects in magnesium diboride , tends to be close to the @xmath77-axis value of the @xmath52 cation and therefore appears , in comparison with in - plane @xmath175 , @xmath176 components , to be more robust to alteration of charge densities .
following the same consideration as in @xcite , this leads us to an understanding regarding why lattice distortions induced by @xmath81 , @xmath86 or @xmath91 ions are so small : the explanation is that the ionic strength changes induced by variations of dynamical charge transfer along the @xmath77-axis are not large enough to drive the relative boron displacements .
in order to understand the leading factors that govern adjustments of the lattice positions during accommodating the foreign cation in @xmath0 , we investigated the induced deformations ( lattice distortions ) in the local geometry of the boron sites .
the first - principles calculations were performed for three different substitutional cations such as divalent @xmath4 and @xmath3 , and tetravalent @xmath5 .
these species are of particular interest because they demonstrate the lowest - observed - droop of the superconducting temperature among most of the impurities of the same formal valences .
inspection of the relaxed lattice structures together with the results of the performed calculations allowed the following conclusions to be drawn : ( i ) at a low doping level , the @xmath8 , @xmath9 and @xmath10 substitutions cause only negligible changes in the structural arrangement of the neighboring boron atoms .
( ii ) such an insignificant impact on the local geometry of the boron layer is mostly attributed to two aspects .
the first is the observation that the existing in - plane @xmath98-@xmath98 covalent bonding is strong enough to keep effectively the boron atoms non - shifted under substitution of magnesium with @xmath8 , @xmath9 or @xmath10 .
the second is that dynamical covalency processes along the cation - boron bonding tend to be tolerant of changes in the host cation and , consequently , can not create a channel of instability in the boron layer .
both aspects are important in view of the fact that the structural balance between local geometry and chemical bonding are related to the electron structure of the impurity and to the bonding connections carried out by the impurity .
( iii ) a role of small deformations of the local geometry of the boron layer induced by the impurity ions may be insufficient to influence alone on superconductivity of the host system .
( iv ) generally , our results emphasize that @xmath0 is the perfect material having a unique combination of an ideal planar layer structure , the predominantly ionic character of the cationic sublattice , and the high degree of chemical bonding stability . in order to investigate the characteristic measures of impurity activity , such as charge states and the impurity - induced changes in the chemical bonding
, we performed the comparison analysis of the underlying charge - transfer processes .
it has been discovered that the replacement of @xmath56 by divalent @xmath4 , @xmath3 or tetravalent @xmath5 influences substantially the character of the charge transfer , so that the distributions of the charge around the impurities and degree of ionicity differ drastically both from that around the host cation and also from each other .
it has been shown that the different cationic nature of the substituents dictates two entirely opposite directions along which the induced changes in the local structure take place .
first , the change in the host cation leads to density contraction the substitutional @xmath8 or @xmath9 demonstrate the considerable increase of the overall charge density due to retaining its valence electrons and the small amount of charge transfer to the boron layers .
particularly , a larger accumulation of a charge and the least charge transfer are observed for the copper ion .
it was argued that this process retracts the certain amount of the delocalized electrons from metallic bonds what causes the observed decrease in the superconducting temperature values as the substitution level grows .
second , the change in the host cation leads to charge delocalization no charge accumulation is observed for @xmath10 .
the @xmath10 incorporation into the @xmath52 site of @xmath0 tends ( a ) to preserve the initial charge transfer channels , and ( b ) by partial donation of @xmath128 electrons to add some amounts of ionicity and covalency into the local cation - anion coupling .
it was suggested that such increasing of covalency character does not favor superconducting properties of the host material .
the stability of charge states related to the impurity cations was tested against pressure : the pressure dependence of the charge transfer degree was found to be very weak .
the theoretically observed effective charges indicate that the substitutional ions do not possess their formal valence ( oxidation state ) , and are characterized by different levels of electron delocalization ; zirconium as an electron donor is indeed stronger than magnesium . as a consequence , by comparison with less positively charged zinc and copper ,
the behavior of zirconium remains much more tolerant ( affine ) to the usual ionic distribution associated with the host magnesium cation . summing up
, one can say that the induced local lattice distortions and changes of the charge transfer degree may serve as structurally sensitive prognostic parameters for signaling how the superconductivity in @xmath0-solid solutions is affected by impurities , for clarifying the functional roles and efficiency of such various factors as a geometry of the boron layer , the electronegativity difference , ionic character of the local environment , relative affinities toward magnesium and boron , etc . , and on this basis for facilitating more precise theoretical predictions in synthesis and characterization .
00 buzea c and yamashita t 2001 r115 ravindran p , vajeeston p , vidya r , kjekshus a and fjellvg h 2001 _ phys .
_ b * 64 * 224509 erwin s c and mazin i i 2003 _ phys .
_ b * 68 * 132505 ivanovskii a l 2003 _ fiz .
tverdogo tela _ * 45 * 1742 engl .
trans . : 2003 _ phys . solid state _ * 45 * 1829 cava r j , zandbergen h w and inumaru k 2003 _ physica _ c * 385 * 8 singh p p 2003 _ bull . mater .
* 26 * 131 bernardini f and massidda s 2006 _ europhys .
_ * 76 * 491 di castro d , ortolani m , cappelluti e , schade u , zhigadlo n d and karpinski j 2006 _ phys .
_ b * 73 * , 174509 kortus j 2007 _ physica _ c * 456 * 54 kuzmenko a b 2007 _ physica _ c * 456 * 63 li w x , zeng r , poh c k , li y and dou s x 2010 135701 ojha n , malik v k , singla rashmi , bernhard c and varma g d 2010 045005 singh p p 2002 _ physica _ c * 382 * 381 moritomo y and xu sh 2001 arxive : cond - mat/0104568v1 kazakov s m , angst m , karpinski j , fita i m and puzniak r 2001 1 tampieri a , celotti g. , sprio s , rinaldi d , barucca g and caciuffo r 2002 497 singh k , mohan r , shelke v , gaur n k and singh r k 2008 _ indian j. pure & appl .
phys . _ * 46 * 420 kalavathi s and divakar c 2005 _ bull .
_ * 28 * 249 feng y , zhao y , yan g , pradhan a k , zhou l , koshizuka n and murakami m 2004 critical current density , flux pinning and microstructure in @xmath0 superconductors _ high temperature superconductivity 1 materials _
ed a v narlikar ( berlin , heidelberg : springer - verlag ) pp . 401 - 430 sun y , yu d , liu z , he j , zhang x , tian y , xiang j and zheng d 2007 appl .
052507 shi l , zhang s and zhang h 2008 27 mchirgui a , ben azzouz f , annabi m , zouaoui m and ben salem m 2005 321 mazin i
i and antropov v p 2003 _ physica _ c * 385 * 49 kristoffel n , rd t and rgo k 2003 _ europhys . lett . _ * 61 * 109 kristoffel n , rd t and rgo k 2003 _ j. supercond . _ * 16 * 517 bussmann - holder a and bianconi a 2003 _ phys .
_ b * 67 * 132509 kresse g and hafner j 1993 _ phys .
_ b * 47 * 558 kresse g and hafner j 1994 _ phys .
_ b * 49 * 14251 kresse g and furthmller j 1996 _ comput .
_ 6 , 15 kresse g and furthmller j 1996 _ phys .
_ b * 54 * 11169 blchl p e 1994 _ phys .
_ b * 50 * 17953 kresse g and joubert d 1999 _ phys . rev .
_ b * 59 * 1758 perdew j p , burke k and ernzerhof m 1996 3865 ; 1997 1396 monkhorst h j and pack j d 1976 _ phys .
_ b * 13 * 5188 bader r f w 1990 atoms in molecules : a quantum theory ( oxford : oxford university press ) http://theory.cm.utexas.edu/vtsttools/bader w. tang , e. sanville , g. henkelman 2009 084204 choi h j , roundy d , sun h , cohen m l and louie s g 2002 _ nature _ * 418 * 758 mitrovi b 2004 9013 barbalace k 1995 - 2011 _ periodic table of elements _ ( environmentalchemistry.com ) switendick a c 1990 electronic structure and charge density of zirconium diboride _ technical report _ sand90 - 1474c ; conf-9008707 an j m and pickett w e 2001 4366 kortus j , mazin i i , belashchenko k d , antropov v p and boyer l l 2001 4656 belashchenko k d , van schilfgaarde m and antropov v p 2001 _ phys .
_ b * 64 * 092503 de la mora p , castro m and tavizn g 2002 _ j. sol .
chem . _ * 169 * 168 monni m , ferdeghini c , putti m , manfrinetti p , palenzona a , affronte m , postorino p , lavagnini m , sacchetti a , di castro d , sacchetti f , petrillo c and orecchini a 2006 _ phys . rev .
_ b * 73 * 214508 kalvins a k 1985 a review of the inorganic and organometallic chemistry of zirconium _ report no 85 - 124-k _ shein
i r and ivanovskii a l 2002 _ fiz .
tverdogo tela _ * 44 * 1752 engl .
2002 _ phys . solid state _ * 44
* 1833 rosner h , an j m , pickett w e and drechsler s - l 2002 _ phys . rev .
_ b * 66 * 024521 kristoffel n 2000 structural distortions and oxygen local dynamic instabilities in superconducting perovskites _ defects and surface - induced effects in advanced perovskites _ ( _ nato science series 3 . high technology _ ) ed g borstel , a krumins and d millers ( dordrecht : kluwer scientific publishers ) pp .
113 - 124 kristoffel n and klopov m 1996 _ phys .
_ b * 54 * 9074 kristoffel n n 1984 _ czech .
_ b * 34 * 1253 kristoffel n and konsin p 1988 _ phys .
status solidi _ ( b ) * 149 * 11 bussmann - holder a , bilz h and benedek g 1989 _ phys . rev .
_ b * 39 * 9214 pishtshev a 2011 _ physica _ b * 406 * 1586 lee k
w and pickett w e 2003 _ phys .
_ b * 68 * 085308 bill a , morawitz h and kresin v z 2003 _ phys . rev .
_ b * 68 * 144519 varshney d , azad m s and singh r k 2004 1446 + varshney d , nagar m , bhatnagar s and varshney m 2010 075016 deng s , simon a and khler j 2005 pairing mechanisms viewed from physics and chemistry _ superconductivity in complex systems _ ( _ structure and bonding vol 114 _ ) ed k a mller and a bussmann - holder ( berlin , heidelberg : springer - verlag ) pp . 103 - 141 ghosez ph and veithen m 2007 096002
-atoms supercell modeling the @xmath57 composition for @xmath61 . the @xmath0 cell containing the impurity cation
is sketched . in - plane and out - of - plane
boron displacements are indicated by arrows .
, scaledwidth=80.0% ] | we report the results of an ab initio modeling of substitutional impurities such as zinc , copper and zirconium ions incorporated into the magnesium sublattice of @xmath0 .
these species are of particular interest because they demonstrate the lowest - observed - droop of the superconducting temperature among most of the impurities of the same formal valences .
the goal of computational studies was to gain an atomistic understanding of how the structural and bonding properties of the local ionic environment are affected by a change in the host cation .
the simulations performed for the given set of substituents indicate that at a low doping level , the induced lattice distortions and additional forces are noticeable small . at the same time , the electron redistributions around the impurities have been found to differ drastically both from that around the host cation and from each other .
results of the first - principles calculations were used to compare and discuss , in the context of the cation properties , specific changes in local charge structures and in chemical bonding caused by the presence of the impurity .
certain trends in substituent effects were found .
it was determined how the substitutional ion modifies the initial charge distribution around the host cation position and therefore , affects both the overall picture of the electronic states involved in the chemical bonding and the intrinsic charge - transfer channels .
our studies demonstrate that the robust structural stability of the boron sublattice of @xmath0 , which preserves the nearest - neighbor environment around the @xmath1 atoms practically unchanged from an impurity incorporation into the magnesium site , has important functional significance
it supports a much weaker dependence of the superconducting temperature on impurity content .
the present work gives insight into structural and electronic features of the impurity caused - processes which contribute to unique material properties of various @xmath0-solid solutions . |
test particle motion in realistic gravitational fields is of obvious astrophysical importance and at the same time it provides reliable evidence of the properties of those gravitational fields .
however , in many actual astrophysical systems the particles are not moving freely but are influenced by ambient matter , electromagnetic fields and radiation . in typical situations , these physical " effects are probably even more important than fine details of the spacetime geometry alone .
the most remarkable conditions , from the point of view of general relativity as well as astrophysics , appear near very compact objects where both the pure gravitational and other physical " effects typically become extraordinarily strong . in the present paper we focus on the motion of test particles in a spherically symmetric gravitational field , under the action of a thomson - type interaction with radiation emitted or accreted by a compact center .
this kind of problem was first investigated by poynting @xcite using newtonian gravity and then in the framework of linearized general relativity by robertson @xcite .
it involves competition between gravity and radiation drag , which may lead to interesting types of motion which do not occur in strictly vacuum circumstances . in particular
, there arises the question of whether equilibrium behavior like circular orbit motion or even staying at rest " are possible in some cases .
theoretical aspects of the poynting - robertson effect as well as its astrophysical relevance in specific situations have been studied by many authors since the original pioneering work .
recently we considered this same effect in the relativistic setting by studying test particles orbiting in the equatorial plane of a schwarzschild or kerr black hole , assuming that the source of radiation is located symmetrically not far from the horizon ( in the case of outgoing flux ) .
we first chose the radiation field to be directed purely radially with respect to the zero - angular - momentum observers so that it too had zero angular momentum @xcite , but then also considered a more general case of radiation having some ( arbitrary ) angular momentum @xcite ( see these papers for a more thorough overview of the references ) . while ingoing radiation might at first seem rather unmotivated , its consideration can give some rough idea about particle motion inside an accretion disk with strong radiation emanating from the disk , at least in the plane of the disk . here
we intend to compare the limiting simpler case of the poynting - robertson effect due to a purely radial ( zero - angular - momentum ) test radiation flux in a schwarzschild background with the treatment using a self - consistent radiation flux in the exact , vaidya spherically symmetric spacetime whose source includes a null dust @xcite
. similarly one could think of comparing the general nonzero angular momentum test flux case in the kerr spacetime background with the rotating kerr - vaidya spacetime @xcite , but the energy - momentum tensor of this latter exact solution was shown not to be interpretable in terms of a null dust alone as in the nonrotating case @xcite , so we will not address that issue here . since the simpler vaidya spacetime contains an arbitrary function @xmath0 which describes a time - dependent mass for the central object , there is a freedom to choose that function in many ways , but we limit our attention to a relatively simple evolution of this function in our applications .
in particular when an appropriate derivative of this function @xmath0 is constant , one models a phase of evolution of the central object in which the central mass changes at a constant rate . in this nonstatic situation , it no longer makes sense to seek radial equilibrium orbits as in the corresponding static test flux case , but one can look for some kind of adiabatically changing equivalent orbits . after summarizing some basic properties of the vaidya spacetime , we write out and reduce the equations describing timelike motion ( with an effective force term describing the particle - radiation interaction ) in schwarzschild - like spherical coordinates with retarded / advanced time .
next we examine the appropriate limit of the vaidya spacetime which leads to the scenario of a test flux in the schwarzschild spacetime , in particular to see how the schwarzschild radial equilibrium orbits fit into the more general situation .
then several numerical examples are shown illustrating typical types of motion .
the gravitational field associated with a spherically symmetric body of variable mass @xmath0 is described by vaidya s nonvacuum solution of the einstein equations @xcite , representing an algebraically special petrov type d spacetime like the schwarzschild metric it generalizes .
the corresponding line element written in schwarzschild - like coordinates is given by [ metrictr ] s^2=-()^2 + + r^2(^2+^2 ^ 2 ) , with @xmath1 , and @xmath2 and @xmath3 denote partial derivatives with respect to @xmath4 and @xmath5 respectively .
a more useful form is obtained by introducing eddington - finkelstein - like coordinates through the transformation to the null coordinate @xmath6 by u= , so that the line element becomes [ metricur ] s^2=-n^2u^2 2u r + r^2(^2+^2 ^ 2 ) , where @xmath7 .
hereafter we will use the eddington - finkelstein - like form ( [ metricur ] ) .
its inverse has non - zero components g^ur = g_ur , g^rr =- g_uu , g^=(g_)^-1 , g^=(g_)^-1 , so the raising or lowering of indices involves the simple relations @xmath8 . in the case of non - positive / non - negative @xmath9 ,
i.e. , for a source loosing / gaining energy , @xmath6 represents retarded / advanced time and one has @xmath10 , respectively .
( outgoing radiation ) while lower signs to @xmath11 ( ingoing radiation ) , so that @xmath12 .
the advanced null coordinate is conventionally denoted by @xmath13 , but we keep @xmath6 in both cases ( distinguishing between them by signs ) . ] we will allow for both possibilities in fact we leave the @xmath14 evolution completely general , with several important specific cases treated as examples .
the metric ( [ metricur ] ) has a curvature singularity at @xmath15 and a coordinate singularity at @xmath16 which corresponds to an apparent ( but not event ) horizon .
actually , on the @xmath16 hypersurface the metric reduces to @xmath17 which is spacelike in both cases , i.e. , with @xmath18 . in the @xmath19
limit the metric ( [ metricur ] ) reduces to the schwarzschild metric written in terms of the retarded / advanced time @xmath20\,\ ] ] for which @xmath21 and @xmath22 are both killing vector fields .
global properties of the vaidya spacetime ( its conformal structure ) are summarized in section 9.5 of @xcite together with the most relevant references .
see also @xcite where different definitions of mass ( section 4.3.5 ) and the distinction between apparent and event horizons ( section 5.1.8 ) are illustrated using the vaidya spacetime as an example
. evaluating the einstein equations @xmath23 for this metric , one finds that the energy - momentum tensor has only one non - zero component in these coordinates [ tmunu ] t_uu = t^rr= > 0 .
such a @xmath24 can be interpreted as a null dust representing pure radiation , " namely @xmath25 , where @xmath26 is a purely radial outgoing / ingoing null vector and @xmath27 depends on the normalization chosen for @xmath26 .
for example , if one takes @xmath28 , then t^rr= ^2= .
this choice for @xmath29 means that @xmath30 ; in particular , the energy of the radiation particles is proportional to @xmath31 . in the corresponding schwarzschild spacetime with this value for @xmath32
, @xmath29 agrees with the choice used in @xcite . as in that case , one checks easily that @xmath29 is tangent to a congruence of affinely parametrized null geodesics since @xmath33 this congruence has nonzero expansion @xmath34 but zero vorticity .
the most natural test observers suitable for physical interpretation are those at rest in the spatial coordinate grid at @xmath35 , @xmath36 , @xmath37 ; their 4-velocity field is [ static - obs ] e_u= _ u. a convenient spatial orthonormal triad tied to this observer congruence is [ framestatic ] e_r = n(_r _ u ) , e_= _ , e_=_. this observer congruence is accelerated and has nonzero expansion and shear , @xmath38 but its vorticity vanishes , ( ) ^^[;]+a()^[^]=0 , so it is hypersurface orthogonal .
in fact these observers follow the @xmath4 coordinate lines orthogonal to the @xmath4 coordinate hypersurfaces in the original coordinate system , and in the schwarzschild case @xmath39 they are just the usual static observers .
when studying motion of massive test particles ( rest mass @xmath40 ) in the vaidya spacetime , one can either restrict attention to geodesics , or allow the particles to interact physically with the radiation in terms of which the energy - momentum tensor is interpreted . a simple way to model
this interaction is to assume that the force on the particle is proportional to the 4-momentum density of radiation observed in the particle s rest frame . denoting the particle s 4-velocity by @xmath41 and acceleration by @xmath42 , this yields the equation of motion [ motion ] ma(u)^=-p(u)^_t^_u^ _ rad(u)^ , where @xmath43 and @xmath44 are particle s proper time and rest mass
, @xmath45 is the effective interaction cross section ( its dimension is length squared ) and @xmath46 is the projector to the particle s instantaneous rest space .
such a force formula is independent of the direction and frequency of radiation ( and the interaction efficiency " @xmath45 is also independent of the momentum - density magnitude ) and can perhaps be adequate in situations when thomson scattering is a dominant interaction mode .
let us divide the equation ( [ motion ] ) by @xmath44 ( denoting @xmath47 ) and write it out in the @xmath48 coordinates .
since the energy - momentum tensor has only one non - zero component @xmath49 ( [ tmunu ] ) , one has [ motion / m ] = -^_u^u^ -(g^u+u^u^u)t_uuu^u .
first , the latitudinal component = -u^r u^+(u^)^2 -t_uu(u^u)^2 u^confirms that the interaction leaves the motion planar : choosing @xmath50 and @xmath51 at some instant , one sees that at that instant @xmath52 too and so the particle remains in the ( equatorial " ) plane . substituting @xmath50 and @xmath51 into the other components of ( [ motion / m ] ) , we have @xmath53
the 4-velocity normalization condition @xmath54 , explicitly -1 = -n^2(u^u)^22u^u u^r+r^2(u^)^2 [ norm - contra ] , enables the simplification of the longest equation ( [ ur - contra ] ) to [ ur - contra - shorter ] = -+(r-3m)(u^)^2 ( u^u)^2 -(u^u u^r1)t_uuu^u .
one could instead use the normalization condition to eliminate one of the three 4-velocity components from all of the equations trying to make the latter a closed system , but unfortunately , even with a favorable " evolution of the mass @xmath55 , exact integration of such a system is almost never possible due to the a priori unknown dependence @xmath56 . in any case , the most important non - schwarzschild " feature ( besides the interaction terms scaled by @xmath57 ) is the third term @xmath58 in the radial equation ( [ ur - contra - shorter ] ) .
it is clearly never positive , so it always increases the static radial pull @xmath59 . in order to balance this
inward pull , the azimuthal velocity @xmath60 in the centrifugal " term
@xmath61 has to be larger than in the schwarzschild field . finally , the last term of eq .
( [ ur - contra - shorter ] ) represents physical interaction of the particle with radiation ; given that @xmath62 , it is seen that only in the case of outgoing radiation ( upper sign ) can this term oppose the gravitational attraction . to interpret the test particle 4-velocity ,
one can express it using the obvious physical tetrad of eqs .
( [ static - obs ] ) , ( [ framestatic ] ) adapted to the
static " observer @xmath63 , u=(u , u ) [ u + ( u , u)^ae_a ] , where @xmath64 is the lorentz factor ( u , u)=(1-_a b(u , u)^a(u , u)^b ) ^-1/2 = nu^u u^r and @xmath65 are defined in eq .
( [ framestatic ] ) .
the @xmath66-velocity tetrad components are related to the corresponding coordinate components by [ eq : firstordervelocities ] ( u , u)(u , u)^r = u^r , ( u , u)(u , u)^ = ru^ , ( u , u)(u , u)^ = ru^. for a particle moving in the equatorial plane @xmath50 , so that @xmath67 , the equations of motion ( [ motion ] ) become @xmath68 + \frac{\gamma n}{r}(\nu^{\hat \phi}){}^2 \pm\frac{\tilde \sigma t_{uu}}{n^2}(1\mp\nu^{\hat r})^2\,,\nonumber \\ \fl\quad \frac{\rmd \nu^{\hat \phi}}{\rmd \tau}&= & \frac{\gamma}{rn^3}\nu^{\hat r}\nu^{\hat \phi } \left[\frac{m}{r}\,n^2\mp m_{,u } ( 1\mp\nu^{\hat r})\right]-\frac{\gamma n}{r}\nu^{\hat r}\nu^{\hat \phi}-\frac{\tilde \sigma t_{uu}}{n^2}\nu^{\hat \phi}(1\mp\nu^{\hat r})\,,\end{aligned}\ ] ] with @xmath69 . to complete this system
one must add the evolution equations for @xmath6 , @xmath5 and @xmath70 , i.e. , [ motionaltre3 ] = ( 1^r ) , = n ^r , = .
in general the above system of equations is only solvable numerically , but we will at least try to reduce it further for several particular , simple cases .
first we will restrict to geodesic motion , then to purely radial ( @xmath37 ) and to circular " ( @xmath35 ) motion , and finally we will check whether equilibrium ( @xmath71 ) between the gravitational pull and the radiation effect is possible .
first we check how the central mass change itself affects the free ( geodesic ) motion of test particles , not taking the interaction with radiation into account ( @xmath72 ) ; this problem was discussed in @xcite .
the most important simplification arises due to the azimuthal symmetry of the field : the specific angular momentum @xmath73 is a constant of the motion as in the schwarzschild spacetime , so one has u^=g^u_= immediately and need not to solve the azimuthal " eq .
( [ uphi - contra ] ) .
equations for the remaining two 4-velocity components read @xmath74 in contravariant form , while the covariant form reduces to @xmath75 since @xmath76 represents the particle s energy , eq .
( [ duu - geod ] ) says that a free particle gains / looses energy if @xmath0 decreases / increases ( see @xcite ) . it must be stressed that this has nothing to do with the radiation ( there is no interaction in the case of geodesics ) , the energy increases / decreases simply because the particle s gravitational binding by the center weakens / strengthens .
formally the geodesic equations only differ from the schwarzschild case by the @xmath77 term in the radial equation , but one must always remember that @xmath0 itself depends on @xmath6 and thus also changes with @xmath43 everywhere . if @xmath79 at some instant , one sees from eq .
( [ uphi - contra ] ) that @xmath80 there as well , so @xmath60 ( and also @xmath81 ) remains zero along the whole world line .
equations for the time and radial components of motion are then @xmath82 substituting for @xmath49 ( [ tmunu ] ) and writing it in terms of @xmath83 , the first equation acquires the form [ duu - radial ] = ( m+ ) . here all the quantities are treated as functions of @xmath43 . in particular , the evolution of the radius @xmath84 is of course not known a priori , and so the system is analytically unsolvable in general .
one can , however , obtain a solution for particular choices of @xmath85 .
for example , for an exponential decay of the mass , @xmath86 , the term in parentheses in ( [ duu - radial ] ) vanishes , which implies @xmath87 . with @xmath88
replaced using the normalization condition , one finds u^r = ( 2u^u)^-1 , which can be solved numerically for @xmath84 . in particular , for the choice @xmath89 , this is exactly solved by @xmath90\ ] ] which represents the growth of @xmath5 from @xmath91 to an asymptotic value @xmath92 .
however , it is clear that this is just one ad hoc , artificial case .
the system is analytically unsolvable even in the geodesic limit ( @xmath72 ) , although it becomes quite compact in that case .
the geodesic form of the radial equation is worth mentioning , in particular : after substituting @xmath93 from the normalization , it reads [ ddr , time - like ] rn^2 = -n^2 -2 m_,u .
the second type of symmetric " motion is that along circular " orbits , namely with @xmath94 .
this only holds permanently ( in order not to speak just of a turning point of the radial motion ) if @xmath94 makes the right hand side of ( [ ur - contra - shorter ] ) zero , [ ur - contra=0 ] ( =) -+(r-3m)(u^)^2 ( u^u)^2 t_uuu^u = 0 .
the azimuthal " eq .
( [ uphi - contra ] ) yields , with @xmath94 , = -t_uu(u^u)^2 u^ = r^2(u^u)^2 u^. since the coefficient of @xmath95 is always negative if nonzero , this means that if the particle interacts with the radiation , @xmath60 is slowed down to zero , which is the radiation drag effect .
circular orbits with @xmath96 thus do not seem to be possible , because the above condition @xmath97 can hardly be constantly satisfied if @xmath60 is variable , unless its variability were exactly compensated by the effect of the change in @xmath98 .
this would require a very special adjustment of the @xmath0 , @xmath60 and @xmath98 evolutions ( for given values of @xmath35 and @xmath57 ) .
substituting for @xmath49 and for @xmath99 from the normalization ( [ norm - contra ] ) , the condition ( [ ur - contra=0 ] ) for @xmath35 becomes @xmath100 \pm\frac{m_{,u}u^u}{r}\left(u^u\mp\frac{\tilde\sigma}{4\pi r}\right).\end{aligned}\ ] ] the positive solution of this last equation reads [ uu - circular ] u^u= , plus the normalization condition @xmath101 must hold .
( remember that the upper signs correspond to @xmath102 and the lower signs to @xmath103 . )
these conditions ensure that at some instant of @xmath6 , the test particle moves along @xmath35 ; the variation of @xmath0 with @xmath6 however makes them @xmath6-dependent , so the respective values of @xmath98 , @xmath60 and @xmath5 change with @xmath6 as well .
it is also possible to use the normalization in the remaining time equation ( [ uu - contra ] ) to obtain = -t_uu(u^u)^3 , where one can in turn substitute from the above @xmath104 condition and rewrite the result in terms of the @xmath6-dependence ( divide by @xmath98 ) , @xmath105 \nonumber \\ & = & \frac{m_{,u}}{rn^2 } \left\{u^u\pm\frac{\tilde\sigma}{4\pi r}\,[n^2(u^u)^2 - 1]\right\}\ , .
\label{duu / du , circular}\end{aligned}\ ] ] in the geodesic limit ( @xmath72 ) the solution ( [ uu - circular ] ) and respective @xmath60 reduce to @xmath106 the latter has to equal @xmath107 , which yields @xmath108 } \nonumber\\ \fl \lefteqn { = \pm m_{,u}r^4-mr^3+(2m^2+\tilde{l}^2\pm\tilde{l}^2 m_{,u})r^2 - 5m\tilde{l}^2 r+6m^2\tilde{l}^2 = 0 \;. } \label{circularity - timelike}\end{aligned}\ ] ] this can be understood either as a quartic equation for @xmath5 , or as a compatibility condition " for @xmath109 , with solution [ m , u - circular - geo ] m_,u= .
the left hand side @xmath110 is never positive , so the condition is only consistent if @xmath111 .
in such a case , its integration gives [ m(u)-compatibility ] m(u)=r , where the integration constant has been chosen so that @xmath112 .
thus one must have a very particular complicated mass function which depends on the special orbit parameters ( radius and angular momentum ) to allow circular geodesic orbits to exist .
for such a special orbit , knowing that @xmath99 must equal @xmath107 , one can of course express @xmath113 in terms of @xmath114 from the normalization , as an alternative to ( [ u^u - circular - geod ] ) : ( u^u)^2 = .
the square of the energy of the particle on this circular geodesic is then ( -u_u)^2=(-g_uuu^u)^2 = = n^2(1 + ) .
finally , it is natural to ask whether it is possible for the particle to remain with @xmath71 at some radii . staying in equilibrium
is a limiting case of purely radial motion ( if in addition @xmath94 ) as well as of purely azimuthal motion ( if in addition @xmath79 ) .
the normalization condition ( [ norm - contra ] ) then implies u^u = n^-1 , the only other constraint being the fulfillment of ( [ dur / dtau=0 ] ) , namely compensation of the radial gravitational pull , modified by the change of the central mass @xmath0 , by the force exerted on the particle through the radiative flux , [ equilibrium - equation ] - ( u^u)=0 . substituting for @xmath98 from above leads to [ equilibrium - equation , explicit ] -m -=0 .
this is a quartic equation for @xmath5 which depends on the function @xmath0 of @xmath6 , which means if @xmath0 is not constant , there is no true equilibrium in general , but in the quasi - stationary case in which @xmath0 changes sufficiently slowly , the roots of this equation will also change sufficiently slowly so as to be called quasi - equilibrium radii that will certainly influence the qualitative behavior of the orbits in a way similar to the actual equilibrium radii in the stationary case in which @xmath0 is constant .
however , one can also express the solution of this condition in terms of @xmath77 : [ equilibrium ] m_,u= .
this equation also gives the condition for a particular limiting case : the quasi - equilibrium can only be permanent , " namely the corresponding radius remains at a given value @xmath5 for every @xmath6 , if @xmath0 evolves with @xmath6 so that the condition is constantly satisfied for that fixed value of @xmath5 .
clearly the condition can only hold with the upper sign , thus for @xmath102 ( outgoing flux ) namely when the interaction term with @xmath57 is larger than the first term @xmath5 in the denominator to make the right hand side negative . a necessary condition for equilibrium to be possible at some given fixed @xmath5 is therefore [ necessary - condition ] > .
the solution of ( [ equilibrium ] ) can be written implicitly as ur ( ) -n = const .
switching off " the physical radiation - particle interaction ( @xmath72 ) , one finds the condition for geodesic equilibrium , m_,u = n^2 ( > 0 ) .
this condition can never be satisfied , because the left hand side is non - positive . this is due to the fact that the change of mass always makes the combined gravitational pull @xmath115 stronger ( more negative ) than the first term alone , irrespectively of the sign of @xmath109 ( see eq .
( [ ur - contra - shorter ] ) with @xmath79 and @xmath72 ) .
let us discuss the special , lower - dimensional types of motion of the previous section in more detail and compare the vaidya results with those obtained in a schwarzschild background with a test radiation flux .
the test - flux case has been treated both in the schwarzschild @xcite and kerr @xcite backgrounds as a relativistic generalization of the classic poynting - robertson effect .
the test particles moving in these spacetimes were subjected to a thomson - type interaction with a superimposed test radiation field , hence their motion was described by eq .
( [ motion ] ) with @xmath25 as in the present paper , but this null dust was taken there to be a test field .
in contrast , in the present discussion the radiation is self - consistent , being a source in the vaidya exact solution of the einstein equations .
since general test motion has to be solved numerically in both cases , we are mainly interested in the comparison of conditions for special types of motion which were obtained analytically .
in particular , we will focus on the possibility of equilibrium locations , as noted in section [ equilibrium - locations ] , or perhaps more appropriately in the present case , quasi - equilibrium locations . in the schwarzschild spacetime with purely radial test flux
, the equilibrium condition takes the form ( see eq .
( 3.22 ) of @xcite ) [ a / m ] = n , a=^2(-k_u)^2 .
( we have added the @xmath116 sign here in order to include both the outgoing- and ingoing - radiation case , but it is immediately clear that it can only hold with outgoing radiation , i.e. , with the plus sign . ) to verify whether this form of the equilibrium condition is consistent with the condition ( [ equilibrium ] ) we derived above for the vaidya field , we must translate the notation of that article . substituting into ( [ a / m ] )
@xmath117 and noting that @xmath118 , the equilibrium condition becomes [ a / m - explicit ] = n m_,u =- n .
this coincides exactly with the condition following from eq .
( [ equilibrium - equation ] ) if the term describing the gravitational effect of mass change ( namely the one proportional to @xmath77 but not containing @xmath57 ) is omitted .
actually , this term corresponds exactly to the @xmath5 term in the denominator of ( [ equilibrium ] ) , so without it that condition reduces exactly to ( [ a / m - explicit ] ) . expressed the other way round ,
switching from the schwarzschild spacetime with a test radiation flux to the exact vaidya spacetime with a self - consistent flux brings that @xmath5 term into the equilibrium condition ( [ equilibrium ] ) .
this term makes the necessary equilibrium @xmath77 more negative , namely bigger in absolute value . in order to understand this , it is convenient to look once more at the equilibrium condition ( [ equilibrium - equation , explicit ] ) , - + - = 0 ( limited to the upper sign since equilibrium is only possible in that case )
. it is seen that the second term , describing the gravitational effect of mass change @xmath109 , falls off more slowly than the third term which describes physical effect of the flux corresponding to that @xmath77 ( and also more slowly than the first , schwarzschild term ) .
therefore , it is difficult to reach equilibrium at very large radii , because the second term dominates there and the third term can only balance it with very large value of @xmath57 .
it is very useful to express the solution of the above equilibrium equation as [ equilibrium - in - r ] r = n(- ) .
it is seen from here that if the central mass were to turn completely into radiation ( @xmath119 , thus @xmath120 ) , there would be two possibilities for the quasi - equilibrium radius : either it decreases to zero together with the mass , or it approaches the value r - r_fin . in particular ,
if @xmath121 vanished in this limit , @xmath122 would be @xmath123 and would surely represent the maximum reached during the entire evolution of @xmath124 , because regardless of how @xmath0 and @xmath5 would behave precisely , one has ( for @xmath125 of course ) @xmath126 and so @xmath127 it is also seen from ( [ equilibrium - in - r ] ) that when @xmath77 is so large that the second term is negligible , the quasi - equilibrium radius evolves according to @xmath128 . to illustrate properties of the equilibrium condition
, we choose a few particular mass functions @xmath14 .
we consider the case where @xmath129 ( which implies exponential decay of @xmath0 ) , the case where the mass @xmath0 decreases linearly , and the case of a hyperbolic tangent mass profile often used in the literature ( see , e.g. , ref .
@xcite ) . with time @xmath6 for exponential mass decrease @xmath131 ( which yields @xmath132 ) with @xmath133 , as given by ( [ equilibrium - in - r , expo ] ) .
the curves ( going from top to bottom along the vertical axis ) correspond to @xmath134 , 0.9 , 0.85 , 0.8 , , 0.1 , 0.05 .
the axes are in the units of @xmath135 .
( the decrease of @xmath5 below @xmath136 does not mean that it is below the apparent horizon radius , because the actual @xmath0 also decreases . ) ] let us choose @xmath137 to be some positive constant ( call it @xmath138 ) , which corresponds to the case when the luminosity at infinity equals the constant fraction @xmath138 of the eddington value ( see section 3.2 of @xcite and eq .
( 2.33 ) in @xcite ) .
the particular behavior of mass which ensures this , @xmath139 , namely exponential decay @xmath140 , makes the equilibrium condition yield [ equilibrium - in - r , expo ] r= .
note that in this case @xmath141 hence the ultimate value of the quasi - equilibrium radius ( reached at @xmath142 here ) is certainly not @xmath123 .
its behavior at final stages of the center s evaporation " can be inferred by linearizing the equilibrium condition ( and then again its solution @xmath5 ) in @xmath0 which yields @xmath143 hence the quasi - equilibrium radius tends to zero exponentially together with @xmath0 . fig .
[ ru - expo ] shows an example of the dependence of the equilibrium curves @xmath144 on the parameter @xmath138 for a given @xmath57 .
@xmath145{fig2a.eps}&\quad \includegraphics[scale=0.55]{fig2b.eps}\\[.4 cm ] \quad\mbox{(a)}\quad & \quad \mbox{(b ) } \end{array}$ ] + another natural possibility is a linear decrease of mass , @xmath146 .
in such a case the equilibrium condition ( [ equilibrium - in - r ] ) reads [ equilibrium - in - r , linear ] r = n(- ) and the ultimate value of the quasi - equilibrium radius is @xmath147 .
this radius is a global maximum of @xmath124 ; actually @xmath124 has an overall tendency to grow .
this is expected since the linear decrease of mass corresponds to the increase of the effective interaction parameter @xmath148 .
the quasi - equilibrium radius in fact need not increase all the time ( for small @xmath57 , it rather decreases initially and later remains more or less constant ) , but finally ( at times @xmath149 ) it always grows with @xmath6 .
the rate of this growth is again obtained by restricting just to linear terms in @xmath0 in the above equation ; this yields @xmath150 the quasi - equilibrium @xmath124 dependence is parameterized by @xmath151 and @xmath57 , @xmath135 playing the role of a scale factor . for illustration
, one can either choose a certain fixed @xmath57 and plot @xmath124 for different possible @xmath151 , but it it more suitable to do it the other way round , because the lifetime of the star " is given by @xmath152 and it is better to have the latter the same for all the curves . see fig .
[ ru - linear ] for an example of the dependence of such quasi - equilibrium curves @xmath153 on @xmath154 , for a given @xmath151 . in accord with formula ( [ r - equilibrium - linear ] ) derived below , the quasi - equilibrium radius first drifts up / down if @xmath155 is bigger / smaller than @xmath156 . figs .
[ ru - expo ] and [ ru - linear ] show that the @xmath124 behavior is rather different for the exponential and linear decrease of @xmath0 with @xmath6 : in absolute measures ( in units of the initial mass @xmath135 ) , the exponential decrease of @xmath0 leads to decrease of @xmath5 , whereas the linear decrease of @xmath0 leads to increase of @xmath5 with @xmath6 .
if expressed in units of the actual mass @xmath0 , the quasi - equilibrium radii of course have stronger tendency to grow in time , and actually they ( slightly ) do so even in the exponential case . on a more general level , this reminds us that it is delicate ( if at all possible ) to compare locations in a non - stationary spacetime at different moments ( and even harder to compare them in different spacetimes , i.e. , with different @xmath14 ) . with time @xmath6 for the mass profile @xmath157 with the following parameter choice : @xmath158 , @xmath159 , @xmath160 and different values of @xmath161\cdot 10 ^ 3 $ ] , with axes in units of @xmath162 .
the black dashed curve corresponds to the apparent horizon . during the transition
phase the quasi - equilibrium radius is even more enhanced for increasing values of @xmath57 , i.e. , when the interaction with the radiation field becomes stronger .
] finally , one can assume that the mass smoothly decreases / increases between some two fixed values @xmath162
, @xmath163 .
this situation can be modeled by the mass profile @xmath164 for the outgoing ( @xmath165 ) and ingoing radiation ( @xmath166 ) cases , where the constant rate parameter @xmath151 governs the time scale of the transition between the two asymptotic schwarzschild spacetimes ( the smaller the value of @xmath151 the longer the transition ) .
the behavior of the quasi - equilibrium radius as a function of @xmath6 is shown in fig .
[ ru - expo2 ] in the outgoing case for selected values of @xmath57 .
interestingly , during the transition phase the quasi - equilibrium radius can reach large values when the interaction with the radiation field is very strong . to compare with the corresponding schwarzschild test radiation case , consider a linear mass decrease @xmath146 with constant and very small @xmath151 ( @xmath167 ) .
the interaction parameter and its dimensionless counterpart are then a= , _
let us stress that @xmath168 is not necessarily small ( in contrast to @xmath151 ) since @xmath57 can be large .
actually the necessary condition for the equilibrium to be at all possible , ( [ necessary - condition ] ) , requires _ 0 > . by rewriting ( [ equilibrium - in - r , linear ] ) as
@xmath169 one sees that the above condition can also be understood as the requirement of positivity of the mass @xmath0 ( which is here ensured at times @xmath170 ) . substituting for @xmath14 in the lapse function yields @xmath171
so the equilibrium condition @xmath172 becomes ( 1-)(n_0 ^ 2 + 2 ) + -_0=0 .
expanding this now only up to terms linear in @xmath151 , one obtains the dimensionless equation n_0(n_0-_0)+ = 0 whose relevant solution ( linearized in @xmath151 ) reads [ r - equilibrium - linear ] r= - . in the limit when even the linear term can be neglected ( @xmath173 ) , we get the familiar schwarzschild equilibrium solution for a test flux , @xmath174 , which only exists when @xmath175 . writing the above solution in dimensionless form ( @xmath176 ) ,
0(1-_0 ^ 2)-_0 ^ 2(1 - 3_0 ^ 2 ) , one sees that if @xmath151 can not be neglected completely but the linear approximation is sufficient , the quasi - equilibrium radius starts from a slightly lower value @xmath177 than in the schwarzschild case and drifts with time @xmath6 towards even smaller values when @xmath178 whereas towards larger values when @xmath179 ; this initial behavior is clearly shown in fig .
[ ru - linear](b ) in the zone before the total mass loss becomes comparable to the initial value and the curves quickly rise . during this initial phase ,
the quasi - equilibrium radius provides a slowly moving target near the corresponding schwarzschild value . note that assuming the mass profile as in eq .
( [ massprofile ] ) also leads to a constant rate of loss of energy that occurs in the test radiation field .
in fact , for small values of @xmath151 , i.e. for a very slow transition between a past asymptotic schwarzschild spacetime with mass @xmath162 and a future asymptotic schwarzschild spacetime with mass @xmath163 , we have in the outgoing case @xmath180 , so that the same considerations as above apply too .
one can also be interested in circular equilibrium orbits , i.e. , with only @xmath181 .
in this case eqs .
( [ motion2 ] ) reduce to @xmath182 + \frac{\gamma n}{r}(\nu^{\hat \phi}){}^2 \pm\frac{\tilde \sigma t_{uu}}{n^2}\ , , \\
\label{motion4due } \frac{\rmd
\nu^{\hat \phi}}{\rmd \tau}&= & -\frac{\tilde \sigma t_{uu}}{n^2}\nu^{\hat \phi}\,.\end{aligned}\ ] ] re - expressing @xmath49 in terms of @xmath183 , the second equation can be solved for @xmath184 leading to ( ) = , where @xmath185 and @xmath186 .
solving then the latter equation for @xmath187 and substituting into eq .
( [ motion4due ] ) gives a first order equation for @xmath85 .
therefore , circular equilibrium exists only if @xmath0 evolves with @xmath6 according to this equation for @xmath188 constant .
this condition selects a particular mass profile . assuming instead a given mass profile , one obtains the evolution of the radius such that the orbit is momentarily circular , i.e. , a quasi - equilibrium state .
[ fig : rdiu_circ ] shows the evolution of such a quasi - equilibrium radius @xmath5 with time @xmath6 for the mass profile ( [ massprofile ] ) in the outgoing case and the corresponding behavior of the azimuthal velocity @xmath187 . in the case of geodesic motion @xmath189
constant and [ nuphigeo ] ( ^)^2=- , according to eqs .
( [ motion4uno])([motion4due ] ) ( see also eqs .
( [ u^u - circular - geod])([u^phi - circular - geod ] ) ) , which is an equation for @xmath5 .
the first term represents the keplerian one . since in the outgoing case
@xmath77 is always negative , the second term in eq .
( [ nuphigeo ] ) is always positive , so that the radius has to be greater than the keplerian one .
if the transition between the two asymptotic schwarzschild spacetimes occurs very slowly , the second term is negligible so that the actual behavior of @xmath124 is practically indistinguishable from the keplerian one .
the effect of the interaction with the radiation field is a growth of the quasi - equilibrium radius for purely azimuthal motion during the transition phase just as in the case of quasi - equilibrium at rest .
@xmath190{fig4a.eps}&\quad \includegraphics[scale=0.3]{fig4b.eps}\\[.4 cm ] \quad\mbox{(a)}\quad & \quad \mbox{(b)}\\ \end{array}$ ] +
@xmath190{fig5a.eps}&\quad \includegraphics[scale=0.3]{fig5b.eps}\\[.4 cm ] \quad\mbox{(a)}\quad & \quad \mbox{(b ) } \end{array}$ ] + @xmath190{fig6a.eps}&\quad \includegraphics[scale=0.3]{fig6b.eps}\\[.4 cm ] \quad\mbox{(a)}\quad & \quad \mbox{(b)}\\ \end{array}$ ] + is shown in the case of outgoing radiation with the same parameter choice as in fig .
[ fig:4 ] , with axes given in units of @xmath162 .
the initial conditions are @xmath191 , @xmath192 , @xmath193 , @xmath194 and @xmath195 $ ] , in ( a ) to ( c ) respectively .
the value @xmath196 corresponds to a circular geodesic in the past asymptotic schwarzschild spacetime with mass @xmath162 . in this case shown in ( b ) , the geodesic orbit exhibits an oscillating behavior while approaching the circular geodesic of the future asymptotic schwarzschild spacetime with mass @xmath163 at keplerian radius @xmath197 and keplerian speed @xmath198 . in contrast , the accelerated particle spirals towards the apparent horizon at @xmath199 and reaches the latter in a finite proper time interval . before falling into the apparent horizon , the value @xmath200 ( and smaller values ) corresponds to a spiraling behavior towards the apparent horizon in both the geodesic and accelerated cases .
the value @xmath201 shows oscillations outside the initial radius , from which the accelerated particles then escape .
further increase of @xmath202 would lead to particle escape in both cases .
, title="fig : " ] + is shown in the case of outgoing radiation with the same parameter choice and initial conditions as in fig .
[ fig:6](b ) , but for different values of the friction parameter @xmath57 . for very small values of @xmath57 the accelerated orbit is close to the geodesic one . as the interaction with the background radiation field becomes stronger , i.e. , for increasing values of @xmath57 , the asymptotic radius decreases to even smaller values .
further increasing of @xmath57 causes the particle to cross the apparent horizon . ]
@xmath190{fig8a.eps}&\quad \includegraphics[scale=0.3]{fig8b.eps}\\[.4 cm ] \quad\mbox{(a)}\quad & \quad \mbox{(b ) } \end{array}$ ] + in order to numerically integrate the equations of motion ( [ motion2 ] ) we must specify the mass function @xmath14 .
we adopt the mass profile ( [ massprofile ] ) , so that the mass smoothly decreases / increases between the value @xmath162 corresponding to the past asymptotic schwarzschild spacetime and the value @xmath163 of the future asymptotic schwarzschild spacetime . another possibility that we will not consider for numerical study would be a mass function which decreases / increases linearly with @xmath6 over a finite interval of time ( see previous section ) . as typical solutions we consider the numerical examples shown in figs .
[ fig:4][fig:8 ] , where the geodesic behavior is compared with the motion of a particle interacting with the background radiation field in the case of outgoing radiation .
the same analysis can be easily repeated for the ingoing case .
we choose @xmath160 and fix the value of the friction parameter obtained by dividing @xmath57 by the mass @xmath162 of the past asymptotic schwarzschild spacetime to be @xmath203 , indicating a strong interaction of the test particle with the background radiation field . for small values of @xmath57 deviations from geodesic motion are not significant .
we have investigated two different conditions for the initial radius of the orbit : @xmath204 and @xmath205 .
the typical feature is that an initially circular orbit in the past asymptotic schwarzschild spacetime spirals inwards if its velocity is smaller than the keplerian one ; instead for greater values both the geodesic and accelerated particles escape outwards .
if the initial velocity equals the keplerian one , instead , the geodesic particle escapes , whereas the accelerated one spirals towards the apparent horizon ( see fig .
[ fig:4 ] ) . increasing the initial value of the radius
enriches the situation , as shown in fig .
[ fig:6 ] .
in fact , the initially circular orbit in the past asymptotic schwarzschild spacetime undergoes a transition to a quasi - circular geodesic in the future asymptotic schwarzschild spacetime if its velocity equals the keplerian one , i.e. the motion turns out to be confined in a region close to such a geodesic orbit , because the path oscillates between a minimum and a maximum radius .
if the interaction is not so strong the contribution of mass variation dominates with respect to that due to the acceleration , leading to an oscillating behavior of the accelerated orbit around an asymptotic radius as in in the case of geodesic motion ( see fig .
[ fig:7 ] ) . finally , if the path is initially radial , it remains radial , and the particle can eventually escape if it is directed outwards with a large enough initial speed ( see fig .
[ fig:8 ] ) .
note that in this case geodesic and accelerated orbits are practically indistinguishable even for strong interaction .
it is worth to recall that in the original works on poynting - robertson effect the main concern was a situation outside normal star , where it is only relevant to consider the outgoing - radiation case . in the present paper , we are also actually mainly interested in ultracompact centre like black hole , when the ingoing - radiation case is also relevant .
namely , a black hole does not itself radiate ( if not taking quantum effects into account ) , it also does not allow any stable accretion configuration ( as a radiating source ) near the horizon , and finally , even if such a source was there , most of its radiation would fall below the horizon .
therefore , at least in the vicinity of the horizon the ingoing radiation occurs more probably than outgoing . in the case of ingoing radiation the dominant effect for both geodesic and accelerated particles
is a push towards the apparent horizon .
an initially circular orbit in the past asymptotic schwarzschild spacetime always spirals inwards , eventually reaching the apparent horizon after a few revolutions in a finite proper time interval .
the coupling with the background radiation field causes accelerated particles to cross the apparent horizon before the corresponding geodesics .
the radial motion is characterized by the same feature as in the case of outgoing radiation .
the vaidya spacetime with a thomson interaction of its null dust with test particle motion provides an arena for the investigation of a poynting - robertson - like effect in a self - consistent way without the requirement that the null dust itself be a test field .
one should admit that using the exact , vaidya solution , where the radiation flux is tied consistently to the mass loss / gain by the centre , has only a theoretical importance in most situations , because the mass change is almost always negligible with respect to the mass itself .
( this may only be false in final stages of black hole evaporation . ) however , the properties of this effect evident in the simpler case of a test radiation field in the schwarzschild spacetime are reflected by those of the appropriate limit of the vaidya case , but the latter case allows one to see how they change under more extreme conditions where the outgoing radiation itself contributes to the gravitational field .
all authors thank icranet for support . os also thanks for support from czech projects gacr-202/09/0772 , msm0021610860 and lc06014 . | motion of massive test particles in the nonvacuum spherically symmetric radiating vaidya spacetime is investigated , allowing for physical interaction of the particles with the radiation field in terms of which the source energy - momentum tensor is interpreted .
this poynting - robertson - like effect " is modeled by the usual effective term describing a thomson - type radiation drag force .
the equations of motion are studied for simple types of motion including free motion ( without interaction ) , purely radial and purely azimuthal ( circular ) motion , and for the particular case of static " equilibrium ; appropriate solutions are given where possible .
the results mainly those on the possible existence of equilibrium positions are compared with their counterparts obtained previously for a test spherically symmetric radiation field in a vacuum schwarzschild background . |
the subject of order statistics deals with the properties and distributions of the ordered random variables ( rvs ) and their functions .
it has found applications in many areas of statistical theory and practice @xcite , with examples including life - testing , quality control , signal and image processing @xcite .
recently , order statistics makes a growing number of appearance in the analysis and design of wireless communication systems ( see for example @xcite ) .
for example , diversity techniques have been used for over the past fifty years to mitigate the effects of fading on wireless communication systems .
these techniques improve the performance of wireless systems over fading channels by generating and combining multiple replicas of the same information bearing signal at the receiver .
the analysis of low - complexity selection combining schemes , which select the best replica , requires some basic results of order statistics , i.e. the distribution functions of the largest one among several random variables .
more recently , the design and analysis of adaptive diversity combining techniques and multiuser scheduling strategies call for some further results on order statistics @xcite . in particular , the joint statistics of partial sums of ordered rvs are often necessary for the accurate characterization of system performance @xcite .
the major difficulty in obtaining the statistics of partial sums of ordered rvs resides in the fact that even if the original unordered rvs are independently distributed , their ordered versions are necessarily dependent due to the inequality relations among them .
recently , the co - author has applied a successive conditioning approach to convert dependent ordered random variables to independent unordered ones @xcite .
that approach , however , requires some case - specific manipulations , which may not always be generalizable . in this paper , we present an unified analytical framework to determine the joint statistics of partial sums of ordered rvs using a moment generating functions ( mgf ) based approach .
more specifically , we extend the result in @xcite , which only derive the joint mgf of the selected individual order statistics and the sum of the remaining ones , and systematically solve for the joint statistics of arbitrary partial sums of ordered rvs .
the main advantage of the proposed mgf - based unified framework is that it applies not only to the cases when all the @xmath0 ordered rvs are considered but also to those cases when only the @xmath1 @xmath2 best rvs are involved . after considering several illustrative examples , we focus on the exponential rv special case and derive the closed - form expressions of the joint statistics .
these statistical results can apply to the performance analysis of various wireless communication systems over generalized fading channels .
the remainder of this paper is organized as follows . in section
ii , we summarize the main idea behind the proposed unified analytical framework , including the general idea and some special considerations .
we then introduce some common functions and useful relations in section iii , which will help make the results in later sections more compact . in section iv and v ,
we present some selected examples on the derivation of joint pdf based on our proposed approach . following this ,
we show in section vi some closed form expressions for the selected examples presented in previous sections under i.i.d .
rayleigh fading conditions .
finally , we discuss some useful applications of these results in section vii .
let @xmath3 be the order statistics obtained by arranging @xmath0 nonnegative i.i.d .
rvs , @xmath4 , in decreasing order of magnitude .
the objective is to derive the joint pdf of their partial sums involving either all @xmath0 or the first @xmath1(@xmath5 ) ordered rvs , e.g. the joint pdf of @xmath6 and @xmath7 or the joint pdf of @xmath8 and @xmath9 .
the proposed analytical framework adopts a general two - step approach : i ) obtain the analytical expressions of the joint mgf of partial sums ( not necessarily the partial sums of interest as will be seen later ) , ii ) apply inverse laplace transform to derive the joint pdf of partial sums ( additional integration may be required to obtain the desired joint pdf ) . to facilitate the inverse laplace transform calculation , the joint mgf from step i )
should be made as compact as possible .
an observation made in @xcite involving the interchange of multiple integrals of ordered rvs becomes useful in the following analysis .
suppose for example that we need to evaluate a multiple integral over the range @xmath10 .
more specifically , let @xmath11 it can be shown by interchanging the order of integration , while ensuring each pair of limits is chosen to be as tight as possible , the multiple integral in ( [ eq : multi_integral ] ) can be rewritten into the following equivalent representations , @xmath12 the general rule is that the integration limits should be selected as tight as possible using the remaining variables . for example , in the first equation of ( [ eq : integral_solution ] ) , the variables are integrated in the order of @xmath13 , @xmath14 , @xmath15 , and @xmath16 .
based on the given inequality condition @xmath10 , the integration limit of @xmath13 should be from @xmath14 to @xmath17 because @xmath14 is the tightest among the remaining rvs .
similarly , the integration limit @xmath18 is from @xmath16 to @xmath17 because @xmath13 and @xmath14 were already integrated out . after obtaining the joint mgf in a compact form
, we can derive joint pdf of selected partial sum through inverse laplace transform .
for most cases of our interest , the joint mgf involves basic functions , for which the inverse laplace transform can be calculated analytically . in the worst case , we may rely on the bromwich contour integral . in most of the case
, the result involves a single one - dimensional contour integration , which can be easily and accurately evaluated numerically with the help of integral tables @xcite or using standard mathematical packages such as mathematica and matlab .
the general steps can be directly applied when all @xmath0 ordered rvs are considered and the rvs in the partial sums are continuous . when these conditions do not hold , we need to apply some extra steps in the analysis in order to obtain a valid joint mgf .
specifically , when only the best @xmath1 ( @xmath5 ) ordered rvs are involved in the partial sums , we should consider the @xmath1th order statistics @xmath19 separately . without this separation ,
we can not find the valid integration limit in calculating the joint mgf . as an illustration , the example in fig .
[ example_1 ] considers 3-dimensional joint pdf of the partial sums of three group of rvs which are @xmath20 , @xmath21 , and @xmath22 , with @xmath23 and @xmath24 . following the proposed approach
, we will derive the 4-dimensional joint mgf in step i ) by considering @xmath25 separately , i.e. the joint mgf of the following four groups of rvs @xmath20 , @xmath21 , @xmath26 , and @xmath27 .
after obtaining the corresponding 4-dimensional joint mgf , we need to perform another finite integration to obtain the desired 3-dimensional joint pdf . when the rvs involved in one partial sum
is not continuous , i.e. separated by the other rvs , we need to divide these rvs into smaller sums .
the example in fig .
[ example_2 ] illustrate this process .
if we consider 3-dimensional joint pdf of @xmath28 , @xmath29 , @xmath30 , @xmath31 , @xmath32 , and @xmath22 , then there are three partial sum of rvs @xmath28 , @xmath29 , @xmath30 , @xmath31 , @xmath32 , and @xmath22 .
note that the first group is split by the second group of rvs ( as such discontinuous ) .
more specifically , the second group , @xmath32 , split the original group , @xmath28 , @xmath29 , @xmath30 , @xmath31 , into two groups as @xmath33 , and @xmath34 .
therefore , we also consider this split group @xmath33 and @xmath34 as two smaller groups . as a result
, we will derive 5-dimensional joint mgf of @xmath35 , @xmath32 , @xmath36 , @xmath26 , and @xmath27 in step i ) .
similarly to the first example , after the joint pdf of the new substituted partial sums are derived with inverse laplace transform in step ii ) , we can transform it to a lower dimensional desired joint pdf with finite integration .
the proposed approach is summarized in the flowchart given in fig .
[ flowchart ] , where we consider three different case separately . in the following sections , we present several examples to illustrate the proposed analytical framework .
our focus is on how to obtain a compact expression of the joint mgfs , which can be greatly simplified with the application of the following function and relations .
in this section , we introduce some common functions and their properties . these results will be used to simplify the derivation of joint mgfs in later sections . 1 . a mixture of a cdf and an mgf @xmath37 : @xmath38 where @xmath39 denotes the pdf of the rv of interest .
note that @xmath40 is the cdf and @xmath41 leads to the mgf . here , the variable @xmath42 is real , while @xmath43 can be complex .
2 . a mixture of an exceedance distribution function ( edf ) and an mgf , @xmath44 : @xmath45 note that @xmath46 is the edf while @xmath47 gives the mgf .
an interval mgf @xmath48 : @xmath49 note that @xmath50 gives the mgf .
note that the functions defined in ( [ eq : cdf_mgf ] ) , ( [ eq : edf_mgf ] ) and ( [ eq : intervalmgf ] ) are related as follows @xmath51 1 .
integral @xmath52 : + based on the derivation given appendix [ ap : a ] , the integral @xmath52 defined as : @xmath53 can be expressed in terms of the function @xmath54 as @xmath55^{\left ( { k - m + 1 } \right)}.\ ] ] 2 .
integral @xmath56 : + following the similar derivation as given in appendix [ ap : a ] , the integral @xmath56 defined as @xmath57 can be expressed in terms of the function @xmath58 as @xmath59^m.\ ] ] 3 .
integral @xmath60 : + based on the derivation given in appendix [ ap : c ] , the integral @xmath60 defined as @xmath61 can be expressed in terms of the function @xmath62 as @xmath63^{\left ( { b - a - 1 } \right ) } \quad\quad\text{for } b > a.\ ] ]
assume the original rvs @xmath64 are i.i.d . with a common arbitrary pdf @xmath65
, the @xmath0-dimensional joint pdf of @xmath66 is simply given by @xcite @xmath67 where @xmath68 .
( pdf of @xmath69 among @xmath0 ordered rvs ) + let @xmath70 for convenience .
we can derive the pdf of @xmath71 $ ] as @xmath72^k } \right\},\end{aligned}\ ] ] where @xmath73 denotes the inverse laplace transform with respect to @xmath74 .
the mgf of @xmath75 is given by the expectation @xmath76 where @xmath77 denotes the expectation operator . by applying ( [ eq : cdf_mgf_multiple ] ) ,
we can obtain the mgf of @xmath78 as @xmath79^k.\end{aligned}\ ] ] therefore , we can derive the pdf of @xmath78 by applying the inverse laplace transform as @xmath80^k } \right\}.\end{aligned}\ ] ] ( joint pdf of @xmath6 and @xmath7 ) + let @xmath81 and @xmath82 for convenience .
we can obtain the 2-dimensional joint pdf of @xmath83 $ ] as @xmath84^{\left ( { k - 1 } \right ) } } \right\}&{\rm for } \ ; m=1 , \ ; z_1\ge \frac{1}{k-1}z_2 \\ \!\ ! \frac{{k!}}{{\left ( { k - m } \right)!\left ( { m - 1 } \right)!}}p\left ( { z_1 } \right)\mathcal{l}_{s_2 } ^ { - 1 } \left\ { { \left [ { c\left ( { z_1 , - s_2 } \right ) } \right]^{\left ( { k - m } \right ) } \left [ { e\left ( { z_1 , - s_2 } \right ) } \right]^{\left ( { m - 1 } \right ) } } \right\ } & { \rm for } \;m\ge2 .
\end{cases}\end{aligned}\ ] ] the second order mgf of @xmath85 $ ] is given by the expectation @xmath86 we show in appendix [ ap : d ] that by applying ( [ eq : cdf_mgf_multiple ] ) , ( [ eq : integral_solution ] ) and ( [ eq : edf_mgf_multiple ] ) , we can obtain the second order mgf of @xmath81 and @xmath82 as @xmath87^{\left ( { k - m } \right ) } } \left [ { e\left ( { \gamma _ { m : k } , \lambda _ 2 } \right ) } \right]^{\left ( { m - 1 } \right)}. \end{aligned}\ ] ] with the mgf expression given in ( [ eq : joint_mgf_1 ] ) at hand , we are now in the position to derive the 2-dimensional joint pdf of @xmath81 and @xmath82 . letting @xmath88 and @xmath89 , we can derive the 2-dimensional joint pdf by applying the inverse laplace transform as @xmath90^{\left ( { k - m } \right ) } \left [ { e\left ( { \gamma _ { m : k } , - s_2 } \right ) } \right]^{\left ( { m - 1 } \right ) } } \right\}\bigg].\end{aligned}\ ] ] based on the inverse laplace transform properties in appendix [ ap : inverse_lt ] , @xmath91 therefore , substituting ( [ eq : inverse_lt_1 ] ) in ( [ eq : joint_pdf_2_u ] ) we can obtain the desired 2-dimensional joint pdf of @xmath81 and @xmath82 .
( joint pdf of @xmath8 and @xmath92 ) + let @xmath93 and @xmath94 for convenience , then we can derive the 2-dimensional joint pdf of @xmath83 $ ] as @xmath95^{\left ( { m - 1 } \right ) } } \right\ } } \nonumber \\ \!\ ! & & \!\ !
\times \mathcal{l}_{s_2}^{-1 } \left\ { { \left [ { c\left ( { \gamma _ { m : k } , -s _ 2 } \right ) } \right]^{\left ( { k - m } \right ) } } \right\}\bigg]\quad \quad { \rm for}\ ; z_1\ge \frac{m}{k - m}z_2.\end{aligned}\ ] ] the second order mgf of @xmath85 $ ] is given by the expectation @xmath96 we show in appendix [ ap : e ] that by applying ( [ eq : cdf_mgf_multiple ] ) and ( [ eq : integral_solution ] ) and then ( [ eq : edf_mgf_multiple ] ) , we can obtain the second order mgf of @xmath75 as @xmath97^{\left ( { k - m } \right ) } \left [ { e\left ( { \gamma _ { m : k } , \lambda _ 1 } \right ) } \right]^{\left ( { m - 1 } \right ) } } .\end{aligned}\ ] ] again , letting @xmath88 and @xmath89 , we can obtain the desired 2-dimensional joint pdf of @xmath93 and @xmath94 by applying the inverse laplace transform .
let us now consider the cases where only the best @xmath1(@xmath98 ) ordered rvs are involved . assuming the original @xmath64 are i.i.d .
rvs with a common arbitrary pdf @xmath65 and cdf @xmath99 , the @xmath1-dimensional joint pdf of @xmath100 is simply given by @xcite @xmath101^{k - k_s } , \ ] ] where @xmath102 .
( pdf of @xmath103 , @xmath104 ) + let @xmath105 for convenience , then we can derive the pdf of @xmath106 as @xmath107 where @xmath108^{\left ( { k - k_s } \right ) } \mathcal{l}_{s_1 } ^ { - 1}\left\ { { \left [ { e\left
( { z_2 , - s_1 } \right ) } \right]^{\left ( { k_s - 1 } \right ) } } \right\}.\ ] ] we only need to consider @xmath19 separately in this case .
let @xmath109 and @xmath110 .
the target second order mgf of @xmath111 $ ] is given by the expectation in @xmath112^{k - k_s } } .\end{aligned}\ ] ] by simply applying ( [ eq : integral_solution ] ) and then ( [ eq : edf_mgf_multiple ] ) to ( [ eq : mgf_of_pure_gsc_integralform ] ) , we can obtain the second order mgf result as @xmath113^{k - k_s } } \left [ { e\left ( { \gamma _ { k_s : k } } , \lambda_1 \right ) } \right]^{k_s - 1}.\end{aligned}\ ] ] again , letting @xmath88 and @xmath89 , we can obtain the 2-dimensional joint pdf of @xmath109 and @xmath110 by applying the inverse laplace transform . finally , noting that @xmath114
, we can obtain the target pdf of @xmath106 with the following finite integration @xmath115 ( joint pdf of @xmath6 and @xmath116 ) let @xmath117 and @xmath118 , then the joint pdf of @xmath119 $ ] can be obtained as @xmath120 + or equivalently @xmath121 to derive the joint pdf of @xmath6 and @xmath116 , we need to consider four cases i ) @xmath122 , ii ) @xmath123 , iii ) @xmath124 and iv ) @xmath125 separately based on our unified framework . while for case iv )
, we can start with the second order mgf of @xmath6 and @xmath116 directly , we should consider substituted groups instead of the original groups for cases i ) , ii ) , and iii ) . more specifically , for cases i ) and iii ) , we need to consider @xmath19 separately as shown in fig .
[ example_3_1 ] and [ example_3_3 ] whereas , for case ii ) , as one of original groups is split by @xmath126 , we should consider substituted groups for the split group instead of original groups as shown in fig . [ example_3_2 ] . as a result
, we will start by obtaining a four order mgf for case ii ) and a three order mgf for case i ) and case iii ) . in all these cases ,
the higher dimensional joint pdf can then be used to find the desired 2-dimensional joint pdf of interest by transformation .
applying the results in ( [ eq : integral_solution ] ) , ( [ eq : cdf_mgf_multiple ] ) , ( [ eq : edf_mgf_multiple ] ) and ( [ eq : intervalmgf_multiple ] ) , we derive in appendix [ ap : g ] the following joint mgf for different cases 1 .
@xmath122 + let @xmath127 , @xmath128 , and @xmath129 , then @xmath130^{\left ( { k - k_s } \right ) } } \nonumber \\ \!\!\!\!\!\!\!\!\!\!\!\ ! & & \!\!\!\ !
\times \int\limits_{\gamma _ { k_s : k } } ^\infty { d\gamma _ { 1:k } p\left ( { \gamma _ { 1:k } } \right)\exp \left ( { \lambda _ 1 \gamma _ { 1:k } } \right)\frac{1}{{\left ( { k_s - 2 } \right)!}}\left [ { \mu \left ( { \gamma _ { k_s : k } , \gamma _ { 1:k } , \lambda _ 2 } \right ) } \right]^{\left ( { k_s - 2 } \right ) } } .
\end{aligned}\ ] ] 2 .
@xmath131 + let @xmath132 , @xmath133 , @xmath134 , and @xmath135 , then @xmath136^{\left ( { k - k_s } \right ) } } \nonumber \\ \!\!\!\!\!\!\!\!\!\!\!\!&&\!\!\!\ ! \times \!\!\!\ !
{ k_s : k } } ^\infty \!\!\ ! { d\gamma _ { m : k } p\left ( { \gamma _ { m : k } } \right)\exp \left ( { \lambda _ 2 \gamma _ { m : k } } \right)\left [ { e\left ( { \gamma _ { m : k } , \lambda _ 1 } \right ) } \right]^{\left ( { m - 1 } \right ) } \left [ { \mu \left ( { \gamma _ { k_s : k } , \gamma _ { m : k } , \lambda _ 3 } \right ) } \right]^{\left ( { k_s - m - 1 } \right ) } } .\end{aligned}\ ] ] 3 . @xmath124 + let @xmath137 , @xmath138 and @xmath139 , then @xmath140^{\left ( { k - k_s } \right ) } } \nonumber \\ \!\!\!\!\!\!\!\!\!\!\!\ ! & & \!\!\!\
! \times \!\!\!\ ! \int\limits_{\gamma _ { k_s : k } } ^\infty \!\!\!\ ! { d\gamma _ { k_s - 1:k } p\left ( { \gamma _ { k_s - 1:k } } \right)\exp \left ( { \lambda _ 2 \gamma _ { k_s - 1:k } } \right)\frac{1}{{\left ( { k_s - 2 } \right)!}}\left [ { e\left ( { \gamma _ { k_s - 1:k } , \lambda _ 1 } \right ) } \right]^{\left ( { k_s - 2 } \right ) } } .\end{aligned}\ ] ] 4 . @xmath125 + let @xmath141 and @xmath142 , then @xmath143^{\left ( { k - k_s } \right ) } } \nonumber \\ & & \times \frac{1}{{\left ( { k_s - 1 } \right)!}}\left [ { e\left ( { \gamma _ { k_s : k } , \lambda _ 2 } \right ) } \right]^{\left ( { k_s - 1 } \right ) } .\end{aligned}\ ] ] starting from the mgf expressions given above , we apply inverse laplace transforms in appendix [ ap : g ] in order to derive the following joint pdfs 1 . @xmath122 @xmath144^{\left ( { k - k_s } \right ) } u\left ( { z_1 - z_3 } \right)\mathcal{l}_{s_2 } ^ { - 1 } \left\ { { \left [ { \mu \left ( { z_3 , z_1 , - s_2 } \right ) } \right]^{\left ( { k_s - 2 } \right ) } } \right\ } , \nonumber \\ & & \text{for } z_3<z_1 , \ ; \left(k_s-2\right)z_3<z_2<\left(k_s-2\right)z_1.\end{aligned}\ ] ] where @xmath145 is the unit step function . 2 . @xmath131 @xmath146^{\left ( { k - k_s } \right ) } u\left ( { z_2 - z_4 } \right ) \nonumber \\ \!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\ ! & & \!\!\!\ !
\times \mathcal{l}_{s_1 } ^ { - 1 } \left\ { { \left [ { e\left ( { z_2 , - s_1 } \right ) } \right]^{\left ( { m - 1 } \right ) } } \right\}\mathcal{l}_{s_3 } ^ { - 1 } \left\ { { \left [ { \mu \left ( { z_4 , z_2 , - s_3 } \right ) } \right]^{\left ( { k_s - m - 1 } \right ) } } \right\ } , \nonumber \\ & & \text{for } z_4<z_2,\ ; z_1>(m-1)z_2\;\text{and}\ ; \left(k_s - m-1\right)z_4<z_3<\left(k_s - m-1\right)z_2 . \end{aligned}\ ] ] 3 .
@xmath124 @xmath147^{\left ( { k - k_s } \right ) } u\left ( { z_2 - z_3 } \right)\mathcal{l}_{s_1 } ^ { - 1 } \left\ { { \left [ { e\left ( { z_2 , - s_1 } \right ) } \right]^{\left ( { k_s - 2 } \right ) } } \right\ } , \nonumber \\ & & \text{for } z_3<z_2 , \ ; z_1>\left(k_s-2\right)z_2.\end{aligned}\ ] ] 4 .
@xmath125 @xmath148^{\left ( { k - k_s } \right ) } \mathcal{l}_{s_2 } ^ { - 1 } \left\ { { \left [ { e\left ( { z_1 , - s_2 } \right ) } \right]^{\left ( { k_s - 1 } \right ) } } \right\ } , \nonumber \\ & & \text{for } z_2 \ge \left(k_s-1\right)z_1 .
\end{aligned}\ ] ] finally , the joint pdf of @xmath6 and @xmath149 can be obtained .
note that ( [ eq : joint_pdf_1 ] ) and ( [ eq : joint_pdf_2 ] ) involve only finite integrations of joint pdfs .
therefore , while a generic closed - form expression is not possible , the desired joint pdf can be easily numerically evaluated with the help of integral tables @xcite or using standard mathematical packages such as mathematica or matlab etc .
( joint pdf of @xmath8 and @xmath9 ) + let @xmath150 and @xmath151 , then we can simply obtain the joint pdf of @xmath152 $ ] as @xmath153 omitted . note again that only the finite integrations of joint pdfs are involved .
the above novel generic results are quite general and apply to any rvs .
we now illustrate in this section some results for the i.i.d .
exponential rv special case , where the pdf and the cdf of @xmath42 are given by @xmath154 and @xmath155 respectively , where @xmath156 is the common average . therefore , ( [ eq : cdf_mgf_multiple ] ) , ( [ eq : edf_mgf_multiple ] ) and ( [ eq : intervalmgf_multiple ] ) specialize to @xmath157^m & = & \frac{{\left [ { \exp \left ( { - \left ( { s_i + \frac{1}{{\bar \gamma } } } \right)z_a } \right ) } \right]^m } } { { \left ( { \bar \gamma } \right)^m \left ( { s_i + \frac{1}{{\bar \gamma } } } \right)^m } } \label{eq : common_function_rayleigh_1 } \\ \left [ { c\left ( { z_a , - s_i } \right ) } \right]^m & = & \sum\limits_{j = 0}^m { \frac{{\left ( { - 1 } \right)^j } } { { \left ( { \bar \gamma } \right)^m } } \binom { m } { j } \left [ { \exp \left ( { - \frac{{z_a } } { { \bar \gamma } } } \right ) } \right]^j \frac{{\left [ { \exp \left ( { - z_a s_i } \right ) } \right]^j } } { { \left ( { s_i + \frac{1}{{\bar \gamma } } } \right)^m } } } \label{eq : common_function_rayleigh_2 } \\ \left [ { \mu\left ( { z_a , z_b , - s_i } \right ) } \right]^m & = & \sum\limits_{j = 0}^m \bigg[{\frac{{\left ( { - 1 } \right)^j } } { { \left ( { \bar \gamma } \right)^m } } \binom { m } { j } \exp \left ( { - \frac{{\left ( { m - j } \right)}}{{\bar \gamma } } z_a } \right ) } \exp \left ( { - \frac{j}{{\bar \gamma } } z_b } \right ) \nonumber \\ & & \times \frac{{\exp \left ( { - \left ( { \left ( { m - j } \right)z_a + jz_b } \right)s_i } \right)}}{{\left ( { s_i + \frac{1}{{\bar \gamma } } } \right)^m } } \bigg ] .
\label{eq : common_function_rayleigh_3}\end{aligned}\ ] ] after substituting ( [ eq : common_function_rayleigh_1 ] ) , ( [ eq : common_function_rayleigh_2 ] ) and ( [ eq : common_function_rayleigh_3 ] ) into the derived expressions of the joint pdf of partial sums of ordered statistics presented in the previous sections , it is easy to derive the following closed - form expressions for the pdfs by applying the classical inverse laplace transform pair given in ( [ eq : inverse_lt_pair ] ) and the property given in ( [ eq : inverse_lt_property ] ) . while some of these results have been derived using the successive conditioning approach previously , we list them here for the sake of convenience and completeness . 1 .
pdf of @xmath69 : @xmath158 2 .
joint pdf of @xmath6 and @xmath7 : @xmath159^{k - 2 } u\left ( { z_2 - j z_1 } \right ) } & \text{for } m=1 , \\ \!\
! \frac{{k!}}{{\left ( { k - m } \right)!\left ( { m - 1 } \right)!\left ( { k - 2 } \right)!\bar \gamma ^k } } \exp \left ( { - \frac{{z_1 + z_2 } } { { \bar \gamma } } } \right ) \\ \!\
\times \sum\limits_{j = 0}^{k - m } { \left ( { - 1 } \right)^j \binom{k - m}{j } \left [ { z_2 - \left ( { m + j - 1 } \right)z_1 } \right]^{k - 2 } } \\ \!\ ! \times u\left ( { z_2 - \left ( { m + j - 1 } \right)z_1 } \right ) & \text{for } m\ge2 . \end{cases}\ ] ] 3 .
joint pdf of @xmath8 and @xmath92 : @xmath160^{m - 2 } u\left ( { z_1 - m\gamma _ { m : k } } \right ) } \\
\times \sum\limits_{j = 0}^{k - m } { \left ( { - 1 } \right)^j \binom{k - m}{j}\left [ { z_2 - j\gamma _ { m : k } } \right]^{k - m - 1 } u\left ( { z_2 - j\gamma _ { m : k } } \right)}\bigg ] , \quad & m\ge2 \\ \frac{{k!}}{{\left ( { k - 1 } \right)!\left ( { k - 2 } \right)!\bar \gamma ^k } } \exp \left ( { - \frac{{z_1 + z_2 } } { { \bar \gamma } } } \right ) \\
\times \sum\limits_{j = 0}^{k - 1 } { \left ( { - 1 } \right)^j \binom{k-1}{j}\left [ { z_2 - jz_1 } \right]^{k - 2 } u\left ( { z_2 - jz_1 } \right ) } , \quad & m=1 . \end{cases}\ ] ] 4 .
pdf of @xmath103 : @xmath161^{k - k_s } \left [ { x - k_s z_2 } \right]^{k_s - 2 } } } dz_2 , \quad & k_s \ge 2\\ \frac{k}{{\bar \gamma } } \exp \left ( { - \frac{{x } } { { \bar \gamma } } } \right)\left [ { 1 - \exp \left ( { - \frac{{x } } { { \bar \gamma } } } \right ) } \right]^{k - 1 } , \quad & k_s = 1 \end{cases}\ ] ] where @xmath162^{m } = \sum\limits_{i = 0}^{m } { \left ( { - 1 } \right)^i \binom{m}{j}\left [ { \exp \left ( { - \frac{{a } } { { \bar \gamma } } } \right ) } \right]^i } .\ ] ] 5 .
joint pdf of @xmath6 and @xmath116 : 1 .
for @xmath122 , @xmath163^{k - k_s } u\left ( { z_1 - z_3 } \right ) \nonumber \\ \!\!\!\ ! & & \!\ !
\times \sum\limits_{j = 0}^{k_s - 2 } \bigg[\left ( { - 1 } \right)^j \binom{k_s-2}{j}\left [ { z_2 - \left ( { k_s - 2 - j } \right)z_3 - jz_1 } \right]^{k_s - 3 } \nonumber \\ \!\!\!\ ! & & \!\ !
\times u\left ( { z_2 - \left ( { k_s - 2 - j } \right)z_3 - jz_1 } \right)\bigg],\end{aligned}\ ] ] 2 . for @xmath131 , @xmath164^{k - k_s } \left [ { z_1 - \left ( { m - 1 } \right)z_2 } \right]^{m - 2 } \nonumber \\ \!\!\!\!\!\!\!\!\!\!\!\ ! & & \!\ !
\times \sum\limits_{j = 0}^{k_s - m - 1 } \bigg[{\left ( { - 1 } \right)^j \binom{k_s - m - 1}{j}\left [ { z_3 - \left ( { k_s - m - 1 - j } \right)z_4 - jz_2 } \right]^{k_s - m - 2 } } \nonumber \\ \!\!\!\!\!\!\!\!\!\!\!\ ! & & \!\ ! \times u\left ( { z_2 - z_4 } \right)u\left ( { z_1 - \left ( { m - 1 } \right)z_2 } \right)u\left ( { z_3 - \left ( { k_s - m - 1 - j } \right)z_4 - jz_2 } \right)\bigg].\end{aligned}\ ] ] 3 . for @xmath124 , @xmath165^{k - k_s } u\left ( { z_2 - z_3 } \right)\left [ { z_1 - \left ( { k_s - 2 } \right)z_2 } \right]^{k_s - 3 } \nonumber \\ \!\!\!\ ! & & \!\!\!\ !
\times u\left ( { z_1 - \left ( { k_s - 2 } \right)z_2 } \right).\end{aligned}\ ] ] 4 . for @xmath125 , @xmath166^{k - k_s }
\left [ { z_2 - \left ( { k_s - 1 } \right)z_1 } \right]^{k_s - 2 } u\left ( { z_2 - \left ( { k_s - 1 } \right)z_1 } \right).\end{aligned}\ ] ]
the above derived joint pdfs of partial sums of ordered statistics can be applied to the performance analysis of various wireless communication systems . in this section ,
we discuss two examples . in conventional parallel multiuser scheduling schemes ,
a particular scheduled user s signal is detected by correlating signals of all scheduled users at the receiver .
therefore , under these practical conditions , every scheduled user is interfering with every other scheduled user .
these effect is called multiple access interference ( mai ) or multiple user interference ( mui ) .
this mui is a factor which limits the capacity and performance of multiuser systems and the level of interference becomes substantial as the number of scheduled users increases . to take into account the effect of mui caused by the other scheduled users , the signal to interference plus noise ratio ( sinr ) which measures the ratio between the useful power and the amount of noise and interference generated by all the other scheduled users is used
the level of interference becomes substantial as the number of the scheduled users increases because the sinr can decrease considerably in these conditions .
this decrease in sinr can lead to a certain reduction of the rate allocated to each scheduled user . with the above motivation in mind , the impact of interference on the performance ( throughput ) of the scheduled users assuming a selection based parallel multiuser scheduling scheme is needed to investigate the total average sum rate capacity and the average spectral efficiency ( ase ) based on the sinr of the scheduled users .
the major difficulty in investigating this total average sum rate capacity and ase resides in the determination of the statistics of the sinr of the m - th scheduled user .
based on our approach , we can derive these results and then be applied to obtain the total average sum rate capacity and the ase .
based on this system model , the statistics of the sinr of the m - th scheduled user can be derived using the joint pdf of snr of @xmath167-th desired scheduled user and the sum of the snrs of the interfering @xmath168 scheduled users among total @xmath0 @xmath169 users .
application to our proposed approach and the correspondent performance results over was presented in @xcite .
minimum selection generalized selection combining ( ms - gsc ) is an adaptive diversity combining scheme @xcite .
the basic idea is to combine a minimum number of best diversity paths among @xmath170 ones such that the combiner output snr is above a certain preselected threshold , denoted by @xmath171 .
the performance analysis of ms - gsc scheme is challenging due to the ordering operation on diversity paths and adaptive combining operation . in @xcite , the average symbol error rate ( ser ) of ms - gsc was calculated as the weighted sum of the conditional average ser given that @xmath167 paths are combined , with the weights being the probabilities of combining @xmath167 paths and the average ser of conventional generalized selection combining ( gsc ) , which always combines @xmath167 best paths , used as the conditional average ser . however , because the receiver with ms - gsc may combine @xmath167 best paths only under the condition that the combined snr of first @xmath172 best paths is below the output threshold , the statistics of combined snr with ms - gsc given that @xmath167 paths are combined is different from that with corresponding conventional gsc . as such
, the average ser result in @xcite should serve as an approximation . to obtain the exact statistics of the combiner output snr with ms - gsc
, we need the joint statistics of the @xmath167th largest rv and the partial sum of the @xmath173 largest rvs @xcite .
specifically , the exact cdf of the combiner output snr with ms - gsc involves the probability @xmath174 $ ] , which can be expressed in terms of the joint pdf of @xmath175 and @xmath176 . with our proposed analytical framework , this joint pdf can be obtained in a systematic fashion , with the generic expression for generalized fading environments given in eq .
this new statistical result allows us accurately evaluate the performance of ms - gsc scheme over general fading channels .
additionally , this proposed method can be applied to the performance analysis of various wireless communication systems with diversity techniques such that the performance analysis of gsc rake receiver with self - interference and so on .
it is easy to see from ( [ eq : closed_form_5 ] ) , ( [ eq : closed_form_6 ] ) , ( [ eq : closed_form_7 ] ) and ( [ eq : closed_form_8 ] ) that the derivations of the pdf from the mgf involve the classical inverse laplace transform pair @xcite @xmath177 and the laplace transform property @xcite @xmath178
in this appendix , we derive eq .
( [ eq : cdf_mgf_multiple ] ) .
let us first consider the case @xmath179 .
noting that @xmath180 , we can rewrite @xmath181 as @xmath182 for the case of @xmath183 , after applying integration by part and ( [ eq : appendix_a_2 ] ) , we have @xmath184 ^ 2 } \right|_0^{\gamma _ { k - 2:k } } - \int\limits_0^{\gamma _ { k - 2:k } } { d\gamma _ { k - 1:k } c\left ( { \gamma _ { k - 1:k } , \lambda } \right)c'\left ( { \gamma _ { k - 1:k } , \lambda } \right)}.\end{aligned}\ ] ] after moving the integration part of the right hand side ( rhs ) to the left hand side ( lhs ) and some manipulation , we can show @xmath185 ^ 2.\ ] ] using ( [ eq : appendix_a_4 ] ) in ( [ eq : appendix_a_3 ] ) , ( [ eq : appendix_a_3 ] ) can be re - written as @xmath186 ^ 2 .\ ] ] similarly for the case of @xmath187 , with the help of integration by part , we can obtain the following @xmath188 ^ 2 } \nonumber \\ & & = \frac{1}{2}\left .
{ \left [ { c\left ( { \gamma _ { k - 2:k } , \lambda } \right ) } \right]^3 } \right|_0^{\gamma _ { k - 3:k } } - \int\limits_0^{\gamma _ { k - 3:k } } { d\gamma _ { k - 2:k } \left [ { c\left ( { \gamma _ { k - 2:k } , \lambda } \right ) } \right]^2 c'\left ( { \gamma _ { k - 2:k } , \lambda } \right)}.\end{aligned}\ ] ] after some manipulation and substitution , we have @xmath189 ^ 3.\end{aligned}\ ] ] this process can be generalized to arbitrary @xmath167 , which leads to the result in eq .
( [ eq : cdf_mgf_multiple ] ) .
in this appendix , we show the derivation of eq.([eq : intervalmgf_multiple ] ) .
let @xmath190 using similar manipulations to the ones used in the previous appendices [ ap : a ] , we can write @xmath191 @xmath192 ^ 2,\ ] ] @xmath193 ^ 3.\end{aligned}\ ] ] using these results , @xmath60 can be found in closed - form as @xmath194^{\left ( { b - a - 1 } \right)}.\end{aligned}\ ] ]
starting with ( [ eq : joint_mgf_1 ] ) , by simply applying ( [ eq : cdf_mgf_multiple ] ) , we can obtain the following result easily @xmath195^{\left ( { k - m } \right)}.\ ] ] by inserting ( [ eq : joint_mgf_3_integralform ] ) into ( [ eq : joint_mgf_2_integralform ] ) , the mgf has the following form : @xmath196^{\left ( { k - m } \right ) } } .
\end{aligned}\ ] ] by applying the integral solution presented in ( [ eq : integral_solution ] ) , we can re - write ( [ eq : joint_mgf_4_integralform ] ) as the following : @xmath197^{\left ( { k - m } \right ) } } \nonumber \\ \!\!\!\!\!\!\!\!\!\!\!\ ! & & \times \!\ ! \int\limits_{\gamma _ { m : k } } ^\infty \!\!\!\ ! { d\gamma _ { m - 1:k } p\left ( { \gamma _ { m - 1:k } } \right)\exp \left ( { \lambda _ 2 \gamma _ { m - 1:k } } \right ) \!\!\!\ ! \int\limits_{\gamma _ { m - 1:k } } ^\infty \!\!\!\ ! { d\gamma _ { m - 2:k } p\left ( { \gamma _ { m - 2:k } } \right)\exp \left ( { \lambda _ 2 \gamma _ { m - 2:k } } \right ) } } \nonumber \\ & & \times \cdots \times \!\!\!\ ! \int\limits_{\gamma _ { 2:k } } ^\infty \!\!\!\
! { d\gamma _ { 1:k } p\left ( { \gamma _ { 1:k } } \right)\exp \left ( { \lambda _ 2 \gamma _ { 1:k } } \right)}.\end{aligned}\ ] ] by simply applying ( [ eq : edf_mgf_multiple ] ) , we can obtain the following result easily @xmath198^{\left ( { m - 1 } \right ) } \end{aligned}\ ] ] by inserting ( [ eq : joint_mgf_6_integralform ] ) into ( [ eq : joint_mgf_5_integralform ] ) , we can obtain the second order mgf of @xmath81 and @xmath82 easily as the following : @xmath199^{\left ( { k - m } \right ) } } \left [ { e\left ( { \gamma _ { m : k } , \lambda _ 2 } \right ) } \right]^{\left ( { m - 1 } \right)}.\ ] ]
starting with ( [ eq : joint_mgf_2 ] ) , by simply applying ( [ eq : cdf_mgf_multiple ] ) , we can obtain the following result easily @xmath200^{\left ( { k - m } \right)}.\end{aligned}\ ] ] by inserting ( [ eq : joint_mgf_8_integralform ] ) into ( [ eq : joint_mgf_7_integralform ] ) , the mgf has the following form : @xmath201^{\left ( { k - m } \right ) } } .\end{aligned}\ ] ] by applying the integral solution presented in ( [ eq : integral_solution ] ) , we can re - write ( [ eq : joint_mgf_9_integralform ] ) as the following : @xmath202^{\left ( { k - m } \right ) } } \nonumber \\ \!\!\!\!\!\!\!\!\!\!\!\ ! & & \times \!\ ! \int\limits_{\gamma _ { m : k } } ^\infty { d\gamma _ { m - 1:k } p\left ( { \gamma _ { m - 1:k } } \right)\exp \left ( { \lambda _ 1 \gamma _ { m - 1:k } } \right ) \cdots \!\!\!\ ! \int\limits_{\gamma _ { 2:k } } ^\infty { d\gamma _ { 1:k } p\left ( { \gamma _ { 1:k } } \right)\exp \left ( { \lambda _ 1 \gamma _ { 1:k } } \right ) } } .
\end{aligned}\ ] ] by simply applying ( [ eq : edf_mgf_multiple ] ) , we can obtain the following result easily @xmath203^{\left ( { m - 1 } \right ) } .\end{aligned}\ ] ] substituting ( [ eq : joint_mgf_11_integralform ] ) in ( [ eq : joint_mgf_10_integralform ] ) , we can obtain the second order mgf of @xmath93 and @xmath94 as @xmath204^{\left ( { k - m } \right ) } \left [ { e\left ( { \gamma _ { m : k } , \lambda _ 1 } \right ) } \right]^{\left ( { m - 1 } \right ) } } .\end{aligned}\ ] ]
in this appendix , we derive the joint pdf of @xmath6 and @xmath116 among @xmath0 ordered rvs by considering four cases i ) @xmath122 , ii ) @xmath131 , iii ) @xmath124 and iv ) @xmath125 separately .
consider first the case ii ) , @xmath131 .
let @xmath132 , @xmath133 , @xmath134 and @xmath135 .
the 4-dimensional mgf of @xmath205 $ ] is given by the expectation @xmath206^{\left ( { k - k_s } \right ) } } .
\end{aligned}\ ] ] with the help of ( [ eq : integral_solution ] ) , ( [ eq : cdf_mgf_multiple ] ) , ( [ eq : edf_mgf_multiple ] ) and ( [ eq : intervalmgf_multiple ] ) , we can easily obtain the 4-dimensional mgf of @xmath132 , @xmath133 , @xmath134 and @xmath135 as @xmath207^{\left ( { k - k_s } \right ) } } \nonumber \\ & & \!\!\!\
! \times \!\!\!\ !
{ k_s : k } } ^\infty { d\gamma _ { m : k } p\left ( { \gamma _ { m : k } } \right)\exp \left ( { \lambda _ 2 \gamma _ { m : k } } \right)\left [ { e\left ( { \gamma _ { m : k } , \lambda _ 1 } \right ) } \right]^{\left ( { m - 1 } \right ) } \left [ { \mu \left ( { \gamma _ { k_s : k } , \gamma _ { m : k } , \lambda _ 3 } \right ) } \right]^{\left ( { k_s - m - 1 } \right ) } } .\end{aligned}\ ] ] having a mgf expression given in ( [ eq : app_joint_mgf_gsc_2 ] ) , we are now in the position to derive the 4-dimensional joint pdf of @xmath132 , @xmath133 , @xmath134 and @xmath135 . letting @xmath88 , @xmath89 , @xmath208 , and @xmath209 we can derive the 4-dimensional pdf of @xmath132 , @xmath133 , @xmath134 and @xmath135 by applying an inverse laplace transform yielding @xmath210^{\left ( { k - k_s } \right ) } } \nonumber \\ \!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\ ! & & \!\!\!\
! \times \!\!\!\ !
\int\limits_{\gamma _ { k_s : k } } ^\infty \!\!\!\ ! { d\gamma _ { m : k } \bigg[p\left ( { \gamma _ { m : k } } \right)\mathcal{l}_{s_2 } ^ { - 1 } \left\ { { \exp \left ( { - s_2 \gamma _ { m : k } } \right ) } \right\}\mathcal{l}_{s_1 } ^ { - 1 } \left\ { { \left [ { e\left ( { \gamma _ { m : k } , - s_1 } \right ) } \right]^{\left ( { m - 1 } \right ) } } \right\ } } \nonumber \\ \!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\ ! & & \!\!\!\ !
\times \mathcal{l}_{s_3 } ^ { - 1 } \left\ { { \left [ { \mu \left ( { \gamma _ { k_s : k } , \gamma _ { m : k } , - s_3 } \right ) } \right]^{\left ( { k_s - m - 1 } \right ) } } \right\}\bigg ] \nonumber \\ \!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\
& = & \!\!\!\ ! \frac{f}{{\left ( { k_s - m - 1 } \right)!\left ( { m - 1 } \right)!}}p\left ( { z_2 } \right)p\left ( { z_4 } \right)\left [ { c\left ( { z_4 } \right ) } \right]^{\left ( { k - k_s } \right ) } u\left ( { z_2 - z_4 } \right ) \nonumber \\ \!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\ ! & & \!\!\!\ !
\times \mathcal{l}_{s_1 } ^ { - 1 } \left\ { { \left [ { e\left ( { z_2 , - s_1 } \right ) } \right]^{\left ( { m - 1 } \right ) } } \right\}\mathcal{l}_{s_3 } ^ { - 1 } \left\ { { \left [ { \mu \left ( { z_4 , z_2 , - s_3 } \right ) } \right]^{\left ( { k_s - m - 1 } \right ) } } \right\}. \end{aligned}\ ] ] with this 4-dimensional joint pdf , letting @xmath211 and @xmath212 we can obtain the 2-dimensional joint pdf of @xmath213 $ ] by integrating over @xmath214 and @xmath215 yielding @xmath216 or equivalently we can obtain the 2-dimensional joint pdf of @xmath213 $ ] by integrating over @xmath217 and @xmath215 giving @xmath218 we now consider the case i ) for which @xmath122 . let @xmath127 , @xmath128 and @xmath129 for convenience . for this case ,
the 3-dimensional mgf of @xmath219 $ ] is given by the expectation @xmath220^{\left ( { k - k_s } \right ) } } .\end{aligned}\ ] ] with the help of ( [ eq : integral_solution ] ) and ( [ eq : intervalmgf_multiple ] ) , we can easily obtain the 3-dimensional mgf of @xmath127 , @xmath128 and @xmath139 as @xmath221^{\left ( { k - k_s } \right ) } } \nonumber \\ \!\!\!\!\!\!\!\!\!\!\!\ ! & & \!\!\!\ !
\times \!\!\!\ ! \int\limits_{\gamma _ { k_s : k } } ^\infty { d\gamma _ { 1:k } p\left ( { \gamma _ { 1:k } } \right)\exp \left ( { \lambda _ 1 \gamma _ { 1:k } } \right)\frac{1}{{\left ( { k_s - 2 } \right)!}}\left [ { \mu \left ( { \gamma _ { k_s : k } , \gamma _ { 1:k } , \lambda _ 2 } \right ) } \right]^{\left ( { k_s - 2 } \right ) } } .
\end{aligned}\ ] ] having an expression from the mgf as given in ( [ eq : app_joint_mgf_gsc_3 ] ) , we are now in the position to derive the 3-dimensional joint pdf of @xmath127 , @xmath128 and @xmath139 . letting @xmath88 , @xmath89 and @xmath208 , we can derive the 3-dimensional joint pdf of @xmath127 , @xmath128 and @xmath139 by applying an inverse laplace transform yielding @xmath222^{\left ( { k - k_s } \right ) } } \nonumber \\ \!\!\!\!\!\!\!\!\!\!\!\ ! & & \!\!\!\
! \times \!\!\!\ ! \int\limits_{\gamma _ { k_s : k } } ^\infty { d\gamma _ { 1:k } p\left ( { \gamma _ { 1:k } } \right)\mathcal{l}_{s_1 } ^ { - 1 } \left\ { { \exp \left ( { - s_1 \gamma _ { 1:k } } \right ) } \right\}\frac{1}{{\left ( { k_s - 2 } \right)!}}\mathcal{l}_{s_2 } ^ { - 1 } \left\ { { \left [ { \mu \left ( { \gamma _ { k_s : k } , \gamma _ { 1:k } , - s_2 } \right ) } \right]^{\left ( { k_s - 2 } \right ) } } \right\ } } \nonumber \\ \!\!\!\!\!\!\!\!\!\!\!\ ! & = & \!\!\!\ ! \frac{f}{{\left ( { k_s - 2 } \right)!}}\int\limits_0^\infty { d\gamma _ { k_s : k } p\left ( { \gamma _ { k_s : k } } \right)\delta \left ( { z_3 - \gamma _ { k_s : k } } \right)\left [ { c\left ( { \gamma _ { k_s : k } } \right ) } \right]^{\left ( { k - k_s } \right ) } } \nonumber \\ \!\!\!\!\!\!\!\!\!\!\!\ ! & & \!\!\!\ !
\times \!\!\!\ ! \int\limits_{\gamma _
{ k_s : k } } ^\infty { d\gamma _ { 1:k } p\left ( { \gamma _ { 1:k } } \right)\delta \left ( { z_1 - \gamma _ { 1:k } } \right)\frac{1}{{\left ( { k_s - 2 } \right)!}}\mathcal{l}_{s_2 } ^ { - 1 } \left\ { { \left [ { \mu \left ( { \gamma _ { k_s : k } , \gamma _ { 1:k } , - s_2 } \right ) } \right]^{\left ( { k_s - 2 } \right ) } } \right\ } } \nonumber \\ \!\!\!\!\!\!\!\!\!\!\!\ ! & = & \!\!\!\ ! \frac{f}{{\left ( { k_s - 2 } \right)!}}p\left ( { z_1 } \right)p\left ( { z_3 } \right)\left [ { c\left ( { z_3 } \right ) } \right]^{\left ( { k - k_s } \right ) } u\left ( { z_1 - z_3 } \right)\mathcal{l}_{s_2 } ^ { - 1 } \left\ { { \left [ { \mu \left ( { z_3 , z_1 , - s_2 } \right ) } \right]^{\left ( { k_s - 2 } \right ) } } \right\ } .\end{aligned}\ ] ] we now consider the case iii ) for which @xmath124 .
let @xmath137 , @xmath138 and @xmath139 .
the 3-dimensional mgf of @xmath219 $ ] is given by @xmath227^{\left ( { k - k_s } \right ) } } .\
\end{aligned}\ ] ] with the help of ( [ eq : integral_solution ] ) and ( [ eq : edf_mgf_multiple ] ) , we can easily obtain the 3-dimensional mgf of @xmath137 , @xmath138 and @xmath139 as @xmath228^{\left ( { k - k_s } \right ) } } \nonumber \\ \!\!\!\!\!\!\!\!\!\!\!\ ! & & \!\!\!\
! \times \!\!\!\ ! \int\limits_{\gamma _ { k_s : k } } ^\infty { d\gamma _ { k_s - 1:k } p\left ( { \gamma _ { k_s - 1:k } } \right)\exp \left ( { \lambda _ 2 \gamma _ { k_s - 1:k } } \right)\frac{1}{{\left ( { k_s - 2 } \right)!}}\left [ { e\left ( { \gamma _ { k_s - 1:k } , \lambda _ 1 } \right ) } \right]^{\left ( { k_s - 2 } \right ) } } .\end{aligned}\ ] ] having a mgf expression as given in ( [ eq : app_joint_mgf_gsc_4 ] ) , we are now in the position to derive the 3-dimensional joint pdf of @xmath137 , @xmath138 and @xmath139 . letting @xmath88 , @xmath89 and @xmath208 we can derive the 3-dimensional joint pdf of @xmath137 , @xmath138 and @xmath139 by applying an inverse laplace transform giving @xmath229^{\left ( { k - k_s } \right ) } } \nonumber \\ \!\!\!\!\!\!\!\!\!\!\!\ ! & & \!\!\!\
! \times \!\!\!\ ! \int\limits_{\gamma _ { k_s : k } } ^\infty { d\gamma _ { k_s - 1:k } p\left ( { \gamma _ { k_s - 1:k } } \right)\mathcal{l}_{s_2 } ^ { - 1 } \left\ { { \exp \left ( { - s_2 \gamma _ { k_s - 1:k } } \right ) } \right\}\mathcal{l}_{s_1 } ^ { - 1 } \left\ { { \left [ { e\left ( { \gamma _ { k_s - 1:k } , - s_1 } \right ) } \right]^{\left ( { k_s - 2 } \right ) } } \right\ } } \nonumber \\ \!\!\!\!\!\!\!\!\!\!\!\ ! & = & \!\!\!\ ! \frac{f}{{\left ( { k_s - 2 } \right)!}}p\left ( { z_2 } \right)p\left ( { z_3 } \right)\left [ { c\left ( { z_3 } \right ) } \right]^{\left ( { k - k_s } \right ) } u\left ( { z_2 - z_3 } \right)\mathcal{l}_{s_1 } ^ { - 1 } \left\ { { \left [ { e\left ( { z_2 , - s_1 } \right ) } \right]^{\left ( { k_s - 2 } \right ) } } \right\}.\end{aligned}\ ] ] with these 3-dimensional joint pdf , letting @xmath211 and @xmath230 we can obtain the 2-dimensional joint pdf of @xmath213 $ ] by integrating over @xmath217 yielding @xmath231 or equivalently we can obtain the 2-dimensional joint pdf of @xmath213 $ ] by integrating over @xmath214 yielding @xmath232 finally , we now consider the case iv ) for which @xmath125 .
let @xmath141 and @xmath142 .
for this case , the 2-dimensional mgf of @xmath85 $ ] is given by the expectation @xmath233^{\left ( { k - k_s } \right ) } } .
\end{aligned}\ ] ] with the help of ( [ eq : integral_solution ] ) and ( [ eq : edf_mgf_multiple ] ) , we can easily obtain the 2-dimensional mgf of @xmath141 and @xmath142 as @xmath234^{\left ( { k - k_s } \right ) } } \left [ { e\left ( { \gamma _ { k_s : k } , \lambda _ 2 } \right ) } \right]^{\left ( { k_s - 1 } \right)}.\ ] ] having an mgf expression as given in ( [ eq : app_joint_mgf_gsc_5 ] ) , we are now in the position to derive the 3-dimensional joint pdf of @xmath141 and @xmath142 .
letting @xmath88 and @xmath89 we can derive the 2-dimensional joint pdf of @xmath141 and @xmath142 by applying an inverse laplace transform yielding @xmath235^{\left ( { k - k_s } \right ) } } \nonumber \\ \!\!\!\!\!\!\!\!\!\!\!\ ! & & \!\!\!\ ! \times \mathcal{l}_{s_2 } ^ { - 1 } \left\ { { \left [ { e\left ( { \gamma _ { k_s : k } , - s_2 } \right ) } \right]^{\left ( { k_s - 1 } \right ) } } \right\ } \nonumber \\ \!\!\!\!\!\!\!\!\!\!\!\ ! & = & \!\!\!\!\frac{f}{{\left ( { k_s - 1 } \right)!}}p\left ( { z_1 } \right)\left [ { c\left ( { z_1 } \right ) } \right]^{\left ( { k - k_s } \right ) } \mathcal{l}_{s_2 } ^ { - 1 } \left\ { { \left [ { e\left ( { z_1 , - s_2 } \right ) } \right]^{\left ( { k_s - 1 } \right ) } } \right\}. \end{aligned}\ ] ] s. choi , m .- s .
alouini , k. a. qaraqe , and h .- c .
yang , `` finger replacement method for rake receivers in the soft handover region , '' _ ieee trans .
wireless commun .
twc-7 , no . 4 , pp . 11521156 , 2008 .
p. lu , h .- c .
yang , and y .- c .
ko , `` sum - rate analysis of mimo broadcast channel with random unitary beamforming , '' in _ proc . of ieee wireless commun . and networking conf .
( wcnc08 ) _ , las vegas , nevada , mar .
2008 .
y. ma and c. c. chai , `` unified error probability analysis for generalized slection combining in nakagami fading channels , '' _ ieee j. select .
areas commun .
_ , vol . 18 , no . 11 , pp . 21982210 ,
november 2000 .
m. z. win and j. h. winters , `` virtual branch analysis of sysmbol error probability for hybrid selection / maximal - ratio combining rayleigh fading , '' _ ieee trans .
_ , vol .
49 , no . 11 , pp . 19261934 , 2001 .
a. h. nuttall and p. m. baggenstoss , `` joint distributions for two useful classes of statistics , with applications to classification and hypothesis testing , '' _ ieee trans .
signal processing _ , submitted for publication .
[ online ] .
available : http://www.npt.nuwc.navy.mil/csf/papers/order.pdf s. s. nam , m .- s .
alouini , and h .- c .
yang , `` impact of interference on the performance of selection based parallel multiuser scheduling , '' in _ proc . of ieee international symposium on information theory and its applications ( isita08 )
_ , auckland , new zealand , dec . 2008 .
s. w. kim , d. s. ha , and j. h. reed , `` minimum selection gsc and adaptive low - power rake combining scheme , '' in _ proc . of ieee int .
symp . on circuits and systems .
( iscas03 ) _ , bangkok , thailand , may 2003 , pp . 357360 .
sung sik nam ( s05 , m09 ) was born in korea .
he received the b.s . and
degrees in electronic engineering from hanyang university , korea , in 1998 and 2000 , respectively .
also he received the m.s .
degree in electrical engineering from university of southern california ( usc ) , los angeles , ca , usa , in 2003 , and the ph.d .
degree at texas a&m university ( tamu ) , college station , tx , usa , in 2009 . from 1998 to 1999 , he worked as a researcher at the electronics & telecommunication research institute ( etri ) , daejeon , korea . from 2003 through 2004 , he worked as a manager at the korea telecom corporation ( kt ) , new business planning office , new business planning team , korea . since june 2009 , he has been with the department of electronic engineering at hanyang university , korea .
his research interests include the design and performance analysis of wireless communications , diversity techniques , power control , multiuser scheduling .
mohamed - slim alouini ( s94 , m98 , sm03 , f09 ) was born in tunis , tunisia .
he received the ph.d .
degree in electrical engineering from the california institute of technology ( caltech ) , pasadena , ca , usa , in 1998 .
he was with the department of electrical and computer engineering of the university of minnesota , minneapolis , mn , usa , then with the electrical and computer engineering program at the texas a&m university at qatar , education city , doha , qatar . since june 2009
, he has been a professor of electrical engineering in the division of physical sciences and engineering at kaust , saudi arabia . , where his current research interests include the design and performance analysis of wireless communication systems .
hong - chuan yang ( s00 , m03 , sm07 ) was born in liaoning , china .
he received the b.e .
degree from the changchun institute of posts & telecomm .
( now part of jilin university ) , changchun , china , in 1995 . during his graduate study at the university of minnesota , usa , dr .
yang received an m.sc .
degree in applied and computational mathematics in 2000 , an m.sc .
degree in electrical engineering in 2001 , and a ph.d .
degree in electrical engineering in 2003 .
yang has been with the department of electrical and computer engineering department of the university of victoria , victoria , b.c .
, canada since september 2003 .
he became an associate professor in july 2007 . from 1995 to 1998 , dr .
yang was a research associate at the science and technology information center ( stic ) of ministry of posts & telecomm .
( mpt ) , beijing , china .
his research work focuses on different aspects of wireless communications , with special emphasis on channel modeling , diversity techniques , system performance evaluation , cross - layer design and energy efficient communications .
yang is a recipient of the doctoral dissertation fellowship ( ddf ) award from the graduate school of the university of minnesota for the 2002 - 2003 academic year .
he is serving as an editor for _ ieee transactions on wireless communications _ and _ journal on communications and networking_. dr .
yang is a registered professional engineer in b. c. , canada . | order statistics find applications in various areas of communications and signal processing . in this paper
, we introduce an unified analytical framework to determine the joint statistics of partial sums of ordered random variables ( rvs ) . with the proposed approach , we can systematically derive the joint statistics of any partial sums of ordered statistics , in terms of the moment generating function ( mgf ) and the probability density function ( pdf ) .
our mgf - based approach applies not only when all the @xmath0 ordered rvs are involved but also when only the @xmath1 @xmath2 best rvs are considered .
in addition , we present the closed - form expressions for the exponential rv special case .
these results apply to the performance analysis of various wireless communication systems over fading channels .
joint pdf , moment generating function ( mgf ) , order statistics , probability density function ( pdf ) , rayleigh fading . |
recently there has been considerable interest in planar systems which exhibit topological phases .
these phases are characterized by topological field theories ( tqfts ) or corresponding conformal field theories ( cfts ) .
it is of great interest to have a clear understanding of the edges of such systems , and of domain walls between regions in different phases . in fractional quantum hall systems ,
where experimental support for the existence of a variety of topological phases is strongest , observations are almost entirely restricted to edge transport , and proposed devices for probing the topological order rely on interference of tunneling currents between edges @xcite . in such experiments
the electron density is usually not constant throughout the sample and islands with different filling fractions form , separated by domain walls . in lattice models with several topological phases ,
one may similarly induce phase boundaries by varying the local couplings . in this letter
, we present a general method to determine the degrees of freedom of boundaries between topological phases and their relation to the bulk degrees of freedom , based on the condensation of bosonic quasiparticles in auxiliary layered systems .
consider a medium in topological phase i supporting a single island in topological phase ii .
a setup like this has been considered in experiments by camino et al .
@xcite for the case where phase i is the fractional quantum hall state at filling fraction @xmath0 and phase ii the fqh state at @xmath1 .
we plan to give a treatment of this situation and more generally , of the hierarchy states described in @xcite and @xcite in a separate publication @xcite . here , we describe how the topological symmetry breaking procedure of @xcite can be applied in such physical settings , and follow up with two different examples .
these involve kitaev s spin model on the honeycomb lattice @xcite and a domain wall between spin polarized and unpolarized non - abelian fractional quantum hall liquids @xcite .
one way to match two different phases on a boundary is by taking the tensor product of the two boundary cfts .
this would correspond to the situation where the two phases are separated by a strip of vacuum . to describe more general interfaces , we consider the geometry sketched in figure [ fig : hallbars ] , where we started with two layers in phases i and iii , which we let partially overlap as indicated . in the middle section
we have a two layer system .
if we bring the layers close we may have some binding between degrees of freedom in i and iii .
in particular , a bosonic composite of excitations from the two layers could occur , and consequently , a condensate of such bosons may form .
this condensation will lead to a different phase for the middle region , which we denote by ii .
this situation can be analysed with the tools we have developed in @xcite .
let us assume we are given cfts or tqfts @xmath2 and @xmath3 describing phases i and iii , or even just the spectrum of topological sectors @xmath4 , the fusion rules and the topological spins @xmath5 for these phases ( the topological spins @xmath6 correspond to the fractional parts of the conformal weights @xmath7 ) .
the topological sectors of the @xmath8 theory will be labeled by pairs @xmath9 of labels from the @xmath2 and @xmath3 theories .
we can now proceed as follows .
\(i ) we can find out which bosonic sectors are possible in the @xmath8 theory .
bosons always have @xmath10 and in the simple case of sectors with quantum dimension @xmath11 , this requirement is all that is needed .
\(ii ) we assume that a condensate forms in one or more of the bosonic channels and use our method @xcite to determine the topological spectrum and fusion rules of the condensed phase .
we denote the residual theory by @xmath12 .
sectors of the @xmath8 theory branch into according to branching rules of the form @xmath13 , where the @xmath14 are integer branching multiplicities .
thus in the new phase , some sectors , or primary fields of the @xmath15 theory that we started off with may branch to the same unique @xmath12-sector and hence will be identified , while others may split into independent @xmath12-sectors .
typically , sectors which are related by fusion with the condensed boson are identified , while sectors which are invariant under fusion with the condensed boson may split .
the condensed sector itself always branches to the vacuum of the @xmath12-theory .
we require that branching and fusion are compatible , i.e. they commute . as a result , branching conserves quantum dimensions .
\(iii ) while the @xmath12-theory is required to have good fusion rules , some @xmath12-sectors will not inherit well defined spin factors or conformal weights from the uncondensed theory , basically because they have nontrivial braiding interaction with the condensed excitation .
the corresponding excitations will pull strings in the condensed medium and will be confined . in effect , this means that they are expelled from the bulk and can propagate only on the boundary of the condensed medium .
\(iv ) the @xmath12-sectors which do inherit well defined topological spins from the uncondensed theory survive in the bulk and define the theory @xmath16 , which is a tqft which describes the fusion and braiding of excitations of phase ii .
we now have a clear picture of the situation before and after condensation . before condensation ,
arbitrary excitations of the system could be labeled @xmath9 , with excitations in phases i resp .
iii labeled @xmath17 and @xmath18 resp . , where @xmath19 denotes the vacuum , or topologically trivial sector .
( we may use this notation even where the second layer is not present . ) after condensation , excitations in phases i and iii are labeled as before , but in the overlap region , we now have phase ii , with bulk excitations described by the unconfined @xmath12-sectors and boundary excitations described by the confined @xmath12-sectors . using the fusion rules of the @xmath12-theory , we can understand all the kinematics of processes that may occur when excitations are moved toward or through the edges . the crucial insight here
is that excitations in all parts of the system can be labeled by sectors of the @xmath12-theory . for the bulk and boundary of phase ii
this is true by definition and for the bulk of phase i we can use the branching rules to obtain the possible @xmath12-sectors from the actual sectors @xmath17 .
the situation is particularly simple if every sector of phase i branches to a unique @xmath12-sector .
this happens under quite general circumstances , notably whenever the theory @xmath2 does not by itself have bosonic sectors .
we can now classify all boundary processes that may occur .
for example any @xmath2-particle that is identified with a non confined @xmath12-particle can pass through the phase boundary unnoticed and vice versa , while a @xmath2-particle that corresponds to a confined @xmath12-particle can not enter the region in phase ii .
on the other hand , @xmath12-particles which are confined in phase ii but which can be obtained from a @xmath2 sector by branching can pass into the area in phase i after being driven out of phase ii , without leaving a trace on the boundary .
hence the true boundary excitations are labeled by the confined @xmath12-sectors which do not correspond to @xmath2-sectors . for processes involving three or more excitations
, we need to use the fusion rules of @xmath12 .
any process allowed by these rules could in principle occur .
for example , a @xmath2 particle corresponding to a confined @xmath12-sector @xmath20 could hit the phase boundary and split into a boundary excitation @xmath21 and a bulk excitation @xmath22 of phase ii , provided that @xmath23 according to the fusion rules of @xmath12 .
the fusion rules of @xmath12 are valid throughout . for instance , if two particles in phase i have a definite joint fusion channel , then this should be preserved even if one of the particles is moved into the region in phase ii .
in fact , the full topological state of this multi - phase system should be characterized by specifying the amplitudes for the @xmath12-fusion channels obtained on successive fusions of all the quasiparticles that are present .
this yields fusion bases analogous to the standard fusion bases of the the topological hilbert spaces of systems described by a single tqft . clearly , to actually perform the fusions involved
, it will usually be necessary to bring the quasiparticles from the bulk of phases i and ii to the boundary .
there are many situations where these general ideas apply .
in particular , coset models form a large class that can be analyzed in terms of bose condensates @xcite .
the construction of these models closely parallels the construction of figure [ fig : hallbars ] , where phase i is a @xmath24 phase , phase iii is a @xmath25 phase with the opposite chirality and in the overlap region we obtain a phase with the topological order of the @xmath26 coset , after condensation of all available bosons .
we continue with two concrete applications of a slightly different , but related type .
kitaev s honeycomb model @xcite is a model of spins living on the sites of a honeycomb lattice and interacting through nearest neighbor ising - like interactions .
the model is exactly solvable and displays two types of phases : three equivalent gapped abelian topological phases with the same topological order as the @xmath27 toric code model and central charge @xmath28 , and a gapless phase , which becomes gapped when a zeeman term is added to the hamiltonian and then displays non - abelian topological order described by the ising tqft at @xmath29 .
the abelian phase has four sectors with @xmath30 fusion rules and the ising model has the well known three sectors labeled @xmath19 , @xmath31 , @xmath32 , with @xmath19 denoting the vacuum and with nontrivial fusion rules given by @xmath33 , @xmath34 and @xmath35 .
@xmath36 @xmath37 we wish to consider a situation with an island in the abelian phase surrounded by a medium in the ising phase . to achieve this
, we start with a large disc in the ising phase ( phase i ) and imagine placing a small disk on top of it , which is in a suitable phase iii so that a bose condensate can form leaving the bulk of the small disc in the @xmath38 phase ( phase ii ) .
not surprisingly one should use an opposite chirality ising model for the small phase iii disc and then condense the @xmath39 field .
this example has been worked out in detail in section x of @xcite .
condensation leads to the identifications @xmath40 , @xmath41 , @xmath42 and @xmath43 , while the remaining field has to split : @xmath44 .
hence , the t - algebra has @xmath45 sectors and one finds that it has @xmath46 fusion rules .
the fields @xmath47 and @xmath48 will be confined because they can not be assigned a consistent conformal weight ( the corresponding identified @xmath49 fields have conformal weights that differ by @xmath50 ) .
the unconfined fields @xmath51 , @xmath52 , @xmath53 and @xmath54 correspond precisely to the fields @xmath55 and @xmath56 of the @xmath38 given in table [ tab : ising ] .
let us now look at the wall in between the phases .
of the three unconfined fields that correspond to particle excitations in the interior bulk , the fermionic @xmath54 excitation can freely move out through the wall into the exterior region and the other two bulk excitations can not .
this corresponds well to the results of @xcite where it was shown that free fermionic excitations occur throughout the phase diagram .
the confined excitations are expelled from the interior .
one , the @xmath47 excitation can move into the exterior region , while the other , the @xmath48 excitation is strictly confined to the wall .
now consider a @xmath31 excitation hitting the boundary . from the @xmath12 theory s fusion rules
, we see that @xmath57 .
hence , the @xmath31-particle can split into a boundary excitation and either an @xmath58 or an @xmath59 type toric code excitation .
this corresponds well with the results of @xcite , where @xmath31-like excitations were exhibited in the toric code using superpositions of @xmath58 and @xmath59 type excitations .
pushing another @xmath31 particle through the phase boundary will allow the confined @xmath48 excitations to annihilate , yielding either @xmath51 or @xmath54 . if the two @xmath31 particles were pair created ( had trivial fusion channel ) , the two toric code particles that form must have fusion channel @xmath19 resp .
@xmath60 , conserving overall @xmath12-charge .
now we turn to the interface between the moore read ( mr ) pfaffian fractional quantum hall state at filling @xmath61 or @xmath62 @xcite and the non - abelian spin - singlet ( nass ) state of ardonne and schoutens at @xmath63 or @xmath64 @xcite .
this was recently considered in @xcite .
we will again realize it as single layer
two layer boundary .
we concentrate on the non - abelian parts of the the mr and nass theories here and leave out the @xmath65 factors ( these can be put back in at any point ) . consider a disc with @xmath66 , corresponding to mr , with on top of that a smaller disc with @xmath67 .
the latter cft is the minimal model with @xmath68 corresponding to a tri - critical ising model .
we give the field content of the @xmath69 and @xmath70 theories in tables [ tab : ising ] and [ tab : m4 - 5 ] .
for the @xmath70 fusion rules we refer to the literature @xcite .
the @xmath72 current is the only bosonic channel in the @xmath73 model and we assume that it condenses . as this is a simple current it is straightforward to see what happens to the various fields in the model .
this is shown in table [ tab : walls ] .
we start with @xmath74 fields . in the second row
we indicate how @xmath75 of these fields become pairwise identified ( because they are equivalent modulo fusion with @xmath72 ) , and how the other two fields split .
we are thus left with the @xmath76 fields of the t - theory , which form an associative fusion algebra .
a detailed analysis of the arguments for this case shows that the fusion rules of the algebra are given by @xmath77 where the nontrivial @xmath27 element corresponds to the either the @xmath78 or the @xmath79 field .
the @xmath12-fields that are not confined form the residual bulk theory ( phase ii ) and correspond to the fields of the nass state of @xcite , as we intended .
their quantum dimensions and topological spins are given in table [ tab : nass ] ) .
the twelve fields of the full t - theory describing the interface are listed at the top of table [ tab : walls ] .
we also give their quantum dimensions , which are consistent with the fusion rules and with the decomposition of the original sectors .
it is not possible to assign unambiguous conformal weights to the confined fields , and therefore there is no consistent braid group representation for the full t - theory .
we recall ( cf .
@xcite ) that this is no problem , since the full t - theory has only a strictly one - dimensional interpretation .
it is now interesting to look more closely at the properties at the walls between the phases . in table
[ tab : walls ] we have explicitly indicated the fields in the different possible interior and exterior regions , as well as those being strictly confined to particular walls , where the latter explicitly depend on what wall we are talking about . more precisely , we give the fields that extend from the wall into the exterior mr ( phase i ) region in the fourth row and the fields that extend from the wall into the interior nass ( phase ii ) region in the fifth row , and we have to conclude that the three remaining fields can only propagate in the wall .
note that we also have to identify the @xmath81 field in the nass region , with the @xmath32 field in the mr region , which means that this field can propagate right through the wall .
it is clear that the fusion rules of the t - theory fix the kinematically allowed channels in which particles which hit the wall , coming from either the interior or the exterior region , can split into particles in the other region plus a wall - excitation .
for instance , from the t fusion rule @xmath82 , we find that if we start with a @xmath83 coming from the interior region , it can split into a @xmath31 going into the mr region and a @xmath84 staying in the wall .
but since @xmath85 , it may also happen that the @xmath83 excitation splits into the two wall - excitations @xmath84 and @xmath86 .
this scenario may also be turned around , two strict boundary excitations may fuse into a state that is not confined .
obviously there is a myriad of possibilities and we refrain from listing them here . a final comment concerns the relaxation of qubits near a wall @xcite .
if we encode a topological qubit in the nass phase , for example in a pair of sigma particles ( which indeed span a two - dimensional hilbert space ) the qubit may relax to the lowest energy state by transferring a neutral excitation to the boundary . for a pair of @xmath87 fields we have the fusion rule @xmath88 , while @xmath89 = @xmath90 .
each of these pairs can relax under emission of a @xmath91 excitation .
a @xmath91 excitation may convert into one of the pairs @xmath92 , @xmath93 , or @xmath94 , which are all strictly confined to the interface .
alternatively we may have @xmath95 where @xmath96 is confined to the wall but the @xmath31 can enter in the mr region . | we investigate domain walls between topologically ordered phases in two spatial dimensions and present a simple but general framework from which their degrees of freedom can be understood .
the approach we present exploits the results on topological symmetry breaking that we have introduced and presented elsewhere . after summarizing the method
, we work out predictions for the spectrum of edge excitations and for the transport through edges in some representative examples . these include domain walls between the abelian and non - abelian topological phases of kitaev s honeycomb lattice model in a magnetic field , as well as recently proposed domain walls between spin polarized and unpolarized non - abelian fractional quantum hall states at different filling fractions . |
one - dimensional systems remain a fascinating field in condensed - matter physics since they constitute prime examples for the breakdown of fermi - liquid theory , which has to be replaced by the luttinger - liquid paradigm.@xcite arguably the most dramatic consequence of this is the absence of electron - like quasiparticles , manifesting itself in the separation of spin and charge degrees of freedom visible for example in angle - resolved photoemission,@xcite transport,@xcite scanning tunneling spectroscopy@xcite or resonant inelastic x - ray scattering@xcite experiments as well as analytical@xcite and numerical studies of several one - dimensional models.@xcite the spectral properties of one - dimensional electron systems have been intensively investigated in the past .
these works considered the gapless luttinger liquid,@xcite gapped systems like mott insulators or charge - density wave states,@xcite luttinger liquids with impurities,@xcite corrections to the luttinger model due to non - linear dispersions@xcite or the momentum dependence of the two - particle interaction,@xcite as well as additional phonon degrees of freedom.@xcite these investigations uncovered universal power - law behaviour at low energies as well as deviations thereof , spin - charge separation visible in the propagation modes , and signatures of these features in various experimental probes . in this article
we consider another situation , namely the microscopic study of the boundary effects on one - dimensional mott insulators .
specifically we numerically study the local density of states ( ldos ) of one - dimensional hubbard models with open , ie , hard - wall , boundary conditions , where the system is at half filling and thus in its mott phase .
a previous field - theoretical analysis@xcite has shown that the fourier transform of the ldos@xcite exhibits clear signatures of propagating spin and charge degrees of freedom , thus providing a way to detect spin - charge separation .
furthermore , an additional boundary potential may lead to the formation of a boundary bound state , which manifests itself as a non - dispersing feature in the ldos .
the aim of our work is to calculate the fourier transform of the ldos directly in the microscopic lattice model using a multi - target@xcite variant of the density matrix renormalisation group ( dmrg ) method @xcite employing an expansion in chebyshev polynomials .
we find our numerical results to be fully consistent with the analytical predictions both qualitatively , ie , concerning the number of dispersion modes and their basic properties , as well as quantitatively with respect to the numerical values of the effective parameters like the mott gap and spin and charge velocities as compared to the exact results obtained from the bethe ansatz.@xcite thus our work provides a microscopic calculation of the fourier transform of the ldos in a gapped , strongly correlated electron system , showing spin - charge separation as well as the formation of a boundary bound state .
this paper is organised as follows : in sec .
[ sec : model ] we present the microscopic models to be analysed and discuss the basic setup . in sec .
[ sec : method ] we give a brief summary of the numerical method we employ to calculate the single - particle green function .
our results for the ldos of the mott insulators with open boundary conditions are discussed in sec .
[ sec : ldos ] . in sec .
[ sec : bbs ] we study the effect of a boundary potential on the ldos , in particular we analyse the properties of the boundary bound state existing for sufficiently strong boundary potentials . in sec .
[ sec : conclusion ] we summarise our results .
in this work we analyse the ldos of the one - dimensional hubbard model @xcite at half filling .
the hamiltonian is given by @xmath0 where @xmath1 and @xmath2 denote the annihilation and creation operators for electrons with spin @xmath3 at lattice site @xmath4 and @xmath5 the corresponding density operators .
the parameters @xmath6 and @xmath7 describe the hopping and repulsive on - site interaction respectively .
furthermore we consider a chain with @xmath8 sites and open boundary conditions .
since the system is assumed to be at half filling , the fermi momentum is given by @xmath9 . as is well known,@xcite in the hubbard model at half filling , ie ,
when there are @xmath8 electrons in the system , the repulsive interaction opens a gap in the charge sector and the system becomes a mott insulator .
using bosonisation the low - energy behaviour of the system is described by the massive thirring model;@xcite the ldos of which in the presence of boundaries has been analysed in refs . .
the main objective of our article is the comparison of the ldos of the hubbard model with the field - theoretical results obtained in the thirring model .
hereby the effective parameters in the field theory , ie , the mass gap and velocities , can be obtained from the exact bethe - ansatz solution of the hubbard model .
this allows us to choose the microscopic parameters such that the expected features of the fourier transformed ldos can be easily resolved in the numerical results .
in addition to the standard hubbard model we also consider its extension including a nearest - neighbour interaction @xmath10 , ie , the hamiltonian is given by@xcite @xmath11 where @xmath12 is the total electron density .
the low - energy regime of the extended hubbard model is still described @xcite by the massive thirring model . however , since is no longer integrable , the explicit relation between the microscopic parameters @xmath6 , @xmath13 and @xmath10 and the field - theory ones is not known .
thus the investigation of the phase diagram of the extended hubbard model at half filling had to be performed by numerical means.@xcite using these results we choose the microscopic parameters such that the system is well inside the mott - insulating phase with an energy gap @xmath14 so that we are able to clearly resolve the interesting features in our numerical results .
in order to determine the ldos we calculate the retarded green function in frequency space using an expansion of the occurring resolvent in chebyshev polynomials.@xcite an alternative numerical approach consists in the expansion of the lehmann representation of the spectral function in chebyshev polynomials , the kernel polynomial method ( kpm ) , see refs . .
in contrast we specifically evaluate the complete ( real and imaginary part ) green functions @xmath15 with @xmath16 here @xmath17 denotes the ground state of the system with energy @xmath18 .
note that since we are interested in the ldos we have already taken the electron creation and annihilation operators to be at the same site @xmath19 with @xmath20 denoting the lattice spacing .
furthermore , since the systems we consider possess spin - rotation invariance we have suppressed the formal spin dependence of the green functions . in eqs . and
we have included the convergence factor @xmath21 , which in the continuum limit should be taken as @xmath22 . in the numerical evaluations it has to be larger than the finite level splitting brought about by the finite system size . at the same time
@xmath21 has to be smaller than any physically relevant energy scale in order to resolve the relevant features of the spectrum . to attain a small value of @xmath21
we employ a chebyshev polynomial expansion approach for the resolvents in and .
more details on this approach can be found in refs . .
the applied chebyshev expansion is based on the representation of the functions @xmath23 in terms of chebyshev polynomials @xmath24 the expansion coefficients are given by @xmath25 where @xmath26 is a function of the artificial broadening @xmath21 which would theoretically allow arbitrarily small @xmath21 .
the chebyshev polynomials @xmath27 are defined by their recursion relation @xmath28 and fulfil @xmath29 as well as @xmath30 in order to apply the expansion , which is only valid for @xmath31 , to the resolvents appearing in the green functions , we first have to rescale the energies . to this end
we run initial dmrg calculations to determine the ground - state energy @xmath18 as well as the smallest and the largest energies of the system with @xmath32 electrons .
this allows us to find the scaling factor @xmath33 and shift @xmath34 such that the operator @xmath35 has a spectrum between @xmath36 and @xmath37 in the sectors with @xmath32 particles .
then the green function can be expressed as @xmath38\,\mu_n^+(x ) , \label{eq : g+expansion}\ ] ] where the chebyshev moments @xmath39\,c_{j,\sigma}^\dagger{\left|\psi_0\right\rangle}\ ] ] ( recall @xmath19 ) can be evaluated recursively via @xmath40 with the recursion relations @xmath41\vert \phi_0^{+}\rangle,\label{eq : iteration2}\\ \vert \phi_{n+1}^{+ } \rangle & = 2[a(h - e_0)-b]\vert \phi_{n}^{+}\rangle -\vert \phi_{n-1}^{+}\rangle.\label{eq : iteration3}\end{aligned}\ ] ] similarly , for the green function we obtain the expansion @xmath42\,\mu_n^-(x ) , \label{eq : g - expansion}\ ] ] where @xmath43\,c_{j,\sigma}{\left|\psi_0\right\rangle}.\ ] ] in the numerical evaluations the sums appearing in and are truncated at @xmath44 .
the moments @xmath45 are calculated iteratively from using dmrg . during the dmrg finite - lattice sweeps we determine each state @xmath46 and include it into a modified density matrix . by performing a singular - value decomposition of this modified density matrix
we ensure that all the states @xmath46 are part of the hilbert space after the dmrg truncation . the moments for @xmath47
are then obtained employing and as @xmath48 and @xmath49 .
finally , we note that the chebyshev moments @xmath45 are typically strongly oscillating with respect to the index @xmath50 .
therefore , the final results oscillate slightly when changing the value of @xmath51 . on the other hand ,
we find small oscillating parts in the spectral function if we choose @xmath51 too small . both effects can be avoided by implementing a smoothing window for the last @xmath52 moments . throughout this article
we use a @xmath53-filter for the last @xmath54 moments .
this way one can obtain a good approximation for the spectral function using a smaller number of moments @xmath51 . previously it was observed@xcite that the number of required chebyshev moments sufficient to approximate the green function is inversely proportional to the width of the spectrum @xmath33 and the desired artificial broadening @xmath21 , ie , @xmath55 . throughout this work use @xmath56 chebyshev moments for the series expansion of the green function .
furthermore , @xmath21 is chosen such that the resulting curves become smooth and artificial features are suppressed .
the ldos is obtained from the retarded green function in the usual way . as was noted by kivelson et al.@xcite in the study of luttinger liquids with boundaries , it is useful to consider the fourier transform of the ldos , as physical properties like the dispersions of propagating quasiparticles can be more easily identified .
since we consider a finite chain of length @xmath8 we analyse @xmath57 , \label{eq : nq}\ ] ] where the momenta @xmath58 take the values @xmath59 , @xmath60 .
we note that the ldos is directly related to the tunneling current measured in scanning tunneling microscopy experiments , thus its fourier transform is experimentally accessible . in the following we focus on the ldos for positive energies
; the ldos for negative energies can be analysed analogously .
the ldos of the low - energy effective field theory of the hubbard models and has been analysed@xcite in refs . .
in the field - theoretical description the momentum regimes @xmath61 and @xmath62 are treated separately . for small momenta @xmath61 the main features of the fourier transform are a strong divergence at @xmath63 as well as a propagating excitation in the gapped charge sector above the mott gap . in contrast , the behaviour at momenta @xmath64 shows a divergence at @xmath65 , a propagating excitation in the charge sector as well as a linearly dispersing excitation in the gapless spin sector . furthermore there exists a critical momentum
above which a second linearly dispersing mode becomes visible .
in addition , it was shown that certain boundary conditions lead to the formation of boundary bound states which manifest themselves as non - propagating features in the ldos .
the main aim of our article is the calculation of the fourier transform of the ldos in the microscopic models and and its comparison to the field - theoretical predictions.@xcite we start with the standard hubbard chain before considering the extended version . in sec .
[ sec : bbs ] we then analyse the effect of additional boundary potentials which give rise to the existence of boundary bound states .
we first consider the fourier transform of the ldos in the standard hubbard model .
the results in the vicinity of @xmath63 and @xmath66 are shown in figs .
[ fig : std_q0 ] and [ fig : std_q2kf ] respectively , where we have chosen a repulsive interaction of @xmath67 corresponding to the dimensionless hubbard parameter @xmath68 and @xmath69 lattice sites . throughout our article
we use the hopping parameter @xmath70 as our unit of energy . as is well known , the hubbard model is exactly solvable by bethe ansatz .
@xcite in particular , the velocities of the spin and charge excitations @xmath71 and @xmath72 as well as the mott gap @xmath73 can be determined analytically ; the results in the thermodynamic limit read @xmath74 where @xmath75 and @xmath76 denote the bessel functions and modified bessel functions of the first kind respectively .
our chosen parameters for the microscopic system correspond to@xcite @xmath77 .
, for interaction @xmath78 ( recall @xmath70 ) , @xmath69 lattice sites , broadening @xmath79 and momenta @xmath80 ( from bottom to top ) .
the curves are constant @xmath58-scans that have been offset along the y - axis by a constant with respect to one another .
@xmath81 is dominated by a strong peak at @xmath82 which is only partially displayed in the figure in order to improve visibility for the other cuts .
we clearly observe the mott gap @xmath73 as well as a dispersing feature indicated by the arrow .
this feature corresponds to propagating charge excitations , it follows the dispersion relation @xmath83 given in with @xmath84 obtained from .,scaledwidth=49.5% ] in fig .
[ fig : std_q0 ] we plot @xmath81 in the vicinity of small momenta @xmath61 . all features ( except for a strong peak at @xmath85 ) appear at energies @xmath86 , clearly showing that the system is in a gapped phase .
the observed energy gap @xmath73 agrees perfectly with the value @xmath87 obtained from the bethe ansatz in the thermodynamic limit .
this suggests that the length of our chain is long enough to avoid significant finite - size effects in our results .
the fourier transform of the ldos for small momenta is dominated by a global maximum at @xmath88 .
this peak is attributed to a spin - density wave pinned at the boundary , it is also well visible in the field - theoretical results.@xcite at low energies above the energy gap we further observe a dispersing feature indicated by the arrow .
this again agrees well with the results from the field theory that predict a gapped , dispersing charge excitation with dispersion relation @xmath89 where @xmath90 and @xmath72 is the velocity of the charge excitations .
the bethe - ansatz solution gives the value @xmath91 , which is in excellent agreement with the velocity observed in the plot .
the physical origin of this dispersing feature is the decay of the electronic excitation into gapped charge and gapless spin excitations . in the process giving rise to the external momentum @xmath92 is taken by the charge excitation propagating through the system and eventually getting reflected at the boundary , while the spin excitation does not propagate and thus possesses zero momentum .
the appearance of @xmath93 in originates from the fact that the charge excitation has to propagate to the boundary and back , thus covering the distance @xmath94 .
in addition , we find a second dispersing feature that seemingly follows the same dispersion relation albeit with a different value for the gap @xmath95 .
this feature is not contained in the field - theory description of the ldos , which solely focuses on the low - energy regime .
furthermore , the field theory makes predictions about the power - law decay of @xmath81 at @xmath63 which , however , can not be resolved in our numerical data . for the observation of such features
we would require a significantly higher resolution , both in energy and momentum .
this can in turn only be achieved by turning to a significantly larger system size and a higher amount of calculated chebyshev moments . for momenta in the vicinity of @xmath96 with @xmath97 ( from top to bottom ) .
all other parameters are as in fig .
[ fig : std_q0 ] .
the curves are constant @xmath92-scans that have been offset along the y - axis by a constant with respect to one another .
we observe two dispersing features ( indicated by the arrows ) at @xmath98 and @xmath99 originating from propagating charge and spin excitations respectively.,scaledwidth=49.5% ] we now turn our attention to momenta in the vicinity of @xmath96 .
we first note that features in this momentum regime originate from umklapp processes coupling left- and right - moving modes which are absent in translationally invariant systems and thus constitute a particularly clean way to investigate the boundary effects . in fig . [
fig : std_q2kf ] we again observe the existence of the mott gap as well as two dispersing features at @xmath98 as defined in and @xmath100 both indicated by the arrows .
the spin velocity observed in the plot is in excellent agreement with the bethe - ansatz result giving @xmath101 .
while the feature adhering to is again due to a propagating charge excitation , the feature following originates from the propagation of spin excitations with the charge excitation possessing zero momentum .
furthermore , we note that in contrast to the field - theoretical prediction we observe only one linearly dispersing mode . in order to understand this we recall that the two linearly dispersing modes
are energetically separated by@xcite @xmath102\approx 0.1\delta\approx 0.08 $ ] , where in the last step we have put in the parameters used in fig .
[ fig : std_q2kf ] . on the other hand ,
our resolution in energy is limited by finite - size effects to about @xmath103 , implying that for the treatable system sizes the two linearly dispersing features can not be separated . for the parameters of fig .
[ fig : std_q0 ] .
the dominant , white peak at @xmath61 is due to the spin - density wave pinned at the boundary .
the solid and dashed lines indicate the holon dispersion around @xmath63 and @xmath65 respectively , the dashed - dotted line represents the spinon dispersion around @xmath65 .
the parameters @xmath73 , @xmath72 and @xmath71 used in the plot were obtained from the bethe ansatz for the bulk system
, ie , there is no free fitting parameter.,scaledwidth=49.5% ] to summarise our results , in fig .
[ fig : modes ] we show a contour plot of the ldos . for comparison
we plot the holon dispersion around @xmath63 and @xmath65 as well as the spinon dispersion around @xmath65 , for which we used the parameters @xmath73 , @xmath72 and @xmath71 obtained from the bethe ansatz for the bulk system .
in particular , we stress that there is no fitting parameter . in conclusion ,
our results are in very good agreement with the features of the ldos predicted by the field - theoretical investigations .
for an extended hubbard model with interaction @xmath104 , @xmath105 , @xmath106 , @xmath79 and momenta @xmath107 ( from bottom to top ) .
the results are qualitatively similar to the ones for the standard hubbard model shown in fig .
[ fig : std_q0 ] , ie , we observe a mott gap @xmath73 , a dispersing feature following ( indicated by the arrow ) and another one at higher energies.,scaledwidth=49.5% ] we have performed the analysis presented in the previous section for the extended hubbard model ( [ eq : ext ] ) at half - filling and @xmath106 lattice sites . since the extended hubbard model is not integrable , there exist no analytical results for the parameters @xmath73 , @xmath72 and @xmath71 .
still , the field theory is expected to qualitatively describe the behaviour of the system in the low - energy limit .
the ldos for momenta in the vicinity of @xmath63 and @xmath65 is shown in figs .
[ fig : hf_q0 ] and [ fig : hf_q2kf ] respectively . in both plots
we have renormalised the energy scale by the gap @xmath108 obtained from the data at @xmath61 .
for an extended hubbard model in the vicinity of @xmath65 with @xmath109 ( from top to bottom ) .
all other parameters are as in fig .
[ fig : hf_q0 ] . similar to the standard hubbard model , at low energies
we observe two dispersing features at and respectively.,scaledwidth=49.5% ] at low energies the dispersing features are qualitatively identical to the ones seen for the standard hubbard model , namely a propagating charge mode for @xmath61 and both a propagating charge and spin mode around @xmath65 .
the only difference is that the charge and spin velocities take the values @xmath110 and @xmath111 respectively , which were determined by comparison with the quasiparticle dispersions and .
the energy gap @xmath73 and charge velocity @xmath72 for the two different momentum regimes agree well .
we thus conclude that the low - energy sector is well described by the field theory .
furthermore , for small momenta we again observe a second charge mode which now seems to have the gap @xmath112 .
having analysed the ldos in the presence of open boundary conditions , we now turn to the investigation of the effect of a boundary chemical potential .
specifically we consider the hubbard model with a boundary potential at site @xmath113 , @xmath114 using bosonisation such a boundary potential is translated into non - trivial boundary conditions for the bosonic degrees of freedom . in particular , certain boundary conditions give rise to the existence of boundary bound states in the gapped charge sector@xcite which manifest themselves@xcite in the ldos as non - propagating features inside the mott gap .
the spectrum of the hubbard chain with boundary potential has been investigated by bedrftig and frahm@xcite using the bethe - ansatz solution .
in particular it was found that a boundary bound state corresponding to a charge bound at the first site exists for @xmath115 . for even smaller boundary potentials , @xmath116 , two electrons in a spin singlet
get bound to the surface . for interaction @xmath117 , boundary potential @xmath118 , @xmath69 lattice sites and broadening @xmath119 .
besides the peak at @xmath63 and the dispersing modes at @xmath86 we observe a non - dispersing feature inside the energy gap at @xmath120 which originates from the boundary bound state.,scaledwidth=49.5% ] the fourier transform of the ldos in the presence of a boundary chemical potential is shown in fig .
[ fig : bbs_q0 ] . besides the peak at @xmath63 due to the pinned charge - density wave and several dispersing modes above the mott gap , we observe
a clear , non - dispersing maximum inside the gap at @xmath121 , which is a manifestation of the boundary bound state in the ldos . in the following we analyse this contribution in more detail by considering the ldos @xmath122 close to the boundary .
, for various values of @xmath123 and broadening @xmath124 .
all other parameters are as in fig .
[ fig : bbs_q0 ] . in the absence of a boundary potential ( thick line )
there is barely any spectral weight inside the energy gap . for @xmath115
the spectral density inside the gap grows continuously but its maximum is still located above the gap . for @xmath125
the maximum is located inside the mott gap , providing a clear manifestation of the boundary bound state .
inset : position @xmath126 of the maximum of @xmath127 as a function of the boundary potential @xmath123 .
we observe that a potential @xmath128 is needed for @xmath129.,scaledwidth=49.5% ] first we analyse the ldos at the boundary site , @xmath127 , which is shown in fig .
[ fig : bbs_x0 ] for several values of the boundary potential @xmath123 using an artificial broadening @xmath124 .
one can clearly see that the maximum of the ldos is shifted towards lower energies for decreasing @xmath123 . for @xmath130
we find a considerable spectral density inside the mott gap @xmath73 ; for @xmath131 the maximum of the ldos is located inside the energy gap as well . from this
we deduce that for @xmath132 there exists a clear boundary bound state contribution to the ldos .
we attribute the deviation to the critical value @xmath133 obtained from the bethe ansatz@xcite to the finite system - size as well as the artificial broadening @xmath21 introduced in our numerical calculations .
this is supported by the dependence of the energy of the maximum in the ldos on the broadening presented in fig .
[ fig : bbs_mu ] , which shows that the energy of the maximum indeed decreases with decreasing @xmath21 . extrapolating the results to @xmath134 and keeping in mind the finite system size as well as the fact that for @xmath135 the contributions from the boundary bound state and the standard continuum at @xmath86 start to significantly overlap ,
we conclude that our results are consistent with the bethe - ansatz solution .
this is further supported by the electron density at the boundary shown in the inset of fig .
[ fig : bbs_mu ] . as a function of the artificial broadening @xmath21 for @xmath117 and @xmath69 . extrapolating to @xmath134 ( indicated by lines )
we find that that the energy of the maximum lies within the mott gap for @xmath136 .
inset : electron density @xmath137 at the boundary showing very good agreement with the bethe - ansatz result.@xcite the dotted vertical lines indicate the positions @xmath138 at which two electrons get bound to the boundary.,scaledwidth=49.5% ] , as a function of the distance to the boundary for @xmath117 , @xmath124 and @xmath69 . for decreasing @xmath123
we observe that the spectral weight gets more and more localised at the boundary.,scaledwidth=49.5% ] finally we consider the space dependence of the ldos when going away from the boundary . as is shown in fig .
[ fig : bbs_exp ] , lowering the boundary potential leads to an increase of the ldos at the boundary , consistent with the formation of a boundary bound state localised at @xmath113 .
however , the system size and energy resolution is not sufficient to unveil an exponential space dependence of the ldos as predicted by the field - theory analysis,@xcite ie , @xmath139 $ ] .
in this work we have performed a numerical study of the ldos of one - dimensional mott insulators with an open boundary . as microscopic realisations of the mott insulator
we have studied the ( extended ) hubbard model at half filling .
the results for the fourier transform of the ldos revealed the existence of the mott gap as well as several gapped and gapless dispersing modes .
these qualitative features were in perfect agreement with the results of field - theoretical calculations@xcite of the ldos in the mott insulator .
furthermore , we extracted quantitative values for the gap and velocities , which , in the case of the integrable hubbard chain , were found to be in excellent agreement with the exact results.@xcite besides open boundary conditions we have also considered the effect of a boundary potential . for sufficiently strong potentials
this results in the formation of a boundary bound state , which manifests itself in the ldos as a non - dispersing feature inside the mott gap . in summary ,
our results show that spin - charge separation and the formation of boundary bound states can be observed in the fourier transform of the ldos amenable to numerical simulations or scanning tunneling spectroscopy experiments .
we thank fabian essler , holger frahm , martin hohenadler and volker meden for useful comments and discussions .
bs and ds were supported by the foundation for fundamental research on matter ( fom ) , which is part of the netherlands organisation for scientific research ( nwo ) , under 14pr3168 .
ps was supported by dfg - sfb 1170 and the erc starting grant topolectrics ( erc - stg-336012 ) .
this work is part of the d - itp consortium , a program of the netherlands organisation for scientific research ( nwo ) that is funded by the dutch ministry of education , culture and science ( ocw ) .
the authors acknowledge support by the state of baden - wrttemberg through bwhpc . c. kim , a. y. matsuura , z .- x .
shen , n. motoyama , h. eisaki , s. uchida , t. tohyama , and s. maekawa , phys .
lett . * 77 * , 4054 ( 1996 ) ; b. j. kim _
et al . _ , nature phys . * 2 * , 397 ( 2006 ) ; m. grioni , s. pons , and e. frantzeskakis , j. phys . : condens .
matter * 21 * , 023201 ( 2009 ) .
m. bockrath , d. h. cobden , j. lu , a. g. rinzler , r. e. smalley , l. balents , and p. l. mcruen , nature * 397 * , 598 ( 1999 ) ; o. m. auslaender , h. steinberg , a. yacoby , y. tserkovnyak , b. i. halperin , k. w. baldwin , l. n. pfeiffer , and k. w. west , science * 308 * , 88 ( 2005 ) ; y. jompol , c. j. b. ford , j. p. griffiths , i. farrer , g. a. c. jones , d. anderson , d. a. ritchie , t. w. silk , and a. j. schofield , _ ibid . _
* 325 * , 597 ( 2009 ) .
p. a. bares and g. blatter , phys .
64 * , 2567 ( 1990 ) ; m. ogata and h. shiba , phys .
b * 41 * , 2326 ( 1990 ) ; k. penc , f. mila , and h. shiba , phys . rev . lett . * 75 * , 894 ( 1995 ) ; m. arikawa , y. saiga , and y. kuramoto , _ ibid . _ * 86 * , 3096 ( 2001 ) .
e. a. jagla , k. hallberg , and c. a. balseiro , phys .
b * 47 * , 5849 ( 1993 ) ; m. g. zacher , e. arrigoni , w. hanke , and j. r. schrieffer , _ ibid . _ * 57 * , 6370 ( 1998 ) ; d. snchal , d. perez , and m. pioro - ladrire , phys . rev . lett . * 84 * , 522 ( 2000 ) ; c. kollath , u. schollwck , and w. zwerger , _ ibid . _
* 95 * , 176401 ( 2005 ) ; c. kollath and u. schollwck , new j. phys . * 8 * , 220 ( 2006 ) ; h. benthien and e. jeckelmann , phys .
b * 75 * , 205128 ( 2007 ) ; p. schmitteckert , in _ high performance computing in science and engineering 07 _ , edited by w. e. nagel , w. jger , and m. resch ( springer , berlin , 2007 ) ; t. ulbricht and p. schmitteckert , epl ( europhysics letters ) * 86 * , 57006 ( 2009 ) ; _ ibid . _
* 89 * , 47001 ( 2010 ) ; a. moreno , a. muramatsu , and j. m. p. carmelo , phys .
b * 87 * , 075101 ( 2013 ) ; k. a. al - hassanieh , j. rincn , e. dagotto , and g. alvarez , _ ibid .
_ * 88 * , 045107 ( 2013 ) .
v. meden and k. schnhammer , phys .
b * 46 * , 15753 ( 1992 ) ; k. schnhammer and v. meden , _ ibid . _ * 47 * , 16205 ( 1993 ) ; j. voit , _ ibid .
_ * 47 * , 6740 ( 1993 ) ; h. benthien , f. gebhard , and e. jeckelmann , phys .
. lett . * 92 * , 256401 ( 2004 ) ; a. abendschein and f. f. assaad , phys .
b * 73 * , 165119 ( 2006 ) .
r. preuss , a. muramatsu , w. von der linden , p. dieterich , f. f. assaad , and w. hanke , phys .
lett . * 73 * , 732 ( 1994 ) ; j. voit , j. phys . : condens .
matter * 8 * , l779 ( 1996 ) ; f. h. l. essler and a. m. tsvelik , phys .
lett . * 88 * , 096403 ( 2002 ) ; _ ibid . _
* 90 * , 126401 ( 2003 ) .
m. fabrizio and a. o. gogolin , phys .
b * 51 * , 17827 ( 1995 ) ; s. eggert , h. johannesson , and a. mattsson , phys .
* 76 * , 1505 ( 1996 ) ; a. e. mattsson , s. eggert , and h. johannesson , phys . rev .
b * 56 * , 15615 ( 1997 ) ; v. meden , w. metzner , u. schollwck , o. schneider , t. stauber , and k. schnhammer , eur .
j. b * 16 * , 631 ( 2000 ) ; s. eggert , phys .
lett . * 84 * , 4413 ( 2000 ) ; s. andergassen , t. enss , v. meden , w. metzner , u. schollwck , and k. schnhammer , phys .
b * 70 * , 075102 ( 2004 ) ; i. schneider , a. struck , m. bortz , and s. eggert , phys .
* 101 * , 206401 ( 2008 ) ; w. metzner , m. salmhofer , c. honerkamp , v. meden , and k. schnhammer , rev . mod . phys . * 84 * , 299 ( 2012 ) .
v. meden , k. schnhammer , and o. gunnarsson , phys .
b * 50 * , 11179 ( 1994 ) ; h. matsueda , t. tohyama , and s. maekawa , _ ibid . _ * 74 * , 241103 ( 2006 ) ; m. hohenadler and f. f. assaad , _ ibid .
_ * 87 * , 075149 ( 2013 ) ; m. weber , f. f. assaad , and m. hohenadler , _ ibid .
_ * 91 * , 235150 ( 2015 ) .
f. h. l. essler and r. m. konik , in _ from fields to strings : circumnavigating theoretical physics ( ian kogan memorial collection ) _ , edited by m. shifman , a. vainshtein , and j. wheater ( world scientific , singapore , 2005 ) , vol . i. r. t. clay , a. w. sandvik , and d. k. campbell , phys .
b * 59 * , 4665 ( 1999 ) ; e. jeckelmann , phys . rev . lett . *
89 * , 236401 ( 2002 ) ; a. w. sandvik , p. sengupta , and d. k. campbell , _ ibid .
_ * 91 * , 089701 ( 2003 ) ; a. w. sandvik , l. balents , and d. k. campbell , _ ibid .
_ * 92 * , 236401 ( 2004 ) ; s. ejima and s. nishimoto , _ ibid .
_ * 99 * , 216403 ( 2007 ) .
a. weie , g. wellein , a. alvermann , and h. fehske , rev .
phys . * 78 * , 275 ( 2006 ) ; a. holzner , a. weichselbaum , i. p. mcculloch , u. schollwck , and j. von delft , phys .
b * 83 * , 195115 ( 2011 ) ; j. hackmann and f. b. anders , _ ibid . _
* 89 * , 045317 ( 2014 ) ; f. a. wolf , i. p. mcculloch , o. parcollet , and u. schollwck , _ ibid . _
* 90 * , 115124 ( 2014 ) .
we note that the refs .
consider the case of a charge - density wave state where the gap appears in the spin sector .
the results for the mott insulator considered in our article are obtained by simply interchanging spin and charge sectors in refs . .
s. ghoshal and a. b. zamolodchikov , int .
phys . a * 9 * , 3841 ( 1994 ) ; _ ibid . _
* 9 * , 4353(e ) ( 1994 ) ; s. skorik and h. saleur , j. phys .
a : math . gen . *
28 * , 6605 ( 1995 ) ; p. mattsson and p. dorey , _ ibid .
_ * 33 * , 9065 ( 2000 ) . | we numerically calculate the local density of states ( ldos ) of a one - dimensional mott insulator with open boundaries , which is modelled microscopically by a ( extended ) hubbard chain at half filling . in the fourier transform of the ldos we identify several dispersing features corresponding to propagating charge and spin degrees of freedom , thus providing a visualisation of the spin - charge separation in the system .
we also consider the effect of an additional boundary potential , which , if sufficiently strong , leads to the formation of a boundary bound state which is clearly visible in the ldos as a non - dispersing feature inside the mott gap . |
the propagation of high - intensity femtosecond laser pulses in a bulk medium is a complex phenomenon accompanied by a dramatic spatiotemporal and spectral rearrangement of the pulse .
this originates from the highly intensive nonlinear and dispersive interaction of the optical pulse with the medium .
the rigorous description of the pulse propagation is based on the ( 3 + 1)d nonlinear schrodinger equation ( nlse ) , in which diffraction , group velocity dispersion ( gvd ) of second order , and kerr nonlinearity of third order form a basic set of optical processes of lowest order [ 1 - 6 ] , which is universal for all media and parameters of the input pulse and belongs to the strongest effects acting on the pulse in a centrosymmetric neutral medium .
additional terms in the nlse are required to describe some particular features or more extreme conditions in the pulse propagation as , _
e.g. _ , self - steepening , space - time focusing , non - instantaneous nonlinearity , higher order dispersion , _ etc . _
[ 7 - 9 ] .
higher orders of nonlinearity and ionization must also be considered for pulses creating significant ionization [ 10 - 12 ] . while the femtosecond pulse behavior strongly depends on the material and pulse parameters ,
some general simple understandings were widely accepted . in this way ,
simultaneous collapse of the pulse in space and time and formation of a spatiotemporal soliton is predicted at negative gvd @xcite , while , in the more common case of positive gvd , the theoretical and experimental studies predict time broadening and splitting of the pulse [ 2 - 5,7,8,10 ] . in a single study ,
initial pulse shortening is found while not supported by experimental evidences , instead , splitting of the femtosecond pulse is shown @xcite .
self - compression ( sc ) , before splitting , of high - intensity femtosecond laser pulses in positive gvd has been discovered in various types of media pure atomic and molecular gases [ 13 - 15 ] and fused silica @xcite .
this allows recognizing the sc as a new phenomenon a kind of a time focusing , or the temporal counterpart of the usual space domain self - focusing ( sf ) @xcite .
the mechanism of both phenomena is different but , as will be shown below , the sc is closely related to the sf .
the generation of compressed ( almost five times in magnitude ) pulses in the experiments in gaseous media is accompanied by a strong increase of the peak intensity , an improvement of the spatiotemporal pulse shape , and a stable propagation of the compressed pulse along a distance that is many times longer than the characteristic length of the strongest pulse rearrangement factor , in this case the nonlinearity @xcite .
the time - spectral studies show that the compressed pulses are transform - limited @xcite .
sc of high - intensity femtosecond laser pulses was also observed in air @xcite , and in ionized noble gasses [ 19 - 22 ] .
recently , sc was also reported in bk-7 glass using negatively chirped pulses @xcite .
generation of long light filament and twofold pulse compression has been observed experimentally in air @xcite .
substantially stronger pulse compression has been achieved in ionized noble gasses [ 19 - 22 ] .
generation of few - cycle pulses through filamentation is achieved experimentally in argon - filled gas cell @xcite .
generation of single - cycle pulse is predicted numerically in argon gas with pressure gradient @xcite .
grate attention has been paid in the study of the femtosecond pulse splitting . from the other side , while the sc is much more interesting for practical use , its exact `` physical mechanism ... is still not fully understood '' @xcite .
recently , we have shown for the first time that sc and intensity gain of high intensity femtosecond laser pulses , before the pulse splitting , can be achieved in a positive gvd medium based on basic set of the lowest order optical processes , diffraction , group velocity dispersion of second order and kerr nonlinearity of third order , providing the gvd of the medium is low enough @xcite .
the physical mechanism of sc in this case was found and referred to as sc in low dispersion regime .
the numerical simulations were performed for pressurized argon assuming sub - millijoule femtosecond laser pulses so as to avoid the effect of the ionization on the pulse propagation .
the lack of ionization substantially simplifies the pulse propagation dynamics . to scale up the above method to higher energy / intensity pulses , media of higher ionization potential have to be used .
here we demonstrate sc of multi - millijoule femtosecond laser pulses in pressurized helium and neon .
the pulse compression is accompanied by a dramatic increase of the peak intensity at almost no loss of energy .
the pressurized helium and neon possess the lowest specific gvd among the known media , which provides an increased degree of sc and peak intensity gain in comparison with argon .
the dynamics of the optical pulse at lack of ionization is relatively smooth and it allows a reliable control on the pulse propagation and constructing a pulse of well predictable parameters in given space target point .
the degree of sc and intensity gain , as well as the space positions of maximal sc and intensity gain of the pulse can be controlled by the pulse energy and/or gas number density / pressure .
we consider a physical model of the high - intensity femtosecond pulse propagation in a centrosymmetric nonlinear medium , which includes the basic set of lowest order optical processes , diffraction , gvd of second order and kerr nonlinearity of third order .
the nlse ( in moving frame ) @xmath0 for the complex field amplitude @xmath1 of a pulse propagating along the @xmath2-direction , including the above mentioned physical processes , is used as a propagation equation within that model . here
, @xmath3 is the magnitude of the wave - vector , @xmath4 is the linear refractive index of the medium at the central wavelength @xmath5 of the pulse , @xmath6 is the transversal laplacian , @xmath7 is the gvd , and @xmath8 is the nonlinear refractive index .
the field is normalized such that @xmath9 is the intensity .
neon and helium are used as nonlinear media .
the nonlinear refractive indexes of neon and helium , @xmath10 p@xmath11/w and @xmath12 p@xmath11/w , have been determined from a comparative study of the relative nonlinear refractive indexes of noble gas media @xcite , taking the nonlinear refractive index of ar , @xmath13 p@xmath11/w , as a reference value .
the gvd of neon and helium has been determined from the sellmeier equation @xcite and found to be @xmath14 pfs@xmath15/cm and @xmath16 pfs@xmath15/cm . in the above expressions @xmath17
is the pressure in atm .
the initial pulse is a linearly polarized chirp - free gaussian @xmath18 in space and time having axial symmetry and 100 fs time duration ( full width at half maximum ( fwhm ) ) of the intensity profile .
the central wavelength of the pulse is tuned at 800 nm .
the proper initial conditions are important in order to remain within the range of validity of the present model .
for such a purpose , special attention is paid to the ionization .
the ionization strongly and non - instantaneously modifies the material parameters and thus , the pulse propagation . also , it is highly nonlinear process , which means that it is very sensitive to the changes of the intensity / field strength and its control is very difficult in practical means , more over the pulse intensity strongly changes during the propagation .
the avoiding the ionization of medium can be achieved by a suitable choice of the input pulse energy and/or the gas number density / pressure so that the peak intensity , raised by the sf , does not surpass the level at which the ionization causes substantial modification of the material parameters .
for instance , the plasma contribution to the refractive index should be kept negligible in comparison to the respective contribution of the cubic nonlinear kerr effect within the whole range of intensity variation .
avoiding the ionization ensures a relatively simple , smooth and , thus , well controllable pulse propagation dynamics . the determination of the absolute value of the ionization rate and , in this way , the plasma density created by the pulse itself is difficult task .
the description of the ionization of atoms and simple molecules by the general ionization theories is not quite satisfactory @xcite .
the ionization rate and the plasma number density predicted , _
e.g. _ , by perelomov - popov - terentev ( ppt ) theory @xcite , which appears to be among the most successful ionization theories @xcite , closely matches the trend of the experimental data over a wide range of intensities from multi - photon to tunnel ionization regimes .
however , the agreement between the experimental data and the theoretically predicted plasma number density , based on the ppt theory , is still within a factor of two .
the discrepancy with the other ionization theories may reach orders of magnitude @xcite . from the other side , precision measurements of the ionization of helium @xcite as well as of the other noble gases
@xcite are performed over a wide counting range .
unfortunately , only the relative value of the ion signal has been linked to the absolute value of the peak intensity of the pulses in these experiments [ 27,29 ] , whereas , due to the ion detection technique , the absolute value of the plasma density remains undetermined . to account for the latter , we require at least one absolute reference point to calibrate the relative plasma number density - peak intensity relationships found from these experimental studies .
then , the precise relative dependence can be used to determine the respective absolute value dependence . to find such a reference point ,
the following approach will be applied .
the intensity of complete ionization of neon and helium atoms will be determined using the simple barrier suppression ionization ( bsi ) model @xcite . within the bsi model , the threshold intensity @xmath19
at which the optical field suppresses the coulomb potential in an atom so as to allow a bound electron to escape freely above the potential barrier ( above barrier ionization ) without tunneling is ( in atomic units ) @xmath20 where @xmath21 is the ionization potential of the initial atom / ion , and @xmath22 is the charge of the created ion . at the ionization potential of neon , @xmath23 ev , and helium , @xmath24 ev , the intensity threshold for complete ionization of the medium is found to be @xmath25 w/@xmath11 and @xmath26 w/@xmath11 , respectively .
the bsi intensity of helium atom so determined is relatively close to the experimentally determined value of @xmath27 w/@xmath11 , beyond which indication of saturation of the ionization can be distinguished ( the slow @xmath28 increase of the experimentally measured ion yield beyond the sited intensity is attributed to the expansion of the gaussian focal volume ) @xcite .
the intensity of the barrier suppression ionization , @xmath19 , corresponds to complete ionization of the medium , and , at @xmath29 , the plasma number density will be equal to the number density of the initial neutral gas ( the plasma recombination will be neglected on the time scale comparable to the pulse duration of 100 fs ) .
the latter is , in fact , the reference point that will be used here in order to find the relationship between the absolute values of the plasma number density and the peak intensity of the pulse .
using such a relationship allows to account for the intensity at which the plasma contribution to the refractive index is still negligible . according to the above considerations ,
the electron / plasma number density @xmath30 at @xmath29 will be equal to the initial number density of the neutral gas , @xmath31 .
we will take tentatively 6 orders of magnitude lower plasma number density than @xmath31 , and the intensity at which such plasma number density is created will be determined approximately from the respective experimental curves , figure 1 of ref.@xcite and figure 5 of ref.@xcite for he , and figure 4 of ref.@xcite for ne . from the sited figures , the intensity corresponding to such a plasma density was found to be about @xmath32 w/@xmath11 and @xmath33 w/@xmath11 for helium and neon , respectively . at the intensity so obtained and the respective plasma number density , the magnitude of , otherwise , negative plasma contribution to the refractive index , @xmath34 @xcite , will be compared to the respective kerr effect contribution to the refractive index , @xmath35 .
this is justified taking into account that the latter quantities determine the magnitude of the contributions of these effects to the respective terms , @xmath36 and @xmath37 , in the nlse @xcite . for 30 atm gas pressure of helium ( the maximal gas pressure used here ) , the plasma contribution to the refractive index at @xmath32 w/@xmath11 ( _ i.e. _ , assuming , as sited above , 6 orders of magnitude lower plasma density than @xmath31 ) is @xmath38 while the kerr effect contribution to the refractive index at the same intensity is @xmath39 .
the same quantities for 17 atm pressure ( around which most of the simulations are performed here ) are @xmath40 and @xmath41 for the case of helium at @xmath32 w/@xmath11 intensity , and @xmath40 and @xmath42 for the case of neon at @xmath33 w/@xmath11 intensity .
as can be seen , the plasma contribution to the modification of the refractive index is more than 2 orders of magnitude lower than the respective kerr effect contribution and neglecting the plasma contribution seems justified for neon and helium at intensities not exceeding @xmath33 w/@xmath11 and @xmath32 w/@xmath11 , respectively . within the above mentioned intensity range we are still in perturbative regime what concerns the neon and the helium atoms .
the relation between two successive terms of the nonlinear polarization for bound - bound transitions is given by the following approximate relation @xcite @xmath43 ( for an isotropic medium the even order nonlinearities are zero ) , where @xmath44 is the time dependent field amplitude , @xmath45 is the bohr radius , and @xmath46 is the energy detuning of the laser photon energy @xmath47 ( at the carrier frequency of the pulse ) from the atomic resonance of energy @xmath48 between the nearest exited atomic state ( coupled to the ground state by a nonzero dipole moment ) and the ground state . for the helium atom ,
the lowest excited level with nonzero dipole moment to the singlet ground state @xmath49 is the singlet @xmath50 state , lying about 21.2 ev above the ground state . at the field strength corresponding to the maximal intensity of @xmath51 w/@xmath11 , the relation between the leading cubic term and the next quintic term is about @xmath52 , or the truncation of the nonlinearity after the cubic term seems justified within the present considerations .
one approach to control the peak intensity below the sited levels is to change the gas pressure while keeping the input energy fixed for given study .
this is also a convenient approach for a real experiment . in view of the above considerations , the propagation equation ( [ eq1 ] ) ( which does not include irreversible losses and non - instantaneous processes ) is appropriate to describe the propagation of femtosecond pulses within the conditions in which the present model is valid .
the lack of substantial losses allows imposing energy conservation condition @xcite @xmath53 where @xmath54 is the input pulse energy , @xmath55 is the pulse energy along the propagation distance , and the integration is taken over the transversal cross - section @xmath56 and the local time @xmath57 of the pulse .
the latter also represents a normalization condition for the complex field envelope @xmath58 along the propagation distance @xmath2 . applying the energy normalization condition for the field envelope allows to perform absolute value calculations .
the nlse ( [ eq1 ] ) is solved numerically in a polar coordinate system ( assuming axial symmetry ) by using a split - step method similar to that applied in @xcite .
nonlinear media of highest ionization potentials pressurized neon and helium , are considered in this study . for better comparison of the pulse compression capabilities of neon and helium ,
completely identical initial conditions for the optical pulse and the medium have been taken for the initial numerical simulations in both media .
our simulations show that at @xmath59 mj input pulse energy and @xmath60 atm pressure of neon or helium , the maximal peak intensity does not surpass @xmath61 w/@xmath11 in neon and @xmath51 w/@xmath11 in helium along the whole propagation distance considered here . consequently , at such initial conditions , the parameters of the pulse fall well within the range of validity of the present model .
the sc in low dispersion regime is inherently accompanied by a gain of the peak intensity of the pulse .
that is why , two parameters have main importance here the magnitude of the peak intensity gain @xmath62 and the magnitude of the sc , @xmath63 . the peak intensity gain will be defined as the relation between the peak intensity of the pulse at given space point @xmath2 , @xmath64 , and the peak intensity of the input pulse , @xmath65 , _
i.e. _ , @xmath66
. the magnitude of the sc will be defined as the relation between the time duration of the pulse ( fwhm ) at given space point @xmath2 , @xmath67 , and the time duration of the input pulse @xmath68 , _ i.e. _ , @xmath69 .
the evolution of the transversal pulse width @xmath70 ( fwhm at @xmath71 plane ) , the normalized peak intensity @xmath72 ( _ i.e. _ , the peak intensity gain ) , and the pulse duration @xmath73 ( fwhm ) of the intensity profile along the propagation distance @xmath2 at 2 mj input pulse energy and 17 atm pressure are shown in figures 1(a),(b),(c ) for neon and helium .
the pulse shows similar behavior in both media , which is also close to the qualitative behavior of the pulse in the case of argon @xcite .
the pulse rearrangement begins with a confinement of the pulse in the transversal direction due to the sf , figure 1(a ) .
the minimal width of the pulse in the transversal direction is reached at @xmath74 m in neon and @xmath75 m in helium .
the sf confines the peripheral energy of the pulse toward the longitudinal axis , which leads to an increase of the peak intensity , figure 1(b ) .
the space positions at which the peak intensity reaches its maximal value is @xmath76 m in neon and @xmath77 m in helium .
the corresponding maximal peak intensity gain at these positions in neon and helium is approximately @xmath78 and @xmath79 times , respectively .
the most interesting and nonintuitive result is that the pulse undergoes sc in time ( _ before splitting _ ) while propagating along a positive gvd medium , figure 1(c ) .
the position of the maximal sc is @xmath80 m in neon and @xmath81 m in helium .
the magnitude of the sc in these positions is @xmath82 times in neon and @xmath83 times in helium .
the intensity gain at the position of maximal sc is approximately @xmath84 times in neon and @xmath85 times in helium .
as can be seen , the positions of the maximal sc and intensity gain do not coincide but are relatively close .
this means that within given propagation distance simultaneous time compression and intensity gain of the pulse can be achieved . in a `` chronological '' order ,
first appears the maximal sc at @xmath86 , next occurs the maximal intensity gain at @xmath87 , and finally the maximal self - focusing at @xmath88 .
once the pulse reaches the maximal sc at @xmath89 , it begins broadening in time from @xmath86 to @xmath87 and the magnitude of the sc reduces to @xmath90 times in neon and @xmath91 times in helium . at the same time , the intensity continues to rise until @xmath92 because the pulse continue to narrow in the transversal direction due to the sf while it reaches the position of maximal sf at @xmath93 .
the propagation of the pulse beyond @xmath87 leads to the usual pulse splitting [ 2 - 8,10,24 ] .
indication of the pulse splitting in figure 1(b ) is the rapid fall of the peak intensity at the central peak point of the pulse at @xmath71 beyond @xmath92 .
the comparative studies of the pulse compression in neon and helium clearly reveals a well expressed trend the pulse propagation dynamics develops more slowly in helium than in neon , while the magnitude of the pulse compression and the intensity gain is substantially higher in helium than in neon at equal other conditions .
this can be explained by the lower gvd of helium .
the latter diminishes the effect of dispersion spreading of the pulse while it effectively compresses in time due to the sf .
in fact , helium possesses the lowest specific gvd ( at equal other conditions ) and the highest ionization potential among the known media .
that is why it appears to be the top `` speed '' medium within the present concept of propagation type compression of high intensity femtosecond laser pulses .
the optical pulse do not tend to form light filaments within the basic set of processes , but starts to diverge once it reaches the position of maximal sf , @xmath93 , figure 1(a ) .
that is why , one may speculate that the generation of light filaments results from the defocusing effect of the above cubic nonlinearity , _
e.g. _ , @xmath94 , and the ionization .
these processes provide an instantaneous dynamical ( due to the @xmath94-nonlinearity ) and non - instantaneous ( due to the ionization ) balance of the @xmath95-sf and lead to formation of more or less stable light filaments [ 14,18,20 ] . for
the experimental conditions considered here , the intensity raised by the @xmath95-sf does not reach high enough values so as to `` switch on '' the @xmath94 and the ionization defocusing , and the @xmath95-sf is restricted only by the diffraction and the pulse splitting due to the dispersion .
the latter factors seem not to be sufficient to form relatively stable light filaments . a general explanation of the sc in low dispersion regime has been found @xcite based on the energy conservation . for the case of localized electromagnetic fields ( pulses ) , the energy conservation has a simple geometrical sense , expressed by eq.([eq3 ] ) . according to it
, the energy integral can be considered as a four - dimensional `` volume '' ( two transversal directions over the cross - section of the pulse , a longitudinal direction the local time , and a `` vertical '' direction the intensity @xmath96 ) , which must be conserved along the propagation coordinate @xmath2 .
an intensive rearrangement of the pulse is triggered by the sf coming from the nonlinear term .
the rapid ( relatively to the dispersive broadening ) development of the sf is the main point in the pulse compression mechanism considered here .
the sf confines the pulse radially from all transversal directions toward the longitudinal axis .
this increases the peak intensity and further accelerates the pulse rearrangement .
the increased peak intensity intensifies the self - phase modulation thus generating new spectral components the `` red''-frequency components are generated on the raising edge and the `` blue''-frequency components are generated on the falling edge of the pulse . in a positive gvd medium , this results in faster propagation of the fore edge and slower propagation of back edge of the pulse , which tends to cause pulse broadening . due to the low dispersion of the rare gases neon and helium ,
the pulse would not substantially expand in time ( within given propagation distance ) if it is ruled solely on the dispersion . at the same time , the strong confinement of the pulse in the transversal direction due to the sf results in a strong increase of the pulse intensity , figure 1(b ) .
the latter leads to an effective , but _
real , _ shortening of the pulse ( based on the standard fwhm characterization of the pulse duration ) before the pulse succeed to broaden substantially due to the dispersion .
this is the mechanism of the pulse compression observed here .
such a concept of femtosecond pulse compression , formulated for the first time in @xcite , has been called _ self - compression in low dispersion regime . _ the pulse compression in low dispersion regime is inherently accompanied by a strong increase of the peak intensity at almost no loss of energy , in accordance to the condition of conservation of the pulse energy .
this is the main advantage of the present pulse compression method .
in fact , the pulse behavior is subject of competition between two processes acting in opposite directions the pulse compression due to the intensity gain as a result of the sf and the pulse broadening due to the dispersion .
the final outcome depends on the relative rates of both trends .
pulse compression along given propagation distance can be achieved if the rate of pulse broadening due to dispersion remains lower that the rate of pulse shortening due to intensity gain forced by the sf . otherwise , pulse broadening and splitting will be observed .
this has been confirmed numerically @xcite , showing that if the gvd exceeds given value ( at equal other conditions ) , only pulse broadening and splitting , but not sc , is observed .
this means that the sf does not necessarily lead to pulse compression [ 33,34 ] .
the pulse compression in low dispersion regime is substantially related with the low value of gvd and the efficiency of this mechanism increases when the gvd diminishes .
this explains why the degree of sc and the intensity gain increases from argon @xcite toward neon and helium , figure 1 .
it is important to underline that the above sc phenomenon is substantially ( 3 + 1)d effect that is based on the sf and can not be observed in the ( 1 + 1)d case at positive gvd ( _ e.g. _ , optical fibers ) , where the sf is not considered .
the latter also shows the mutual relation between the different dimensions in the high - intensity femtosecond pulse dynamics .
although the pulse collapses in all dimensions in space and time , the confinement mechanism of the pulse within the basic set of optical processes is still incomplete .
this is because confinement `` force '' in this case , _
i.e. _ , the sf , exists only in the transversal space dimensions but not in the time dimension at positive gvd .
the sc in the time dimension results from the retarded dispersion broadening of the pulse due to the low gvd of the medium .
consequently , the stable pulse propagation in the positive gvd medium , as has been observed experimentally @xcite , requires additional process(es ) to stabilize the optical pulse .
the present mechanism of pulse compression is based on the geometrical rearrangement of the pulse , in which the nonlinearity by means of the sf plays a leading active role , whereas the gvd , while it is also very important , has a secondary role , _
i.e. _ , to promote or to hamper the sc .
that is why , such kind of sc is expected to work not only for initially chirp - free pulses , but for ( regularly ) chirped pulses and pulses having complicated , even irregular , frequency distortions . in the latter cases , the efficiency of such sc mechanism is subject of additional studies .
the specified _ basic set _ of optical processes is universal for all centrosymmetric media and experimental conditions . in particular ,
the basic lowest order processes are common for cases without or with ionization .
that is why , the mechanism of sc described here should also play important role for the cases with substantial ionization of the medium .
to what extend the proposed here mechanism of sc plays important role for the cases beyond the _ basic set , _ besides on the particular experimental conditions , depends also on existing of other , if any , accompanying process(es ) of comparatively strong action as those from the basic set . among such additional processes ,
the most important are expected to be the ionization and above - cubic nonlinearities @xcite .
the contribution of the ionization and the above - cubic nonlinearities ( in the case of higher intensity pulses ) to the present mechanism , as well as the efficiency of the latter , is subject of additional studies .
no principle limitations seem to exist for the present mechanism of pulse compression to work for much shorter pulses then the ones used here but more advanced propagation equation @xcite have to be used in that case . a kind of time shortening can also be found within the pulse splitting phenomenon ,
even if the initial pulse has newer reached shortening but broadens continuously and finally splits . at given stage of development of the split pulse , each of two sub - pulses , in which the initial pulse splits , may become shorter than the initial pulse .
the split pulse , however , is usually considered as deteriorated pulse , having reduced energy and peak intensity in each subpulse , and such a kind of compression seems useless for the practical application .
the sc before splitting , as it is observed here , must be clearly distinguished from the derivation of shorter subpulses within the pulse splitting phenomenon . , where pulses of maximal intensity are generated , versus the input energy of the pulse at fixed 17 atm pressure of helium.,width=316 ] for comparison with the results for 2 mj energy pulses , compression of higher energy 5 mj - pulses were also studied in neon and helium , keeping the other conditions same as above . in this case , we found lower degree of the peak intensity gain , @xmath97 times in neon and @xmath98 times in helium , which results in lower degree of pulse compression , @xmath99 times in neon and @xmath100 times in helium . whereas the initial peak intensity is higher for 5 mj input energy , the maximal peak intensity at @xmath87 does not substantially increases when the pulse energy increases from 2 mj to 5 mj .
thus , the maximal peak intensity at @xmath87 for the case of 2 mj input pulse energy is @xmath101 w/@xmath11 in neon and @xmath102 w/@xmath11 in helium , while for the case of 5mj these values are @xmath103 w/@xmath11 and @xmath104 w/@xmath11 , respectively .
the pulse compression is also lower at higher input energy .
consequently , the simple increase of the pulse energy does not necessarily lead to increase of the sc and the intensity gain and a proper optimization of the parameters of the initial pulse and the medium is required in order to achieve the highest performance from the pulse rearrangement process . in view of the non - intuitive sc behavior of the high - intensity femtosecond pulses versus the input pulse energy and medium density , a more detailed study on that subject is required and will be presented below . whereas the propagation of high intensity femtosecond laser pulses in the heavier rare gases ( argon , krypton and xenon ) has been widely investigated [ 10,11,13 - 15,19 - 22,24 ] , the lighter rare gases ( neon and , especially helium ) remain insufficiently explored as propagation media . from the other side , the above results show the superior self - compression and intensity gain capabilities of helium in comparison with that of neon and argon @xcite within the present pulse compression concept .
that is why , the further studies will be focused namely on helium . in the following ,
the dependence of the main parameters of the pulse on the input pulse energy and helium pressure will be shown .
the development of the maximal peak intensity at @xmath92 and the respective magnitude of the peak intensity gain versus the input pulse energy at fixed helium pressure , @xmath60 atm , are shown in figure 2 .
as one can see , the absolute value of the peak intensity increases with the input pulse energy , while , at the same time , the magnitude of the peak intensity gain , @xmath62 , decreases .
the increase of the peak intensity with the pulse energy is more rapid in the low energy scale and tends to saturate at high energies .
the space position , @xmath87 , where a pulse of the maximal peak intensity is created , as a function of the same values of the input pulse energies , as in figure 2 , is shown in figure 3 . the space position @xmath87 , where the pulse has maximal intensity gain , changes in a kind of reciprocal way with the pulse energy .
at relatively low energy , say @xmath105 mj , the position of the maximal peak intensity is very far from the entrance of the medium , and it rapidly approaches the entrance when the pulse energy increases , figure 3 . , where pulses of minimal pulse duration are generated , versus the input energy of the pulse at fixed 17 atm pressure of helium.,width=307 ] the dependence of the minimal pulse duration at @xmath89 and the respective magnitude of the pulse compression versus the input pulse energy at fixed helium pressure , @xmath60 atm , are shown in figure 4 .
the space positions , @xmath86 , where a pulse of minimal pulse duration is created , at the same input pulse energies , as in figure 4,is shown in figure 5 .
in contrast to the pulse intensity , figure 2 , the minimal pulse duration increases almost linearly with the input pulse energy . what is also interesting , the degree of the sc , @xmath106 , decreases with the increase of the pulse energy . at the same time , the positions @xmath86 where maximally compressed pulses are created , change in a similar reciprocal way in figure 5 as for the case of @xmath87 in figure 3 . at
the lowest energy used , @xmath105 mj , @xmath86 is very far from the entrance of the medium and it very rapidly moves toward the entrance when the energy increases , figure 5 . , where such pulses are generated , versus the gas pressure of the medium at fixed @xmath107 mj input energy of the pulse .
the coefficients of the intensity gain @xmath62 at the respective values of gas pressure are also given in the figure.,width=336 ] , where such pulses are generated , versus the gas pressure of the medium at fixed @xmath107 mj input energy of the pulse .
the coefficients of the self - compression @xmath106 at respective values of gas pressure are also given in the figure.,width=336 ] the dependence of the maximal peak intensity at @xmath92 as well as the position @xmath87 , where such pulses are generated , from the helium pressure at fixed input pulse energy of @xmath107 mj is shown in figure 6 .
the dependence of the minimal pulse duration at @xmath89 as well as the position @xmath86 , where such pulses are generated , on the helium pressure at fixed input pulse energy of @xmath107 mj is shown figure 7 .
the coefficients of the intensity gain , @xmath62 , and the sc , @xmath106 , at the respective gas pressure , are also given in the figures .
these studies show that the strongest degree of sc and intensity gain is achieved at lowest pressure , @xmath108 atm , which also means lowest value of the total gvd of the medium .
the peak intensity at the lowest helium pressure in figure 6 , as well as at the highest energy pulses in figure 3 , surpasses the accepted here value of maximal intensity of @xmath32 w/@xmath11 ( for the case of helium ) within about 10% , and these results are shown for completeness of the study .
further reduction of the pressure results in peak intensity which is substantially beyond the specified value , and , for @xmath107 mj pulses , the simulations were not performed below that pressure .
the evolution of the spatiotemporal pulse shape with the variations of gas pressure at fixed @xmath107 mj input pulse energy has been also investigated .
a `` time view '' of the spatiotemporal pulse shape taken at the position of maximal sc , @xmath89 , is shown in figure 8 .
the parameters of the predicted pulses at the specified space points are also given in the figure .
these results demonstrate the rearrangement capabilities of the present pulse formation method at strong nonlinear interaction with the medium and low dispersion regime .
if the medium ends around @xmath86 or within @xmath109 interval , a strongly compressed single femtosecond pulse of multi - millijoule energy and dramatically increased peak intensity will be generated .
input pulses of parameters used in these simulations can be generated by the present oscillator - chirped pulse amplifier titan : sapphire laser systems , which makes the results of this study suitable for practical applications .
, versus the gas pressure of the medium.,width=432 ] all results found here directly follow from the solution of the nlse . from the other side ,
the behavior of the pulse in regime of sc can be physically explained if we consider the interplay between the most active and important processes , in this case the sf and the dispersion . in the nlse , they are described by the terms @xmath110 ( in the space domain ) and @xmath111 , respectively , within such an approach , the slower development of the pulse compression in helium than in neon , figure 1 , can be explained with the lower nonlinear refractive index @xmath8 of helium and , at same peak intensity , smaller nonlinear term @xmath110 .
this makes the process of sf in helium more gradual than in neon . from the other side ,
the higher value of intensity gain , and , as a result higher value of sc , in helium can be explained by the lower value of gvd of helium . while more slowly , the nonlinear term succeeds to sf the pulse in helium at almost same spot diameter , figure 1(a ) . at the same time
, the lower gvd in helium prevents the pulse from substantial time spreading .
thus , the common action of both terms `` squeeze '' stronger the pulse energy in the spatiotemporal domain , which results in a higher increase in the `` vertical '' direction , _
i.e. _ , intensity gain , and , thus stronger sc , figure 1 .
the higher intensity gain and pulse compression at low input energy , figure 2 and figure 4 , also looks naturally if it is considered within the interplay between above two terms .
the low input energy for pulses of same initial pulse duration and same size in transversal direction means low input intensity and field amplitude .
the latter also means low steepness of the pulse fronts and , thus , relatively low value of the time derivatives . as a final result , this diminishes the expanding rate due to the dispersion term @xmath111 in which , the second time derivative determines dynamically the magnitude of the expanding rate
this explains why at low input energy , the magnitude of the sc is higher .
the above factors support the opposite trend with the increase of the input energy of the pulse .
the higher input energy means higher input peak intensity and , consequently , higher value of the field time derivatives and higher magnitude of the dispersion term .
the latter intensifies the process of the dispersion spreading and splitting , thus preventing the pulse to achieve higher levels of intensity gain and sc ( in comparison with the lower energy case ) .
the results in figure 3 and figure 5 also look transparent within the framework of such interplay . at low energy , but equal other conditions , the peak intensity and the field amplitude of the initial pulse is relatively low , which results in low rate of development of the self - focusing by the cubic term of the nlse , @xmath110 .
that is why , the position where the maximal sf , and , as a result , the maximal intensity gain and time compression occur is far from the entrance point of the medium , figures 3,5 .
increasing the input pulse energy , and thus , the input peak intensity , intensifies the sf and the pulse needs shorter distance , @xmath87 and @xmath86 , to develop the maximal intensity gain and sc , respectively .
finally , the increased peak intensity gain , figure 6 , and , as a result , stronger sc ( shorter pulse durations ) , figure 7 , with reduction of gas pressure can be explained by the reduced value of the gvd of the medium , and in this way reduced value of the whole dispersion term at lower pressure .
this diminishes the dispersion spreading and promotes the intensity gain and the sc .
of course , the reduced pressure reduces also the ( total ) nonlinear refractive index of the medium , and , in this way , the whole nonlinear term .
the reduction of the nonlinearity results in slower rate of development of the sf and extends distances @xmath87 and @xmath86 to generate pulses of maximal intensity and sc , see the data of @xmath87 and @xmath86 in figures 6 and 7 .
the numerical simulations show that the net effect of the reduced dispersion dominates , and the final effect is increased pulse compression . the existence of sc before the splitting has crucial importance for the application of this effect .
for example , it allows , to certain degree , to overcome , or , at least , to relax the requirement for existing of stable soliton - like pulse propagation for some applications .
the theoretical and experimental studies show that such pulses do not exist within the specified basic set of optical processes and experimental conditions , considered here . for many applications , however , it is not necessary to have a stable pulse along the whole propagation distance but only a pulse of specified parameters at given space point .
such point will be called space _
target point .
_ the generation of such pulses can be considered as a satisfactory alternative to the soliton generation at conditions not supporting solitons . in this way
, the _ soliton concept _ ( where it is not applicable ) can be replaced by a _
controllable guidance concept , _ which we introduce here .
the controllable guidance concept seems to be applicable in broad range of conditions , in which the pulse propagation dynamics can be unambiguously predicted .
the absence of ionization in our case plays important role in the realization of such a concept because it ensures a smooth and irreversible ( along the propagation distance of interest ) pulse propagation dynamics .
the latter consists in a single sc event , reaching maximal efficiency at @xmath86 , followed by slight broadening of the pulse along a short distance ( from @xmath86 to @xmath87 ) , while the peak intensity continuously increases ( reaching maximal value at @xmath92 ) , and , finally , initiation / development of splitting of the pulse at / beyond @xmath92 , figure 1 .
although the pulse evolves along the whole propagation distance , the sc mechanism together with the smooth pulse propagation dynamics facilitates the generation of a pulse of well predictable parameters at a desired space target point .
it is based on the fact that the magnitude of the time compression and the intensity gain as well as the space position of the strongest time `` focusing '' @xmath86 and the strongest intensity gain @xmath87 , can be controlled by the pulse ( _ e.g. _ , energy ) , figures 2 - 5 , and the medium ( density / pressure ) , figures 6 , 7 , parameters .
the pulse shape , figure 8 , also changes smoothly and could be a subject to control by the initial parameters .
such a control , however , has few shortcomings . at first , it is a passive kind of control and , once launched in the medium , the pulse is no more subject to external control . in the second place ,
such a control is not complete because the parameters of the compressed pulse can not be controlled independently . in any case , however , the investigations toward the controllable guidance concept look perspective . intensity profiles of the tsunami pulse at the space position of maximal time compression , at @xmath60 atm of helium . @xmath112 time shape ( dashed line )
is also shown for comparison.,width=144 ] a very interesting feature of the pulse propagation can be distinguished from the results in figures 1 and 8 . as can be seen , the initial pulse has relatively long pulse duration and relatively low peak intensity .
the parameters of the compressed pulse are most interesting for the practical applications around the points of maximal pulse compression , @xmath86 , and/or the maximal intensity gain , @xmath86 , which play role of target points .
far from given space target point , the pulse develops slowly and the peak intensity increases almost linearly with the propagation distance , figure 1(b ) . approaching the target point ,
the dynamical rearrangement of the pulse rapidly speeds up and its intensity increases in an exploding manner , figure 1(b ) , whereas the pulse duration shortens gradually , figure 1(c ) .
this is also illustrated by figure 8 if compare the initial pulse and some of the compressed pulse , say , at the lowest pressure
@xmath108 atm .
such a behavior of an optical pulse closely resembles the behavior of the water tsunami wave , approaching the shore .
that is why , such a compressed optical pulse whose intensity dramatically surges around the target points will be called _ optical tsunami pulse . _
the time shape of the intensity profile of the compressed pulse at @xmath86 and @xmath60 atm of helium is shown in figure 9 .
although the input pulse has gaussian temporal shape , the intensity time profile of the tsunami pulse is very close to @xmath113 pulse shape , also shown for comparison in figure 9 .
the latter represents the temporal shape of the fundamental soliton of the ( 1 + 1)d nlse in the case of medium of negative gvd @xcite .
insubstantial deviation of the shape of the tsunami pulse from that one of the fundamental soliton exists only around the wings of the pulse .
consequently , in accordance to the controllable guidance concept , a strongly compressed femtosecond optical tsunami pulse of huge intensity and soliton - like time intensity profile can be generated in given space point and the parameters of that pulse can be controlled within given limits by the input pulse and the medium parameters .
self - compression of multi - millijoule femtosecond laser pulses has been found as a result of the interplay of the diffraction , group velocity dispersion of second order and kerr nonlinearity of third order in lighter noble gases , neon and helium , at conditions of positive group velocity dispersion .
these three processes are sufficient to trigger the self - compression at a low magnitude of positive group velocity dispersion of the medium . a comparative study of the self - compression in neon and helium has been performed and it confirms the importance of the low value of group velocity dispersion for the operation of the self - compression mechanism .
the pressurized helium appears to be the best medium for compression the high energy femtosecond laser pulses within the pulse compression concept considered here .
the parameters of the compressed pulse can be controlled within certain limits by the initial pulse and the medium parameters .
a generation of compressed pulse of strongly increased intensity and soliton - like temporal profile in the vicinity of given space target point , called optical tsunami , is predicted .
y. silberberg , _ opt
. lett . _ * 15 * , 12821284 ( 1990 ) .
p. chernev and v. petrov , _ opt .
_ , * 17 * , 172174 ( 1992 ) . j. e. rothenberg , _ opt
_ , * 17 * , 583595 ( 1992 ) . j. k. ranka , r. w. schirmer , and a. l. gaeta , _ phys . rev
_ , * 77 * , 3783 ( 1996 ) . s. a. diddans , h. k. eaton , a. a. zozulya , and t. s. clement , _ opt . lett .
_ , * 23 * , 379381 ( 1998 ) . s. a. diddans , h. k. eaton , a. a. zozulya , and t. s. clement , _ ieee , j. selected topics of quant .
_ , * 4 * , 306316 ( 1998 ) . j. k. ranka , and a. l. gaeta , _ opt .
_ , * 23 * , 534 ( 1998 ) .
a. zozulya , s. diddams , a. g. van engen , and t. s. clement , _ phys .
_ , * 82 * , 1430 , ( 1999 ) . t. brabec and f. krausz , _ phys
_ , * 78 * , 3282 ( 1997 ) .
m. mlejnek , e. m. wright , and j. v. moloney , _ phys .
e _ , * 58 * , 4903 ( 1998 ) .
m. nurhuda , a. suda , m. hatayama , k. nagasaka , and k. midorikawa , _ phys .
a _ , * 66 * , 023811 ( 2002 ) .
s. tzortzakis , l. berge , a. couairon , m. franco , b. prade , and a. mysyrowicz , _ phys .
_ , * 86 * , 5470 ( 2001 ) . i. g. koprinkov , a. suda , p. wang , and k. midorikawa , _ jap . j. appl . phys .
_ , l 978 ( 1999 ) .
i. g. koprinkov , a. suda , p. wang , and k. midorikawa , _ phys .
_ , * 84 * , 3847 ( 2000 ) . i. g. koprinkov , a. suda , p. wang , and k. midorikawa , in _ soliton - driven photonics , _ a. d. boardman and a. p. sukhorukov ( eds ) , pp .
355358 , kluwer academic publishers , 2001 .
i. g. koprinkov , a. suda , p. wang , m. hatayama , and k. midorikawa , in _ proc . of
the nonlinear guided waves2001 _
, m. segev and f. lederer ( eds . ) , pp.232234 , 2001 .
the initial pulses used in these experiments are not transform - limited while not , at the same time , regularly frequency modulated , or chirped .
instead , they possess complicated frequency distortions ( sometimes , such pulses are also generally called `` chirped '' ) , acquired within the process of amplification , that can not be removed by purely dispersive pulse compressor ( say , diffraction grating pulse compressor ) , as the regularly chirped pulses .
that is why , sc of such pulses can not be achieved by passing though a purely dispersive linear medium , which is a well known trivial effect , but results from a complex nonlinear pulse rearrangement process @xcite .
this is especially clear from the pulse compression in solid bulk material , where the dispersion of the material remains constant and the sc results solely from an increase of the peak intensity , see ref.@xcite of this work .
s. tzortzakis , b. lamouroux , a. chiron , s.d .
moustaizis , d. anglos , m. franco , b. prade , and a. mysyrowicz , _ opt .
commun . _ * 197 * , 131 ( 2001 ) .
n. l. wagner , e. a. gibson , t. popmintchev , i. p. christov , m. m. murnane , h. c. kapteyn , _ phys .
_ , * 93 * , 173902 ( 2004 ) .
c. p. hauri , a. guandalini , p. eckle , w. kornelis , j. biegert , and u. keller , _ opt . express _ ,
* 13 * , 7541 ( 2005 ) .
couairon , m. francko , a. mysyrowics , j. biegert , and u. keller , _ opt .
_ , * 30 * , 2657 ( 2005 ) .
s. skupin , g. stibenz , l. berge , f. lederer , t. sokollik , m. schnurer , n. zhavoronkov , and g. steinmeyer , _ phys .
e _ * 74 * , 056604 ( 2006 ) .
j. liu , x. chen , j. liu , y. zhu , y. leng , j. dai , r. li , z. xu , _ opt . express _
, * 14 * , 979 ( 2006 ) . i.g .
koprinkov , m.d .
todorov , m.e .
todorova , and t.p .
todorov , _
b : at . mol .
phys _ , * 40 * , f231 ( 2007 ) .
h. nishioka , w. odajima , k. ueda , and h. takuma , _ opt .
_ , * 20 * , 2505 ( 1995 ) .
a. dalgarno and a. e. kingston , _ proc .
royal soc .
london , ser .
a _ * 259 * , 424 ( 1960 ) . s. f. larochelle , a. talebpour and s. l. chin , _ j. phys .
b : at . mol .
_ , * 31 * , 1215 ( 1998 ) . a. m. perelomov , v. s. popov , and m. v. terentev , _ sov . phys .
* 23 * , 924 ( 1966 ) .
b. walker , b. sheehy , l. f. dimauro , p. agostini , k. j. schafer , and k. c. kulander , _ phys .
_ , * 73 * , 1227 ( 1994 ) .
s. augst , d. d. meyerhofer , d. strickland , and s. l. chin , _ j. opt
b _ * 8 * , 858 ( 1991 )
. g. marcus , a. zigler , and z. henis ,
b _ * 16 * , 792 ( 1999 ) .
t. brabec and f. krausz , _ rev .
_ , * 72 * , 545 ( 2000 )
. g. l. mcallister , j. h. marburger , and l. g. deshazer , _ phys .
_ , * 21 * , 16481649 ( 1968 ) . j. h. marburger and w. g. wagner , _ ieee j. quantum electron .
_ , * qe-3 * , 415 ( 1967 ) . g. p. agrawal , _ nonlinear fiber optics _ , ( academic press , new york , 1994 ) . | self - compression of multi - millijoule femtosecond laser pulses and dramatic increase of the peak intensity are found in pressurized helium and neon within a range of intensity in which the ionization modification of the material parameters by the pulse is negligible .
the pulse propagation is studied by the ( 3 + 1)-dimensional nonlinear schrodinger equation including basic lowest order optical processes diffraction , group velocity dispersion of second order , and kerr nonlinearity of third order .
smooth and well controllable pulse propagation dynamics is found .
constructing of compressed pulse of controllable parameters at given space target point can be achieved by a proper chose of the pulse energy and/or gas pressure . |
the purpouse of this paper is twofold . on the one hand ,
it is intended to be a complement of the arenou & luri paper in this volume , presenting an example of application of a parametric method for luminosity calibrations . on the other hand ,
a new result for the cepheid pl relation using hipparcos data is presented , and the lmc distance modulus is deduced from it .
our working sample was formed by selecting all classical cepheids contained in the hipparcos catalogue ( esa , sp-1200 ) , even first overtone pulsators .
the resulting sample contains 238 stars .
all data ( including periods ) were taken from the hipparcos catalogue , except the radial velocities , that were taken from the _ hipparcos input catalogue _
( turon et al . 1992 ) .
as described in arenou & luri ( this volume ) parametrical methods can be applied to obtain unbiased luminosity calibrations . in our case , to obtain the pl relation of classical cepheids we have used the lm method .
this method , using maximum - likelihood estimation and specifically designed for luminosity calibrations , is fully described in luri et al .
its main advantages are : * all the available information for each star can be used : apparent magnitude , position , trigonometric parallax , proper motions and radial velocities as well as other relevant astrophysical parameters ( color index , period , etc . ) . *
it has been designed to fully exploit hipparcos data for luminosity calibrations .
* * the observational selection criteria used to construct the sample of stars are taken into account .
this is essential to obtain unbiased results ( brown et al . , 1997 arenou & luri , this volume ) . * * the effects of observational errors are also taken into account , thus avoiding biases ( arenou & luri , this volume ) .
* it gives at the same time the kinematical properties and spatial distribution of the sample .
* it is able to handle inhomogeneous samples composed of groups of stars with different luminosity , kinematical or spatial characteristics . *
it provides an individual distance estimate with optimal statistical properties ( arenou & luri , this volume ) .
as described in arenou & luri ( this volume ) , to apply a parametric method ( the lm method in our case ) it is necessary to model the physics and observational errors of the sample as well as the censorship applied for its construction . each distribution ( luminosity , velocity , spatial and error distribution ) and the observational selection is described by a p.d.f . depending on a set of parameters .
the values of these parameters are determined by applying the lm method to the sample .
in our case we have used the modeling described in this section to describe our sample of hipparcos cepheids .
* linear mean pl relation : @xmath1 * gaussian dispersion ( @xmath2 ) of the individual absolute magnitude around the mean relation + actually , the real intrinsic distribution has a top - hat shape , but combined with the errors in the estimation of the interstellar absorption it is reasonably well approximated by a gaussian . * interstellar absorption from the 3d model by arenou et al .
( 1992 ) * velocity ellipsoid : @xmath3 * galactic rotation : oort - lindblad model at first order with @xmath4 , @xmath5 and @xmath6 ( kerr & lynden - bell , 1986 ) . * exponential disk : @xmath7 as we are dealing with an all - sky sample , we can safely assume that the astrometrical observational errors of the stars are uncorrelated and follow a gaussian distribution ( arenou & luri , this volume ) .
we have also assumed gaussian errors for radial velocity but we have considered negligible the errors in apparent magnitude , position and period . as we have included all the classical cepheids contained in the hipparcos catalogue ,
the only observational selection is the one of the catalogue itself .
the main censorship of the catalogue is in apparent magnitude : the catalogue is complete up to the hipparcos magnitude @xmath8 ( the survey ) and its magnitude limit is @xmath9 .
we have modeled this censorship with a selection function giving completeness up to a certain apparent magnitude @xmath10 ( estimated by the method togheter with the rest of parameters ) and a linear decrease of the completeness up to the limiting magnitude ( figure [ fig_s ] ) . on the other hand ,
we have considered the censorship on @xmath11 not significant on average . as pointed out in sec .
[ sec_sample ] , our sample is a mixture of two types of cepheids : fundamental pulsators and overtone pulsators .
we have classified the stars in these two types using the fourier coefficients obtained from the hipparcos light curves : 204 were classified as fundamental pulsators and 34 as overtone pulsators .
we have also made two supplementary hypothesis : 1 .
the slopes of the pl relations for both groups have been fixed to the values for the smc obtained by the eros team ( sasselov et al . , 1997 ) :
@xmath12 ( fundamental ) and @xmath13 ( overtones ) .
only the zero points of the pl relations have been determined , @xmath14 ( fundamental ) and @xmath15 ( overtones ) .
2 . we have assumed that both groups have the same kinematic and spatial distributions ( i.e. the same velocity ellipsoid and scale height ) .
the application to our sample of the lm method with the modeling described above gives the following estimates for the zero points of the pl relations : cepheids pulsating on the fundamental mode : : : @xmath16 cepheids pulsating on the 1st overtone : : : @xmath17 on the other hand , the values obtained for the kinematical and spatial distribution parameters are @xmath18 @xmath19 , @xmath20 @xmath19 and @xmath21 @xmath22 .
the lmc distance modulus can be calculated from the difference of zero points between our pl relation and the eros apparent magnitude pl relation for the lmc ( sasselov et al .
however , the difference of zero points has to be corrected of the effects of interstellar absorption and metallicity and thus two more assumptions are needed : * we have assumed a value of @xmath23 for the lmc extinction ( from freedman et al . , 1994 ) .
then , we have calculated the absorption from this value using the ratio @xmath24 * we have assumed a correction of @xmath25 to take into account the effects on the zero points of the metallicity difference between the lmc and our galaxy ( laney & strobie , 1994 ) this result is in agreement with the short " distance scale obtained from ( for instance ) rr - lyrae ( luri et al .
1998 ) but not compatible with the long " distance scale from other hipparcos results for cepheids ( feast & catchpole , 1997 ) .
please notice that the error estimation does not take into account the error in the value of @xmath27 for the lmc . | the distance modulus of the large magellanic cloud ( lmc ) is the first step on the determination of the cosmic distance scale . in this paper
it is obtained from a new calibration of the classical cepheid period - luminosity ( pl ) relation .
our new calibration is obtained by applying the lm method a parametric method designed to obtain luminosity calibrations from astrometric information to the hipparcos data for classical cepheids .
the resulting distance modulus is @xmath0 . |
the cumulative connection constants ( * ccc * ) @xmath0 were introduced by the authors of @xcite : @xmath1 where @xmath2 and pascal - like array @xmath3 of connection constants @xmath4 `` connects '' two polynomial sequences ( note : @xmath5 is a polynomial sequence if @xmath6 ) .
+ `` the connection constants problem '' i.e. combinatorial interpretations , algorithms of calculation , recurrences and other properties - is one of the issues of the binomial enumeration in finite operator calculus ( foc ) of rota - roman and others and in its afterwards extensions ( see : abundant references in @xcite-@xcite,@xcite ) .
important knowledge on recurrences for @xmath3 one draws from @xcite - consult also the navima group program : + http://webs.uvigo.es/t10/navima/ .
+ + + the cumulative connection constants ( * ccc * ) as defined above appear to represent combinatorial quantities of primary importance - as shown by examples below .
a special case of interest considered in @xcite is the case of monic persistent root polynomial sequences @xcite characterized by the following conditions : @xmath7 the pair of persistent root polynomial sequences and hence corresponding connection constants array are then bijectively labeled by the pair of root sequences : @xmath8=\left\ { r_{k}\right\}_{k\geq 1},\;\;\;[s]=\left\ { s_{k}\right\}_{k\geq 1}.\ ] ] in this particular case of persistent root polynomial sequences the authors of @xcite gave a particular name to connection constants : `` generalized lah numbers '' - now denoted as follows @xmath9 .
the generalized lah numbers do satisfy the following recurrence @xcite : @xmath10 @xmath11 from which the recurrence for cumulative connection constants ( * ccc * ) follows @xmath12 where @xmath9 .
+ the clue examples of @xcite are given by the following choice of a pair of root sequences : + + + @xmath13=\left\ { r_{k}\right\}_{k\geq 1}=\left\{0,0,0,0,0,0,\ldots\right\},\ ; [ s]=\left\ { s_{k}\right\}_{k\geq 1}=\left\{0,0, ... ,0, ... \right\}= [ 0]$ ] + + + + @xmath14=\left\ { r_{k}\right\}_{k\geq 1};\;\ ; r_{1 } = 0 , r_{2 } = 2 , r_{k } = 1- ( r_{k-1})^{2}$ ] for @xmath15 and @xmath16 = [ 0]$ ] .
we shall supply now several examples of connection constants arrays and corresponding cccs - fibonomial case included - altogether with their combinatorial interpretation .
( help ! - for notation - help - see : appendix ) .
+ * example 1 .
* @xmath17 @xmath18 here @xmath19 is the translation operator ( @xmath20 ) and the recurrence ( [ c ] ) is obvious .
+ * example 2 .
* @xmath21 denote stirling numbers of the second type .
then @xmath22 . let @xmath23 , then + @xmath24 we easily get from ( [ c ] ) the inspiring identity @xmath25 * example 3 . * @xmath26 $ ] are stirling numbers of the first kind
. then @xmath27 x^{k},\;n\geq 0 $ ] .
@xmath28=n!$ ] and the obvious recurrence @xmath29 coincides with ( [ c ] ) .
* example 4 .
* @xmath30 , @xmath31 @xmath32 where @xmath33 denotes the lattice of all subspaces of @xmath34 - ( the @xmath35-th dimensional space @xmath34 over galois field @xmath36 ) - see : @xcite .
+ according to konvalina @xmath37 is at the same time the number of all possible choices of objects from exponentially weighted boxes @xcite .
+ as @xmath38-gaussian polynomials @xmath39 constitute a persistent root polynomial sequence - the generalized lah numbers - which are now are @xmath40 must satisfy the recurrence equation @xmath41 where @xmath42=\left\ { q^{k-1}\right\}_{k\geq 1}$ ] is the root sequence determining @xmath43 . see appendix 2 .
hence the recurrence ( [ c ] ) for the number @xmath0 of all possible choices of objects from exponentially weighted boxes @xcite is of the form @xmath44 because @xmath45 in another @xmath38-umbral notation ( see : appendix 2 ) @xmath46 occasionally note that using the mbius inversion formula ( see : @xcite - the @xmath38-binomial theorem corollaries in section 5 - included ) we have @xmath47 i.e. the identity
@xmath48 `` the mbius @xmath38-inverse example '' is then the following .
* example 4. * @xmath49 , where ( written with help of generalized shift operator @xcite - analogously to the use of @xmath50 above in example 1 . ) we have ( see appendix 2 ) : @xmath51 compare with @xmath52 in @xcite ( pp .
240 - 241 ) and note that @xmath53 is the number of all possible choices of objects from exponentially weighted boxes @xcite .
it is then combinatorial - natural to consider also another triad . * example 4 .
* @xmath54 .
* example 5 .
* this is the fibonomial example : @xmath55 where @xmath56 are called fibonomial coefficients ( see a3 ) .
using the mbius inversion formula - analogously to the example 4 case - we get @xmath57 where mbius matrix @xmath58 is the unique inverse of the incidence matrix @xmath59 defined for the `` fibonacci cobweb '' poset introduced in @xcite ( see : a4 for explicit formula of @xmath59 from @xcite ) . for combinatorial interpretation of these fibonomial coefficients
see : appendix 4 .
interpretation of @xmath0 stated below results from combinatorial interpretation of fibonomial coefficients .
combinatorial interpretation of ccc @xmath0 in this case is then the following : @xmath60_number of such subposets of the fibonacci cobweb poset @xmath61 which are cobweb subposets starting from the level @xmath62 and ending at the level labeled by @xmath63 . _ as in the @xmath38-umbral case of the example 4 we have `` @xmath64-umbral '' representation of the fibonomial ccc @xmath0 ( see a3 ) .
ccc - recurrences from examples 1 - 4 were an easy game to play .
all these four cases of ccc might be interpreted on equal footing due to the ingenious scheme of one unified combinatorial interpretation by konvalina @xcite . as for the other cases the `` exercise '' of finding the recurrences for ccc - up to the fibonomial case
- might be the slightly harder task . in this last case with fibonomial coefficients
one may derive the well known recurrence @xmath66 @xmath67 or equivalently @xmath68 @xmath67 also by combinatorial reasoning @xcite ( it might be a hard exercise - not to check it but to prove it another way - a use of @xcite - being recommended ) .
derivation of the recurrence for fibonomial ccc @xmath0 - we leave as an exercise .
@xmath38-binomial and fibonomial cases are specifications of the @xmath69-sequence choice in the so called @xmath69- calculus notation @xcite-@xcite in conformity with @xcite referring back to works of brenke and boas on generalized appell polynomials ( see abundant references in @xcite ) recall now that the jackson @xmath77- derivative is defined as follows @xcite-@xcite : @xmath78 - and is also called the @xmath38-derivative . the consequent notation for the generalized shift operator @xcite-@xcite,@xcite
is : @xmath79 then we identify @xmath80 .
( ward in @xcite does not use any subscript like @xmath38 or @xmath64 or @xmath69 . in straightforward analogy consider now
the _ fibonomial coefficients _ @xmath83 , @xmath84 where @xmath85,@xmath86;@xmath87 and linear difference operator @xmath88 acting as follows : @xmath89 - we shall call the @xmath64-derivative . then in conformity with @xcite and with notation as in @xcite
one writes : [ [ combinatorial - interpretation - of - fibonomial - coefficients ] ] combinatorial interpretation of fibonomial coefficients @xcite ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ this fibonacci cobweb partially ordered infinite set @xmath61 is defined as in @xcite via its finite part - cobweb subposet @xmath95(_rooted at @xmath96 level subposet _ ) to be continued ad infinitum in an obvious way as seen from the figure of @xmath97 below .
it looks like the fibonacci tree with a specific `` cobweb '' .
it is identified with hasse diagram of the partial order set @xmath61 .
if one defines this poset @xmath61 with help of its incidence matrix @xmath59 representing @xmath61 uniquely then one arrives at @xmath59 with easily recognizable staircase - like structure - of zeros in the upper part of this upper triangle matrix @xmath59 .
this structure is depicted by the figure 2 where : empty places mean zero values ( under diagonal ) and in filled with - - ... - parts of rows ( above the diagonal ) `` - '' stay for ones .
the picture below is drawn for the sequence @xmath98 , where @xmath99 are fibonacci numbers .
1 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + 1 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + 1 * * - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + 1 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + 1 * * - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + 1 * * - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + 1 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + @xmath1001 * * - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + 1 * * - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + 1 * * - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + 1 * * - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + 1 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - + @xmath1011 * * - - - - - - - - - - - - - - - - - - - - - - - - - - + 1 * * - - - - - - - - - - - - - - - - - - - - - - - - - + 1 * * - - - - - - - - - - - - - - - - - - - - - - - - - + 1 * * - - - - - - - - - - - - - - - - - - - - - - - - - + 1 * * - - - - - - - - - - - - - - - - - - - - - - - - - + 1 * * - - - - - - - - - - - - - - - - - - - - - - - - - + 1 * * - - - - - - - - - - - - - - - - - - - - - - - - - + 1 - - - - - - - - - - - - - - - - - - - - - - - - - + @xmath1021 * * - - - - - - - + 1 * * - - - - - - - + 1 * * - - - - - - - + 1 * * - - - - - - - + 1 * * - - - - - - - + 1 * * - - - - - - - + 1 * * - - - - - - - + 1 * * - - - - - - - + 1 * * - - - - - - - + 1 * * - - - - - - - + 1 * * - - - - - - - + 1 * * - - - - - - - + 1 - - - - - - - + @xmath103 ......................... + _ and so on _ the characteristic function @xmath106 = @xmath107 of the partial order relation @xmath108 of the graded * @xmath64 * -cobweb poset @xmath61 - or in brief just @xmath59 - might be expressed by kronecker delta as here down ( @xcite ( 2003 ) and see also [ 12,40,18,16 ] ) * for any natural numbers valued sequence @xmath64 * - including the motivating example with fibonacci sequence @xmath64 denominated graded cobweb poset @xmath61 for which it reads ( @xmath109 ) : * .
* @xmath95 depending on the context is also to be viewed on as representing the set of maximal chains .
it is naturally a particular case of the layer of graded poset called cobweb poset [ 1 - 23 ] - for the reader convenience - see * * with basic ponderables . for the sake of coming right now combinatorial interpretation , all layers on one hand are viewed on as subposets while on the other hand are also to be viewed on as representing the set of maximal chains .
consider now the following behavior of a _
`` sub - cob '' _ moving from any given point of the @xmath62 level of the poset up .
it behaves as it has been born right there and can reach at first @xmath119 points up then @xmath120 points up , @xmath121 and so on - thus climbing up to the level @xmath122 of the poset @xmath61 .
it can see - as its great ancestor at the root @xmath96-th level and potentially follow one of its own accessible finite subposet @xmath95 .
one of many @xmath95 s rooted at the @xmath123-th level might be found .
how many ?
let @xmath124 .
the number of subposets equipotent to subposet @xmath95 rooted at any * fixed * point at the level labeled by @xmath62 and ending at the @xmath35-th level labeled by @xmath63 is equal to @xmath125 let @xmath126 be a natural numbers valued sequence with @xmath127 ( or @xmath128 being exceptional as in case of fibonacci numbers ) .
any such sequence uniquely designates both @xmath64-nomial coefficients of an @xmath64-extended umbral calculus as well as @xmath64-cobweb poset introduced by this author ( see : the source [ 19 ] from 2005 and earlier references therein ) .
if these @xmath64-nomial coefficients are natural numbers or zero then we call the sequence @xmath64 - the @xmath64-*cobweb admissible sequence*. let @xmath135 then the @xmath136 set of hasse sub - diagram corresponding maximal chains defines biunivoquely the layer @xmath137 as the set of maximal chains nodes and vice versa - for these * graded * dags ( kodags included ) . * theorem * [ 20 ] + ( kwaniewski ) _ for @xmath64-cobweb admissible sequences @xmath64-nomial coefficient @xmath138 is the cardinality of the family of _ equipotent _ to @xmath139 mutually disjoint maximal chains sets , all together * partitioning * the set of maximal chains @xmath140 of the layer @xmath141 , where @xmath142 .
_ for february 2009 readings on further progress in combinatorial interpretation and application of the presented author invention i.e. partial order posets named cobweb posets and their s corresponding encoding hasse diagrams kodags see [ 19 - 29,31 - 42 ] and references therein
. for active presentation of cobweb posets see [ 40 ] .
this electronic paper is an update of [ 41 ] let @xmath219 .
let @xmath220 .
let @xmath221 be the graded partial ordered set ( poset ) i.e. @xmath222 and @xmath223 constitutes ordered partition of @xmath221 .
a graded poset @xmath221 with finite set of minimal elements is called * cobweb poset * _ iff _
@xmath224 @xmath225 the definition 6.1 . is the reason for calling hasse digraph @xmath229 of the poset @xmath230 a * * * * dag as in professor * * * * azimierz * * * * uratowski native language one word * mplet * means * complete ensemble * - see more in [ 23 ] and for the history of this name see : the internet gian - carlo polish seminar * subject 1 .
odags and kodags in company * ( dec . 2008 ) . *
acknowledgments * thanks are expressed here to the now student of gdask university maciej dziemiaczuk for applying his skillful tex - nology with respect to most of my articles since three years as well as for his general assistance and cooperation on kodags investigation .
maciej dziemiaczuk was not allowed to write his diploma with me being supervisor - while maciej studied in the local bialystok university where my professorship till 2009 - 09 - 30 comes from .
l. colucci , o.m .
dantona , c. mereghetti : _ fibonacci and lucas numbers as cumulative connection constants " _ , the fibonacci quaterly 38.2 ( 2000 ) , pp.157 - 164 e.damiani,o.m .
dantona , g. naldi _ on the connection constants_,studies in appl .
85(4)(1991 ) , pp .
289 - 302 a. di bucchianico , d. loeb : _ sequences of binomial type with persistent roots _ , j. math .
39 - 58 goldman j. rota g - c .
: _ on the foundations of combinatorial theory iv ; finite - vector spaces and eulerian generating functions _ studies in appl .
49 1970 , pp .
239 - 258 j.konvalina : _ generalized binomial coefficients and the subset - subspace problem _ , adv . in appl .
21 ( 1998 ) , pp .
228 - 240 j. konvalina : _ a unified interpretation of the binomial coefficients , the stirling numbers and the gaussian coefficients _ , the american mathematical monthly vol .
107 , no 10 , ( 2000 ) , pp.901 - 910 ward m. : _ a calculus of sequences _ , amer.j
58 ( 1936 ) pp .
255 - 266 a. k. kwaniewski : _ main theorems of extended finite operator calculus _ , integral transforms and special functions vol 14 , no 6 , ( 2003),pp .
499 - 516 a.k.kwaniewski : _ towards @xmath69-extension of finite operator calculus of rota _ , rep .
phys.48 no 3 ( 2001)pp . 305 - 342 a.k.kwaniewski :
_ on simple characterisations of sheffer @xmath69-polynomials and related propositions of the calculus of sequences _ , bulletin de la soc . des sciences et de lettres de lodz ; 52 , ser . rech .
36 ( 2002 ) 45 - 65 .
arxiv : math.co/0312397 s.r.roman : _ more on the umbral calculus with emphasis on the q - umbral calculus _ , j. math .
107(1985 ) , pp . 222 - 254 a.k.kwaniewski : _ information on combinatorial interpretation of fibonomial coefficients_,inst .
uwb / preprintno**52**/november/*2003 * , @xmath235 , bull .
42 ( 2003 ) pp .
39 - 41 arxiv : math.co/0402291 v 1 22 feb 2004 a. k. kwaniewski : _ on duality triads _ , bull .
42 ( * 2003 * ) pp.11 - 25 arxiv : math.gm/0402260 v 1 feb .
2004 ; a. k. kwaniewski : _ the second part of on duality triads paper _ bull .
42 ( * 2003 * ) pp.27 -37 arxiv : math.gm/0402288 v 1 feb .
2004 a.k.kwaniewski : _ combinatorial interpretation of the recurrence relation for fibonomial coefficients _ , bulletin de la societe des sciences et des lettres de lodz ( 54 ) serie : recherches sur les deformations vol .
44 ( 2004 ) pp .
23 - 38 , arxiv : math/0403017v2 [ v1 ] mon , 1 mar 2004 02:36:51 gmt a. krzysztof kwaniewski , _ on cobweb posets and their combinatorially admissible sequences _ , adv .
studies contemp .
18 ( 1 ) , 2009 vol .
18 no 1 , 2009 17 - 32 .
arxiv : math/0512578v4 [ v1 ] mon , 26 dec 2005 20:12:57 gmt a. krzysztof kwaniewski , _ more on combinatorial interpretation of fibonomial coefficients _ , inst .
comp . sci.uwb/preprint**no56**/november/*2003 * , bulletin de la societe des sciences et des lettres de lodz ( 54 ) serie : recherches sur les deformations vol . 44 ( 2004 )
23 - 38 , arxiv : math/0402344v2 [ v1 ] 23 feb 2004 mon , 20:27:13 gmt a. krzysztof kwaniewski , _ cobweb posets and kodag digraphs are representing natural join of relations , their dibigraphs and the corresponding adjacency matrices _ , arxiv:0812.4066v1 [ v1 ] sun , 21 dec 2008 the 19 pages chistmas eve article ewa krot _ a note on mbius function and mbius inversion formula of fibonacci cobweb poset _ , bulletin de la societe des sciences et des lettres de lodz ( 54 ) , serie : recherches sur les deformations vol . 44 ( 2004 ) , 39 - 44 arxiv : math/0404158v2 , [ v1 ] we d , 7 apr * 2004 * 10:23:38 gmt [ v2 ] we d , 28 apr 2004 07:37:14 gmt ewa krot _ the first ascent into the incidence algebra of the fibonacci cobweb poset _ , advanced studies in contemporary mathematics 11 ( 2005 ) , no . 2 , 179 - 184 , arxiv : math/0411007v1 [ v1 ] sun , 31 oct _ 2004 _ 12:46:51 gmt
* remark 1 * the matrix elements of @xmath143 were given in 2003 ( [ 18 ] kwaniewski ) using @xmath144 labels of vertices in their `` natural '' linear order : + * 1 .
* set @xmath145 , + * 2 .
* then label subsequent vertices - from the left to the right - along the level @xmath123 , + * 3 .
* repeat 2 . for @xmath146 until @xmath147 ; @xmath148 inspired [ 8,9,13 ] by gauss @xmath149 finite geometries numbers and in the spirit of knuth `` notationlogy '' [ 30 ] we shall refer here also to the upside down notation effectiveness as in [ 21 - 24,20 ] or earlier in [ 8,9,10,13 ] , ( consult also the appendix in [ 33 ] ) .
* as for * that upside down attitude @xmath150 being much more than `` just a convention '' to be used substantially in what follows as well as for the sake of completeness - * let us * quote it as the principle according to kwasniewski [ 42 ] where it has been formulated as an `` * * of course * * '' principle i.e. simultaneously trivial and powerful statement .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * the upside down notation principle . * + * 1 .
let * the statement @xmath151 depends only on the fact that @xmath64 is a natural numbers valued statement .
then * if one proves that @xmath152 is true - the statement @xmath153 is also true .
formally - use equivalence relation classes induced by co - images of @xmath154 and proceed in a standard way .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ in order to proceed further let us now recall - rewrite purposely here kwaniewski @xmath155 - formula for @xmath59 function of * arbitrary * cobweb poset so as to see that its algorithm rules authomatically make it valid for all @xmath64-cobweb posets where @xmath64 is any natural numbers valued sequence i.e. with .
@xmath156 stays for the incidence algebra of the poset @xmath157 over the commutative ring @xmath105 .
@xmath164 @xmath165 + * note .
* @xmath166 `` produces the pacific ocean of '' in the whole upper triangle part of a would be incidence algebra @xmath167 matrix elements + * note .
* @xmath168 cuts out i.e. thus producing `` zeros @xmath64-la scala staircase '' in the delivered by @xmath166 .
+ this results exactly in forming rectangular triangles : @xmath169 of them at the start of subsequent stair and then down to one till - after @xmath169 rows passed by one reaches a half - line of @xmath170 which is running to the right- right to infinity and thus marks the next in order stair of the @xmath64- la scala .
+ the @xmath59 matrix explicit formula was given for arbitrary graded posets with the finite set of minimal in terms of natural join of bipartite digraphs in snack = the * * s**ylvester * * n**ight * * a**rticle on * * k**odags and * * c**obwebs = [ 21 ] . + what was said is equivalent to the fact that the cobweb poset coding la scala is of the natural join operation origin while thus producing @xmath59 matrix [ 23,21,24 ] which is of the form ( quote from snack = [ 21 ] , see : subsection 2.6 . )
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the explicit expression for zeta matrix @xmath171 of cobweb posets via known blocks of zeros and ones for arbitrary natural numbers valued @xmath64- sequence was given in ( here ) [ 23 ] due to more than mnemonic efficiency of the up - side - down notation being applied ( see [ 23 ] and references therein ) . with this notation inspired by gauss and replacing @xmath123 - natural numbers with `` @xmath172 '' numbers one gets @xmath173\ ] ] where @xmath176 stays for @xmath177 matrix of ones i.e. @xmath178_{ij }
= 1 $ ] ; @xmath179 and @xmath180 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ in the @xmath181 formula from [ 23 ] @xmath182 denotes the boolean product , hence - used for boolean powers too .
we readily recognize from its block structure that @xmath64-la scala is formed by of block - diagonal matrices @xmath183 which sacrifice these their to constitute the @xmath123-th subsequent stair in the @xmath64-la scala descending and descending far away down to infinity .
thus the cobweb poset coding la scala is due to the natural join origin of @xmath59 matrix . because of that the above last expression of the @xmath59 expressed in terms of @xmath187 may be still simplified [ for the sake of verification and portraying via computer simple program implementation ] .
namely the following is true : it was remarked by my * gdask university * student * maciej dziemiaczuk * - that my @xmath161 ( equivalent ) expressions are valid according to him only for @xmath64 = fibonacci sequence . in view of the upside down notation principle
if any of these is valid for any particular natural numbers valued sequence @xmath64 it should be true for all of the kind .
prove the claim : + _ dziemiaczuk formula is equivalent to kwaniewski formulas_. + - what is against ?
+ - what is for ?
my `` for '' is the socratic method question .
why not use the arguments in favor of @xmath191 * remark 2 .
krot choice .
* while the above is established it is a matter of simple observation by inspection to find out how does the the mbius matrix @xmath206 looks like . using in [ 25,26 ] this author example and expression for @xmath59 matrix
this has been accomplished first ( see also [ 26 ] ) for fibonacci sequence and then - via automatic extension - the same formula was given for @xmath64 sequences as above by my former student in her recent articles [ 27,28 ] .
here is her formula for the cobweb posets mbius function ( see : ( 6 ) in [ 28 ] ) .
+ @xmath207 now bearing in mind the upside down notation principle * for all * @xmath64-cobweb posets with @xmath208 ( as it should be for natural numbers valued sequences ) we may now rewrite the above in coordinate grid @xmath209 as below .
the further relevant references of the same author see [ 27 - 29 ] . the above rewritten mbius function formula is of course literally * valid for all * natural numbers valued sequences @xmath64 .
consult : the recent note `` on characteristic polynomials of the family of cobweb posets '' [ 29 ] .
the author of [ 25 ] introduces parallely also another form of @xmath59 function formula and since now on -except for [ 27]- in subsequent papers [ 26,28,29 ] their author uses the formula for @xmath59 function in this another form . namely - this other form formula for @xmath59 function in the present authors grid coordinate system description of the cobweb posets was given by krot in her note on mbius function and mbius inversion formula for fibonacci cobweb poset [ 23 ] with @xmath64 designating the fibonacci cobweb posets . in [ 24 ] the formula for the mbius function for fibonacci sequence @xmath64 was rightly treated as literally * valid for all * natural numbers valued sequences @xmath64 and the alternative .
of course the same concerns the @xmath59 function formulas - the former and the latter .
here is this other latter form formula for @xmath59 function ( see : ( 7 ) in [ 25 ] or ( 1 ) in [ 27 ] ) .
+ let @xmath210 and @xmath211 where @xmath212 , @xmath213 while @xmath214 . | in this note we present examples of cumulative connection constants - new fibonomial ones included . all examples posses combinatorial interpretation .
+ this presentation is an update of a.k.kwaniewski _ fibonomial cumulative connection constants _ * bulletin of the ica vol .
44 ( 2005 ) , 81 - 92 .
* ams classification numbers : 06a06 , 05b20 , 05c75 , 05a40 this electronic paper is affiliated to the internet gian - carlo polish seminar : _ http://ii.uwb.edu.pl/akk/sem/sem_rota.htm _ |
numerical simulations of gluons on a space - time lattice provide at present the most reliable means of studying glueballs .
glueball correlation functions are , however , notoriously difficult quantities to measure in monte carlo simulations : since the masses of these states are rather high and their creation operators have large vacuum fluctuations , the signal - to - noise ratio falls extremely rapidly as the separation between the source and sink is increased .
because of this , reliable studies of glueballs on fine lattices have required prohibitively large computer resources .
thus , the development of more efficient simulation techniques in lattice qcd is crucial to establishing a detailed description of glueballs and their interactions .
the objective of this work is to examine the effectiveness of using an improved , anisotropic lattice action to reduce the computational effort needed to determine the glueball spectrum in quenched qcd .
improved actions allow access to continuum physics on coarser lattices than possible using the simple wilson discretization .
coarse lattice simulations are more efficient for several reasons : for a given physical volume , much fewer lattice sites are needed ; the alleviation of critical slowing down permits the faster generation of statistically independent gauge - field configurations ; glueball operator smearing is faster due to the decreased number of links and a decrease in the number of smearing iterations required ; glueball wavefunctions extend over much fewer lattice sites on a coarse lattice , making the variational technique far more effective when using a feasible number ( a dozen or so ) of basis operators .
however , for glueball mass calculations , the coarseness of the temporal lattice spacing is a severe drawback .
as the masses in lattice units of the states of interest are so large , the number of correlator time intervals which can be measured is reduced greatly @xcite .
a straightforward solution to this problem which preserves the computational advantages of coarse lattices@xcite is to make use of anisotropic lattices in which the temporal spacing is much smaller than that in the spatial directions .
this enables us to exploit the enhanced signal - to - noise of the correlation functions at smaller temporal separations . a natural scale for
the temporal lattice spacing should be the inverse of the energy of the states of interest ; thus , for glueballs , a temporal cut - off larger than 1.5 gev allows resolution from accessible statistics of the correlator over a few time - slices .
meanwhile , the scale for the spatial lattice should be set by the size of the wavefunction of the state ; a spatial grid separation in the range @xmath2 fm would seem reasonable . since we propose to use lattices
in which the temporal lattice spacing is small , improvement of the discretization in this direction is not needed .
thus , a lattice action which couples only nearest - neighbor time - slices can be used .
the transfer matrix corresponding to such an action is hermitian and positive definite ; all of our effective masses must converge to their plateau values monotonically from above .
this ensures the validity of variational techniques which minimize the effective masses at small temporal separations .
such techniques are very effective in diminishing the excited - state contributions to the glueball correlation functions and are crucial for efficient extraction of ground - state masses . in this paper , we demonstrate the increased efficiency of glueball simulations using these actions on anisotropic lattices .
we present results for the masses of three of the lighter su(3 ) glueball states , the scalar ( @xmath3 ) , the tensor ( @xmath0 ) , and the pseudovector ( @xmath1 ) . the masses of the first excited states in the scalar and tensor channels were also examined .
ten simulations at lattice spacings ranging from 0.2 to 0.4 fm were performed , enabling reliable extrapolations to the continuum limit ( although the mass of the scalar glueball was somewhat problematic ) .
the results are compared to previous simulation data obtained using the wilson action and we find that more accurate determinations of the tensor and pseudovector glueball masses have been achieved .
a comparison of efficiencies is also made .
lastly , finite volume effects are shown to be small .
the new action used in our simulations is described in sec .
[ sec : action ] .
the details of the glueball simulations , including the construction of the glueball operators , the generation of the gauge - field configurations , and the analysis of the monte carlo data , are given in sec .
[ sec : simulation ] . the hadronic scale @xmath4 is used to relate our results at different values of the coupling @xmath5 and the aspect ratio @xmath6 .
the determination of this scale in terms of the lattice spacing using the static potential is outlined in sec .
[ sec : r0as ] .
[ sec : resdiscuss ] contains our results and discussion : the glueball mass measurements are presented in detail ; finite volume effects are studied ; extrapolations of the masses at finite spacing to the continuum limit are undertaken ; the conversion of our results into physical units is described ; and a comparison of efficiencies with wilson action simulations is made .
our conclusions are given in sec .
[ sec : conclusions ] , along with an outline of future work .
our glueball mass determinations rely on numerical simulations of glueballs on a euclidean space - time lattice with spatial and temporal spacings @xmath7 and @xmath8 , respectively . the improved gluonic action used in this study
is given by @xcite @xmath9 where @xmath10 , @xmath11 is the qcd coupling , @xmath12 and @xmath13 are mean link renormalization parameters , @xmath6 is the aspect ratio ( @xmath14 at tree level in perturbation theory ) , and @xmath15 , with @xmath16 denoting the path - ordered product of link variables along a closed contour @xmath17 on the lattice .
@xmath18 includes the sum over all spatial plaquettes on the lattice , @xmath19 indicates the temporal plaquettes , @xmath20 denotes the product of link variables about planar @xmath21 spatial rectangular loops , and @xmath22 refers to the short temporal rectangles ( one temporal link , two spatial ) .
explicitly , @xmath23,\\ \omega_{\rm tp } & = & \sum_x\sum_{i } \textstyle{\frac{1}{3}}{\rm re tr } \left[1-u_t(x)u_i(x\!+\!\hat t)u^\dagger_t(x\!+\!\hat\imath)u^\dagger_i(x ) \right],\\ \omega_{\rm sr } & = & \sum_x\sum_{i\neq j } \textstyle{\frac{1}{3}}{\rm re tr } \left[1-u_i(x)u_i(x\!+\!\hat\imath)u_j(x\!+\!2\hat\imath ) u^\dagger_i(x\!+\!\hat\imath\!+\!\hat\jmath ) u^\dagger_i(x\!+\!\hat\jmath ) u^\dagger_j(x)\right],\\ \omega_{\rm str } & = & \sum_x\sum_{i } \textstyle{\frac{1}{3}}{\rm re tr } \left[1-u_i(x)u_i(x\!+\!\hat\imath)u_t(x\!+\!2\hat\imath ) u^\dagger_i(x\!+\!\hat\imath\!+\!\hat t ) u^\dagger_i(x\!+\!\hat t ) u^\dagger_t(x)\right],\end{aligned}\ ] ] where @xmath24 labels the sites of the lattice , @xmath25 are spatial indices , and @xmath26 is the parallel transport matrix in the gluon field from site @xmath24 to @xmath27 .
this action , intended for use with @xmath28 , has @xmath29 discretization errors .
the @xmath30 errors can be removed by the addition of counterterms which couple next - nearest - neighbor time - slices , but this introduces spurious high - energy modes which can cause considerable problems for our glueball simulations .
these unphysical states appear in perturbation theory as additional poles in the gluon propagator .
their detrimental effects on the glueball correlation functions have been previously demonstrated@xcite .
although these spurious states do not affect the asymptotic behavior of the glueball correlators , they do appreciably change the correlators at short temporal separations and can seriously hinder attempts to reduce excited - state contamination to hasten the onset of asymptotic behavior .
since our glueball mass measurements rely heavily on the reduction of such excited - state contributions to the glueball correlation functions , the use of an action which is free of spurious lattice modes is crucial .
the action given in eq .
[ eq : impaction ] couples only link variables on neighboring time - slices , which ensures that all of our effective masses converge to their plateau values monotonically from above and so validates the variational techniques employed . it is now known that perturbation theory by itself does not reliably determine the couplings in an improved action in lattice gauge theory . hence , the interaction strengths in @xmath31 have been determined using a judicious combination of perturbation theory and mean field theory .
mean field theory is introduced by separately renormalizing the spatial and temporal link variables : @xmath32 and @xmath33 , where @xmath12 and @xmath13 denote the renormalization factors for the spatial and temporal links , respectively .
the mean - link parameters @xmath13 and @xmath12 are best determined by guessing input values for use in the action , measuring the mean links in landau gauge in a simulation , then readjusting the input values accordingly and tuning until the input values match the measured values .
the determination of these renormalization factors is described in more detail in refs .
however , when @xmath8 is significantly smaller than @xmath7 , we expect the mean temporal link @xmath13 to be very close to unity since @xmath34 in perturbation theory .
hence , to simplify matters , we set @xmath35 . we introduce further simplifications by using a convenient and gauge - invariant definition for @xmath12 in terms of the mean spatial plaquette given by @xmath36 , where @xmath37 denotes the spatial plaquette .
this eliminates the need for gauge fixing , yields values for @xmath12 which differ from those found using the landau gauge definition by only a few per cent , and significantly speeds up the tuning process . at finite coupling @xmath11
, the anisotropy @xmath38 is renormalized away from its input value @xmath6 .
measurements of this renormalization have been made using the static potential extracted from correlations along the different spatial and temporal axes of the lattice @xcite . without mean - link improvement
, this renormalization can be as large as @xmath39 .
when the action includes mean - link corrections , this renormalization is found to be small , typically a few per cent .
we used @xmath40 in all of our calculations , accepting the small radiative corrections to the anisotropy as finite lattice spacing errors , which vanish in the continuum limit .
glueballs may be labeled by their total ( integral ) spin @xmath41 and their symmetries under spatial inversion and charge conjugation . however , on a cubic lattice , glueballs are characterized by their transformation properties under the cubic point group , combined with parity and charge conjugation operations .
the cubic group , @xmath42 , has 24 elements that fall into five conjugacy classes , and thus , the dimensions of the irreducible representations ( irreps ) are 1 , 1 , 2 , 3 and 3 . these irreps are labeled @xmath43 , @xmath44 , @xmath45 , @xmath46 , and @xmath47 , respectively . including parity and charge conjugation symmetry operations ,
there are 20 irreps ( labeled by @xmath48 , where @xmath41 now denotes an irrep of @xmath42 ) . in this study ,
four of the irreps which generate light ( @xmath49 gev ) glueball states were simulated : the @xmath50 , @xmath51 , @xmath52 , and @xmath53 channels . of particular interest are the @xmath51 and @xmath52 irreps whose combined five rows correspond to the five polarization states of the tensor ( @xmath0 ) glueball which become degenerate as continuum rotational invariance is restored .
this then gives information on the magnitude of lattice artifacts at finite cut - off .
the mass of a glueball , @xmath54 , having a given @xmath48 can be extracted from the large-@xmath55 behavior of a correlation function @xmath56 , where @xmath57 denotes the lattice irrep corresponding to the @xmath48 of interest and @xmath58 is a gauge - invariant , translationally - invariant , vacuum - subtracted operator capable of creating a glueball out of the qcd vacuum @xmath59 .
as the temporal separation @xmath55 becomes large , this correlator tends to a single decaying exponential @xmath60 , where @xmath61 is the mass of the lowest - lying glueball which can be created by the operator @xmath62 . in order to extract @xmath61
, the correlator @xmath63 must be determined for @xmath55 sufficiently large that @xmath63 is well approximated by its asymptotic form .
however , the signal - to - noise ratio in any monte carlo determination of @xmath63 falls exponentially fast with respect to @xmath55 .
thus , it is crucial to use a glueball operator for which @xmath63 attains its asymptotic form as quickly as possible .
if @xmath64 denotes the glueball state of interest , this means that we must choose an operator for which the overlap @xmath65^{1/2}$ ] is as near to unity as possible .
for such an operator , the signal - to - noise ratio is also optimal@xcite . in order to construct such operators
, we exploited the smearing @xcite and variational techniques which have been used with success in earlier wilson action simulations . in each of the @xmath48 channels of interest ,
glueball operators were constructed on each time - slice in a sequence of three steps .
first , smeared links @xmath66 and fuzzy superlinks @xmath67 were formed .
secondly , a set of basic operators @xmath68 were constructed using linear combinations of gauge - invariant , path - ordered products of the @xmath66 and @xmath67 matrices about various closed spatial loops ; each such linear combination was designed to be invariant under spatial translations and to transform irreducibly under the symmetry operations of the cubic point group according to the irrep of interest .
lastly , the glueball operators @xmath69 were formed from linear combinations of the basic operators , @xmath70 , where the coefficients @xmath71 were determined using the variational method . each of these three steps is described below .
operators constructed out of smeared links and fuzzy superlinks have dramatically reduced mixings with the high frequency modes of the theory .
thus , the use of spatially - smoothed links is an important part of reducing excited - state contamination in the glueball correlation functions .
two smoothing procedures were used : a single - link procedure and a double - link procedure . in the single - link procedure , every spatial link @xmath72 on the lattice is replaced by itself plus a sum of its four neighboring ( spatial ) staples , projected back into su(3 ) : @xmath73 where @xmath74 denotes the projection @xcite into su(3 ) . here , we denote this mapping of the spatial link matrices into the smeared link variables by @xmath75 . in the double - link procedure , new superlinks @xmath76 of length @xmath77
are built using neighboring staples which connect sites separated by a distance twice that of the length of the source link variables : @xmath78 and we denote this mapping by @xmath79 .
both procedures can be applied recursively ; smeared links can be smeared again and fuzzy links of increasing length @xmath80 can be constructed .
a _ smoothing scheme _
, @xmath81 , is defined as a composition of single - link mappings and double - link mappings .
six different smoothing schemes were used .
the simplest scheme used was the composition of two single - link smearings : @xmath82 . to simplify notation
, we write this as @xmath83 .
we also used the compositions of four and six single - link mappings : @xmath84 and @xmath85 . in the other three smoothing schemes ,
the application of several single - link smearings , followed by one final iteration of double - link fuzzing was used : @xmath86 , @xmath87 , and @xmath88 .
only one iteration of the fuzzing procedure which results in links connecting sites separated by @xmath77 was found to be useful for the range of coarse @xmath7 values explored here .
for the finer lattices ( @xmath89 and @xmath90 ) , an extra four initial iterations of single - link smearing were used in all six smoothing schemes to enhance ground - state overlap . to simplify matters , the same values for the two parameters @xmath91 and @xmath92
were used in all smearing and fuzzing iterations .
these values were chosen to minimize excited - state contamination in the glueball correlation functions .
a crude optimization was done in a set of low statistics runs and the optimal values @xmath93 and @xmath94 were then used in all the glueball simulations .
the second step in the construction of our glueball operators was the formation of a set of basic operators @xmath68 using linear combinations of gauge - invariant , path - ordered products of the @xmath66 and @xmath67 matrices about various closed spatial loops .
combinations which were hermitian , invariant under spatial translations , and transformed irreducibly under the operations of the cubic point group according to the irrep of interest were constructed . for a more detailed exposition of this construction , see ref .
@xcite . in each channel ,
a large set of prototypes were programmed , and a short simulation was then performed to determine the coefficients of each operator in the variational ground state . in each channel , the four operators with the highest of these contributions were then chosen for use in the production runs .
the paths in this optimal set are illustrated in fig .
[ glueops ] . in the glueball simulations ,
these wilson loops were measured on the link variables from the six smoothing schemes , yielding a total of @xmath95 basic operators @xmath68 in each of the four channels . finally , @xmath69 was formed from a linear combination of the basic operators , @xmath96 .
the coefficients @xmath71 were determined using the variational method .
first , the @xmath97 correlation matrix was computed in the glueball simulations : @xmath98 where @xmath99 denotes a vacuum - subtracted operator @xmath100 .
note that @xmath101 is independent of @xmath55 .
the coefficients @xmath71 were then determined by minimizing the effective mass @xmath102,\ ] ] where the time separation for optimization was fixed in all cases to @xmath103 .
let @xmath104 denote a column vector whose elements are the optimal values of the coefficients @xmath71 .
then requiring @xmath105 for all @xmath106 yields an eigenvalue equation : @xmath107 the eigenvector @xmath108 corresponding to the largest eigenvalue @xmath109 then yields the coefficients @xmath110 for the operator @xmath111 which best overlaps the lowest - lying glueball @xmath54 in the channel of interest .
operators which overlap excited glueball states can also be constructed using the other eigenvectors of eq .
[ eq : variation ] .
in particular , the operator @xmath112 expected to best overlap the first - excited glueball state @xmath113 was obtained from the eigenvector corresponding to the second largest eigenvalue of eq .
[ eq : variation ] . the elements of the correlator matrix given in eq .
[ eq : gluecorr ] were estimated using the monte carlo method .
ten separate glueball simulations were performed on dec alpha - workstations .
configuration ensembles were generated using both cabibbo - marinari ( cm ) pseudo - heatbath and @xmath114 sub - group over - relaxation ( or ) methods .
link variables were updated in serial order on the lattice .
we define a _ compound sweep _ as one cm updating sweep followed by three or sweeps . in the glueball simulations
, three compound sweeps were performed between measurements , and the measurements were averaged into bins of 100 in order to reduce data storage requirements ( except for the @xmath115 , @xmath116 run in which 40 configurations were included in each bin ) . in all ten simulations , 100 bins were obtained .
our ensembles were tested for residual autocorrelations during the analysis phase by over - binning by factors of two and four ; in all cases , the statistical error estimates remained unchanged .
values for the mean link parameter @xmath12 were determined self - consistently as previously described .
this tuning procedure required a minimal amount of computational effort and provided thermalized configurations for later computations .
the improved action simulation parameters used are given in table [ tab : runparams ] .
for the data - fitting phase , the large @xmath117 correlator matrices in each channel were reduced using the coefficients @xmath108 and @xmath118 to smaller @xmath119 matrices @xmath120 for @xmath121 : @xmath122 the ground - state correlator @xmath123 was fit for @xmath124 using a single exponential @xmath125 where @xmath126 was the temporal extent of the periodic lattice , to obtain an estimate of the mass @xmath61 ( in terms of @xmath127 ) of the lowest - lying glueball in each channel . to determine the mass @xmath128 of the first - excited glueball and another estimate of @xmath61 , the @xmath119 correlator was also fit for @xmath124 using the form @xmath129 various fit regions @xmath130 to @xmath131
were used in order to check for consistency in the extracted values for the masses .
best fit values were obtained using the correlated @xmath132 method .
error estimates were calculated using a @xmath133point bootstrap procedure ; in all cases , error estimates were very close to being symmetric about the central best - fit values and were thus averaged to simplify presentation .
in order to convert the glueball masses as measured in our simulations into physical units , we must set the scale by determining the lattice spacing @xmath8 for each @xmath5 and @xmath6 we consider . to do this
, we must first choose one physical quantity to use as a reference .
this reference quantity must then be measured on the lattice in terms of @xmath8 . the experimentally - known value for the reference quantity is then used to extract the lattice spacing .
a quantity which can be easily and accurately determined both experimentally and in numerical simulations is an ideal choice for such a reference .
the mass of a low - lying particle is typically used for setting the scale . in our case
, however , there are no unambiguous experimental determinations of the glueball masses , so instead , we must look for another purely gluonic quantity .
the hadronic scale parameter @xmath4 defined in terms of the force between static quarks by @xmath134_{r = r_0}=1.65 $ ] , where @xmath135 is the static - quark potential , is an attractive possibility .
it can be measured very accurately on the lattice .
the advantages in using @xmath4 to set the scale have been enumerated in ref .
@xcite . from phenomenological potential models ,
one finds @xmath136 fm .
a disadvantage in using @xmath4 is that its physical value must be deduced indirectly from experiment , and there is some ambiguity in doing this , as will be discussed below .
however , in the absence of a better gluonic reference , we have chosen @xmath4 to set the scale . in this section ,
we outline the determination of @xmath4 in terms of @xmath7 . in order to determine @xmath4 in terms of the lattice spacing , we need accurate measurements of the static - quark potential .
we extracted @xmath135 for various spatial separations @xmath137 , both on and off the axes of the lattice , from the expectation values of wilson loops @xmath138 in the standard manner : @xmath139 + { \rm excited\ state\ contributions}.\ ] ] in the monte carlo evaluation of the wilson loops , measurements were taken after every four compound sweeps ( as defined in sec . [ sec : simulation ] ) .
the measurements of the wilson loops were done independently of the glueball mass studies using separate ensembles of configurations . to minimize contamination from excited states ,
the wilson loops were constructed from iteratively smeared spatial links .
the single - link smearing method described previously was used .
a given smearing scheme is specified not only by the parameter @xmath91 , but also by the total number of smearing iterations , denoted by @xmath140 .
two different choices of the smearing parameter were used in all cases : one smearing was chosen to work well for small @xmath141 , the other to work well for large @xmath142 .
separate measurements for each smearing were taken ; cross correlations were not determined .
the statistical noise in the evaluation of @xmath138 was reduced dramatically , especially for large temporal separations , by constructing the wilson loops , whenever possible , from thermally - averaged temporal links @xcite .
the thermal averaging was accomplished using the cabibbo - marinari pseudo - heatbath method ( 40 updates ) .
other relevant run parameters are given in table [ potrun ] .
the values of the potential @xmath135 were extracted from the wilson loop measurements by fitting @xmath138 to the exponential form @xmath143 $ ] in the range @xmath144 , for each @xmath137 .
the plateau region from @xmath130 to @xmath131 was chosen separately for each @xmath137 in order to minimize the uncertainty in the extracted values for @xmath135 while maintaining a good quality @xmath145 of fit .
best fit values were determined using the standard @xmath146-square test , taking into account temporal correlations among the @xmath138 .
the covariance matrix in @xmath132 was determined using the jack - knife procedure , and estimates for the uncertainties in the extracted values for @xmath135 were computed using the bootstrap method .
binning of the data was done as a crude check that our measurements were statistically independent .
the results of a typical fit are shown in fig .
[ veffmass ] , which is an effective mass plot for @xmath135 for @xmath147 .
the effective mass for @xmath135 is a function of @xmath55 , defined as @xmath148 $ ] , which tends to the true mass as @xmath55 becomes large .
once a suitable plateau region in the effective mass was established for each @xmath135 , the hadronic scale @xmath149 could be determined .
we found that the on - axis potential @xmath135 for the range of @xmath137 values studied here using coarse lattices fit a coulomb plus linear form @xmath150 very well ( with qualities of fit ranging from @xmath151 to @xmath152 ) .
we , therefore , used this form to interpolate @xmath135 and the force between static quarks .
simultaneous fits of the wilson loops for the on - axis potential to the form @xmath153 $ ] were done , taking into account all correlations among the @xmath154 for both different @xmath55 and @xmath142 .
different regions in @xmath55 were used for different @xmath142 values ; the plateau regions determined previously were used .
only the on - axis potentials were used ; this prevented the covariance matrix in the @xmath132 to be minimized from getting too large .
this covariance matrix was evaluated using the jack - knife method ; uncertainties in the fit parameters @xmath155 , @xmath156 , @xmath157 , and @xmath158 were obtained using the bootstrap method .
once we had an ensemble of bootstrap estimates for these fit parameters , the ratio @xmath149 and its bootstrap uncertainty were then determined using @xmath159 note that to compute @xmath149 , we need the ratio @xmath38 since our fits yielded estimates of @xmath160 only .
we used the input value @xmath6 since we know that its renormalization is small .
results for @xmath149 are given in table [ r0values ] .
using the results in table [ r0values ] , we can now express all energies measured in simulations in terms of @xmath4 .
for example , in fig .
[ potential ] , we show the potential , including off - axis inter - quark separations , expressed in terms of @xmath4 .
lattice spacing errors are seen to be small .
to allow clear resolution of the scaling properties of the low - lying glueball masses in the improved action , two sets of simulations were performed at two different anisotropies : six lattice spacings for an aspect ratio @xmath116 and four spacings for @xmath161 were studied .
the input parameters used in these simulations are given in table [ tab : runparams ] .
the results of fitting the variationally - optimized correlators @xmath63 to the functions given in eqs .
[ eq : ggcorrelator ] and [ eq:2by2correlator ] are summarized in tables [ tab : emone]-[tab : emten ] .
effective mass plots for the two smallest-@xmath7 simulations are presented in figs .
[ fig : b26-r3-effmass - a1pp]@xmath162[fig : b26-r3-effmass - t2pp ] and figs .
[ fig : b24-r5-effmass - a1pp]@xmath162[fig : b24-r5-effmass - t2pp ] for @xmath116 and @xmath161 , respectively . for each channel in each of the ten simulations , it was possible to find a fit region @xmath163 in which the correlation function was well described by its asymptotic form as indicated by the quality of fit .
in other words , convincing plateaux were observed in all effective masses . the most impressive plateau , observed in the @xmath50 channel for @xmath164 and @xmath161 , spanned ten time - slices . in most cases ,
the onset of the plateau occurred when the source and sink operators were separated by only one time - step .
the overlaps with the lowest - lying states were also found to be extremely good , better than @xmath165 in most cases and often consistent with unity .
this clearly demonstrates the effectiveness of the link - smearing and variational techniques in diminishing excited - state contamination .
fits using @xmath166 or @xmath167 were also done to check for agreement with the @xmath168 results . at time separations
for which the ground state could be reliably observed , the off - diagonal elements of the reduced correlation matrices @xmath120 were found to be consistent with zero within statistical uncertainty .
this suggests that the link - smearing , variational method also gives an excellent construction of the first - excited state in each channel .
our best estimates for the glueball masses in terms of @xmath127 are indicated in boldface in each of the tables [ tab : emone]-[tab : emten ] .
these estimates are summarized in table [ tab : final ] .
masses for the first - excited states are also indicated in the @xmath169 fits listed in the tables [ tab : emone]-[tab : emten ] . in this work ,
we were concerned with the magnitude of discretization errors in the glueball mass determinations from coarse lattice simulations using an improved action . in order to evaluate these errors
, we had to eliminate uncertainties from all other sources .
the increased efficiency of simulations on coarse , anisotropic lattices allowed us to reduce statistical errors to the acceptable level of about @xmath170 .
the only remaining source of uncertainty we had to address was finite volume .
the masses of particles confined in a small box with periodic boundary conditions can differ appreciably from their infinite volume values ; finite volume effects can also induce a splitting in the masses of the @xmath45 and @xmath47 tensor polarizations .
finite volume effects on the scalar glueball mass have been analyzed before@xcite , but the effects on the tensor and the pseudovector are less well known . in order to ascertain the effects on our glueball masses of simulating in a finite volume , four extra simulations were performed for @xmath164 , @xmath116 using lattices of spatial extent @xmath171 and @xmath172 .
the temporal extent was held fixed at 24 grid points . for each of these volumes ,
the mean - field renormalization parameter @xmath12 was recalculated .
the @xmath173 lattice was the only simulation that required any change in this parameter , and in this case , the effect was small ; @xmath12 increased by only @xmath174 .
the results from the @xmath175 and @xmath176 runs for the glueball masses in terms of @xmath127 are given in table [ tab : finvolvals ] .
note that the results from the @xmath177 lattice differ very little from those from the @xmath178 lattice , suggesting that our lattice volumes are sufficiently large to ensure that finite volume errors are negligible . for the @xmath173 lattice
, no plateaux in the effective masses for the @xmath50 , @xmath51 and @xmath52 channels were observed ; the mass in the @xmath53 channel was found to be @xmath179 .
the operators used in these runs were the same as those constructed for the large volume runs and thus , were not optimized to give large overlaps with the light torelon states present in small volumes .
it is likely that this effect was responsible for the poor overlap of our operators with the scalar and tensor states on the @xmath173 lattice .
the properties , such as the mass , of a glueball confined in a small box with periodic boundary conditions differ from those in an infinite volume .
the modification of the mass of a particle due to finite volume effects has been estimated in ref .
@xcite : @xmath180 , \label{eq : luscherfinvol}\ ] ] where @xmath181 is the dimensionless length scale @xmath182 , @xmath183 is the infinite - volume mass of the scalar glueball , and @xmath184 is related to the strength of an effective triple scalar glueball interaction vertex . the mass shift given in eq .
[ eq : luscherfinvol ] is valid for sufficiently large @xmath181 and arises from the exchange of scalar glueballs across the periodic boundaries of the lattice .
finite volume errors in our glueball masses measured on an @xmath178 lattice at @xmath164 , @xmath116 ( where @xmath185 fm , similar to the volumes used in the other nine simulations ) can be estimated by fitting the form given in eq . [
eq : luscherfinvol ] to the masses in table [ tab : finvolvals ] .
let @xmath186 and @xmath187 , then @xmath188 . the @xmath50 fit was done first using the function @xmath189 , where @xmath190 and @xmath191 were the fitting parameters .
the best fit value for @xmath190 was then used in the fits to the results for the other irreps ; to simplify matters , the uncertainty in @xmath190 was neglected in these fits .
the @xmath53 fit also included the energy estimate extracted from the @xmath192 simulation .
the results of these fits are summarized in table [ tab : finvolfits ] ; the estimates of the finite volume errors are listed in the final column of this table and are given by @xmath193 using eq .
[ eq : luscherfinvol ] . in all cases
, these errors were insignificant compared to statistical errors ; this means , for example , that any differences between the large - volume masses in the @xmath47 and @xmath45 channels must be due purely to discretization errors .
it is interesting to note that our estimate of @xmath191 agrees well with the value @xmath194 found in ref .
@xcite .
the glueball mass estimates in terms of @xmath127 were combined with the determinations of the hadronic scale @xmath149 .
the results are shown in figs .
[ fig : scaling ] , [ fig : tensor ] , and [ fig : scalarscaling ] . in these figures , the dimensionless product of @xmath4 and the glueball
mass estimates are shown as functions of @xmath195 .
solid symbols indicate results from the @xmath116 simulations , while open symbols are used for the results from the @xmath161 runs . in fig .
[ fig : scaling ] , the lowest - lying masses in each of the channels @xmath50 , @xmath51 , @xmath52 , and @xmath53 are compared with results from small-@xmath7 wilson action simulations . the lowest - lying and first - excited masses in the @xmath51 and @xmath52 channels are shown in fig .
[ fig : tensor ] , and the ground - state and first - excited state in the @xmath50 channel are depicted in fig .
[ fig : scalarscaling ] . to extract physical predictions ( for the pure - gauge theory )
, the curves in these plots must be extrapolated to the continuum limit @xmath196 .
discretization errors are given by the deviations of the finite-@xmath7 results from these limiting values .
the lowest - lying states in the @xmath51 and @xmath52 channels correspond to the five polarizations of the tensor @xmath0 glueball in the continuum .
differences between the @xmath51 and @xmath52 masses are a measure of violations of rotational symmetry due to finite spacing artifacts . in fig .
[ fig : scaling ] , such violations are seen to be small for our less coarse lattices and become appreciable as the spacing gets very large .
discretization errors in the @xmath52 exceed those of the @xmath51 ; on our coarsest lattices , finite spacing errors are only a few per cent for the @xmath51 channel , but about @xmath197 in the @xmath52 channel . in the @xmath51 channel ,
the @xmath116 results differ very little from those using the higher aspect ratio @xmath161 , suggesting that the @xmath30 errors are negligible . however , small differences between the results from the two anisotropies are visible in the @xmath52 channel .
one expects that @xmath30 errors will decrease as @xmath6 is increased .
since the @xmath52 discretization errors are slightly larger for the @xmath161 runs , @xmath30 errors can account for this difference only if such errors offset the @xmath198 errors .
the leading discretization errors in the tensor glueball masses are expected to be @xmath199 .
however , we have already argued that the results in fig . [
fig : scaling ] imply that the @xmath30 errors are negligible . since the action included mean - field correction factors , we also expected that @xmath198 errors would dominate over @xmath200 errors and in our continuum limit extrapolations , we assumed that this was true unless the fit provided compelling evidence to the contrary .
although we expected the leading discretization errors to be @xmath198 , the following three functions were used in our continuum limit extrapolations : @xmath201 where @xmath202 , @xmath203 and @xmath204 are best - fit parameters .
the results of these fits are given in table [ tab : extrap-3 ] for the @xmath116 data and table [ tab : extrap-5 ] for the @xmath161 simulations .
comparing the values of @xmath205 , one sees that the fitting function @xmath206 was preferred for both the @xmath116 and @xmath161 results , although only marginally so for @xmath116 .
given this fact and our expectation that @xmath206 should best describe the leading discretization effects , we took @xmath204 from the @xmath206 fits as our continuum limit estimates ( indicated in boldface in tables [ tab : extrap-3 ] and [ tab : extrap-5 ] ) .
these four estimates are in very good agreement not only with one another , but also with the wilson action estimates .
these fits using @xmath206 are shown in fig .
[ fig : scaling ] . for our final estimate of the tensor glueball mass
, we performed a simultaneous fit with the four data sets ( two irreps and two anisotropies ) using four separate @xmath206 functions but constraining the intercept parameter @xmath204 to be the same for all four fitting functions .
this yielded @xmath207 with @xmath208 , in agreement with the wilson action estimate @xmath209 , obtained by fitting all of the wilson action measurements shown in fig .
[ fig : scaling ] to @xmath210 .
we also examined the discretization errors in the masses of the first - excited glueball states in the @xmath51 and @xmath52 channels .
these are shown in fig .
[ fig : tensor ] .
there are several reasons for interpreting these data as different polarizations of a spin - two excited state : the two irreps extrapolate to the same continuum limit value ; if the @xmath52 state were spin three , then there would be a degeneracy with the @xmath211 and @xmath212 channels and this was not observed ( these results will be presented elsewhere ) ; if they were polarizations of a spin - four state , then again , a similar level must also be found in the @xmath211 channel .
the degeneracy between the two irreps and the weak finite - volume dependence of their energies also makes an interpretation of this state as a torelon pair or a two - scalar - glueball scattering state unlikely , although the mass of this level lies close to twice the mass of the scalar glueball .
continuum limit extrapolations were performed using the three functions of eqs .
[ eq : fitconst]-[eq : fita4 ] ; measurements from the two largest @xmath7 spacings for @xmath116 and the single largest spacing for @xmath161 were excluded from these fits .
the results are given in tables [ tab : extrap-3 ] and [ tab : extrap-5 ] .
again , we expected @xmath206 to provide the most reliable extrapolations to the @xmath213 limit ; this was confirmed by the fact that @xmath206 yielded @xmath51 and @xmath52 continuum limits in best agreement with each other .
differences found between the @xmath116 and @xmath161 extrapolations were not statistically significant .
the fits using @xmath206 are shown in fig .
[ fig : tensor ] .
our final determination of the mass of the excited tensor glueball , obtained from a constrained set of four @xmath206 , similar to the ground - state extraction , was @xmath214 , where @xmath215 .
this mass has not been reliably determined in any previous simulations .
finite spacing errors in the mass of the @xmath53 pseudovector glueball were also studied .
these were found to be small and are shown in fig . [
fig : scaling ] .
the results from the different anisotropies are in good agreement .
extrapolations to the @xmath213 limit were done using the three functions @xmath210 , @xmath216 , and @xmath206 ; the results of these fits are summarized in tables [ tab : extrap-3 ] and [ tab : extrap-5 ] .
the continuum limits obtained from fits to the @xmath116 and @xmath161 data agreed only for the constant fit form @xmath210 .
the fits to @xmath216 and @xmath206 yielded slope parameters ( @xmath202 and @xmath203 ) with large relative errors and opposite signs for the different anisotropies .
hence , the function @xmath210 was used to extrapolate to the continuum limit . due to the very good agreement between the @xmath116 and @xmath161 results ,
all ten data points were used in our extrapolation fit .
our estimate from this fit ( shown in fig .
[ fig : scaling ] ) was @xmath217 , where @xmath218 , in agreement with the extrapolation @xmath219 using @xmath210 of the wilson action results shown in fig . [
fig : scaling ] .
in contrast to the tensor and pseudovector , the scalar glueball mass showed significant finite spacing errors ( see fig . [
fig : scaling ] ) , even for our less coarse lattices . as @xmath7 was increased ,
the scalar mass first decreased , reached a minimum near @xmath220 , then gradually increased . near the minimum , the mass was about @xmath221 lower than estimates of the continuum limit from small-@xmath7 wilson action computations ; a @xmath222 discretization error was observed in the result from our smallest @xmath7 simulation .
although the magnitudes of these errors were significant , they were smaller than those obtained using the wilson action by a factor of two . in order to extrapolate to the continuum limit
, an appropriate fitting function was needed .
the leading discretization errors were expected to be @xmath199 .
however , there were no distinguishable differences between the @xmath116 and @xmath161 results , suggesting that the @xmath30 errors were negligible , leaving us to consider @xmath223 effects . by inspection
, one sees that the fitting form @xmath206 , which neglects one - loop @xmath200 effects , can not describe the data , in contrast to the data for the tensor and pseudovector glueballs . as @xmath213
, we expect the coupling @xmath224 to vanish as @xmath225 , where @xmath226 is an appropriate scale parameter .
hence , we were led to consider the following four - parameter fitting function : @xmath227 } + c_4\ ( a_s / r_0)^4 .
\label{eq : scalarfit2}\ ] ] however , it was not known how reliably the leading perturbative behavior of @xmath224 would describe the true cut - off dependence of the coupling over the large range of spacings considered .
taking this into account and inspecting the behavior of the actual data , we decided to also consider the following simpler quadratic form : @xmath228 both of these functions were fit to the mass measurements from all ten simulations ; the results of these fits are summarized in table [ tab : scalarextrap ] .
the function @xmath229 yielded a slightly better fit and a continuum limit for the scalar glueball mass of @xmath230 .
this fit is shown in fig .
[ fig : scaling ] .
an extrapolation of existing wilson action data using @xmath216 yielded @xmath231 . given the quality of the scalar glueball mass estimates using the wilson action
, this slight discrepancy raises doubts concerning the reliability of the extrapolation using @xmath229 ; mass estimates using the improved action for a few values of @xmath7 smaller than those considered here would be needed to resolve this discrepancy .
one explanation for the @xmath222 discretization errors in the scalar glueball mass is that the scalar glueball is extremely small . however , there is evidence@xcite that the presence of a critical endpoint of a line of phase transitions ( not corresponding to any physical transition found in qcd ) in the fundamental - adjoint coupling plane is responsible for lowering the scalar glueball mass near the crossover region in the wilson action .
it is possible that the scalar glueball mass in the improved action used here may be similarly influenced .
if so , the fact that this effect appears to be less pronounced for this action suggests the possible existence of other perturbatively - improved actions in which the scalar glueball mass is even less affected by scaling violations .
we are currently searching for such actions .
discretization errors in the mass of the first - excited state in the @xmath50 channel were also found to be significant , as shown in fig .
[ fig : scalarscaling ] .
the mass of this state is nearly twice that of the lowest - lying scalar glueball , suggesting that this state may simply be two glueballs .
given the significant discretization errors in the single glueball scaling data , one would expect similar systematic errors in the two glueball state .
the absence of any level of similar mass in all other channels justifies the spin - zero interpretation of this state .
considering the difficulties encountered in extrapolating the lowest - lying scalar to the continuum limit , we made no serious attempt to determine the continuum limit of this first - excited state . however , the result of a fit using @xmath232 is included in table [ tab : scalarextrap ] .
the possible interpretation of this level as a two glueball system might be strengthened by a more precise finite - volume study . in order to convert our glueball mass computations into physical units , we must specify the value of the hadronic scale .
the hadronic scale @xmath4 has a precise definition in terms of the static - quark potential .
however , the static - quark potential can not be directly measured in an experiment ; it must be deduced indirectly from other observables .
we decided to use a variety of different physical quantities to deduce @xmath4 .
in table [ hadronictable ] , estimates of @xmath233 using the results from various quenched lattice simulations are shown .
for each computation , the quantity used to set the lattice spacing , such as the mass of the @xmath234 or the @xmath235 splitting in heavy quarkonia , is indicated .
the determination of @xmath233 from @xmath236 was accomplished using values of @xmath237 given in ref .
@xcite for the wilson gluonic action at various values of @xmath5 , interpolating where necessary .
note that due to quenching effects , @xmath4 varies with the quantity used to set the scale .
the entries in the last column of table [ hadronictable ] can not be considered as different measurements of a single quantity and thus , strictly speaking , their weighted average has no statistical meaning .
the last column of the table is meant to illustrate the range of values one obtains for @xmath233 when using various scale setting quantities .
we expect that the value of @xmath233 appropriate for the low - lying glueballs should lie somewhere within this range .
the simple average @xmath238 mev of the determinations in table [ hadronictable ] was taken as our estimate of the hadronic scale . for our final continuum mass estimates of the tensor glueball
, we found @xmath239 mev ( where the first error is statistical and the second is from uncertainties in the determination of @xmath4 ) .
it is interesting to note that our mass estimate lies within @xmath240 of the mass of the @xmath241 resonance @xcite , reported to have quantum numbers @xmath242 . in order to make a direct comparison with experiment ,
however , corrections to our result from light quark effects and mixings with nearby conventional mesons must be taken into account .
our estimate of the mass of the first - excited glueball in the tensor channel was @xmath243 mev ; for the pseudovector state , we found a mass of @xmath244 mev .
our estimate from the fit using @xmath229 for the mass of the scalar glueball was @xmath245 mev ; however , we regard the continuum limit extrapolation for this state as being less reliable than those for the other glueballs .
a quantitative comparison of efficiencies is difficult to make .
there are many factors which affect the overall efficiency of a monte carlo simulation . certainly , the number of link updates is an important factor .
the speed of an update is , of course , platform- , action- , and algorithm - dependent . on the dec alpha - workstations we used , a cm update using an improved action required twice as much time as for the wilson action
; the improved - action or updating time was three times longer .
critical slowing down and thermalization are also contributing factors , but a crucial issue is the reduction of excited - state contaminations in the glueball correlators . in our coarse lattice simulations , we found that current methods for constructing good glueball operators were very effective in hastening the onset of plateaux in the effective masses .
given these difficulties in assessing the efficiency of a glueball simulation , we decided to make our comparisons based simply on the number of link updates @xmath246 and the fractional error @xmath247 attained in the final mass estimates .
since the error in a monte carlo estimate decreases with the number of measurements @xmath248 as @xmath249 , we expect that the reciprocal product of the number of link updates and the square of the fractional error is approximately proportional to the efficiency of a simulation ; we denote this quantity by @xmath250 .
an interesting comparison to make is between simulations at a small lattice spacing , such as @xmath251 fm , using the wilson action and improved - action simulations at a spacing @xmath252 fm .
such a comparison is relevant because the discretization errors in these cases are of comparable magnitude , excepting the scalar glueball mass . for the @xmath253 wilson action run in ref .
@xcite by the gf11 collaboration using gauge invariant glueball operators , a total of @xmath254 link updates were performed , an error of @xmath255 was achieved in the scalar glueball mass , and a fractional error of @xmath256 was obtained for the tensor mass . using @xmath257 fm and @xmath258 ,
the lattice volume in this simulation was @xmath259 .
for the same value of @xmath5 in ref .
@xcite by the ukqcd collaboration , @xmath260 link updates were made and fractional errors @xmath261 , @xmath262 , and @xmath263 were obtained for the scalar , tensor , and pseudovector masses , respectively .
the lattice volume was @xmath264 for this simulation . in our @xmath115 ,
@xmath116 run , @xmath265 link updates were performed and @xmath266 , @xmath267 , and @xmath267 errors in the scalar , tensor , and pseudovector masses were achieved .
our lattice volume was @xmath268 since @xmath269 . for our @xmath164 ,
@xmath161 run , the number of link updates was @xmath270 and the errors obtained in the scalar , tensor , and pseudovector masses were @xmath267 , @xmath271 , and @xmath272 , respectively . using @xmath273 , our lattice volume for this run was @xmath274 .
thus , the ratios of the @xmath275 values for our @xmath115 and @xmath164 simulations to those of the gf11 run were @xmath276 and @xmath277 , respectively , for the scalar glueball mass , and @xmath278 and @xmath279 for the tensor glueball mass . the ratios of the @xmath115 and @xmath164 @xmath275-values to those of the ukqcd run were @xmath280 and @xmath281 , respectively , for the scalar mass , @xmath282 and @xmath283 for the tensor , and @xmath284 and @xmath285 for the pseudovector . considering our ten simulations together , a total of @xmath286 link updates were performed .
for the wilson action simulations of refs .
@xcite , an estimated @xmath287 and @xmath288 link updates were required , respectively , to generate continuum limit results whose statistical uncertainties were larger ( for the tensor and pseudovector states ) than those quoted here : the statistical error on our estimate for @xmath289 was about five times smaller than that from the extrapolation of the wilson action results , and for @xmath290 , the uncertainty was twenty times smaller , implying that about @xmath291 times greater statistics would be required for similar accuracy .
thus , in total , the anisotropic lattice simulations were certainly more than 1000 times more efficient .
the above discussion illustrates the computational advantages of extracting non - scalar glueball masses from simulations on coarse , anisotropic lattices ( @xmath252 fm ) using an improved action instead of lattices for which @xmath251 fm .
the excellent overlaps achieved from our variational calculations demonstrate another advantage of simulating on coarse spatial lattices .
the glueball wavefunctions extend over only a few points of the lattice when @xmath292 fm .
thus , variational calculations using a feasible number of basis functions ( a dozen or so ) can yield very good approximations to the glueball wavefunctions
. this will be especially important for future decay constant calculations and determinations of mixings with non - glueball states .
these advantages have already enabled us to study the more massive glueball states which have yet to be simulated reliably using the wilson action ; these results will be reported elsewhere .
these advantages are less clear for the scalar glueball mass due to the presence of @xmath222 discretization errors at @xmath252 fm . using the action of eq .
[ eq : impaction ] , simulations at one or more lattice spacings smaller than 0.2 fm would be needed to firmly establish the continuum limit .
a more attractive approach would be to use an action for which discretization errors in the scalar glueball mass at @xmath252 fm are negligibly small .
the search for such an action is currently underway .
we have demonstrated the advantages of using anisotropic lattices and an improved gluonic action for simulating glueballs .
ten simulations at lattice spacings ranging from 0.2 to 0.4 fm were performed using dec alpha - workstations , and the results were extrapolated to the continuum limit .
results for the masses of the scalar ( @xmath3 ) , the tensor ( @xmath0 ) , and the pseudovector ( @xmath1 ) glueballs in @xmath293 pure - gauge theory were presented in terms of the hadronic scale @xmath233 .
the continuum limits for the tensor and pseudovector glueball masses were obtained with uncertainties of less than @xmath170 , significantly improving upon previous estimates from wilson action simulations carried out with the aid of super - computer resources .
extrapolation of the scalar glueball mass to the continuum limit was hampered by uncertainties in choosing the fitting function and discretization errors which were @xmath294 even for our smallest lattice spacings ; although uncomfortably large , these finite - spacing errors were half as large as those obtained using the wilson action . finite volume errors in our results were shown to be negligible .
the masses of the first excited states in the scalar and tensor channels were also examined .
our results show that spatially - coarse , anisotropic lattice simulations are an effective means of studying gluonic systems .
the techniques exploited here are sufficiently powerful to overcome the difficulties which plague monte carlo calculations involving gluonic excitations .
these methods should be useful for studying the spectrum of heavier glueball states .
data for the masses of all twenty lattice irreps of the cubic group ( including parity and charge conjugation ) are currently being accumulated in order to survey the spectrum of @xmath293 glueball states below 4 gev comprehensively .
we shall report on these results in the near future .
we also plan to use the techniques outlined in this paper to determine various glueball matrix elements and decay strengths , to investigate the mixings of glueballs with conventional hadronic states , and to study mesonic states containing excited glue ( the so - called hybrid mesons ) .
the size of discretization errors in the scalar glueball mass was the only disappointing aspect of this work ; we are currently investigating a new class of lattice actions with the hope of reducing these lattice artifacts for @xmath252 fm to the level of a few per cent .
we would like to thank peter lepage , julius kuti , chris michael , terrence draper , and keh - fei liu for helpful discussions .
this work was supported by the u.s .
doe , grant no .
de - fg03 - 90er40546 .
mp is grateful to the university of kentucky center for computational sciences for financial support . c. morningstar and m. peardon , nucl .
b ( proc . suppl . ) * 47 * , 258 ( 1996 ) . c. morningstar , nucl .
. suppl . ) * 53 * , 914 ( 1996 ) . c. morningstar and m. peardon , nucl .
b ( proc . suppl . )
* 53 * , 917 ( 1997 )
. m. alford _
et al . , _ in preparation
. f. brandstaeter , a. kronfeld , and g. schierholz , nucl . phys . * b345 * , 709 ( 1990 ) . m. albanese _ et al . , _ phys .
b * 192 * , 163 ( 1987 ) .
m. teper , phys .
b * 183 * , 345 ( 1986 ) .
y. liang _ et al .
_ , phys .
b * 307 * , 375 ( 1993 ) .
b. berg and a. billoire , nucl .
* b221 * , 109 ( 1983 ) .
r. sommer , nucl .
phys . * b411 * , 839 ( 1994 ) . g. parisi _ et al .
_ , phys .
* 128b * , 418 ( 1983 ) . g. schierholz , nucl .
b ( proc . suppl . ) * 9 * , 244 ( 1989 )
. m. lscher , commun . math
. phys . * 104 * , 177 ( 1986 ) .
u. heller , phys .
b * 362 * , 123 ( 1995 ) .
a. patel _ et al .
, _ phys .
* 57 * , 1288 ( 1986 ) .
bali and k.s .
schilling , int .
j. mod .
* c4 * , 1167 ( 1993 ) and private communication ; h. wittig , nucl .
b ( proc . suppl . )
* 42 * , 288 ( 1995 ) .
barnett _ et al .
_ , phys .
d * 54 * , 1 ( 1996 ) .
et al . , _ phys .
lett . * 56 * , 107 ( 1986 ) .
h. chen _ et al .
_ , nucl .
b ( proc . suppl . ) * 34 * , 357 ( 1994 ) .
bali _ et al .
_ , phys .
b * 309 * , 378 ( 1993 ) .
p. de forcrand and k - f .
liu , phys .
lett . * 69 * , 245 ( 1992 ) . c. michael and m. teper , nucl .
* b314 * , 347 ( 1989 ) .
forcrand _ et al .
_ , phys .
* 160b * , 137 ( 1985 ) . c. davies _ et al . , _ phys .
d * 50 * , 6963 ( 1994 ) . c. davies _ et al .
, _ phys . rev .
d * 52 * , 6519 ( 1995 ) .
t. bhattacharya _ et al .
, _ phys .
d * 53 * , 6486 ( 1996 ) . f. butler _ et al .
, _ nucl .
phys . * b430 * , 179 ( 1994 ) . s. aoki _ et al .
, _ nucl .
b ( proc . suppl . )
* 53 * , 355 ( 1997 )
. s. aoki _ et al .
, _ nucl .
b ( proc .
suppl . ) * 53 * , 349 ( 1997 ) .
el - khadra , nucl .
b ( proc . suppl . )
* 34 * , 141 ( 1994 ) . c. bernard , j. labrenz , and a. soni , phys .
d * 49 * , 2536 ( 1994 ) .
allton _ et al . , _ phys .
b * 284 * , 377 ( 1992 ) . ( 30,20 ) ( 10.00 , 7.00 ) ( 2 , 1 ) 8 ( 18.00 , 11.00 ) ( 1 , 0)10 ( 20.00 , 7.00 ) ( 2 , 1 ) 8 ( 14.00 , 7.00 ) ( 1 , 0 ) 6 ( 2.00 , 3.00 ) ( 2 , 1 ) 8 ( 2.00 , 3.00 ) ( 0 , 1)10 ( 2.00 , 13.00 ) ( 2 , 1 ) 8 ( 10.00 , 11.00 ) ( 0 , 1)6 ( 10.00 , 11.00 ) ( 1,-1)4 ( 30,20 ) ( 10.00 , 7.00 ) ( 2 , 1 ) 8 ( 18.00 , 11.00 ) ( 1 , 0)10 ( 20.00 , 7.00 ) ( 2 , 1 ) 8 ( 14.00 , 7.00 ) ( 1 , 0 ) 6 ( 2.00 , 3.00 ) ( 2 , 1 ) 8 ( 2.00 , 3.00 ) ( 0 , 1)10 ( 2.00 , 13.00 ) ( 2 , 1 ) 8 ( 10.00 , 11.00 ) ( 0 , 1)6 ( 10.00 , 11.00 ) ( 1,-1)4 ( 30,20 ) ( 14.00 , 7.00 ) ( 1 , 0)10 ( 16.00 , 3.00 ) ( 2 , 1 ) 8 ( 6.00 , 3.00 ) ( 1 , 0)10 ( 6.00 , 3.00 ) ( 0 , 1)9 ( 5.00 , 12.50 ) ( 2 , 1)9 ( 14.00 , 7.00 ) ( 0 , 1)10 ( 13.00 , 6.00 ) ( 1 , 0)11 ( 16.00 , 2.00 ) ( 2 , 1 ) 8 ( 5.00 , 2.00 ) ( 1 , 0)11 ( 5.00 , 2.00 ) ( 0 , 1)10.5 ( 6.00 , 12.00 ) ( 2 , 1 ) 7 ( 13.00 , 6.00 ) ( 0 , 1)9.5 ( 30,20 ) ( 10.00 , 7.00 ) ( 2 , 1 ) 8 ( 18.00 , 11.00 ) ( 1 , 0)10 ( 20.00 , 7.00 ) ( 2 , 1 ) 8 ( 14.00 , 7.00 ) ( 1 , 0 ) 6 ( 2.00 , 3.00 ) ( 2 , 1 ) 8 ( 2.00 , 3.00 ) ( 0 , 1)10 ( 2.00 , 13.00 ) ( 2 , 1 ) 8 ( 10.00 , 11.00 ) ( 0 , 1)6 ( 10.00 , 11.00 ) ( 1,-1)4 ( 30,20 ) ( 14.00 , 9.00 ) ( 2 , 1 ) 4 ( 18.00 , 11.00 ) ( 1 , 0)10 ( 20.00 , 7.00 ) ( 2 , 1 ) 8 ( 10.00 , 7.00 ) ( 1 , 0)10 ( 2.00 , 3.00 ) ( 2 , 1 ) 8 ( 2.00 , 3.00 ) ( 0 , 1)10 ( 2.00 , 13.00 ) ( 2 , 1 ) 8 ( 10.00 , 11.00 ) ( 0 , 1)6 ( 14.00 , 9.00)(-2 , 1)4 | monte carlo results for the low - lying glueball spectrum using an improved , anisotropic action are presented .
ten simulations at lattice spacings ranging from 0.2 to 0.4 fm and two different anisotropies have been performed in order demonstrate the advantages of using coarse , anisotropic lattices to calculate glueball masses .
our determinations of the tensor ( @xmath0 ) and pseudovector ( @xmath1 ) glueball masses are more accurate than previous wilson action calculations . epsf |
in recent years there has been a growing interest in flux instabilities and catastrophic flux penetration events in superconductors @xcite . in particular
, one finds that in many superconductor films flux may enter abruptly in the form of magnetic dendrites .
while the phenomenon has been observed in various materials such as nb , nb@xmath1sn , yni@xmath0b@xmath0c , nbn , yba@xmath0cu@xmath1o@xmath2 ( induced by laser pulses ) , and patterned pb films @xcite , it has been most widely studied @xcite in mgb@xmath0 .
this interest stems in part from the fact that dendrites are omnipresent in mgb@xmath0 films below 10 k and do not need triggering or patterning to occur , and in part from their debilitating effect on the critical current of this material @xcite , otherwise very promising for many applications @xcite .
it is now generally believed that the dendrites occur as a result of a thermo - magnetic instability , whereby i ) motion of vortices releases energy and leads to local heating , and ii ) increased temperature leads to a local decrease of the pinning force , enabling enhanced vortex motion .
if the released heat is not carried away fast enough , this constitutes a feedback mechanism which induces a thermo - magnetic runaway .
recent experimental work @xcite on mgb@xmath0 has indeed suggested that such a thermo - magnetic mechanism is a feasible explanation for dendritic instabilities .
lending further support to this picture , it has been shown theoretically @xcite that the instability will develop into a highly non - uniform pattern if the thermal diffusivity in the superconductor is much smaller than the magnetic diffusivity .
finally , the predictions of these models for the threshold instability field were recently @xcite found to quantitatively agree with experiments on mgb@xmath0 and nb films .
but even as the fundamental mechanism seems to be understood , there are many open questions regarding details in dendritic nucleation and evolution .
one of them is the interplay between dendrites of flux and anti - flux . while it was previously shown that coexisting flux and anti - flux helps the nucleation of dendritic avalanches , @xcite it has never been observed how dendrites of opposite polarity interact when they both penetrate a virgin sample .
one geometry where such a situation may be realised is that of a planar ring .
@xcite in zero - field cooled ( zfc ) , circular superconductor rings exposed to perpendicular fields , shielding currents flow around the ring in the same direction everywhere @xcite .
these currents lead to an enhanced field at the outer edge , and a field of opposite polarity at the inner edge . in this paper we present results of a magneto - optical ( mo ) investigation of thin film mgb@xmath0 rings showing a rich variety of dendrite behaviour .
the paper is structured as follows : the experimental details are described in section [ exp ] .
section [ results ] presents our results , with a discussion of our observations given in section [ discussion ] .
mgb@xmath0 films were grown by pulsed laser deposition on sapphire substrates .
details on sample fabrication can be found elsewhere @xcite .
using photolithography , two films , 500 nm thick , were patterned into circular rings of different size .
the lateral dimensions of the larger sample were @xmath3 mm and @xmath4 mm , and the smaller sample @xmath5 mm and @xmath6 mm . for observations we used a standard magneto - optical ( mo ) imaging set - up with a leica polarisation microscope @xcite , a liquid helium flow cryostat from oxford instruments , a 12-bit retiga - exi fast digital camera from qimaging , and a computer running labview to aqcuire data and control the applied field .
the magnetic sensor was a mirror coated @xmath7 m thick faraday rotating ferrite garnet film placed directly on top of the sample . to avoid suppression of dendrites by the metallic mirror layer @xcite on the indicator films , we used small , insulating ugelstad spheres ( monodisperse with diameter @xmath8 m ) as spacers between the film and the sample . in a polarisation microscope
the image light intensity is described by malus law @xmath9 where @xmath10 is the local faraday rotation of the polarisation ( the signal ) , @xmath11 is an offset angle from exactly crossed polariser and analyser , @xmath12 is the residual intensity at full extinction ( @xmath13 ) caused by imperfections in the optical components , and @xmath14 is the intensity at maximum opening . allowing the offset angle @xmath11 to be a non - zero value ( typically a few degrees ) brings two important benefits : firstly the image contrast is improved , and secondly we can distinguish between opposite field directions . in our images
bright pixels correspond to positive field , while negative field show up as dark pixels . in the present report
we have estimated the field - vs - intensity relation using image pixels away from the sample .
the experiments consisted of ramping the applied field slowly to a maximum level , and then slowly back to zero .
the applied field was controlled by computer with a ramp rate of 0.1 mt / s .
images were recorded at frequent and regular intervals during the ramp .
mt . the numbers next to dendrites indicate the order in which they appeared . the profile plot is obtained by averaging vertically within the rectangle .
the kink in the profile inside the central hole is an artefact caused by the presence of zig - zag domains in the mo indicator film.,title="fig : " ] mt . the numbers next to dendrites indicate the order in which they appeared .
the profile plot is obtained by averaging vertically within the rectangle .
the kink in the profile inside the central hole is an artefact caused by the presence of zig - zag domains in the mo indicator film.,title="fig : " ] figure [ critical_state ] shows the flux distribution in the large ring , initially in the zfc state , at an applied field @xmath15 mt . also shown is a flux density profile across the ring averaged over the rectangle indicated in the mo image . at this moderate applied field
the profile shows smooth flux penetration from the edges in agreement with critical state calculations @xcite and previous experiments @xcite on yba@xmath0cu@xmath1o@xmath2 .
notice the large positive field at the outer edge and the smaller negative field at the inner edge . in the mo image in figure [ critical_state ]
we also see several tree - like flux structures .
each distinct tree has grown extremely fast .
they have been labelled with a number indicating the order in which they appeared .
the first of them , labelled 1 , appeared at 3.4 mt . it is seen how the dendrites become larger with increasing applied field .
however , at these low fields all of them terminate far from the inner rim , where no activity is seen apart from a steadily increasing negative field . the first dendritic structures to almost reach the inner edge appear when the applied field reaches 7 mt .
increasing the field further , anti - flux dendrites appearing as dark fingers , eventually nucleate at the inner edge , see figure [ field_cycle_up ] a ) where @xmath16 mt . the zoomed view in b ) shows the details surrounding the two dominating bright structures .
most importantly , the anti - flux dendrites all originate at a point close to a bright finger tip .
the two large bright dendrites grew at different times , but in both cases the associated dark dendrites appeared in the same image in the sequence . while it seems clear that the anti - dendrites have grown after the dendrites , the two events take place in a very short timespan .
images c ) and d ) of figure [ field_cycle_up ] show the flux distribution in the small ring at 16.3 mt .
the overall features are essentially the same . in the images
one can see a few dark dendrites that have grown from the inner rim , tracking the core of some bright dendrites .
again associated dark and bright dendrites grow simultaneously within our temporal resolution .
in fact , it is a general feature of all our experiments that the anti - dendrites occur in conjunction with a bright dendrite - they coincide both temporally and spatially .
is shown in the zoomed - in view of the image .
dark and bright dendrites are weaved together in what appears to have been multiple avalanche events .
the image has been background corrected by subtracting an image acquired on the virgin sample at 15 mt . ] furthermore , we observe that while the bright dendrites branch multiple times , the anti - dendrites always consist of just one long finger . in addition
, the dark dendrites usually find a bright branch of a tree from the outer edge and trace that branch quite closely .
the image of the small ring in figure [ field_cycle_up ] d ) shows how close this tracing can be .
the same observations apply also to field cooled ( fc ) samples . in figure [ field_cycle_up_fc ]
the images show dark fingers which grow deep into bright trees that originate at the outer edge . just as for the zfc experiments , the anti - dendrites are temporally and spatially strongly correlated with bright dendrites .
an interesting detail can be seen in the zoom view of figure [ field_cycle_up_fc ] , where dendrites and anti - dendrites are stacked on top of each other as if they were weaved together .
returning to the zfc experiments , the behaviour is different when we decrease the applied field from its maximum value .
the field at the outer and inner edges then decreases and increases , respectively .
flux of opposite polarity - dark on the outer edge , bright on the inner - penetrates the sample , with a regular penetration being interrupted by dendritic structures , see figure [ field_cycle_down ] . in order to illustrate more clearly the dynamical aspects , image
b ) displays the difference between subsequent images in the sequence . where the flux density is unchanged , pixels are gray .
dark pixels indicate that flux has left or anti - flux has entered .
dendritic structures , dark from the outer and bright from the inner edges , have appeared in both images , meaning they nucleated at the same time . however , unlike the behaviour in increasing field , we find that i ) the tips of the structures are far apart , ii ) neither of them come close to the opposite edge , and iii ) both are branching .
we have found that the anti - dendrites forming in increasing field appear to be triggered by bright dendrites approaching the inner rim .
there are at least two possible triggering mechanisms : i ) a non - local magnetic coupling where the negative field at the inner edge is enhanced by the sudden appearance of a bright dendrite , and ii ) a local thermal mechanism where heat associated with the bright dendrite tip facilitates the nucleation and growth of an anti - dendrite .
the enhancement of the negative field is demonstrated in figure [ j_dendrite ] , showing the field ( left ) and current ( right ) maps close to a long dendrite which almost comes across to the center hole .
the current map shows a significantly increased current density between the finger tip and the inner edge .
one can understand this increase by considering the current flow around dendrites .
figure [ j_dendrite ] shows that the current density is large along flux fingers , but very small in the dendrite cores , implying that little current flows across the dendrite .
indeed , previous work on mgb@xmath0 has indicated that the current density is in fact maximum along dendrites , @xcite thus making them effective barriers against additional current . as a result , the meissner state currents otherwise flowing throughout the ring become concentrated near the inner edge , resulting in a regional increase in the magnitude of the negative edge field .
an anti - dendrite can form provided the magnitude exceeds a threshold value , since a typical feature of both conventional @xcite and dendritic @xcite flux jumps is the existence of a threshold field which must be exceeded for an avalanche to occur .
furthermore , the abrupt character of the field increase will lead to a large electric field which also helps the nucleation @xcite of an anti - dendrite .
in addition to this non - local effect , a bright dendrite that reaches the negative flux region near the inner edge will also induce a local temperature increase .
this is because the core of the bright dendrite is itself a region of increased temperature , and because the ensuing flux annihilation releases heat .
the resulting elevated temperature in a small area helps trigger an anti - dendrite , much like the laser pulse triggering @xcite on yba@xmath0cu@xmath1o@xmath2 . with
mo imaging it is very difficult to determine how far from the inner edge a bright dendrite actually stops when an anti - dendrite has grown on top of it . from our experiments we are unable to tell whether bright and dark regions have been in contact prior to the nucleation of the anti - dendrite , and thus whether the thermal trigger mechanism is feasible .
this important issue is open for future study using ultra - fast mo imaging techniques @xcite .
once anti - dendrites have been triggered they tend to trace the bright dendrite whose appearance triggered them .
we believe this tracing is assisted by the flux - anti - flux attraction , heating as a result of flux annihilation , and possibly the residual heat in the core of the bright dendrite .
these three effects help contain the anti - dendrite tip within the bright finger , and hence also lead to the observed suppression of branching . for low applied fields
the magnitude of the field at the outer edge is larger than that at the inner edge . in consequence , the threshold field
is reached sooner at the outer edge and dendrites nucleate there first .
this fact has profound implications for the dynamics at the inner edge . before the negative field at the inner edge reaches the threshold , bright dendrites perforate the ring , bringing positive flux from the outside to the central hole .
after the first perforation event , new events are frequent and increase the average flux density at the inner edge to positive values , so the net effect of increasing the applied field is to increase the field at the inner edge as well .
this is demonstrated in figure [ descending ] , where the left mo image shows the flux distribution at peak applied field , @xmath17 mt .
while the positive inner edge field explains why no dark denrites form in increasing applied field , one needs to examine the images in figure [ descending ] more closely to understand why bright dendrites form when @xmath18 is subsequently decreased .
of particular importance are regions where the field is large and positive at the inner edge , while the flux density inside the superconducting material is small , see e.g the encircled area in the images . in such regions
the field gradient inside the superconducting material is the opposite of what one would see in the case of regular penetration .
moreover , as is shown in the right image in figure [ descending ] , where @xmath18 has been decreased to 12.6 mt , the flux change is initially uniform at the two edges , meaning that the already positive edge field in the encircled region has increased further .
thus a modest decrease in applied field is sufficient to induce the bright dendrite we see in the right image in figure [ descending ] .
the result is that while the perforations _ impede _
the nucleation of dark dendrites in increasing @xmath18 , they _ facilitate _ nucleation of bright dendrites in decreasing @xmath18 .
flux dendrites which nucleate at the outer edge of a superconducting mgb@xmath0 ring lead to unexpected flux penetration at the inner edge .
in particular , we have found that when increasing the applied field to an intermediate level , i ) dendrites and anti - dendrites nucleate at the outer and inner edges of the rings , respectively ; however , all anti - flux dendrites are triggered by large flux dendrites ; ii ) anti - dendrites do not branch , instead they find a finger of the triggering positive dendrite and trace it closely .
the triggering can occur either due to a locally enhanced magnetic or electric field , or due to a local temperature elevation in the negative flux near the inner edge .
ultra - high temporal resolution is needed in order to conclusively decide which of the two mechanisms is dominant .
we further found that iii ) for larger applied field very large dendrites perforate the rings , bring flux into the center hole , and ultimately reverse the field profile near the inner edge ; and iv ) the reversed field profile leads to prolific nucleation of flux dendrites at the inner edge when the applied field is subsequently reduced . | magneto - optical imaging was used to observe flux dendrites with opposite polarities simultaneously penetrate superconducting , ring - shaped mgb@xmath0 films . by applying a perpendicular magnetic field , branching dendritic structures
nucleate at the outer edge and abruptly propagate deep into the rings . when these structures reach close to the inner edge , where flux with opposite polarity has penetrated the superconductor , they occasionally trigger anti - flux dendrites .
these anti - dendrites do not branch , but instead trace the triggering dendrite in the backward direction .
two trigger mechanisms , a non - local magnetic and a local thermal , are considered as possible explanations for this unexpected behaviour .
increasing the applied field further , the rings are perforated by dendrites which carry flux to the center hole .
repeated perforations lead to a reversed field profile and new features of dendrite activity when the applied field is subsequently reduced . |
rare decays of heavy mesons received considerable interest in the recent years as possible grounds for testing the standard model in flavor changing neutral current processes which known to occur only at one loop level @xcite .
exclusive decays of @xmath9-mesons to the non - charmed final states ( which are examples of fcnc ) are somewhat easier to observe experimentally
. however theoretical calculations of exclusive rates include modeling of the hadronization of the final @xmath10-quark and the resulting predictions vary by the order of magnitude from one to another .
the inclusive radiative decay @xmath11 was recently observed by cleo collaboration @xcite and is consistent with the @xmath12 expectations .
it is now generally accepted that @xmath1 corrections significantly change the decay rate for this process so it is important to bring all of the possible corrections under control to reduce theoretical uncertainties associated with @xmath1 .
most work on the inclusive @xmath0 decay proceeds by calculating radiative corrections to the quark decay @xmath4(see for example @xcite ) . however , there are also several diagrams where a radiative correction involves a gluon which interacts with the spectator quark .
they generate effects which are not accounted for in spectator model calculations , and the inclusive rate @xmath11 could in principle differ from that of @xmath4 at order @xmath2 .
if one is interested in a complete analysis of this decay mode , these non - spectator diagrams must be included .
this is especially true if one is interested in the photon spectrum , since reactions which include the spectator quark produce a softer photon spectrum than does the free quark decay .
the calculation presented below fills this gap and completes the calculation of the process of inclusive radiative decay of heavy meson at the level @xmath13 in this paper , we present a @xmath1 based calculation which estimates these non - spectator effects .
there are 12 diagrams which we consider , shown in fig 1 .
the weak vertices are calculated in @xmath1 perturbation theory ( some vertices already exist in the literature ) , and the accmm model @xcite is used to calculate the decay rates and spectra .
the interference of these diagrams with the leading @xmath4 amplitude only occurs for small values of the momentum of the spectator quark .
this portion is then particularly sensitive to the bound - state wavefunction , but fortunately the interference is relatively small .
the main non - spectator effect comes from the region of phase space where the spectator quark recoils significantly , and hence are second order in @xmath2 .
the plan of the paper is as follows . in section 2 ,
we organize the calculation and recall some of the information which in known about the gluonic vertices .
section 3 is devoted to most of the details about the calculation of the modifications to the decay rate and photon spectrum .
the impact of the results is described in section 4 . before proceeding ,
let us resolve a nomenclature ambiguity .
it is conventional to call calculations which focus on the @xmath4 transition by the phrase `` spectator model '' and to refer to the light quark in the b hadron as the `` spectator '' quark .
we will continue to use the latter phrase for the light quark in the b meson although , in the diagrams which we consider , it is no longer a spectator but is an active participant in the weak decay .
we will call the @xmath14 diagrams the spectator process and conversely will label those which actively involve the spectator quark as non - spectator transitions .
there are quite a number of diagrams at order @xmath2 which lead to modifications to @xmath11 , as shown in fig 1 .
let us divide these into five classes : \1 .
@xmath1-corrections to the single quark process @xmath4 .
`` gluonic - penguin '' ( bremstrahlung ) diagrams which include @xmath15-vertex with photons emitted from external quark legs . \3 .
`` triangle '' diagrams in which both the gluon and photon are emitted from weak vertex due to a quark loop .
hard gluon spectator exchange diagrams with gluons emitted from @xmath16 or @xmath17-lines .
diagram ( @xmath18 ) involves the @xmath19-vertex and and is suppressed by a power of @xmath20 .
we will not consider this diagram further .
the calculation proceeds by evaluating each of the diagrams , adding them together coherently , integrating over the initial particles momentum using a momentum space wavefunction and summing over the final momenta and spins .
the effective vertices are often momentum dependent and depend on the momentum flow in the diagram .
since we are dealing with inclusive decays , the details of hadronization of the final @xmath17-quark state are not relevant for our discussion ( all hadronic final states are summed over ) , so we have to model only the initial state .
this substantially reduces the model - dependence of the final result .
however we do have to carefully account for interference effects between the many diagrams , as different diagrams contribute to different portions of the final particles phase space . for the @xmath4 process
, the spectator quark emerges with only low momentum reflecting its momentum in the bound state wavefunction .
however in the other diagrams , the final quarks carry a typical momentum of a fraction of the b mass .
interference between the two classes only occurs in a portion of the available phase space .
we address this region specifically below . for the initial state wavefunction we use the accmm model .
in this calculation we shall encounter two types of vertices : `` penguin '' and `` triangle '' .
1.penguin vertex ( fig.2a ) .
this vertex was calculated quite a while ago @xcite and for the @xmath15-interactions turned out to be [ penguin ] _ ^a=- v_bt v_ts ( f_1^g ( x ) ( k^2 _ - k_/k ) ( 1 + _ 5 ) - m_b f_2^g ( x ) i _
k^(1-_5 ) ) where @xmath21 for @xmath22 .
@xmath23 are the inami - lim functions , @xmath24 , using unitarity of ckm matrix we get @xmath25 and @xmath26 for @xmath27,@xmath28 .
only the magnetic - moment type vertex contributes to the real photon emission in the @xmath4 , so [ onshell ] _ = v_biv_is m_b f_2^ ( x ) i _
q^(1-_5 ) @xmath29 .
we refer to @xmath30 being the momentum of the gluon and @xmath31 - momentum of the photon . in the gluon vertex
, we shall drop the small contribution from @xmath32 , which , however , might be sizable for gluonic penguin - mediated hadronic processes @xcite . \2 .
`` triangle '' vertex ( fig.2b ) .
the calculation of this vertex reduces to the calculation of the set of usual triangle diagrams with certain form - factors ( a complete parameterization of the diagrams of this type was first given in @xcite ) and is interesting in the sense that light ( spectator quark ) current couples to @xmath33 current through @xmath34-tensors .
the vertex itself was recently calculated by d.wyler and h.simma in @xcite and takes the form [ triangle ] _
^a=- v_bi v_is ( i f_1 s + i f_2 s + i f_3 s ) ^s ( 1 + _ 5 ) where @xmath35 with @xmath36 and @xmath37,@xmath38 , @xmath17 is a total momentum of internal quarks .
these vertices will be used to calculate the rate and spectrum in the next section .
we define the total decay rate as _
tot=_q+_sc+_mix where @xmath39 represents the leading contributions from @xmath4 ( diagrams of the class @xmath40 ) , @xmath41 gives the corrections which involve the spectator quarks , and @xmath42 gives the mixing of the two . in powers of the strong coupling constant , @xmath42 is of order @xmath2 and @xmath41 is of order @xmath43 .
in order to account for bound state effects , we will treat the decay of a @xmath9-meson as a @xmath44 scattering process , assuming that the scattering occurs from the bound state and model this state with the accmm - model @xcite . in calculating the decay rate ( cross section @xmath44 )
we average over the spins of the initial particles which , in principle , leads to the prediction for @xmath45 for the `` mixture '' of @xmath9 and @xmath46-particles ( one can , at the expense of a considerable increase in complexity of the calculation , separate @xmath9 and @xmath46 by introducing polarization operators for projecting out different spin combinations ) .
thus the decay rate is defined as [ rate ] d= |t_fi|^2 d where factor of @xmath47 in the denominator comes from the spin averaging and [ phase ] d=(2)^4 ^ 4(p_f - p_i ) we also treat spectator quark as massless to simplify the following calculations .
the final phase space integration is performed numerically .
the introduction of interactions with the light quark obviously makes the kinematics of the process more involved - the final state is no longer a simple two - body final state which results in delta - function - like peak in the photon energy distribution . instead
, the peak is now extended down to lower photon energies making it possible for the photon to populate the energy interval from @xmath48 to @xmath49 . moreover , a difference between the decay spectra of neutral and charged mesons also arises .
the question that naturally appears here is whether it is possible to have an interference of the spectator and non - spectator mechanisms .
since the @xmath4 transition is of order @xmath50 and the non - spectator matrix elements are first order in @xmath2 this would provide corrections of order @xmath2 . if there were not any momentum associated with the bound state , the spectator quark would remain precisely at rest in the @xmath4 transition and therefore would not interfere with the diagrams where momentum was transferred to the spectator quark .
however if bound state effects are present the spectator quark will always carry some momentum ( typically of order 1 gev or less ) and the two classes of diagrams can interfere . since this occurs only over one portion of the phase space , we can say that the interference effects are really of order @xmath51 . in the bound state ,
the initial momenta are described by a momentum distribution @xmath52 ( which is given by the wave function of quarks inside of the meson).the interference takes place with a probability proportional to a wave - function overlap .
we now turn to the calculation of this overlap .
let @xmath53 to be an operator responsible for the transitions via mechanism @xmath40 and @xmath54 - via mechanisms @xmath55 .
then , the mixing term arises as a result of @xmath56 where we have used completeness relation for @xmath57-states .
we are interested in the last term .
let us write down the contributions from different classes of diagrams in turn .
@xmath58 here @xmath34 is the polarization vector of the photon @xmath59 are the charges of b - quark and spectator respectively , and @xmath60 is given by ( [ penguin ] ) .
note that @xmath61 in the denominator due to gluon propagator is to be cancelled by @xmath61 in @xmath15-vertex , so @xmath62 is a local operator .
this operator ( convoluted with the lowest order non - spectator graph , along with its charge conjugate part ) gives a contribution to the shape of the photon spectrum .
the result of convolution can be represented as @xmath63 with @xmath64 , @xmath65 and analogously for @xmath66 and @xmath67 , @xmath68 , @xmath69 , @xmath70 is a color factor coming from the averaging over initial colors and summing over the final ones ( taking into account the fact that initial and final states are color - singlets ) .
also @xmath71 the calculation of the contribution from the diagrams of the class @xmath72 is more involved mainly because of the complicated structure of the vertex ( [ triangle ] ) .
the amplitude squared reads t_iii^2= h_^ l^w_= -h^_l^with @xmath73 we will be working explicitly with components of this current in order to get @xmath74 . using various @xmath75-matrix identities @xmath76 , \nonumber
\\ \gamma_\alpha \gamma_\nu \gamma_\theta = g_{\alpha \nu } \gamma_\theta + g_{\nu \theta } \gamma_\alpha - g_{\alpha \theta } \gamma_\nu - i \epsilon_{\alpha
\nu \theta \beta } \gamma_\beta \gamma_5\end{aligned}\ ] ] we arrive at h^=2i q^p^|b(p ) _ i f^_i s \{g _ _ _ s - g _ _ _ s + i _ _ _ s } ( 1+_5 ) b(p ) note that the term symmetric under @xmath77 disappears .
next , using explicit form of @xmath78 , identities ^ _ 5 = - ^ _ to get rid of @xmath34-couplings and discarding terms proportional to @xmath79 ( on - shell photons ) , we write @xmath80 using ( [ de ] ) we obtain the contributions from class @xmath72 . finally , class @xmath81 gives the following @xmath82 let us also mention that some range of momentum transferred to the spectator gives a contribution to the decay rate which is highly suppressed - in the accmm - model , for instance , decay rate looks like = d^3p ^*_b(p ) _ qm _
b(p ) with @xmath83 and @xmath84 . note that this expression reduces to the conventional accmm - averaging if the momentum flow between heavy quark and spectator lines is discarded .
it is clear that if the magnitude of @xmath85-momentum transferred to the spectator line is large , the factor of wave function suppression is large as well which results in the exponential fall - off of the @xmath2-spectrum as @xmath86 .
this is the reason why non - spectator @xmath2 corrections turn out to give much smaller impact on the line shape of the spectrum than those at order @xmath43 , in fact , 1.0 10 ^ -2 let us now consider corrections of order @xmath43 .
in fact they can simply be obtained by squaring the sum of @xmath87 presented in the previous section .
we shall present only few of them emphasizing their possible experimental signatures and highlighting some steps of derivations .
* class i*. diagrams of class @xmath40 give contributions to @xmath39 and have been calculated elsewhere @xcite .
this contribution to the branching ratio @xmath88 varies from @xmath89 to @xmath90 .
the major uncertainties at present come from the choice of the renormalization scale in the effective operators and the km elements , and so this branching ratio can be predicted more accurately in the future as the km elements are more precisely measured . *
class ii*. diagrams of this class give a new contribution to @xmath45 which comes from the emission of the photon from the spectator quark .
the matrix elements for diagrams ( @xmath91 ) and ( @xmath92 ) ( fig .
1 ) can be rewritten as t_fi= h_l_^^and are given by ( [ t2 ] ) .
taking the hermitian conjugate and multiplying by ( [ t2 ] ) gives @xmath93 for which we can write [ master ] |t_fi|^2= h _
w^ where @xmath94 where as before we set @xmath95 and @xmath96 is a color factor . to obtain @xmath96
we have used the fact that @xmath9-meson is a superposition of the color - singlet states , so c^2_= tr ( ) = evaluating the traces we obtain for @xmath97 : @xmath98 where the last term vanishes upon contraction with @xmath99 for which we can write down ( taking into account contraction with @xmath100 ) : @xmath101 for the diagrams ( @xmath91 ) and ( @xmath92 ) .
these diagrams might be interesting from the experimental point of view since they give rise to different contributions from charged and neutral @xmath9-mesons ( the amplitude is proportional to the charge of spectator @xmath102 ) , so one , in principle , can observe this charge asymmetry a= ~0.3 10 ^ -3 however the small size of our prediction means that it will be difficult to make this measurement .
* class iii*. calculation of the diagrams of this class involves the effective vertex ( [ triangle ] ) for the sum of two diagrams - 8 and 9 .
as before , the amplitude squared can be written in the form @xmath103 expanding @xmath104 as @xmath105 we get @xmath106 thus , the problem is the calculation of @xmath107 .
this calculation , which requires the explicit form for the @xmath108-functions and involves expansions of the products of @xmath34-tensors , results in a quite lengthy final expression .
however , restricting our attention only to the case of real photons ( @xmath109 ) and taking advantage of the symmetry of @xmath110 under @xmath111 ( implying further convolution with `` light '' tensor @xmath110 ) simplify the result enormously since there are only four combinations of @xmath108 s that give non - vanishing contributions to the problem of interest .
they are @xmath112 combining all of the terms together and contracting @xmath113 to @xmath110 one obtains the contribution of the class iii diagrams to the @xmath45 which is to be computed numerically . the resulting analytical expression for
the amplitude is too long to be presented here . *
class iv*. finally for the class of diagrams responsible for the @xmath16(@xmath17)-spectator hard gluon exchange , using the same technique as above , we write for @xmath114 and @xmath115 tensors @xmath116 here the `` heavy '' current is h^a_= ( @xmath117 is given by ( [ onshell ] ) .
evaluating the traces and making use of the facts that @xmath118 is symmetric under @xmath119 and @xmath109 , we obtain @xmath120 \nonumber\end{aligned}\ ] ] with @xmath121 and @xmath122 , here @xmath123 is a charge of the internal quarks .
contracting @xmath124 with @xmath118 gives the expression for the diagrams of class @xmath81 . in the final calculation of the magnitude of the spectator effects
, the dominant contribution comes from the class @xmath81 diagrams @xcite . in evaluating the matrix elements ,
we have taken the @xmath16-quark spinors to be non - relativistic .
the interferences between the different classes of diagrams involving the spectator quark have also been taken into account , although we have not displayed the ( lengthy ) formulas explicitly above .
some of the complicated sums over many dirac matrices were evaluated numerically with the help of @xmath125 .
substituting @xmath93 and ( [ phres ] ) into ( [ rate ] ) , and performing numerical @xmath126-dimensional monte - carlo integration over @xmath127 we obtain photon energy distribution presented in fig.3 . integrating over @xmath128 gives for the decay rate :
@xmath129 given the recent measured by @xmath130 value for @xmath131 @xcite we conclude that spectator corrections give an overall effect of about @xmath7 on the inclusive decay rate @xmath132 .
the diagrams which involve the spectator quark are in general non - local and involve momentum transfers to the spectator which are comparable to the energy released in the decay .
this is a quite different configuration from the @xmath4 transition where the spectator quark is nearly at rest . despite the inclusion of the bound state momentum of the spectator ,
the interference of the two classes of diagrams turns out to be very small .
we have evaluated the remaining effects of the spectator diagrams , which then amount to effects at order @xmath43 . * * since we are calculating @xmath1-corrections associated with bound - state effects it is convenient to work in the rest frame of @xmath9-meson .
let us adopt accmm - model for use in this particular frame . in this coordinate system ( the center of mass frame for the system `` heavy quark - spectator '' )
: @xmath133 we treat the spectator as a massless particle , so e_u = p_u = p , e_b= as usual in the accmm model the effective @xmath16 quark mass is momentum dependent and is given by @xmath134 .
also [ eu ] e_s=== let us note that in this frame the decay products @xmath135 lay in the same plane , thus the angle parameters are @xmath136-the angle between the directions of @xmath85-momenta of @xmath17-quark and photon and the relative angle between @xmath85-momenta of the initial ( `` spectator '' ) @xmath137-quark and normal to the decay plane .
let us recall that ^4(p_f - p_i)=(e_f - e_i ) ^3(_f-_i ) therefore , ( [ phase ] ) can be integrated over @xmath138 to get d ~ ( m_b - e_u-e_s- ) with @xmath139 .
next , evaluating the @xmath140-integral ( keeping in mind that @xmath141 is given by ( [ eu ] ) ) yields @xmath142 making use of a familiar @xmath143-function relation ( f(x))=_x = x_0 ( x - x_0 ) to perform integration over @xmath144 we arrive at the equation for the roots of @xmath144 : m_b - e_u- - = 0 which can be solved to get e_u_0=. this gives for the @xmath143-function : ( e_f - e_i)= ( e_u - e_u_0 ) hence , the phase volume is [ phres ] d ~d d _ q d _u with @xmath145 . | in @xmath0 there are perturbative @xmath1 corrections at order @xmath2 not only to the single quark line process @xmath3 but also coming from a set of diagrams where the weak interaction vertex involves a gluon which interacts with the spectator quark in the b hadron .
we discuss the impact of the full set of these diagrams .
these can influence the decay rate and also the line shape of the photon spectrum as they favor a softer photon energy than does the pure spectator decay @xmath4 .
a subset of these diagrams generate differences in the decay rates for charged and neutral b s , @xmath5 vs. @xmath6 which can be searched for experimentally , although our calculation indicates that the effect will be hard to observe .
bound state effects turn out to be required to give rise to spectator corrections to the decay rate at order @xmath2 .
the resulting contribution to the total rate is rather small ( of about @xmath7 ) , although the effect is somewhat larger on portions of the photon spectrum . + @xmath8 + * umhep-423 * /#1#1 -0.5em / epsf |
since its introduction by @xcite , the deep boltzmann machine ( dbm ) has been one of the most ambitious attempts to build a probabilistic model with many layers of latent or hidden variables .
the dbm shares some characteristics with the earlier deep belief network ( dbn ) @xcite .
both models may be viewed as multi - layer or `` deep '' extensions of the popular restricted boltzmann machine ( rbm ) . however , unlike the dbn , commonly used dbm approximate inference schemes implement true feedback mechanisms where the inferred activations of high - level units can influence the activations of lower - level units .
thus alternative interpretations of an input are able to compete at all levels of the model simultaneously .
this sort of sophistication in inference has the potential to lead to more robust and globally coherent inferences , which are likely to translate to better performance when the model is employed in tasks such as classification . despite its success
, the dbm has not yet entirely fulfilled this potential and the dbn remains the more popular model paradigm for deep probabilistic models .
one potential explanation of why this is so could be tied to the difficulties one encounters when attempting to estimate the model parameters from data .
straightforward applications of gradient - based methods such as stochastic maximum likelihood ( sml ) @xcite ( also known as persistent contrastive divergence @xcite ) appear to fall in poor local minima , and fail to adequately explore the space of model parameters .
one solution , as presented in @xcite , is based on a greedy layer - wise pretraining strategy which appears to overcome these poor local minima .
each layer is trained as an rbm with the latent activations of the layer below as its input .
these layers are then recombined by rescaling the weights to account for the doubling of inputs into each intermediate layer .
while this procedure seems to work well , it involves many steps and is more complicated than would be ideal .
also , more importantly , the necessity of layer - wise pretraining makes it very difficult for the organization of upper layers to influence the topological organization of lower layers .
for example , in cases where we would like to learn a dbm that is not fully connected between each layer , it would potentially be very desirable for the global connectivity pattern ( eg .
local receptive fields ) to influence the pattern of learned filters at all layers .
layer - wise pretraining precludes this possibility .
if , on the other hand , one could jointly train all layers of a deep boltzmann machine simultaneously , then the pattern of activation of the upper layer units would have an opportunity to influence the weights trained at the lower layers via their effect on lower - layer units activations . in this paper , we describe a simple scheme for joint training of all layers of a deep boltzmann machine . our strategy is based on the observation that the poor local minima in which sml falls are characterized by high variance in the norms of the weight vectors , particularly those between the data and the first layer of hidden units .
our solution is to simply add a regularization term to the standard maximum likelihood training criterion that penalizes large differences in the norms of the weight vectors , both within a layer and across neighbouring layers .
our regularization is based on the intuition that successful models are those where all units contribute roughly equally to the representation of the data .
this is a common principle that has previously been applied in the form of sparsity penalties for rbm and dbn training @xcite , where all hidden units are encouraged to be active over an equal proportion of the data .
our application of this principal is somewhat different , we do not explicitly control the activation of the hidden units , instead we encourage equal influence of each _ active _ hidden unit .
we demonstrate , with experiments , both the failure of standard sml training for dbms and the effectiveness of our regularized - sml dbm training strategy .
a boltzmann machine is defined as a network of symmetrically - coupled binary stochastic units ( random variables ) .
these stochastic units can be divided into two groups : ( 1 ) the _ visible _ units @xmath0 that represent the data , and ( 2 ) the _ hidden _ units @xmath1 that mediate dependencies between the visible units through their mutual interactions .
the pattern of interaction is specified through the energy function : @xmath2 where @xmath3 are the model parameters which respectively encode the visible - to - visible interactions , the hidden - to - hidden interactions , the visible - to - hidden interactions , the visible self - connections , and the hidden self - connections ( also known as biases ) . to avoid over - parametrization
, the diagonals of @xmath4 and @xmath5 are set to zero .
the energy function specifies the probability distribution over the joint space @xmath6 $ ] , via the boltzmann distribution : @xmath7 where the partition function @xmath8 ensures that the density normalizes : @xmath9 this joint probability distribution gives rise to the set of conditional distributions of the form : @xmath10 in general , inference in the boltzmann machine is intractable .
for example , computing the conditional probability of @xmath11 given the visibles , @xmath12 , requires marginalizing over the rest of the hiddens which implies evaluating a sum with @xmath13 terms : @xmath14 however with some judicious choices in the pattern of interactions between the visible and hidden units , more tractable subsets of the model family are possible .
the restricted boltzmann machine ( rbm ) is likely the most popular subset of boltzmann machines .
they are defined by restricting the interactions in the boltzmann energy function , in eq .
[ eq : boltzmann_energy ] , to only those between @xmath15 and @xmath16 , i.e. for @xmath17 , @xmath18 and @xmath19 . as such
, the rbm can be said to form a _
graph with the visibles and the hiddens forming two layers of vertices in the graph . with this restriction
, the rbm possesses the useful property that the conditional distribution over the hidden units factorizes given the visibles : @xmath20 likewise , the conditional distribution over the visible units given the hiddens also factorizes : @xmath21 this conditional factorization property of the rbm immediately implies that most inferences we would like make are readily tractable . for example , the conditional independence of the hiddens implies that posterior marginals of the form @xmath22 are immediately available .
importantly , the tractability of the rbm does not extend to its partition function , which still involves sums with exponential number of terms .
it does imply however that we can limit the number of terms to @xmath23 .
usually this is still an unmanageable number of terms and therefore we must resort to approximate methods to deal with its computation .
learning in the rbm is also rendered much more tractable in comparison to the general boltzmann machine .
typically , learning involves finding a set of model parameters that approximately maximizes the log likelihood of a training dataset : @xmath24 this can be accomplished via gradient ascent .
the gradient of the log likelihood of the data for the rbm is given by : @xmath25 where we have the expectations with respect to @xmath26 in the `` clamped '' condition , and over the full joint @xmath27 in the `` unclamped '' condition . in training , we follow the stochastic maximum likelihood ( sml ) algorithm ( also know as persistent contrastive divergence or pcd ) @xcite , i.e. , performing only one or a few updates of an mcmc chain between each parameter update .
the deep boltzmann machine ( dbm ) is another particular subset of the boltzmann machine family of models where the units are again arranged in layers .
however unlike the rbm , the dbm possesses multiple layers of hidden units as illustrated in figure [ fig : dbm ] . with respect to the boltzmann energy function of eq .
[ eq : boltzmann_energy ] , the dbm corresponds to setting @xmath28 and a sparse connectivity structure in both @xmath5 and @xmath29 .
we can make the structure of the dbm more explicit by specifying its energy function .
for the 2-layer model it is given as : @xmath30 with @xmath31 .
the dbm can also be characterized as a bipartite graph between two sets of vertices , formed by the units in odd and even - numbered layers ( with @xmath32 ) .
a key point of departure from the rbm is that the posterior distribution over the hidden units ( given the visibles ) is no longer tractable , due to the interactions between the hidden units .
@xcite resort to a mean - field approximation to the posterior .
specifically , in the case of the 2-layer model , we wish to approximate @xmath33 with the factored distribution @xmath34 .
such that the kl divergence @xmath35 is minimized or equivalently , that a lower bound to the log likelihood is maximized : @xmath36 maximizing this lower - bound with respect to the mean - field distribution @xmath37 yields the following mean field update equations : @xmath38 iterating eq .
( [ eq : mf1]-[eq : mf2 ] ) until convergence yields the parameters used to estimate the `` variational positive phase '' of eq .
[ eq : variational_gradient ] : @xmath39 \nonumber \\ & = \mathbb{e}_{q_v } \left [ - e_{\mathrm{dbm}}(v , h^{(1)},h^{(2 ) } ) - \log q_v(h^{(1)},h^{(2 ) } ) \right ] - \log z(\theta ) \nonumber \\ \label{eq : variational_gradient } \frac{\partial \mathcal{l}(q_{v } ) } { \partial \theta } & = -\mathbb{e}_{q_v } \left [ \frac{\partial e_{\mathrm{dbm}}(v , h^{(1)},h^{(2)})}{\partial \theta } \right ] + \mathbb{e}_{p } \left [ \frac{\partial e_{\mathrm{dbm}}(v , h^{(1)},h^{(2)})}{\partial \theta } \right]\end{aligned}\ ] ] note that this variational learning procedure leaves the `` negative phase '' untouched .
it can thus be estimated through sml or contrastive divergence @xcite as in the rbm case . despite the intractability of inference in the dbm
, its training should not , in theory , be much more complicated than that of the rbm .
the major difference being that instead of maximizing the likelihood directly , we instead choose parameters to maximize the lower - bound on the likelihood given in eq .
[ eq : variational_lower_bound ] .
the sml - based algorithm for maximizing this lower - bound is as follows : 1 .
clamp the visible units to a training example .
2 . iterate over eq .
( [ eq : mf1]-[eq : mf2 ] ) until convergence .
3 . generate negative phase samples @xmath40 , @xmath41 and @xmath42 through sml .
4 . compute @xmath43 using the values obtained in steps 2 - 3 .
5 . finally , update the model parameters with a step of gradient ascent . while the above procedure appears to be a simple extention of the highly effective sml scheme for training rbms , as we demonstrate in sec .
[ sec : exp ] , this procedure seems vulnerable to falling in poor local minima which leave many filters effectively dead ( not significantly different from its random initialization with small norm ) .
[ sec : layerwise ] the failure of the sml joint training strategy was noted by @xcite . as a far more successful alternative
, they proposed a greedy layer - wise training strategy .
this procedure consists in pre - training the layers of the dbm , in much the same way as the deep belief network : i.e. by stacking rbms and training each layer to independently model the output of the previous layer .
a final joint `` fine - tuning '' is done following the above sml - based procedure .
while greedy layer - wise training has been shown to be reasonably successful , a means of joint training a dbm would be highly desireable . not only
would it be simpler , but it would open the door to local receptive field learning and more general architectures where we would like the top - down pattern of connectivity to influence the learning of lower - level features . in this section we detail our simple proposal for a means to jointly train all layers of a dbm . our strategy is based on the observation that standard sml - based joint training tends produce high variance in the weight vectors associated with each hidden unit particularly across the first - layer weight vectors that connect the visible units to the first hidden layer units .
our proposal is thus to regularize the maximum likelihood objective , in order to encourage units to have similar norms ( i ) within a given layer and ( ii ) to a lesser extent , across neighbouring layers .
we thus introduce an additional parameter @xmath44 for each layer of the dbm , representing the average norm of weight vectors of the @xmath45-th layer .
we then add a regularization term to our objective @xmath46 , which penalizes deviations from this mean - value using a squared - error penalty .
@xmath44 s belonging to adjacent layers are further constrained to be close ( again in the @xmath47 sense ) .
this gives rise to the following regularized objective : @xmath48 one can think of this regularization term as a spring , which ensures that the system evolves jointly from an initial random weight configuration ( where the columns of @xmath49 have small norm ) to a good model of the input distribution @xmath50 , which undoubtedly requires weight vectors with larger norms .
to evaluate the effect of our regularization term , we trained deep boltzmann machines on the pervasive mnist dataset @xcite .
all our models were trained for @xmath51 updates , using minibatches of size @xmath52 , varying the learning rate in @xmath53 . [
[ direct - approach ] ] direct approach + + + + + + + + + + + + + + + our first model was a 3-layer dbm with [ 500,500,1000 ] hidden units in the first , second and third layers ( respectively ) .
the resulting filters are shown in figure [ fig : bad_filters ] .
as we can see , many of the first - layer filters are not significantly different from their random initialization , with very small norm . based on the difference between these results and the successful training of an rbm
, we speculate that the top - down interactions prevent the first layer from learning useful filters .
we have confirmed that the high - frequency `` noise '' filters are actually the ones with lowest norm .
it seems as though early top - down input is influencing the activations of these hidden units , perhaps directly through some form of suppression , or indirectly , by reinforcing the activation of the subset of filters which become useful early on . while we plot filters for 3-layer networks , the same results hold for networks of depth two .
[ [ joint - training ] ] joint training + + + + + + + + + + + + + + using our regularized objective , we train a 2-layer dbm with 500 hidden units in each layer
. we set the hyper - parameters of eq.[eq : regll ] as follows : @xmath54 , @xmath55 and @xmath56 .
furthermore , we dampen the learning rate on the parameters @xmath57 and @xmath58 by a factor of a thousand , in order for the layer norms to evolve more slowly . a random subset of filters is shown in figure [ fig : good_filters ] . the effect of our regularization term is drastic : the vast majority of first layer filters seem to train successfully and resemble the pen - stroke detectors characteristic of features learned on mnist . while more difficult to interpret , the second layer filters are also clearly superior to the ones of figure [ fig : bad_filters ] , combining lower - level features ( pen strokes ) into more global digit - like objects .
we have introduced a simple regularization scheme that appears to prevent sml from falling into a poor local minimum of our objective function .
the regularizer encourages units to learn filters which have similar norms , both within and across adjacent layers .
we have empirically demonstrated the failure of joint training of all layers of a deep boltzmann machine which leaves many filters in the lower layer of a 2-layer dbm with very small norm and , consequently , little contribution to modeling the data .
we have also empirically demonstrated the success of our regularization scheme in simultaneously training both layers of a 2-layer dbm .
while we have not shown it here , our scheme also appears to work well for dbms with more than 2 layers . in future work
, we would like to determine how many layers can be learnt simultaneously using similar basic norm - regularizaiton schemes .
the authors would like to acknowledge the support of the following agencies for research funding and computing support : nserc , calcul qubec and cifar .
we would also like to thank the developers of theano @xcite . | the deep boltzmann machine ( dbm ) has been an important development in the quest for powerful `` deep '' probabilistic models . to date , simultaneous or joint training of all layers of the dbm has been largely unsuccessful with existing training methods .
we introduce a simple regularization scheme that encourages the weight vectors associated with each hidden unit to have similar norms .
we demonstrate that this regularization can be easily combined with standard stochastic maximum likelihood to yield an effective training strategy for the simultaneous training of all layers of the deep boltzmann machine . |
the purpose of a heat engine is to transform an amount of heat @xmath0 , extracted from a hot reservoir at temperature @xmath1 , into an amount of work @xmath2 .
the efficiency @xmath3 for doing so is at most equal to carnot efficiency : @xmath4 , with @xmath5 . here
@xmath6 is the temperature of a second , cold reservoir @xmath7 , in which the remaining energy @xmath8 is deposited .
the equality is reached for reversible operation , implying that the corresponding power output is zero .
curzon and ahlborn were amongst the first to study the question of efficiency at maximum power @xcite . by considering a simple modification of the carnot engine and after applying the so - called endo - reversible approximation ( neglecting dissipation in the auxiliary system ) , they found the following beautiful expression for the efficiency at maximum power : @xmath9 . while this formula appears to describe rather well the efficiency of actual thermal plants , and is close to the efficiency at maximum power for several model systems , it is neither an exact nor a universal result , and it is neither an upper nor a lower bound .
it has therefore come as a surprise that a number of universal predictions can be made about the expansion of the efficiency at maximum power in terms of @xmath10 . in the regime of linear response ,
i.e. , at first order in @xmath10 , it is found that the efficiency at maximum power is at most half of the carnot efficiency @xcite . in other words ,
the ca efficiency is an upper bound at the level of linear response .
the proof was given for systems operating under steady - state conditions .
the upper bound is reached for so - called strongly coupled systems , i.e. , systems in which the heat flux and the work - producing flux are proportional to each other .
more recently , it has been shown that the quadratic coefficient , equal to @xmath11 , is also universal for strongly coupled systems in the presence of an additional left - right symmetry in the system @xcite .
furthermore , the universality of the coefficients is a direct consequence of the time - reversibility of the underlying physical laws .
the coefficient @xmath12 derives from the symmetry of the onsager matrix .
the coefficient @xmath11 can be seen as the implication of onsager symmetry at the level of nonlinear response .
the above universality predictions have been confirmed in a number of steady - state model systems involving classical particles @xcite , fermions @xcite , and bosons @xcite .
universality has also been observed in variants of the ca model based on carnot cycles performed in finite time , even though finding the optimal driving protocol maximizing work extraction can be notoriously difficult @xcite .
the connection with the steady - state analysis has been clarified by identifying the onsager coefficient for a finite - time carnot cycle in the linear regime @xcite .
furthermore , agreement with the universal quadratic coefficient has also been observed in a carnot cycle based on a ( classical ) brownian particle in an harmonic trap @xcite . in this paper
we provide a full analysis of a thermal engine undergoing a carnot cycle , with the auxiliary system consisting of a single - level quantum dot that is switched between a hot and a cold reservoir .
we show that the efficiency at maximum power is again consistent with the above discussed universality .
furthermore , ca efficiency at maximum power is obtained exactly at all orders in @xmath10 in the limit of low dissipation , which is distinct but similar to the case considered in the original ca paper .
our heat engine model consists of a single - level quantum dot interacting with a metallic lead through a tunneling junction . the quantum dot is assumed to have a single energy level @xmath13 near the fermi level of the lead while other levels in the dot do not contribute to the processes described below .
the state of the system is specified by the occupation probability @xmath14 of having an electron in the dot .
the lead plays the role of a thermal bath at temperature @xmath15 and chemical potential @xmath16 .
electrons are assumed to thermalize instantaneously upon tunnelling into the lead .
when the energy level @xmath13 is modulated by an external agent according to a given protocol , a certain amount of energy , positive or negative , flows into the system in the form of work and/or heat . in the case of an occupied level , an amount of work equal to @xmath17 is delivered to the system , where the subscripts @xmath18 and @xmath19 refer to final and initial values . when the electrons at energy level @xmath13 tunnel in ( out ) , an amount of heat equal to @xmath20 ( @xmath21 ) is extracted from the bath .
the basic problem that we address is the finite - time performance of this engine as it runs through the following four standard stages of a carnot cycle ( also see figure [ fig : model ] ) : * _ isothermal process _ + the quantum
dot is in contact with a cold lead at temperature @xmath6 and chemical potential @xmath22 .
the energy level is raised from @xmath23 to @xmath24 according to a certain protocol during a time interval of duration @xmath25 .
both work and heat are exchanged during this process .
* _ adiabatic process _
+ the quantum dot is disconnected from the cold lead , and the quantum level is shifted from @xmath24 to a new level @xmath26 . since the quantum dot is thermodynamically isolated , the population of the level does not change during this process .
hence , there is no heat exchange .
however , the change of the energy level releases a corresponding amount of work .
we assume that the operation time of this step is very short , in particular negligibly small compared to that of the isothermal processes . * _ isothermal process _
+ the dot is connected to the hot lead at temperature @xmath1 and chemical potential @xmath27 .
the energy level is lowered from @xmath26 to @xmath28 based on another protocol during a time interval of length @xmath29 . just as in step
i , both heat and work are exchanged . * _ adiabatic process _ + the system is again disconnected from the lead and the level is restored from @xmath28 to the initial level @xmath23 , at the cost of a corresponding amount of work .
afterwards , the dot is reconnected to the cold lead . again
, we assume that the operation time of this process is negligibly small .
the above procedure defines one cycle of the thermal engine , requiring a total time @xmath30 .
the protocols in steps i and iii must be designed in such a way that the thermodynamic state of the system , in our case the occupation probability @xmath31 of the quantum level , returns to the same initial value after every cycle .
since there is no change in occupation probability during the adiabatic stages ii and iv , the change in occupation probability from , say @xmath32 to @xmath33 , during process i , must necessarily be compensated by a change back from @xmath33 to @xmath32 during process iii .
the time evolution of the occupation probability @xmath14 for the state of the quantum dot in contact with a lead at temperature @xmath34 ( @xmath35 ) obeys the following quantum master equation : @xmath36 , \label{eq : master}\ ] ] where the @xmath37 and @xmath38 are transition rates . in the wide - band approximation ,
these rates are given by [ eq : rates ] @xmath39}+1 } \label{eq : rate1}\\ \omega_b & = & \frac{c}{{\mathrm{e}}^{+\beta[\varepsilon(t)-\mu(t)]}+1 } \label{eq : rate2},\end{aligned}\ ] ] where @xmath40 is a rate constant .
noting that raising the energy level is equivalent to lowering the chemical potential , we introduce an effective energy level @xmath41 . the master equation ( [ eq : master ] )
can now be rewritten as @xmath42 the effective level varies along the carnot cycle as @xmath43 .
note that the change in the chemical potential is included in the jump of the effective level during processes ii and iv .
we next turn to the thermodynamic description of the model .
we use the convention that heat entering the system is ( like work ) positive . the internal energy of the system at time @xmath44 is @xmath45 where @xmath46 the rate of change of the internal energy , @xmath47 , is the sum of two parts , namely , a work flux @xmath48 and a heat flux @xmath49 , @xmath50 note that the particle exchange contributes to the heat flux [ last term in eq .
( [ heatflux ] ) ] .
when the energy level is below the chemical potential , the direction of heat flow is opposite to the direction of tunnelling .
the net total work and net total heat during the process of duration @xmath51 are obtained as functionals of the occupation probability , @xmath52 & = & \int_{0}^{\tau } \dot{\epsilon}(t ) p(t ) { \mathrm{d}t } \label{eq : w } \\
\mathcal{q}[p(\cdot ) ] & = & \int_{0}^{\tau } \epsilon(t ) \dot{p}(t ) { \mathrm{d}t } \label{eq : q}.\end{aligned}\ ] ] for cyclic processes , we have @xmath53 and hence @xmath54 . for mathematical simplicity ,
we evaluate power using net heat instead of net work : @xmath55
our goal is to maximize the power output and to evaluate the corresponding efficiency .
power is a complicated functional of the time - dependent protocols in stage i and iii , and an exact analytical analysis looks difficult at first sight .
the optimization can however be done in two steps .
first , we fix parameter values , @xmath25 , @xmath29 , @xmath32 and @xmath33 and maximize the power with respect to the functional space of @xmath56 .
since the total operation time is fixed , we just need to maximize the heat .
next , we further maximize the power with respect to the remaining degrees of freedom @xmath25 , @xmath29 , @xmath32 and @xmath33 .
the problem of maximizing heat or minimizing work for a single - level quantum dot moving between given initial and final energy states has already been analysed in detail in @xcite .
we reproduce the crucial steps of this analysis for self - consistency . to find the protocol that maximizes the heat , we do not search directly for the optimal schedule @xmath56 , but identify the optimal occupation probability @xmath14 .
this is done by expressing @xmath56 in terms of @xmath14 and @xmath57 , and rewriting the heat , eq .
( [ eq : q ] ) , as a functional of @xmath14 and @xmath57 : @xmath58=\int_{0}^{\tau } \mathcal{l}(p,\dot{p } ) { \mathrm{d}t},\ ] ] where @xmath59 \dot{p}(t).\ ] ] the extremum is found via the standard euler - lagrange method , leading , after integration , to @xmath60 } = k. \label{eq : k}\ ] ] here @xmath61 is the constant of integration . solving the quadratic equation for @xmath62
, we obtain two first order odes , @xmath63 the upper sign ( @xmath64 ) should be used for upward processes in which the quantum level is raised and the lower sign ( @xmath65 ) for downward processes .
it is worth mentioning a useful symmetry between electrons and holes .
we are using the state of an electron , @xmath56 and @xmath14 , to describe the state of system .
instead , we can also use the state of holes , @xmath66 and @xmath67 . if @xmath14 is a solution for an upward process
, then @xmath67 is a solution for a downward process with @xmath66 .
hence we do not need to calculate the downward process separately , as it follows from this symmetry . before turning to the solution of the differential equation ( [ eq : ode ] )
, we examine the physical meaning of the constant @xmath61 . eliminating @xmath62 in eq .
( [ eq : k ] ) by using the master equation ( [ eq : master2 ] ) , the resulting quadratic equation for @xmath14 leads to the relation @xmath68 .
\label{eq : p}\ ] ] this equation indicates that when @xmath69 , @xmath14 is the equilibrium distribution associated with the instantaneous value of the energy , implying that @xmath69 corresponds to the quasi - static limit ( @xmath70 ) . as @xmath61 increases , @xmath14 deviates from the equilibrium distribution . we conclude that @xmath61 measures how far the state of the system deviates from the quasi static limit .
we will use this insight below to obtain a perturbative solution for small dissipation by assuming that @xmath61 is small .
next we proceed to solve eq .
( [ eq : ode ] ) .
separation of the variables @xmath31 and @xmath44 leads to the following explicit result for the upward processes : @xmath71-f[p(0);k],\ ] ] where @xmath72 \\ & + \frac{1}{2 } \ln \left [ \frac{2p+k+\sqrt{k^2 + 4kp(1-p)}}{2(1-p)+k+\sqrt{k^2 + 4kp(1-p ) } } \right ] .
\label{eq : f } \end{split}\ ] ] for the downward processes , we need to use @xmath73 .
the value of @xmath61 is determined by the boundary conditions : @xmath74-f[p(0);k ] .
\label{eq : boundary}\ ] ] note that @xmath61 depends solely on the operation time @xmath51 , the probabilities @xmath32 and @xmath33 , and the tunneling rate @xmath40 but not on temperature .
unfortunately , the function ( [ eq : f ] ) is quite complicated so we can not obtain an analytical expression for @xmath61 . in general we need to solve for it numerically .
however , an exact perturbative solution is possible , cf . the next section
. having thus obtained the optimal @xmath14 with @xmath61 determined by ( [ eq : general ] ) , we insert this expression in eq .
( [ eq : q ] ) to obtain the corresponding maximum heat for the optimal upward processes , @xmath75\\ & = \mathcal{s}[p(\tau);k]-\mathcal{s}[p(0);k ] = \delta\mathcal{s } , \label{eq : optq } \end{split}\ ] ] where @xmath76\\ & - \sqrt{k } \arcsin \left [ \frac{1 - 2p}{\sqrt{k+1 } } \right ] \\ & - \ln \left [ \frac{2(1-p)-k-\sqrt{k^2 + 4kp(1-p)}}{2 } \right ] . \end{split}\ ] ] for the downward processes , @xmath77 is replaced by @xmath78 . out of equilibrium
, @xmath79 is different from the system entropy @xmath80 .
indeed , @xmath81 is the entropy flow and is related to the system entropy change @xmath82 via the always - positive entropy production @xmath83 .
it is only in the quasi - static limit , where @xmath84 and thus @xmath85 , that @xmath77 reduces to @xmath86 .
we are now ready to apply the above results to our heat engine . to make the connection with the left / right symmetry required for the universality of the coefficient in the quadratic term , cf .
the discussion in the introduction , it will be of interest to consider an asymmetry in the rate constant : we will use the subscripts @xmath87 and @xmath88 for the rate constant @xmath40 when in contact with the cold and hot reservoir , respectively . recalling that processes i and iii are upward and downward processes , respectively
, the boundary condition ( [ eq : boundary ] ) leads to [ eq : exactk ] @xmath89 which determine the integration constants @xmath90 and @xmath91 , respectively . substituting @xmath90 and @xmath91 into eq .
( [ eq : optq ] ) , we obtain the amount of heat that enters the system during the processes i and iii : [ eq : exactq ] @xmath92 = t_c \delta \mathcal{s}_c\\ q_h & = t_h [ \mathcal{s}(1-p_0;k_h)-\mathcal{s}(1-p_1;k_h ) = t_h \delta \mathcal{s}_h , \end{aligned}\ ] ] which leads to the efficiency of the engine @xmath93 in the quasi - static limit , @xmath94 and @xmath95 , one has @xmath96 , hence @xmath97 , so that eq .
( [ eq : exact_eta ] ) reduces to carnot efficiency .
the above results provide the required optimization with respect to the schedules .
it remains to perform the optimization with respect to the remaining degrees of freedom @xmath25 , @xmath29 , @xmath32 and @xmath33 .
in general , this can only be done numerically since eq .
( [ eq : general ] ) only provides an implicit equation for the time - dependence of the optimal schedule .
we are , however , mainly interested in the verification of universal features of efficiency at maximum power .
we therefore proceed with a perturbative analysis for which analytic solutions can be obtained .
the deviation from carnot efficiency can be investigated using the theory of linear irreversible thermodynamics where @xmath98 is assumed to be smaller than the temperatures @xmath1 and @xmath6 of the reservoirs .
however , for finite time thermodynamics , a different kind of expansion , directly related to the irreversibility caused by finite operation time , is more natural . as mentioned in the previous section
, @xmath61 is a direct measure of the deviation from the quasi - static limit .
hence , it is natural to expand thermodynamic quantities in @xmath61 .
since this is an expansion about the reversible case of zero dissipation , we will refer to this as the limit of weak dissipation .
we expand eq .
( [ eq : f ] ) in a series in @xmath99 .
the leading term is @xmath100 with this approximation , we are able to solve eq .
( [ eq : exactk ] ) for @xmath61 to obtain @xmath101 where @xmath102 the present expansion is thus valid under the following condition of weak dissipation : @xmath103 note that it can easily be satisfied in our model since the right hand side is bounded by @xmath104 .
once we find the value of @xmath61 , the remaining calculation is straightforward .
equation ( [ eq : general ] ) leads to the optimal protocols : [ eq : optprotocol ] @xmath105\\ p_h(t)&=&\frac{1}{2}\left[1+\sin\left ( \frac{t}{\tau_h}|\phi_1-\phi_0| + \phi_1\right)\right].\end{aligned}\ ] ] expanding in a taylor series with respect to @xmath99 , eq .
( [ eq : s ] ) is approximated by the two lowest order terms as @xmath106 inserting the value of @xmath61 , we obtain the maximum heat [ eq : optq ] @xmath107 where @xmath108 is the reversible entropy change .
the second term on the right hand side is the irreversible heat , which has to be small under the condition ( [ eq : condition1 ] ) of weak dissipation . in the quasi - static limit ( @xmath70 ) , the second term vanishes and the efficiency ( [ eq : exact_eta ] ) reaches the carnot efficiency , as expected . when the operation time is too short , the irreversible heat becomes dominant and the net heat becomes negative .
equation ( [ eq : optq ] ) indicates that positive power can be obtained only if @xmath109 this inequality is consistent with the condition of the asymptotic expansion ( [ eq : condition1 ] ) and can thus be satisfied even for a large temperature difference .
so far , we have maximized the power only for the fixed operation times @xmath25 and @xmath29 and the boundary values @xmath32 and @xmath33 of the occupation probabilities .
now we further maximize the power @xmath110 with respect to the operation times .
it is easy to find that the power is a maximum when [ eq : opttau ] @xmath111
this optimization reflects the usual competition with the denominator of the power preferring faster operation whereas the numerator suggests a slower schedule to stay closer to carnot efficiency . for the asymptotic expansion to be valid , the optimal operation times
must satisfy the condition ( [ eq : condition1 ] ) of weak dissipation .
that is , for the process iii the following inequality must be satisfied : @xmath112 this can be achieved in two ways .
the first one corresponds to the usual condition for linear irreversible thermodynamics , @xmath113 .
the alternative is @xmath114 . in this limit ,
our result remains valid even for large temperature differences . with the optimized operation times ( [ eq : opttau ] ) ,
the resulting power is written as a function of @xmath32 and @xmath33 : @xmath115 where @xmath116 the power reaches its maximum when @xmath117 takes a maximum value , @xmath118 at @xmath119 or @xmath120 . at these conditions , the optimal operation time ( [ eq : opttau ] ) and the maximum heat ( [ eq : optq ] ) both vanish .
however , the power remains finite .
this final optimization thus leads to a singular and unrealistic situation .
we note , however , that since eq .
( [ eq : d ] ) does not depend on the system parameters , the efficiency does in fact not depend on this final optimization step .
therefore , we proceed to evaluate efficiency without further reference to optimal occupation probabilities . using the maximum heat ( [ eq : optq ] ) and the optimal time ( [ eq : opttau ] ) , we finally obtain the following remarkable result for the efficiency at the maximum power : @xmath121 with @xmath122 . when @xmath123 , the efficiency ( [ eq : eta_maxp ] ) exactly coincides with the curzon - ahlborn efficiency @xmath124 .
note also that the efficiency is bounded below by @xmath125 for @xmath126 and bounded above by @xmath127 for @xmath128 .
these limits can be realized without violating the condition of weak dissipation .
we have calculated the efficiency @xmath129 at maximum power of a carnot cycle with a single level quantum dot as the operational device .
our calculation is in agreement with known universality properties . in particular , the efficiency at maximum power is equal to half of the carnot efficiency in the regime of linear response , @xmath130 . in the case of a left / right symmetry , corresponding to equal exchange rate coefficients @xmath131 of the dot with the heat reservoirs , the coefficient of the quadratic term is also given by its universal value @xmath11 , @xmath132 .
however , we need to stress that this result was obtained not by an expansion in @xmath10 but in the limit of weak dissipation .
in fact , this calculation adds a new perspective concerning the occurrence of curzon - ahlborn efficiency itself . indeed , in the presence of left
/ right symmetry , the efficiency is actually _ exactly _ equal to the ca efficiency @xmath133 , in the limit of _ weak dissipation_. this limit is reminiscent of the original derivation of ca efficiency , and is in the present model formally similar to the assumption of a linear conduction law between reservoir and quantum dot .
however the concept of weak dissipation is more general .
it remains to be explored whether this observation implies a wider range of validity of ca efficiency .
in particular , it could explain why observed efficiencies at maximum power are not very different from ca efficiency in a wide range of systems under operational conditions far from linear response .
in general , heat absorbed by the system can be written in the following form : @xmath134 where @xmath135 is irreversible entropy production .
since the irreversible entropy production is expected to be small for slow processes , we can expand @xmath135 is an asymptotic series @xmath136 with the presence of @xmath137 or @xmath138 , the irreversible entropy production takes place even at the quasistatic limit and thus the carnot efficiency is not attainable .
here we assume that the irreversible entropy production vanishes at the quasi static limit ( @xmath139 ) .
note that @xmath140 and @xmath141 based on the second law of the thermodynamics . keeping only the first term ,
power is given by @xmath142 which reaches the maximum at @xmath143 the efficiency at the maximum power is then @xmath144 when @xmath145 , the efficiency coincides the curzon - ahlborn efficiency . *
the process is slow enough that only the second order and higher terms are negligible .
( @xmath146 ) * there is no irreversible heat flow at the quasi static limit .
( @xmath139 ) * the coefficients of the asymptotic expansion is the same between two isothermal processes .
( @xmath147 ) f. curzon and b. ahlborn , am .
43 , 22 ( 1975 ) . c. van den broeck , phys .
95 , 190602 ( 2005 ) ; adv .
135 , 189 ( 2007 ) .
m. esposito , k. lindenberg , and c. van den broeck , phys .
lett . * 102 * , 130602 ( 2009 ) .
z. c. tu , j. phys .
a * 41 * , 312003 ( 2008 ) .
m. esposito , k. lindenberg , and c. van den broeck , europhys .
lett . * 85 * , 60010 ( 2009 ) .
b. rutten , m. esposito , b. cleuren , phys . rev .
b * 80 * , 235122 ( 2009 ) .
t. schmiedl and u. seifert , phys .
108301 ( 2007 ) . t. schmiedl and u. seifert , europhys .
lett . * 81 * , 20003 ( 2008 ) .
h. then and a. engel , phys .
e * 77 * , 041105 ( 2008 ) .
a. gomez - marin , t. schmiedl , and u. seifert , j. chem .
* 129 * , 024114 ( 2008 ) .
y. izumida and k. okuda , phys . rev .
e * 80 * , 021121 ( 2009 ) .
y. izumida and k. okuda , prog .
. suppl . * 178 * , 163 ( 2009 ) .
y. izumida and k. okuda , epl * 83 * , 60003 ( 2008 ) .
m. esposito , r. kawai , k. lindenberg and c. van den broeck [ 0909.3618 ] . | we evaluate the efficiency at maximum power of a quantum - dot carnot heat engine .
the universal value of the coefficients at the linear and quadratic order in the temperature gradient are reproduced .
curzon - ahlborn efficiency is recovered in the limit of weak dissipation . |
the casimir effect @xcite is now universally known as one of the most extensively studied manifestations of zero - point oscillations .
it attracts considerable attention in quantum field theory , gravitation and cosmology , atomic physics and optics , condensed matter physics , and in applications to nanotechnology ( see monographs @xcite ) .
the fundamental theory describing both the van der waals and casimir forces between two semispaces with planar boundaries was developed by lifshitz @xcite . at present
the lifshitz theory is generalized for material bodies with curved boundary surfaces ( see , e.g. , refs .
@xcite ) . over a long period of years
only a very limited experimental information about the casimir force was available .
recently , however , a significant advance has been made in the experimental study of this phenomenon reflected in review @xcite . in principle , the theoretical and experimental progress taken together allows computation and measurement of casimir forces between bodies of complicated geometrical shape made of different materials . in order for such computation to be made , one should know the reflection amplitudes on the boundary surfaces at all matsubara frequencies . at first sight this should present no problems because the reflection amplitudes can be either measured directly or calculated using measured material properties , such as frequency - dependent dielectric permittivities . at this point , however , difficulties emerge unexpectedly which are connected with the role of relaxation properties of charge carriers in the casimir effect . beginning in 2000 ,
the influence of relaxation on the thermal casimir force was hotly debated .
bostrm and sernelius @xcite have noticed that the substitution of the dielectric permittivity of the drude model with nonzero relaxation parameter into the lifshitz formula results in a decreasing magnitude of the casimir free energy and force as a function of temperature over a wide region of separations .
this is in contradiction with the case of ideal metal plates or plate materials described by the nondissipative plasma model , where the magnitudes of the casimir free energy and force are monotonously increasing functions of the temperature @xcite . at large separations ( or in high temperature limit ) the magnitudes of the casimir free energy and force between metal plates calculated using the drude model are by a factor of 2 less then the same quantities calculated for ideal metals or metals described by the plasma model .
it was also shown that for metals with perfect crystal lattices the lifshitz theory combined with the drude model violates the nernst heat theorem @xcite . for metals with impurities leading to a nonzero relaxation at zero temperature nernst s theorem
was shown to be followed @xcite .
this , however , does not solve the problem because perfect crystal metal is an idealized model of a truly equilibrium system with nondegenerate ground state for which the laws of thermodynamics must be satisfied . a large thermal effect in the casimir force between metals at separations of a few hundred nanometers predicted by the lifshitz theory combined with the drude model
was experimentally excluded @xcite by indirect dynamic measurements of the casimir pressure in the configuration of a sphere above a plate of a micromachined oscillator .
later this experiment was repeated for two more times with increased precision .
the exclusion of the thermal effect predicted by the drude model was confirmed at a 95% confidence level @xcite and at a 99.9% confidence level @xcite .
the same measurement data were found to be consistent with the lifshitz theory combined with the plasma model .
it is important to keep in mind that the comparison of experiment with theory in refs .
@xcite was based on the use of an approximate method , the so - called _ proximity force approximation _ ( pfa )
@xcite because at that time for sphere - plate configuration an exact theory was not available .
at the moment there is a conceptual possibility to compute the thermal casimir force between a sphere and a plate made of real metals with no use of the pfa @xcite , but the region of experimental parameters is not yet achieved due to computational difficulties . when it is considered that the drude model correctly describes the relaxation of conduction electrons at low frequencies , the contradiction with basic laws of thermodynamics and disagreement with the experimental data outlined above are puzzling .
these problems were dramatized by the demonstration that the inclusion of dc conductivity of dielectric ( or dielectric - type semiconductor ) materials into the model of dielectric response in the lifshitz theory also results in a violation of the nernst theorem @xcite . from the experimental side , it was shown that the measurement data of the experiment on optical modulation of the casimir force between au sphere and si plate with light @xcite excludes the lifshitz theory taking into account the dc conductivity of dielectrics at a 95% confidence level .
a similar result was obtained from the measurement of the casimir - polder force between the bose - einstein condensate of @xmath0rb atoms and sio@xmath1 plate @xcite . here , the lifshitz theory taking dc conductivity of sio@xmath1 into account was experimentally excluded at a 70% confidence level @xcite .
it is pertinent to note that while the comparison of the optical modulation experiment with theory uses the pfa , the measurement results for the casimir - polder force were compared with the exact lifshitz formula for atom - wall interaction .
one can summarize that experiments with metals , semiconductors and dielectrics exclude the influence of dissipation of conduction electrons on the casimir force ( the statement on the opposite in the introduction to ref .
@xcite is a typo ) .
the conflict between such a fundamental theory , as the lifshitz theory , thermodynamics and the experimental data of several experiments is a problem of great concern .
because of this , a lot of attempts to resolve this problem has been undertaken .
specifically , it was even suggested @xcite to modify the lifshitz theory by including into consideration screening effects and diffusion currents .
it was noted , however , that the modifications proposed do not alleviate contradictions with thermodynamics and the experimental data ( discussion on this subject can be found in refs .
@xcite ) . on the other hand ,
it was suggested @xcite to modify the planck distribution law by taking into account `` saturation effects '' .
there were also attempts to soften contradictions with thermodynamics by reformulating the problem @xcite and by finding additional statistical arguments in favor of the drude model @xcite . in ref .
@xcite , in addition to the usually used exponential screening , the so - called _ algebraic screening _ in atom - wall interaction was considered . as a result ,
linear in temperature thermal correction to the casimir - polder force at short separations was predicted similar to that predicted by the drude model .
note that the algebraic screening is connected with nonanalytic terms in the small wavenumber expansion of the dielectric permittivity .
it was finally suggested @xcite that for the resolution of the problem some concepts of statistical physics related to the theoretical description of the interaction of classical and quantum fluctuating fields with matter might need a reconsideration .
the prospects for a pure theoretical resolution of the above problems seem dim at the moment .
in this situation any additional experimental evidence could be very useful . in this paper
we analyze the possibility to measure thermal effects in the casimir force on the basis of already created and used experimental setups . we stress that in the experiments @xcite mentioned above the large thermal correction to the casimir force , as predicted by the drude model ,
was excluded .
however , these experiments were not of sufficient precision to measure the thermal effect for metals predicted by the plasma model or for dielectrics with dc conductivity omitted ( till the moment the thermal effect was measured in the casimir - polder force alone @xcite ) .
keeping in mind that there is some confusion in the literature concerning the use of the pfa , we present two ( not equivalent ) formulations of the pfa and clarify which of them was really used in the comparison between experiment and theory .
we especially analyze the calculation results for the thermal correction to the casimir force in sphere - plate configuration found using the pfa and explain when they are meaningful .
the obtained conclusions are confirmed by the comparison with exact results for the thermal contribution to the casimir force between a sphere and a plate .
we show that at the moment it is not possible to measure small thermal effects in the casimir force ( or its gradient ) in sphere - plate geometry , as predicted by the lifshitz formula combined with the plasma model .
we also discuss the configuration of two parallel plates in both dynamic and static regimes . according to our results , small thermal corrections to the gradient of the casimir pressure in the dynamic regime
is suppressed .
the only way to measure small thermal effects in the casimir pressure is suggested by the configuration of two parallel plates in the static regime .
the paper is organized as follows . in sec .
ii two different formulations of the pfa are considered and applied to the casimir force in sphere - plate configuration .
section iii discusses the same subject with respect to the gradient of the casimir force between a sphere and a plate . in sec .
iv the relationship between the pfa and the exact results is presented for a sphere and a plate made of ideal metal .
the applicability of the pfa to describe thermal corrections to the casimir force between a sphere and a plate made of real metals and the possibility to measure the thermal effect are considered in sec .
v. in sec .
vi we show that the static casimir configuration of two parallel plates is preferential for the observation of a small thermal effect in the casimir pressure .
section vii contains our conclusions and discussion .
different authors vary somewhat in the meaning of the term `` pfa '' .
in fact the term `` proximity force theorem '' ( later changed for pfa ) was introduced in ref .
@xcite where the so - called _ derjaguin method _
@xcite was applied in order to calculate the force acting between curved surfaces by using the known force per unit area of plane parallel plates . in the derjaguin method ,
the unknown force between the elements of curved surfaces is approximately replaced with a known force per unit area of plane surfaces at respective separations . in application to a sphere of radius @xmath2 above a plane surface of a plate @xmath3 the derjaguin method represents the force between them in the form @xmath4 here , @xmath5 is the element of plate area , @xmath6 is the projection of the sphere onto the plate , @xmath7 is the shortest separation between the sphere and the plate , and @xmath8 is the force per unit area of two plane parallel plates at a separation @xmath9 at temperature @xmath10 ( i.e. , the casimir pressure ) . choosing the origin of a cylindrical coordinate system on the plane @xmath3 under the sphere center , the coordinate @xmath9 of any point on the sphere is given by @xmath11 .
then eq .
( [ eq1 ] ) leads to @xmath12 keeping in mind that the thermal casimir pressure is connected with the free energy per unit area of two parallel plates as @xmath13 and integrating by parts in eq .
( [ eq2 ] ) , one arrives at @xmath14 this generalizes eq . ( 20 ) of ref .
@xcite related to the nonretarded case . further simplification of eq .
( [ eq4 ] ) can be achieved when it is assumed that the free energy @xmath15 is a quickly decreasing function of @xmath9 and drops to zero on the characteristic length of about the sphere radius @xmath2 .
let us consider the case of the free energy of the casimir interaction given by the lifshitz formula , @xmath16 here , @xmath17 is the boltzmann constant , the prime near the summation sign multiplies the term with @xmath18 by 1/2 , @xmath19 is the projection of the wave vector on the plane of plates , and @xmath20 where @xmath21 with @xmath22 are the matsubara frequencies .
the reflection coefficients @xmath23 for the two polarizations of the electromagnetic field , transverse magnetic ( @xmath24 ) and transverse electric ( @xmath25 ) , are expressed in terms of the dielectric permittivity along the imaginary frequencies , @xmath26^{1/2}. \nonumber\end{aligned}\ ] ] in order to simplify eq .
( [ eq4 ] ) , we use an expansion in power series in eq .
( [ eq5 ] ) : @xmath27 then from eq .
( [ eq7 ] ) one finds @xmath28 .
\label{eq8}\end{aligned}\ ] ] when @xmath2 goes to infinity with @xmath7 and @xmath10 fixed , the contribution to @xmath29 of the second term in square brackets on the right - hand side of eq .
( [ eq8 ] ) vanishes as @xmath30 .
this means that for large @xmath2 the value of @xmath29 is determined by the first term in square brackets and can be considered as independent on @xmath2 .
hence , rewriting eq .
( [ eq4 ] ) in the form @xmath31 , \label{eq9}\ ] ] one concludes that in the limit of large @xmath2 it holds @xmath32 where @xmath33 is some constant .
let us now consider the behavior of the quantity @xmath34 in the limiting case @xmath35 keeping @xmath2 fixed .
this is a nonrelativistic limit where @xcite @xmath36 with the hamaker constant defined by @xmath37 ^ 2 e^{y}-1\right\}^{-1}. \label{eq11}\ ] ] from eq .
( [ eq10 ] ) it follows @xmath38 when @xmath7 vanishes .
thus eq .
( [ eq9 ] ) can be rewritten as @xmath39 , \label{eq12}\ ] ] where @xmath40 is scarcely affected by the sphere radius @xmath2 .
for all models of dielectric permittivity used to describe metals the magnitude of @xmath40 is less than unity over wide ranges of experimental parameters ( see below the results of numerical computations ) . under the condition @xmath41 which is usually valid in experiments on measuring the casimir force one may neglect the term of order @xmath42 on the right - hand side of eq . ( [ eq12 ] ) and arrives at @xmath43 just this equation was called the `` proximity force theorem '' in ref .
@xcite and heavily used in the comparison of experiment with theory in all measurements of the casimir force in sphere - plate geometry ( see , e.g. , refs .
@xcite ) . as can be seen from the above derivation , eq .
( [ eq13 ] ) follows from eq .
( [ eq1 ] ) under some conditions , but is not equivalent to it .
because of this , to avoid confusion , ref
. @xcite suggested to call ( [ eq1 ] ) the _ most general formulation _ of the pfa and ( [ eq13 ] ) the _ simplified formulation _ of the pfa . keeping in mind that the used in experiments simplified formulation is obtained by disregarding contributions of order @xmath42
, it would be meaningless to attribute physical meaning to any terms of order @xmath42 in the thermal casimir force @xmath44 calculated by using eq .
( [ eq13 ] ) ( see discussion in secs . iv and v )
. one further version of the pfa discussed in the literature @xcite is connected with the choice of parallel surface elements in eq .
( [ eq1 ] ) . in the original derjaguin method @xcite used by us the surface elements representing the sphere are parallel to the surface of the plate @xmath3 . in this case
@xmath6 is a part of the plane @xmath3 ( the so - called _ plate - based _ pfa ) .
if , however , the lower half of the sphere is chosen as @xmath6 @xcite , the plane elements representing the sphere are tangential to it and respective twin elements of the plate are tilted by different angles with respect to the plane @xmath3 ( the so - called _ sphere - based _ pfa ) . in both cases ,
the distance between the elements is measured along the normal to @xmath6 .
it was shown @xcite that both the plate - based and sphere - based pfa lead to coinciding results in the zeroth order of @xmath42 , i.e. , to eq .
( [ eq13 ] ) .
the form of the function @xmath40 in eq .
( [ eq12 ] ) for both versions of the pfa is , however , different .
basing on this , refs .
@xcite considered the results for the casimir energies and forces obtained by pfa as ambiguous .
the differences between the two versions of the pfa were treated as some `` error bars '' inherent to this approximate method .
keeping in mind , however , that the comparison of experiment with theory in sphere - plate geometry is based not on the general formulation of the pfa ( [ eq1 ] ) , but on an unambiguous simplified formulation ( [ eq13 ] ) , the discussion of inherent to pfa errors is immaterial .
in fact , when speaking about the casimir force , only the zeroth order in @xmath42 results in any of the pfa formulations are of physical significance and only under the condition @xmath41 .
the situation changes drastically when , instead to the casimir force , the pfa is applied , for instance , to the gravitational force . according to refs .
@xcite , the most general formulation of the pfa ( [ eq1 ] ) leads to an exact result for the force between a sphere and a plate for all conservative volumetric forces , particularly for the gravitational force .
in so doing , however , only the original derjaguin choice of @xmath6 ( plate - based ) must be used .
this makes the plate - based version of the pfa preferable in comparison with the sphere - based version .
now we present the results of numerical computations for the function @xmath40 defined in eq .
( [ eq12 ] ) in typical regions of experimental parameters .
computations were performed by using the lifshitz formula ( [ eq5 ] ) and eq .
( [ eq8 ] ) for a sphere and a plate made of au .
the dielectric properties of au were described by using three different models . as a crude approximation ,
the model of ideal metal bodies was used leading to @xmath45 , @xmath46 .
a frequently used description obtains the dielectric permittivity @xmath47 by means of the kramers - kronig relation @xmath48 where @xmath49 is taken from tables of the optical data for au @xcite extrapolated to low freuqencies by means of the drude model @xmath50 here , the plasma frequency and the relaxation parameter of au are given by @xmath51ev , @xmath52ev @xcite . as one more alternative description used in the literature
we have applied the generalized plasma - like model @xcite @xmath53 where @xmath54 are the resonant frequencies of the oscillators describing core electrons , @xmath55 are the relaxation frequencies , and @xmath56 are the oscillator strengths .
the values of all these parameters for au can be found in @xcite .
note that there are proposals in the literature for alternative dielectric functions taking into account the effect of spatial nonlocality ( see , for instance , refs .
however , for metallic test bodies these dielectric functions predict precisely the same thermal effect as the drude model approach @xcite . because of this
, they do not require a special consideration here . in fig .
1 the computational results for @xmath40 are shown as a function of ( a ) separation at room temperature @xmath57k and ( b ) temperature at a separation @xmath58 m . in both cases a sphere with @xmath59 m radius
the dotted , dashed and solid lines demonstrate the results obtained using ideal metals , au described by the optical data extrapolated by the drude model and au described by the generalized plasma - like model , respectively . as can be seen in fig .
1(a ) , in the region of separations from 100 nm to @xmath60 m it holds @xmath61 .
it can be easily verified that in the limiting case @xmath35 the function @xmath62 for ideal metals ( the dotted line ) and @xmath63 for real au independently of the model used for its description ( the dashed and solid lines ) . from fig .
1(b ) it is seen that at any temperature from absolute zero to 300k the magnitudes of @xmath40 remain less than unity and the difference between various models of dielectric properties disappears with vanishing temperature .
now we consider the measure of dependence of the function @xmath40 on the sphere radius .
the computational results for @xmath40 as a function of @xmath64 are shown in fig . 2 as the three groups of lines ( dotted , dashed and solid for ideal metals and real au described using the drude and plasma models as explained above )
numerated 1,2 , and 3 for fixed separations @xmath65 , 2 , and @xmath60 m , respectively .
computations are done at room temperature @xmath57k for @xmath42 varying from @xmath66 to 0.05 .
for example , as is seen in fig . 2 , at @xmath67 m ( the group 1 of lines )
there is almost no dependence of @xmath40 on @xmath2 for radii larger than @xmath68 m . for @xmath58 m
( the group 2 of lines ) @xmath40 does not depend on @xmath2 for radii @xmath69 m .
we remind that typical sphere radii are @xmath59 m in the experiments @xcite and @xmath70 m in the experiments @xcite , performed at separations from less than 100 nm to a few hundred nanometers . in fig .
2 it is seen also that the relative magnitude of the function @xmath40 computed using different models of dielectric permittivity of au depends on separation . for example , at @xmath67 m @xmath71 computed using the model of ideal metals ( the dotted line ) is less than using the generalized plasma - like model ( the solid line ) and using the tabulated optical data extrapolated by the drude model ( the dashed line ) . at separations @xmath72 and @xmath60 m the magnitude of @xmath40 computed using the model of ideal metals is sandwiched between lines computed using the two models of real au . the size of possible corrections to the simplified formulation of the pfa ( [ eq13 ] ) has been investigated experimentally by measuring the casimir force between an au - coated plate and five au - coated spheres with different radii using a micromachined oscillator @xcite .
in so doing spheres with radii @xmath73 , 31.4 , 52.3 , 102.8 and @xmath74 m were used .
the obtained constraint @xmath75 for @xmath76 nm , @xmath57k is in very good agreement with the computational results in fig .
thus , the experimental data are in favor of the simplified formulation of the pfa .
in many experiments using the sphere - plate configuration separation distance between the test bodies was varied harmonically ( see , e.g. , refs .
@xcite ) . in this case
not the casimir force but its gradient is the physical quantity immediately connected with the frequency shift of a micromachined oscillator .
the gradient of the casimir force acting between a sphere and a plate can be found by differentiating eq .
( [ eq4 ] ) with respect to @xmath7 and taking into account eq .
( [ eq3 ] ) , @xmath77 using eq .
( [ eq7 ] ) this can be rearranged to the form @xmath78 , \label{eq18}\ ] ] where the following notation is introduced : @xmath79 .
\label{eq19}\end{aligned}\ ] ] similar to sec .
ii , it can easily be shown that at large @xmath2 , with @xmath7 and @xmath10 fixed , it holds @xmath80 . on the other hand , in the nonrelativistic limit @xmath81 where @xmath82 is defined in eq .
( [ eq11 ] ) . from this
it follows that @xmath83 when @xmath7 vanishes and @xmath2 is kept constant .
thus , eq . ( [ eq18 ] ) can be rewritten in an equivalent form @xmath84 , \label{eq21}\ ] ] where @xmath85 is scarcely affected by @xmath2 .
below we demonstrate that in wide ranges of experimental parameters @xmath85 is a very slowly varying function and @xmath86 . because of this , under the experimental condition @xmath41 we can neglect the term of order @xmath42 in eq .
( [ eq21 ] ) and arrive at the equality @xmath87 which is the simplified formulation of the pfa for a gradient of the casimir force in sphere - plate configuration . equation ( [ eq22 ] ) was used for the comparison of experiment with theory in dynamic measurements of the casimir pressure @xcite .
we have performed numerical computations of the quantity @xmath85 as a function of sepation and temperature for a sphere of fixed radius @xmath59 m . in fig .
3(a ) the computational results for @xmath85 as a function of @xmath7 are presented at @xmath57k .
figure 3(b ) shows @xmath85 as a function of @xmath10 at @xmath58 m . in both cases dotted , dashed and
solid lines indicate the use of ideal metal surfaces and au surfaces described by the drude and plasma model approaches . as can be seen in fig .
3(a ) , within the separation region from 100 nm to @xmath60 m the function @xmath88 varies between 0.31 and 0.48 .
when the temperature increases from absolute zero to @xmath57k , @xmath88 remains sandwiched between 0.31 and 0.39 [ see fig .
this demonstrates that possible corrections to the simplified formulation of the pfa ( [ eq22 ] ) do not exceed @xmath42 .
it can be easily seen that the magnitude of @xmath85 almost does not depend on the radius of the sphere . in fig .
4 we present the computational results for @xmath85 as a function of @xmath64 at @xmath57k .
the groups of lines numbered 1 and 2 are for the separations @xmath65 and @xmath60 m , respectively . the dotted , dashed and solid lines have the same meaning as in figs .
13 . as is seen in fig . 4
, for the group of lines 1 there is no dependence on @xmath42 over the whole range from @xmath66 to 0.05 .
thus , at @xmath67 m , @xmath85 does not depend on @xmath2 for @xmath89 m . for the group of lines 2 , @xmath85 does not depend on @xmath2 for @xmath90 . at @xmath91 m this leads to @xmath92 m . from fig .
4 it also follows that relative magnitudes of the function @xmath85 computed using different models of dielectric properties of metal depend on separation . the experimental constraint on the magnitude of the correction to eq .
( [ eq22 ] ) was obtained in ref .
@xcite by using several spheres with different radii ( see sec .
ii ) . in the separation region
@xmath76 nm at @xmath57k it was shown that @xmath93 at a 95% confidence level .
this is in very good agreement with the computational results in fig . 4 and provides the experimental confirmation for the simplified formulation of the pfa in application to the gradient of the thermal casimir force between a sphere and a plate .
in what follows we apply the pfa to calculate the thermal casimir force between a sphere and a plate made of ideal metal and discuss the possibility to describe the thermal correction using this approximate method . using the notation ( [ eq8 ] ) , eq .
( [ eq4 ] ) for the thermal casimir force can be presented in the form @xmath94 for ideal metals , using the dimensionless variables @xmath95 in eq .
( [ eq8 ] ) , the quantity @xmath96 is given by @xmath97 where @xmath98 is a polylogarithm function . equation ( [ eq25 ] ) can be presented in an equivalent form convenient for the transition to the low - temperature limit using the poisson summation formula @xcite . according to this formula ,
if @xmath99 is the fourier transform of a function @xmath100 , i.e. @xmath101 then it follows that @xmath102 now we return to eq .
( [ eq25 ] ) and put @xmath103 keeping in mind that @xmath104 and using eq .
( [ eq26 ] ) , one obtains @xmath105 here , we have introduced a new variable @xmath106 and the notation @xmath107 , where @xmath108 is the effective temperature related to the separation distance between the sphere and the plate .
note that in terms of this notation it holds @xmath109 .
taking into account that the quantity @xmath110 is an even function of its argument and using eqs .
( [ eq27 ] ) and ( [ eq29 ] ) , we arrive at @xmath111 where we exchanged the order of integrations with respect to @xmath112 and to @xmath113 . equation ( [ eq30 ] ) can be represented in the form @xmath114 , \label{eq31}\ ] ] where @xmath115 direct calculation leads to @xmath116 where @xmath117 .
substituting eq .
( [ eq33 ] ) into eq .
( [ eq31 ] ) , we obtain @xmath118 .
\nonumber\end{aligned}\ ] ] in the low - temperature limit it holds @xmath119 , @xmath120 and eq .
( [ eq34 ] ) results in @xmath121 , \nonumber\end{aligned}\ ] ] where @xmath122 is the riemann zeta function . the quantity @xmath123 in eq .
( [ eq23 ] ) is given by eq .
( [ eq34 ] ) where @xmath7 is replaced with @xmath124 , @xmath125 is replaced with @xmath126 and @xmath127 is defined from @xmath128 the sphere radius introduces a second characteristic temperature into the problem which is much less than @xmath129 in experimentally relevant situations . as an example , for typical spheres used in experiments on measuring the casimir force @xmath59 m and @xmath130 nm resulting in @xmath131k and @xmath132k .
thus , for experiments performed at room temperature , @xmath57k , it holds @xmath119 , but @xmath133 , i.e. , a high - temperature regime with respect to the sphere radius , but a low - temperature regime with respect to the separation between the sphere and the plate . nevertheless , we begin with the case of extremely low temperature with respect to both parameters , @xmath134 , which is achieved only well below 1k . in this case
the quantity @xmath123 is also given by eq .
( [ eq35 ] ) where @xmath7 is replaced with @xmath124 and @xmath129 is replaced with @xmath127 .
we also take into account that under the condition @xmath119 the free energy between two plane parallel plates in eq .
( [ eq23 ] ) is given by @xcite @xmath135 .
\label{eq37}\ ] ] substituting eq .
( [ eq37 ] ) and eq .
( [ eq35 ] ) ( with the characteristic temperatures @xmath129 and @xmath127 ) into eq .
( [ eq23 ] ) , we arrive at the result @xmath136 . \label{eq38}\ ] ] here , we have included the main @xmath10-independent term of order @xmath42 and the first nonvanishing term depending on temperature .
the latter is of order @xmath137 and is multiplied by the third power of the small parameter @xmath42 . for typical experimental parameters mentioned above @xmath138 .
if one puts @xmath139k [ the highest temperature where eq .
( [ eq38 ] ) is applicable ] the @xmath10-dependent term on the right - hand side of eq .
( [ eq38 ] ) appears to be of order @xmath140 .
equation ( [ eq38 ] ) can be identically rewritten in the form @xmath141 where the thermal correction to the casimir force is given by @xmath142 it is interesting to compare this result obtained using the general formulation of the pfa with the exact result for the thermal correction in sphere - plate configuration @xcite . in the functional determinant representation
the exact free energy of the casimir interaction in sphere - plane geometry can be written as @xmath143 , \label{eq41}\ ] ] where the explicit form of the matrix @xmath144 in ideal metal case is contained in ref .
@xcite . in the limit of extremely low temperature , @xmath145 , and small separation distances , @xmath41 , ref .
@xcite arrives at the following exact result for the temperature - dependent part of the free energy @xmath146 from this equation , the exact thermal correction to the casimir force in the limiting case under consideration is the following @xmath147 it can be seen from the comparison of eqs .
( [ eq40 ] ) and ( [ eq43 ] ) that although the most general formulation of the pfa gives the correct dependence of the thermal correction on the fourth power of @xmath10 , the numerical coefficient before it is underestimated by a factor of 4.48 . at the same time , the application of the simplified formulation of the pfa ( [ eq13 ] ) in this limiting case results in an incorrect dependence of the thermal correction on @xmath10 ( the third power of @xmath10 instead of the fourth ) .
these results are not surprising . as explained in sec .
ii , the pfa is applicable only under the condition @xmath41 and leads to reliable predictions which are of zeroth order in the small parameter @xmath42 .
because of this , any contribution either to the free energy or to the force which relative magnitude is numerically of about @xmath42 or smaller calculated using the pfa must be disregarded as physically meaningless . keeping in mind that in the limiting case of extremely low temperature , @xmath145 , the relative magnitude of the thermal correction is much less than @xmath42 , the thermal effect in the casimir force in this case
should be considered as unobservable .
a completely different type of situation occurs under the condition @xmath133 which is well satisfied in all experiments on measuring the casimir force performed at room temperature ( we remind that for the sphere of @xmath59 m radius @xmath131k ) . here ,
irrespective of whether the inequality @xmath119 or @xmath148 is valid , the exact expression ( [ eq41 ] ) leads in the zeroth order of the small parameter @xmath42 to the simplified formulation of the pfa for the thermal correction to the casimir force between a sphere and a plate @xcite : @xmath149 in the case @xmath119 the thermal correction to the casimir free energy for two plane parallel plates , @xmath150 , is contained in eq .
( [ eq37 ] ) : @xmath151 thus , under a condition that @xmath133 the simplified formulation of the pfa is applicable not only to the total quantities @xmath152 and @xmath153 [ see eq .
( [ eq13 ] ) ] , but separately to the zero - temperature and thermal contributions to these quantities as well .
this invalidates the statement of ref .
@xcite that for a sphere above a plate at @xmath154 the pfa is inapplicable to each contribution alone made on the basis of the worldline approach .
the measurement data of all experiments performed in the configuration of a rather large sphere in close proximity to a plane plate ( i.e. , under the condition @xmath41 ) were compared with the theory using the pfa . in static experiments ( e.g. , in refs .
@xcite ) eq . ( [ eq13 ] ) was used for this purpose . in dynamic experiments @xcite the form of the pfa in eq .
( [ eq22 ] ) was employed . as was explained in secs .
ii and iii , in both cases terms of order @xmath42 , as compared with unity , were disregarded .
this raises the question if this neglection imposes some constraints on the possibility to observe thermal effects in the casimir force .
below we analyze this question using both theoretical approaches to the calculation of thermal casimir force discussed in sec .
we start with the drude model approach using the tabulated optical data @xcite for @xmath49 extrapolated to low frequencies by means of eq .
( [ eq15 ] ) .
the thermal effect can be characterized by the relative thermal correction to the casimir force @xmath160 and the casimir pressure @xmath161 we compute the relative correction ( [ eq48 ] ) by eqs .
( [ eq5 ] ) and ( [ eq13 ] ) over the separation region from 0.1 to @xmath60 m . the relative correction ( [ eq49 ] ) is computed over the same separation region by using eqs .
( [ eq3 ] ) and ( [ eq5 ] ) .
we remind that the measurement data for the quantity @xmath162 in dynamic experiments are obtained by using eq .
( [ eq22 ] ) from the gradient of the casimir force @xmath163 which in its turn was recalculated from the measured frequency shift .
thus , in static experiments using sphere - plate configuration , the pfa in the form of eq .
( [ eq13 ] ) is part of the theory , whereas in dynamic experiments the pfa in the form of eq .
( [ eq22 ] ) allows to convert the experimental data for the force gradient into the data for the equivalent casimir pressure . in fig .
5 we present the computational results in percents for the relative thermal corrections @xmath164 ( the solid line 1 ) and @xmath165 ( the solid line 2 ) at @xmath57k as a function of separation in the region ( a ) from 0.1 to @xmath60 m and ( b ) from 0.1 to @xmath159 m . in the same figure ,
the short - dashed and long - dashed lines show the quantity @xmath42 in percents taken with the same sign as the thermal corrections as a function of separation for @xmath166 and @xmath167 m , respectively .
these lines demonstrate the size of typical errors arising from the use of the pfa .
as can be seen in fig . 5 , the relative thermal corrections for both the casimir force and equivalent casimir pressure are rather large . in the region from 1 to @xmath168 m they take negative values and achieve 35% and 47% , respectively . within the separation region from 0.1 to @xmath159 m the magnitudes of thermal corrections to the force and to the pressure increase from 1.5% and 0.7% to 23% and 16% , respectively . in the same separation region
the error due to the use of the pfa varies from 0.1% to 1% for the sphere of @xmath169 m radius and from 0.07% to 0.7% for the sphere of @xmath167 m radius .
thus , the use of the pfa in the drude model approach does not impose any constraints on the possibility to measure large thermal effects predicted in this approach .
the answer to the question whether or not the thermal corrections to the casimir force in a sphere - plate configuration can be measured depends not only on the errors due to the application of the pfa , but also on the size of the total experimental error .
thus , in the static experiment @xcite performed by means of an atomic force microscope the total relative experimental error of the casimir force @xmath44 determined at a 60% confidence level at separations 100 and 200 nm is equal to 3.5% and 22.4% , respectively @xcite .
these are larger than the relative thermal correction @xmath164 in fig .
because of this , in the experiment of refs .
@xcite the predicted thermal correction can not be either confirmed or excluded .
a completely different type of situation occurs in dynamic experiments of refs .
@xcite performed by means of a micromachined oscillator .
thus , in the most precise experiment of refs .
@xcite the total relative experimental error of the casimir pressure @xmath170 determined even at a higher , 95% , confidence level varies from 0.19% to 0.9% and to 9% when separation increases from 160 to 400 and to 750 nm , respectively . as can be seen in fig .
5(b ) , the magnitude of the relative thermal correction @xmath165 remains much larger than the total experimental error over the entire range of experimental separations .
that is the reason why , when no evidence of the predicted thermal corrections was found , refs .
@xcite arrived at the conclusion that the drude model approach is experimentally excluded at a 95% confidence level ( within a narrower separation region from 210 to 620 nm this approach was excluded at a 99.9% confidence level @xcite ) .
it has been speculated in the literature that the experimental exclusion of the drude model approach might be not warranted because of some theoretical and experimental uncertainties .
specifically , it was stressed @xcite that the drude parameters @xmath171 and @xmath172 used for the extrapolation of the optical data to low frequencies may vary . besides , ref .
@xcite surmised that measurements of absolute separations in ref .
@xcite ( which data were also used to exclude the drude model approach ) contain an unaccounted systematic error which could bring the data away from the theory .
here , we make a test of both these opportunities with respect to the theory - experiment comparison in refs .
@xcite . in fig .
6 the differences between the theoretical casimir pressures computed using the drude model approach , as explained above but with account of surface roughness , and the mean experimental casimir pressures measured in refs .
@xcite are indicated as dots .
the theoretical pressures for the lower set of dots are computed with the conventional drude parameters of sec .
ii ( @xmath51ev , @xmath52ev ) .
the theoretical casimir pressures for the upper set of dots are computed with the alternative drude parameters @xcite , @xmath173ev , @xmath174ev , where the value of the plasma frequency is most different from the conventional one .
the differences @xmath175 with @xmath176 computed using all other alternative drude parameters listed in ref .
@xcite are sandwiched between the two sets of dots shown in fig .
the solid lines in fig .
6 indicate the borders of the confidence intervals @xmath177 $ ] computed in ref .
@xcite at a 95% confidence level . by the construction of @xmath177 $ ]
, 95% of dots must belong to these intervals .
however , as is seen in fig . 6 , almost all dots in both sets lie outside the solid lines demonstrating that the theoretical approach using the drude model is excluded by the data at a 95% confidence level over the entire measurement range ( as was mentioned above , within a bit narrower reparation region the exclusion at a 99.9% confidence level holds ) .
it is pertinent to note also that the use of optical data of ref .
@xcite for the first absorption bands ( which are slightly different from the data of ref .
@xcite used in our computations ) does not change this conclusion .
now we consider the possibility to bring the experimental data in agreement with the drude model approach by assuming that there is an unaccounted systematic error in the determination of absolute separations .
note that this systematic error ( if any ) is separation - independent .
this is because in the setup used the differences between the values of separations where the casimir pressures were measured are fixed interferometrically to high precision ( see ref .
@xcite for details ) . to find absolute separations
, one should know the initial absolute separation which is determined from the electrostatic calibrations .
if one admits that electrostatic calibrations contain some uncertainty , the initial separation would be burdened with some unaccounted constant systematic error which is translated to all separations . in an attempt to place the dots within the limits of the confidence intervals , in fig .
7(a ) we plot the differences @xmath175 with two sets of the drude parameters mentioned above as a function of separation , but with all separation distances decreased by @xmath178 nm .
( note that the increase of separations results in even larger deviations of dots from the confidence intervals than in fig .
this corresponds to the assumption that an unaccounted systematic error in the measurement of separation distances in refs .
@xcite is equal to 1 nm ( remind that in actual fact the total error in the measurement of separations in refs .
@xcite is equal to 0.6 nm ) . as can be seen in fig .
7(a ) , the shift of separations for 1 nm does not furnish the desired result for the lower set of dots at separations above 230 nm and is not helpful for the upper set of dots at any separation .
the decrease of all separations for 3 nm [ see fig .
7(b ) ] expels the lower set from the confidence intervals at separations below 230 nm , but not yet includes these dots into the confidence interval at @xmath179 nm . in fig .
7(c ) we illustrate the largest decrease of separations for 6 nm . here , all dots related to the lower set are still outside the confidence intervals at @xmath180 nm , but the dots of both sets are already outside of them at shortest separations .
note that similar to fig .
6 the differences @xmath175 computed with all other alternative drude parameters are sandwiched between the two sets of dots shown in each of figs .
7(a ) , 7(b ) , and 7(c ) .
the same is correct when the optical data ( not just the drude parameters ) from different sources are used . as shown in ref .
@xcite , the use of some alternative optical data instead of the data of ref .
@xcite would decrease the magnitudes of theoretical casimir pressures and , thus , only increase discrepances between the predicted and experimental casimir pressures .
this means that no systematic error in the measurement of absolute separations can bring the experimental data of refs .
@xcite in agreement with the drude model approach using any values of drude parameters and any sets of the optical data .
from the above it follows that the sensitivity of dynamic experiments by means of micromachined oscillator is quite sufficient to registrate the thermal correction to the casimir pressure , as predicted by the drude model approach . what is more , the application of the pfa in the form of eq .
( [ eq22 ] ) in this case is warranted because both the total quantities @xmath181 and @xmath170 and contained in them thermal corrections are much larger than @xmath42 for the experimental parameters .
the lack of any observation effect for the thermal correction to the casimir pressure , as predicted by the drude model approach , means that this approach is experimentally inconsistent .
we are coming now to the question is it possible to observe small thermal effect in the casimir force , as predicted by the plasma model approach in sphere - plate configuration . in this case
the relative thermal correction to the force , @xmath182 , and to the pressure , @xmath183 , are expressed once again by eqs .
( [ eq48 ] ) and ( [ eq49 ] ) where @xmath184 and @xmath185 are replaced with @xmath186 and @xmath187 computed using the dielectric permittivity of the generalized plasma - like model ( [ eq16 ] ) . in fig .
8 the computational results ( in percents ) are presented for @xmath182 ( the solid line 1 ) and for @xmath183 ( the solid line 2 ) at @xmath57k as a function of @xmath7 in the region ( a ) from 0.1 to @xmath60 m and ( b ) from 0.1 to @xmath159 m .
the short - dashed and long - dashed lines show the quantity @xmath42 ( in percents ) for @xmath166 and @xmath167 m , respectively . as can be seen in fig . 8 , the thermal corrections computed using the plasma model approach are positive and increase monotonously with the increase of separation . at @xmath188
m their magnitudes are much less than the magnitudes of the thermal corrections computed using the drude model approach ( see fig . 5 ) . as a result , @xmath182 lies below the short - dashed line related to the static experiments .
the relative correction to the casimir pressure , @xmath183 , lies below the long - dashed line related to the dynamic experiments at @xmath189 m .
this means that one can not attribute physical meaning to such small values of the thermal corrections @xmath182 and @xmath183 computed using the pfa [ eqs .
( [ eq13 ] ) and ( [ eq22 ] ) ] which disregards contributions to the force and to the pressure of order @xmath42 .
the magnitudes of the thermal corrections @xmath182 and @xmath183 are markedly larger than @xmath42 shown by the short - dashed ( long - dashed ) lines only at separations @xmath190 m ( @xmath191 m ) .
however , either in the static experiment @xcite on measuring the casimir force @xmath44 or in the most precise dynamic experiment @xcite on measuring the casimir pressure @xmath170 , over the entire separations regions @xmath192 m and @xmath193 m , the magnitudes of @xmath182 and @xmath183 fall far short of the total relative experimental error .
it is not surprising , then , that thermal effects were not observed in these experiments .
one more possibility for future measurements is to use spheres of larger radius .
this allows to make @xmath42 smaller and , thus , make thermal corrections @xmath182 and @xmath183 computed using the pfa physically meaningful over the entire range of separations @xmath192 m .
in addition , in static experiments the casimir force @xmath44 is larger for a sphere of larger @xmath2 resulting in a smaller relative experimental error . for spheres of centimeter - size radii , however , inavoidable deviations from perfect spherical shape is a problem of great concern which prevents accurate electrostatic calibration @xcite .
it is therefore unlikely that small thermal effect in the casimir force , as predicted by the plasma model approach , will be observed in the configuration of a sphere above a plate .
we conclude this section with a brief discussion of recent exact results for the thermal casimir force between a sphere and a plate described by simple plasma and drude models without account for interband transitions of core electrons @xcite and by the generalized plasma - like and drude - like models @xcite .
the key question is whether or not these results support computations performed using the pfa . as was discussed in sec .
iv , for ideal metals the exact results in the zeroth order in @xmath42 coincide with the pfa if @xmath133 .
this is the case for both the total casimir force or casimir pressure and separately for the thermal corrections .
unfortunately , the exact computations of refs . @xcite were performed in regions far away from the values of experimental parameters and outside the region where the pfa is applicable .
for example , in ref .
@xcite the sphere radii were chosen to be @xmath194 m and computations were performed at separations from 0.5 to @xmath195 m . for a sphere radius @xmath196 m the computations of ref .
@xcite were made from @xmath197 to @xmath195 m .
thus , in all cases considered it holds @xmath198 .
remind that for the experimental parameters of refs .
@xcite @xmath42 varies from 0.00063 to 0.003 and for the parameters of refs .
@xcite from 0.0011 to 0.005 .
according to the results presented in secs .
ii and iii , the pfa in the form of eqs .
( [ eq13 ] ) and ( [ eq22 ] ) is applicable only at @xmath199 . because of this , the speculation of ref .
@xcite that at small separations the drude and plasma models lead to casimir force values much closer than predicted by the pfa made on the basis of computations in the region of @xmath200 is unjustified .
the computational results of ref .
@xcite at @xmath201 m , @xmath57k clearly demonstrate that deviations between the exact theory and the pfa vanish with decreasing @xmath42 .
thus , from fig . 3 of ref .
@xcite it can be seen that relative deviation of the quantity @xmath202 computed exactly from that computed using the pfa decreases from 9.2% to 2.5% when the sphere radius increases from @xmath203 m ( @xmath204 ) to @xmath205 m ( @xmath206 ) .
in such a manner exact computations confirm that the computational error in the ratio @xmath202 arising from using the pfa for real metals ( 0.025 for @xmath206 ) is smaller that @xmath42 .
for the experimental values of @xmath42 mentioned above this error would be much less than the total experimental error .
because of this it can be stated with certainty that the comparison of the performed experiments with the exact theory instead of the pfa ( when the exact will become available ) will not lead to any changes with respect to already obtained conclusions .
both papers @xcite claim that for large separations in the sphere - plane geometry the drude model leads to a force a factor of 3/2 smaller than the plasma model ( instead of a factor of 2 as it holds for two parallel metal plates ) . in such a general form
this formulation is , however , somewhat misleading as it does not cover all relevant limiting cases .
in fact it is correct only in the case when @xmath207 @xcite , i.e. , outside the application region of the pfa . if , however , we consider large separations @xmath208
, i.e. , @xmath148 , and large sphere @xmath209 ( which means that the pfa is applicable ) , then the exactly calculated ratio @xmath210 . the same result is obtained using the pfa . as shown in ref .
@xcite by means of exact computations , there is one more limiting case , @xmath211 , @xmath212 and @xmath148 , where @xmath213 .
this result is quite natural because a small sphere above a plate can be modelled as an atom characterized by some effective dynamic polarizability @xcite . keeping in mind that at large separations only the static polarizability is relevant which is common for the plasma and drude models , both the free energies and forces computed using these models coincide in the limit of large @xmath7 .
thus , in sphere - plate geometry the ratio @xmath214 can take different values , 3/2 , 2 , and 1 , in the large distance limit depending on the parameters of the problem . in the experimental situations ,
however , where @xmath41 and @xmath215 , in the limit of large distances , @xmath216 , it holds @xmath210 like for two parallel plates . in ref .
@xcite the exact computations of the free energies @xmath217 and @xmath218 were performed for a sphere of @xmath205 m radius at @xmath57k and 77k within the separation region from 0.265 to @xmath219 m .
this corresponds to the values of @xmath42 varying from 0.053 to 19 .
once again , only near the smallest @xmath42 considered the pfa becomes applicable , whereas for the comparison with recent experiments more than one order of magnitude smaller ratios @xmath42 are required . at the shortest separation @xmath220 m ( @xmath221% ) from fig . 5 of ref .
@xcite one obtains the exact value for the ratio @xmath222 at @xmath57k .
the same ratio computed using the pfa is equal to 1.14 leading to a relative 2.6% deviation between the pfa and exact results which is less than one half of @xmath42 .
this confirms that for real metals in sphere - plate geometry the pfa results for both the thermal casimir force and free energy go to the respective exact results with decreasing @xmath42 .
the original casimir configuration of two parallel plates is the only one where the comparison of experiment with theory does not require either the pfa or more sophisticated exact computational methods which can be effectively used so far solely in some restricted ranges of parameters .
the only experiment of this kind was performed @xcite in the modern stage of the casimir force measurements .
this is an experiment of dynamic type where one of the plates was oscillating at the natural oscillator frequency . as a result ,
the shift of this frequency under the influence of the casimir pressure was measured and recalculated into the gradient of the casimir pressure .
the experiment @xcite reported 15% measure of agreement between the data and theory using the model of ideal metals ( see review @xcite for details ) .
the gradient of the casimir pressure @xmath223 is obtained from eqs .
( [ eq3 ] ) and ( [ eq5 ] ) by differentiation with respect to separation @xmath224 the thermal correction to the gradient of the casimir pressure can be characterized by the quantity @xmath225 we will add additional indices @xmath226 , @xmath227 and @xmath228 and calculate the quantities @xmath229 , @xmath230 and @xmath231 depending on the used model of dielectric properties of metal described in sec .
the computational results for the relative thermal correction ( [ eq51 ] ) to the gradient of the casimir pressure as a function of separation at @xmath57k are presented in fig .
, the separation region from 0.5 to @xmath60 m is considered because at shorter separations it would be hard to experimentally keep the plates parallel .
the dashed , solid , and dotted lines represent results computed by using the tabulated optical data extrapolated to low frequencies by the drude model with conventional parameters , the generalized plasma - like model , and the model of ideal metals , respectively . as can be seen in fig .
9 , for ideal metal plates and plates described by the generalized plasma - like model the relative thermal correction to the gradient of the casimir pressure monotonously increases with the increase of separation . at separations @xmath232 m , where the experiment is feasible , @xmath233% and @xmath234% .
so small magnitudes of @xmath235 make it impossible to observe the thermal corrections due to the generalized plasma - like model in the measurements of the gradient of the casimir pressure . by contrast , the relative thermal correction @xmath236 computed using the drude model approach decreases monotonously with increase of separation from 0.5 to @xmath237 m where it achieves the minimum value of 53.3% . at @xmath238 m it holds @xmath239% .
thus , the proposed experiment on measuring the gradient of the casimir pressure @xcite might provide one more confirmation for the exclusion of the drude model approach , but is not well adapted for the measurement of the thermal correction , as predicted by the generalized plasma - like model .
note that according to the authors of ref .
@xcite their effort to measure the casimir force is not yet successful due to the use of al surfaces .
the point is that al is the subject of quick oxidation even in high vacuum .
for this reason , in the first precise measurement of the casimir force @xcite , al surfaces were coated with thin transparent layers of au / pd .
the use of such layers , however , complicates the comparison between experiment and theory . as a result , starting from 2000 almost all experiments on the casimir force between metallic test bodies @xcite used au coated surfaces .
presently another setup with parallel au surfaces , instead of al , is under construction @xcite .
it can be shown also analytically that measurements of the gradient of the casimir pressure are unsuitable for the detection of small thermal corrections .
for example , for ideal metal plates in the limit of low temperatures @xmath119 it holds @xcite @xmath240 .
\label{eq52}\ ] ] taking into account that the second contribution on the right - hand side of eq .
( [ eq52 ] ) does not depend on separation , the main contribution to the gradient of the casimir pressure in the low - temperature limit takes the form @xmath241 as can be seen on the right - hand side of eq .
( [ eq53 ] ) , there are no terms in powers of @xmath242 and the thermal correction is exponentially small . equation ( [ eq53 ] ) reproduces the exactly calculated @xmath243 at @xmath57k with an error of less than 1% up to the separation distance @xmath238 m .
the absence of power - type thermal corrections in the gradient of the casimir pressure for ideal metals explains why the thermal effect in this case is so small . at low temperatures , the analytical representation for the gradient of the casimir pressure
can be obtained for real metals as well .
thus , for metals described by the generalized plasma - like model at separations above @xmath244 m considered here the influence of interband transitions can be neglected .
then the thermal correction in the gradient of the casimir pressure is obtained as the negative second derivative of eq .
( [ eq46 ] ) leading to @xmath245 , \label{eq54}\end{aligned}\ ] ] where the exponentially small terms are omitted . here
, the leading thermal correction is of power - type , but it contains the first and third power of small parameters . equation ( [ eq54 ] ) provides a better than 1% approximation for the exactly computed @xmath246 over the range of separations below @xmath247 m .
similar low - temperature results for metals described by the drude model can be obtained by calculating the negative second derivative of eq .
( [ eq47 ] ) @xmath248 this approximate expression reproduces the values of @xmath249 with an error less than 1% at separations below @xmath159 m . to conclude this section , for the configuration of two parallel plates in the dynamic regime the relative thermal correction to the gradient of the casimir pressure computed using the plasma model approach achieves 20% only at @xmath91 m ( see fig .
the same relative size of the thermal correction computed using the plasma model but to the casimir pressure is achieved at a shorter separation @xmath250 m [ see the solid line 2 fig .
thus , the static experiment with two parallel plates of sufficiently large area might be preferable for the registration of small thermal corrections at relatively large separations . although at the same separation the relative thermal correction to the casimir force in a sphere - plate configuration is larger [ 43% in accordance with the solid line 1 in fig .
8(a ) at @xmath250 m ] , small values of forces for spheres of relatively small radii make its measurement impossible .
as was discussed above , the use of centimeter - size spheres meets difficulties due to unavoidable deviations from sphericity . in this respect
the increase of plate area and , thus , the increase of the casimir force , may make possible to decrease the experimental error at separations of a few micrometers to about 10% , i.e. , below the characteristic size of small thermal effect predicted by the plasma model approach .
in the foregoing we have analyzed the possibility to measure thermal effects in the casimir force between metals in the configurations of a sphere above a plate and two parallel plates . in connection with the sphere - plate configuration ,
a detailed study of the reliability of the pfa to describe the thermal casimir force was performed .
we have considered the two formulations of the pfa , the most general , derjaguin , formulation @xcite and the simplified formulation of ref .
@xcite which was used for the comparison of many experiments on the measuring of the casimir force with theory .
we show in sec .
ii that for a large sphere , i.e. , under the condition @xmath41 , the magnitude of the relative difference between thermal casimir forces @xmath44 computed for real metals by using the two formulations of the pfa is less than @xmath42 .
this holds regardless of what approach to the description of charge carriers ( drude or plasma model ) is employed .
a similar result is obtained in sec .
iii for the gradient of the thermal casimir force in sphere - plate configuration , @xmath181 . here
again the relative difference computed using the two formulations is less than @xmath42 .
the obtained results are compared with the experimental data of ref .
@xcite demonstrating that for large spheres of different radii at @xmath57k the relative differences of the measured @xmath251 , @xmath181 and those computed using the simplified formulation of the pfa are also less than @xmath42 . from this
we can conclude that the simplified formulation of the pfa is a good approximation for the calculation of thermal casimir forces in sphere - plate configuration for spheres of large radii . in sec .
iv special attention is paid to the case of an ideal metal sphere above an ideal metal plate . in this case a simple analytic expression for the difference between the two formulations of the pfa is obtained . the thermal correction to the casimir force computed using the pfa is considered both in the limiting case of very low temperatures , @xmath145 , and in the case @xmath133 .
the latter includes the region of parameters where all measurements of the casimir force in sphere - plate configuration have been performed .
the results for the thermal correction obtained using the pfa are compared with available exact results .
it is emphasized that in the experimental region @xmath133 the exact results of ref .
@xcite in the zeroth order of @xmath42 coincide with the simplified formulation of the pfa .
this provides a theoretical basis for the application of the pfa in the measurements of the thermal casimir force for the comparison between experiment and theory .
the applicability of the pfa and the possibility to observe the thermal correction to the casimir force in the configuration of a sphere above a plate made of real metals are discussed in sec .
v. here , the relative thermal corrections to the casimir force and to the casimir pressure are computed in the framework of both the drude model and the plasma model approaches .
the obtained results are compared with the error introduced from the use of the pfa and with typical experimental errors of force and pressure measurements .
new evidence is provided that some unaccounted systematic errors in the experiment of refs .
@xcite can not bring the measurement results in agreement with the prediction of the drude model approach .
it is argued that the configuration of a sphere above a plate is well suited to exclude the large thermal correction derived within the drude model approach but scarcely can be used for the experimental observation of small thermal effects , as predicted from the plasma model approach .
the reason is that the respective thermal correction @xmath252 is several times larger than the error of pfa , @xmath42 , only at sufficiently large separations , where the error of force or equivalent pressure measurements exceeds @xmath252 .
the exact computational results for the thermal correction to the casimir force between real metals described by the drude and plasma models are compared with the pfa results .
it is shown that the region of parameters where the exact results are available is outside the area of application of the pfa .
we demonstrate that at relatively large separations the ratio @xmath214 can go to different limiting values ( 2 , 3/2 or 1 ) depending on the relationship between the parameters @xmath7 , @xmath2 , @xmath253 and @xmath10 .
this generalizes the results of refs .
the performed comparison enables us to conclude that the available exact results for real metals converge to the pfa results when @xmath254 . for the experimental values of @xmath42 varying from 0.00063 to 0.005 the errors due to the use of the pfa instead of the exact methods are much less than the experimental error of force and pressure measurements in all experiments performed to the present day . in the configuration of two parallel plates ,
we have calculated the relative thermal correction to the gradient of the casimir pressure which is measured in the dynamic regime ( sec .
vi ) . we show that a small thermal effect , as predicted in the plasma model approach , is additionally suppressed in the gradient of the casimir pressure .
because of this , the original casimir configuration of two parallel plates of sufficiently large area used in the static regime remains the most prospective for measurement of small thermal effect in the casimir force .
the authors acknowledge helpful discussions with m. bordag . g.l.k . and v.m.m .
are grateful to the institute for theoretical physics , leipzig university , where this work was performed , for kind hospitality .
they were supported by deutsche forschungsgemeinschaft , grant no .
was also supported by the grant of the russian ministry of education p184 .
99 h. b. g. casimir , proc .
akad . wet . *
51 * , 793 ( 1948 ) .
w. milonni , _ the quantum vacuum _ ( academic press , san diego , 1994 ) . v. m. mostepanenko and n. n. trunov , _ the casimir effect and its applications _
( clarendon , oxford , 1997 ) .
m. krech , _ the casimir effect in critical systems _ ( world scientific , singapore , 1994 )
. k. a. milton , _ the casimir effect _ ( world scientific , singapore , 2001 )
. m. bordag , g. l. klimchitskaya , u. mohideen , and v. m. mostepanenko , _ advances in the casimir effect _
( oxford university press , oxford , 2009 ) .
e. m. lifshitz , zh .
* 29 * , 94 ( 1956 ) [ sov .
jetp * 2 * , 73 ( 1956 ) ] .
i. e. dzyaloshinskii , e. m. lifshitz , and l. p. pitaevskii , usp .
nauk * 73 * , 381 ( 1961 ) [ adv .
* 10 * , 165 ( 1961 ) ] .
t. emig , n. graham , r. l. jaffe , and m. kardar , phys .
d * 77 * , 025005 ( 2008 ) .
o. kenneth and i. klich , phys .
b * 78 * , 014103 ( 2008 ) .
g. l. klimchitskaya , u. mohideen , and v. m. mostepanenko , rev .
phys . * 81 * , 1827 ( 2009 ) .
m. bostrm and b. e. sernelius , phys .
lett . * 84 * , 4757 ( 2000 ) .
m. bordag , b. geyer , g. l. klimchitskaya , and v. m. mostepanenko , phys .
lett . * 85 * , 503 ( 2000 ) . c. genet , a. lambrecht , and s. reynaud , phys .
a * 62 * , 012110 ( 2000 ) .
v. b. bezerra , g. l. klimchitskaya , and v. m. mostepanenko , phys .
a * 66 * , 062112 ( 2002 )
. v. b. bezerra , g. l. klimchitskaya , v. m. mostepanenko , and c. romero , phys .
a * 69 * , 022119 ( 2004 ) .
j. s. hye , i. brevik , s. a. ellingsen , and j. b. aarseth , phys .
e * 75 * , 051127 ( 2007 ) .
i. brevik , s. a. ellingsen , j. s. hye , and k. a. milton , j. phys . a : math
. theor . * 41 * , 164017 ( 2008 ) .
r. s. decca , e. fischbach , g. l. klimchitskaya , d. e. krause , d. lpez , and v. m. mostepanenko , phys .
d * 68 * , 116003 ( 2003 ) .
r. s. decca , d. lpez , e. fischbach , g. l. klimchitskaya , d. e. krause , and v. m. mostepanenko , ann .
( n.y . ) * 318 * , 37 ( 2005 ) .
g. l. klimchitskaya , r. s. decca , e. fischbach , d. e. krause , d. lpez , and v. m. mostepanenko , int . j. mod
a * 20 * , 2205 ( 2005 ) .
r. s. decca , d. lpez , e. fischbach , g. l. klimchitskaya , d. e. krause , and v. m. mostepanenko , phys .
d * 75 * , 077101 ( 2007 ) .
r. s. decca , d. lpez , e. fischbach , g. l. klimchitskaya , d. e. krause , and v. m. mostepanenko , eur .
j. c * 51 * , 963 ( 2007 ) .
j. blocki , j. randrup , w. j. swiatecki , and c. f. tsang , ann .
* 105 * , 427 ( 1977 ) .
a. canaguier - durand , p. a. maia neto , a. lambrecht , and s. reynaud , phys .
lett . * 104 * , 040403 ( 2010 ) .
r. zandi , t. emig , and u. mohideen , phys .
b * 81 * , 195423 ( 2010 ) .
b. geyer , g. l. klimchitskaya , and v. m. mostepanenko , phys .
d * 72 * , 085009 ( 2005 ) .
b. geyer , g. l. klimchitskaya , and v. m. mostepanenko , int .
a * 21 * 5007 ( 2006 ) .
g. l. klimchitskaya , b. geyer , and v. m. mostepanenko , j. phys .
a * 39 * , 6495 ( 2006 ) .
b. geyer , g. l. klimchitskaya , and v. m. mostepanenko , ann .
( n.y . ) * 323 * , 291 ( 2008 ) . f. chen , g. l. klimchitskaya , v. m. mostepanenko , and u. mohideen , optics express * 15 * , 4823 ( 2007 )
. f. chen , g. l. klimchitskaya , v. m. mostepanenko , and u. mohideen , phys .
b * 76 * , 035338 ( 2007 ) .
j. m. obrecht , r. j. wild , m. antezza , l. p. pitaevskii , s. stringari , and e. a. cornell , phys . rev .
* 98 * , 063201 ( 2007 ) .
g. l. klimchitskaya and v. m. mostepanenko , j. phys . a : math
. theor . * 41 * , 312002(f ) ( 2008 ) .
l. p. pitaevskii , phys .
* 101 * , 163202 ( 2008 ) .
d. a. r. dalvit and s. k. lamoreaux , phys .
* 101 * , 163203 ( 2008 ) .
b. geyer , g. l. klimchitskaya , u. mohideen , and v. m. mostepanenko , phys .
* 102 * , 189301 ( 2009 ) .
l. p. pitaevskii , phys .
lett . * 102 * , 189302 ( 2009 ) .
r. s. decca , e. fischbach , b. geyer , g. l. klimchitskaya , d. e. krause , d. lpez , u. mohideen , and v. m. mostepanenko , phys
102 * , 189303 ( 2009 ) .
d. a. r. dalvit and s. k. lamoreaux , phys .
lett . * 102 * , 189304 ( 2009 ) . v. m. mostepanenko , r. s. decca , e. fischbach , b. geyer , g. l. klimchitskaya , d. e. krause , d. lpez , and u. mohideen , int .
a * 24 * , 1721 ( 2009 ) .
d. a. r. dalvit and s. k. lamoreaux , j. phys . :
series * 161 * , 012009 ( 2009 ) . b. e. sernelius , phys .
a * 80 * , 043828 ( 2009 ) .
f. intravaia and c. henkel , phys .
* 103 * , 130405 ( 2009 ) .
p. r. buenzli and ph .
a. martin , phys .
e * 77 * , 011114 ( 2008 ) .
g. bimonte , phys .
a * 79 * , 042107 ( 2009 ) .
f. cornu and ph .
a. martin , j. phys . a : math
* 42 * , 495001 ( 2009 ) . v. m. mostepanenko and g. l. klimchitskaya , int .
a * 25 * , 2302 ( 2010 ) .
b. v. derjaguin , kolloid . z. * 69 * , 155 ( 1934 ) .
b. e. sernelius and c. e. romn - velzquez , phys .
a * 78 * , 032111 ( 2008 ) .
e. m. lifshitz and l. p. pitaevskii , _ statistical physics _ , part .
ii ( pergamon press , oxford , 1980 ) .
s. k. lamoreaux , phys .
* 78 * , 5 ( 1997 ) .
u. mohideen and a. roy , phys .
rev . lett . * 81 * , 4549 ( 1998 ) .
a. roy , c .- y .
lin , and u. mohideen , phys .
d * 60 * , 111101(r ) ( 1999 ) .
b. w. harris , f. chen , and u. mohideen , phys .
a * 62 * , 052109 ( 2000 ) .
h. b. chan , v. a. aksyuk , r. n. kleiman , d. j. bishop , and f. capasso , science , * 291 * , 1941 ( 2001 ) .
h. b. chan , v. a. aksyuk , r. n. kleiman , d. j. bishop , and f. capasso , phys .
* 87 * , 211801 ( 2001 ) .
f. chen , g. l. klimchitskaya , u. mohideen , and v. m. mostepanenko , phys .
lett . * 88 * , 101801 ( 2002 ) .
f. chen , g. l. klimchitskaya , u. mohideen , and v. m. mostepanenko , phys .
a * 66 * , 032113 ( 2002 ) .
f. chen , g. l. klimchitskaya , v. m. mostepanenko , and u. mohideen , phys .
lett . * 97 * , 170402 ( 2006 ) .
chiu , g. l. klimchitskaya , v. n. marachevsky , v. m. mostepanenko , and u. mohideen , phys .
b * 80 * , 121402(r ) ( 2009 ) .
chiu , g. l. klimchitskaya , v. n. marachevsky , v. m. mostepanenko , and u. mohideen , phys .
b * 81 * , 115417 ( 2010 ) .
r. s. decca , e. fischbach , g. l. klimchitskaya , d. e. krause , d. lpez , and v. m. mostepanenko , phys .
d * 79 * , 124021 ( 2009 ) .
a. scardicchio and r. l. jaffe , nucl .
b * 704 * , 552 ( 2005 ) .
h. gies and k. klingmller , phys .
d * 74 * , 045002 ( 2006 ) .
a. weber and h. gies , arxiv:1003.3420v1 .
e. fischbach , g. l. klimchitskaya , d. e. krause , and v. m. mostepanenko , eur .
j. c * 68 * , 223 ( 2010 ) . , ed .
e. d. palik ( academic , new york , 1985 ) .
a. lambrecht and s. reynaud , eur .
j. d * 8 * , 309 ( 2000 ) .
v. b. svetovoy and r. esquivel , phys .
e * 72 * , 036113 ( 2005 ) .
b. e. sernelius , phys .
b * 71 * , 235114 ( 2005 ) .
d. e. krause , r. s. decca , d. lpez , and e. fischbach , phys .
lett . * 98 * , 050403 ( 2007 ) .
g. jourdan , a. lambrecht , f. comin , and j. chevrier , europhys .
85 * , 31001 ( 2009 ) . m. bordag and i. pirozhenko , phys .
d * 81 * , 085023 ( 2010 ) .
b. geyer , g. l. klimchitskaya , and v. m. mostepanenko , int .
a * 16 * , 3291 ( 2001 ) .
f. chen , g. l. klimchitskaya , u. mohideen , and v. m. mostepanenko , phys .
a * 69 * , 022117 ( 2004 ) .
v. b. svetovoy , p. j. van zwol , g. palasantzas , and j. th .
m. de hosson , phys .
b * 77 * , 035439 ( 2008 ) .
w. j. kim , a. o. sushkov , d. a. r. dalvit , and s. k. lamoreaux , phys .
a * 81 * , 022505 ( 2010 ) .
b. geyer , g. l. klimchitskaya , and v. m. mostepanenko , phys .
b * 81 * , 245421 ( 2010 )
. r. s. decca , e. fischbach , g. l. klimchitskaya , d. e. krause , d. lpez , u. mohideen , and v. m. mostepanenko , phys .
a * 79 * , 026101 ( 2009 ) .
j. f. babb , g. l. klimchitskaya , and v. m. mostepanenko , phys .
a * 70 * , 042901 ( 2004 ) .
g. bressi , g. carugno , r. onofrio , and g. ruoso , phys .
* 88 * , 041804 ( 2002 ) .
p. antonini , g. bimonte , g. bressi , g. carugno , g. galeazzi , g. messineo , and g. ruoso , j. phys .
: conf . ser . *
161 * , 012006 ( 2009 ) . | we analyze the possibility to measure small thermal effects in the casimir force between metal test bodies in configurations of a sphere above a plate and two parallel plates . for sphere - plate geometry used in many experiments
we investigate the applicability of the proximity force approximation ( pfa ) to calculate thermal effects in the casimir force and its gradient .
it is shown that for real metals the two formulations of the pfa used in the literature lead to relative differences in the obtained results being less than a small parameter equal to the ratio of separation distance to sphere radius . for ideal metals the pfa results for the thermal correction
are obtained and compared with available exact results .
it is emphasized that in the experimental region in the zeroth order of the small parameter mentioned above the thermal casimir force and its gradient calculated using the pfa ( and thermal corrections in their own right ) coincide with respective exact results . for real metals
available exact results are outside the application region of the pfa .
however , the exact results are shown to converge to the pfa results when the small parameter goes down to the experimental values .
we arrive at the conclusion that large thermal effects predicted by the drude model approach , if existing at all , could be measured in both static and dynamic experiments in sphere - plate and plate - plate configurations . as to the small thermal effects predicted by the plasma model approach
, the static experiment in the configuration of two parallel plates is found to be the best for its observation . |
it is now possible to grow a large variety of semiconductor nanocrystals and also control their sizes to obtain mono - dispersed particles .
@xcite a large number of ii - vi @xcite and iii - v @xcite semiconductor nanocrystals have been prepared over the past two decades .
these quantum dots are good candidates for electronic and optical devices @xcite due to their reduced dimensions , enabling one to reduce the size of electronic circuitry .
also , due to the increased oscillator strengths in these nanocrystals as a result of quantum confinement , @xcite these are expected to have higher quantum efficiencies in applications such as light emission .
this is a direct consequence of a greater overlap between the electron and the hole wavefunctions upon size reduction .
moreover , one can tune these properties to suit a specific application by merely changing the size of the nanocrystals .
for example , the band gap of cdse can be varied from 1.9 ev to 2.7 ev by changing the size of the particle from 5.5 nm to 2.3 nm .
@xcite alongwith the band gap of the particle , the photoluminescence can also be varied through the red to the blue region of the visible spectrum .
@xcite this quantum size effect can be explained qualitatively by considering a particle - in - a - box like situation where the energy separation between the levels increases as the dimensions of the box are reduced .
thus , one observes an increase in the band gap of the semiconductor with a decrease in the particle size on a more quantitative footing , various different theoretical approaches have been employed to account for the variation in the electronic structure of nanocrystallites as a function of its size .
the first explanation for the size - dependence of electronic properties in nanocrystals was given by efros and efros .
@xcite it is based on the effective masses of the electron ( @xmath0 ) and the hole ( @xmath1 ) .
known as the effective mass approximation ( ema ) , it is solved by taking various choices for the electron and hole wavefunctions and solving the effective mass equation variationally . in most ema calculations ,
the confining potentials for the electron and the hole have been assumed infinite .
@xcite therefore , the electron and the hole wavefunctions vanish at and beyond the surface of the nanocrystal , without the possibility of any tunnelling . in the strong confinement regime , where @xmath2 , the nanocrystal radius , is much smaller than @xmath3 , the bohr exciton radius , brus proposed @xcite the following expression for the bandgap of the finite sized system @xmath4 where , @xmath5 is the bulk band gap .
the second term is the kinetic energy term containing the effective masses , @xmath0 and @xmath1 , of the electron and the hole , respectively .
the third term arises due to the coulomb attraction between the electron and the hole , and the fourth term due to the spatial correlation between the electron and the hole which is generally small compared to the other two terms .
ema calculations have also been reported where a finite confining potential was used to account for the passivating agents that coat the surface of the nanocrystals in order to arrest their growth .
finite potential calculations are shown to improve the description for cds nanocrystals to a large extent .
@xcite another improvement to the single band ema is the inclusion of multiple bands for describing the hole effective mass .
this is prompted by the fact that the top of the valence band for ii - vi semiconductors is comprised of triply degenerate bands at the @xmath6 point and thus is better defined using a multi - band theory . to account for this degeneracy ,
einevoll @xcite and nair _ et al . _
@xcite have used the effective bond - orbital model for the hole wavefunction , while the electron is described by a single - band ema .
finite barrier heights and the electron - hole coloumb attraction are included in the calculation and exciton energies are obtained variationally in an iterative hartree scheme . the multi - band and finite potential ema methods explain the experimental results reasonably well , but lack the predictive capabilities desirable of a theoretical model , since the finite potentials need to be adjusted to match the experimental results in each specific case .
psuedopotential calculations have also been carried out to study the variation of electronic structure with the nanocrystal size .
@xcite recently , the semi - empirical pseudopotential method has been employed to calculate the electronic structure of si , cdse @xcite and inp @xcite nanocrystals .
the atomic pseudopotentials are extracted from first principles lda calculations on bulk solids .
thus , the wave functions are lda - like while the band structures , effective masses , and deformation potentials are made to match experimental results .
this method provides a reasonable description of the electronic structure of the nanocrystals .
however , major computational efforts and difficulties do not allow one to calculate the properties of large sized nanocrystals
. the tight - binding ( tb ) scheme has been employed by a number of researchers over the past decade .
@xcite this method enjoys several advantages over the other methods discussed above , explaining its popularity . compared to ema , both pseudopotential method and the tight - binding approach provide a substantial improvement in the accuracy of the results .
the tight - binding method has the further advantage of being significantly less demanding in terms of computational efforts , besides providing a simple physical picture in terms of the atomic orbitals and hopping interactions defined over a predetermined range .
a detailed analysis of the first - principle electronic structure calculations can lead to a judicious tight - binding scheme that is minimal in terms of the dimension of the hamiltonian matrix and yet is highly accurate due to the use of a physical and realistic basis .
@xcite the earliest such tb parametrization was provided by vogl _
et al . _
@xcite who used a tb model with the @xmath7 orbital basis in order to describe the electronic structure of bulk semiconductors .
the @xmath8 orbital was employed in an _ ad hoc _ manner in addition to the @xmath9 orbital basis in order to improve the tb - fit to the @xmath10 band dispersions .
subsequently , this tb model was used by lippens and lannoo @xcite to calculate the variations in the band gap for the corresponding semiconducting nanocrystals as a function of the size .
though their results are in better agreement compared to the infinite potential ema , the @xmath7 tb model tends to underestimate the band gap .
the main problem with the @xmath7 model appears to be a failure to reproduce even the lowest lying conduction band within that scheme .
@xcite improvements in the nearest neighbor @xmath7 model have been carried out by including the spin - orbit coupling and the electron - hole interaction .
@xcite however , to account for the conduction bands ,
the inclusion of @xmath11 orbitals becomes necessary .
@xcite this has been shown in the case of inp @xcite nanocrystals , a iii - v semiconductor , that tb model with the @xmath12 orbital basis for the anion and the @xmath9 basis for the cation with next nearest neighbor interactions , for both the anion and the cation , gives excellent agreement with the experimental data . in a recent work
, we have shown that the @xmath12 orbital basis for both the cation and the anion and the inclusion of the next nearest neighbor interactions for the anions provide a very good description of the electronic structure of bulk ii - vi semiconductors .
@xcite this model is shown to describe accurately the band gap and the band dispersions for both the valence and the conduction bands over the energy range of interest .
therefore , this improved model and the parametrization should provide a good starting point for calculating the electronic properties of corresponding nanocrystals , provided the model and parameters are transferable from the bulk to the cluster limit .
@xmath13 calculations for a cds cluster of about 16 diameter @xcite as well as results of ref .
@xcite suggest that the present scheme is of sufficient accuracy down to about 16 , though the applicability of this approach may be limited for still smaller sized clusters . in order to explore the possibility of utilizing it effectively
, we have used this model for calculating the band gap variation over a wide range of sizes for a@xmath14b@xmath15 semiconductor nanocrystals , with a = cd or zn and b = s , se or te , comparing the calculated results with the experimental data from the literature .
the present results show a good agreement with experimental results , where ever available .
the appropriate minimal tb model for the bulk electronic structure of group ii - vi semiconductors was developed in ref .
@xcite by analyzing the atomic wave - function characters of the various bands .
this established @xmath12 basis with the cation - anion and anion - anion interactions as the suitable model .
the tight - binding electronic parameters , namely the orbital energies and the hopping strengths , were determined by fitting the @xmath10 band dispersions to the band dispersions obtained from the tight - binding hamiltonian , given by @xmath16 where , the electron with spin @xmath17 is able to hop from the orbitals labelled @xmath18 with onsite energies equal to @xmath19 in the @xmath20 unit cell to those labelled @xmath21 in the @xmath22 unit cell , with a hopping strength @xmath23 ; the summations @xmath18 and @xmath21 running over all the orbitals considered on the atoms in a unit cell , and @xmath24 and @xmath25 over all the unit cells in the solid .
we use exactly the same model with the parameter strengths given in ref .
@xcite to calculate the electronic structure of corresponding nanocrystals as a function of the size .
we build the cluster shell by shell , starting from a central atom . for the tetrahedrally coordinated compounds in the zinc - blende structure ,
the central atom , say the cation , is surrounded by a shell of four anions . in turn
each of these anions is coordinated by four cations , one of them being the central cation .
the other three cations form a part of the next shell .
the clusters are generated in this manner by successive addition of shells . assuming a spherical shape of the cluster
, the diameter , @xmath11 , is given by @xmath26^{\frac{1}{3}}\ ] ] where , @xmath27 is the lattice constant and @xmath28 the number of atoms present in the nanocrystal .
table i lists the number of atoms present upto a given shell and the diameter of the nanocrystal for various a@xmath29b@xmath15 compounds .
the hamiltonian matrix for any given sized cluster is obtained from eq . 2 with the same atomic orbital basis and electronic parameter strengths as given in ref .
@xcite and is diagonalized to obtain the eigenvalue spectra for the nanocrystal .
direct diagonalization methods are practical only for cluster sizes containing less than @xmath301500 atoms . for larger clusters
, we use the lanczos iterative method .
@xcite the lanczos algorithm uses a starting basis function @xmath31 which can be a linear combination of the atomic orbitals , @xmath32 s , _
i.e. _ @xmath33 once the starting basis function has been generated , a new basis function @xmath34 is generated by applying the hamiltonian and then making the resulting function orthogonal to @xmath31 .
@xmath35 then onwards the subsequent basis functions can be generated by using the recursion formula @xmath36 where , @xmath37 = @xmath38 , @xmath39 = @xmath40 + with @xmath41 = 0 and @xmath42 = 0 . by construction ,
each basis function is orthogonal to the previously generated basis functions . here
the @xmath37 s are the diagonal elements , while @xmath43 s are the off - diagonal terms of the hamiltonian matrix .
diagonalization of this tridiagonal matrix is less time consuming and gives the eigenvalue spectrum for the clusters .
we choose the starting seed vector to be a particular orbital of an atom .
the eigen spectrum thus obtained is composed of only those orbitals that couple with the seed vector .
thus , taking each orbital of every atom in the cluster we obtain the entire density of states . due to the underlying symmetry in the nanocrystal we need not perform calculations for all atomic orbitals as seed vectors , but only those with distinct symmetries .
the band gap for a particular sized nanocrystal is then calculated by subtracting the energy of the top of the valence band ( tvb ) from that of the bottom of the conduction band ( bcb ) .
however , the determinations of the tvb and the bcb become ambiguous due to the presence of dangling bonds at the surface of the nanocrystals .
these non - bonded states lie in the band gap region of the nanocrystals .
these surface states need to be either selectively disposed off @xcite or passivated @xcite in order to remove the midgap states .
once the surface states are passivated , the band gap can be easily determined . in the present work
, we have passivated the surfaces of the nanocrystals in order to remove the mid - gap states from the calculations .
the various steps involved in the calculations for the variation of the band gap with size are quite similar for the different a@xmath29b@xmath15 compounds studied here .
we therefore use the case of cds as an example to illustrate all the steps and various considerations , prior to presenting comprehensive results for all the systems together at the end .
[ doscomparison]a shows the first - principle results for the density of states ( dos ) for cds bulk obtained from the linearized muffin - tin orbital ( lmto ) method with the atomic sphere approximation ( asa ) .
it should be noted that the parameters appearing in the tb hamiltonian ( eq . 1 ) were determined @xcite by a least - squared - error approach in order to obtain dispersions at high symmetry points and a few other @xmath44-points along the symmetry directions in the brillouin zone . since we are eventually interested in the density of states which involves an integration over the entire momentum space , we have explicitly verified in each case that the dos calculated within the tb approach is very similar to the one obtained from the lmto - asa method .
we illustrate this point with the help of dos calculated within the tb model for cds with parameter strengths from ref .
@xcite ; this tb dos is shown in fig .
[ doscomparison]b with the same energy scale as in fig .
[ doscomparison]a .
we notice an excellent agreement of the tb dos with the lmto dos over the entire range of the energy considered .
while the electronic structure of small sized nanocrystals is known to be pronouncedly dependent on the size , larger sized nanocrystals are expected to resemble the bulk in terms of their electronic structures ; evidently in the limit of the large size , the electronic structure of the nanocrystal must smoothly evolve into that of the bulk .
it is known that the quantum confinement effect is generally small for a nanocrystal with typical size larger than the excitonic radius .
the excitonic diameter of cds is about 58 .
@xcite we consider a cds cluster of 76 containing 9527 atoms that is considerably larger than the excitonic diameter . in fig .
[ doscomparison]c we show the dos for this large cds cluster .
the dos of the nanocrystal indeed resembles the bulk dos closely , as is evident in fig .
[ doscomparison ] , apart from the discrete nature of the dos arising from the finite size of the nanocrystallite system .
as discussed in the previous section , the dangling orbitals on the surface atoms appear within the band gap region , complicating the identification of the band gap .
[ znsovercds]a shows the normally obtained dos for a 46 cds nanocrystal ; the corresponding inset shows an expanded view of the band gap region .
as one can clearly see in the expanded view , there are many states spread out over an energy range appearing between the valence band and the conduction band due to the aforementioned dangling bonds within the band gap region . as already discussed , different authors approached the problem of dangling bonds or its removal from the dos in different ways .
for example , lippens and lannoo @xcite got rid of the dangling bonds by removing the unconnected orbitals on the surface atoms in order to obtain the band gap free of the mid - gap states . in spirit
, this approach is similar to the infinite potential barrier on the surface of the nanocrystal assumed in the infinite potential ema .
akin to the finite potential ema , we choose to passivate the surface with a layer of atoms , whose electronic parameters are so chosen that the hopping interactions between the surface atoms and the passivating atoms are stronger compared to those in the bulk of the nanocrystal .
specifically , we choose only the @xmath45 orbital basis on the passivating atoms with the tight - binding hopping parameters about 2 - 3 times larger than that of a - b interactions .
the corresponding dos of the passivated nanocrystals of cds is shown in fig .
[ znsovercds]b . in the main frame of the figures , the unpassivated case in the upper panel and the passivated case in the lower panel
appear almost identical , suggesting that the intrinsic electronic structure of the nanocrystals remain largely unaffected by the passivation . in order to illustrate the effect of passivation on the midgap states , we show an expanded view of the band gap region between the tvb and the bcb in the inset to fig . [ znsovercds]b .
this inset shows that the surface passivation is indeed effective in removing the midgap states , present in the inset to fig .
[ znsovercds]a , illustrating the unpassivated case .
most often , the total band gap variation as a function of the size of the nanocrystal is reported in the literature .
@xcite this is primarily motivated by the fact that this quantity , @xmath46 , is easily determined by experimental uv - visible absorption spectroscopy , that is a routine characterization tool .
however , it is to be noted that the total change in the band gap of any material is simultaneously contributed by shifts of the valence and the conduction band edges away from each other . in general
, the shift of the top of the valence band is not the same as that of the bottom of the conduction band .
moreover , there are recent studies , though few in number @xcite that report the individual shifts in tvb and bcb as a function of the size employing various forms of high energy spectroscopies , such as the photoemission and the x - ray absorption spectroscopies .
thus , it is desirable to compute these shifts of the individual band edges with the size of the nanocrystallite .
the variation of tvb ( circles ) and the bcb ( squares ) with respect to the bulk values are calculated for different sized passivated nanocrystals and shown in fig .
[ cdsbandedge ] .
as expected , the shifts of the band edges decrease smoothly to zero for large sized nanocrystals in every case .
we find that the shift in the bcb is in general much larger compared to the shift in the tvb for any given size of the nanocrystal ; this indicates that the shifts in the total band gap as a function of the nanocrystal size are always dominated by the shifts of the conduction band edge in these systems .
a larger shift for the bcb is indeed expected in view of the fact that the band edge shifts are related inversely to the corresponding effective masses ( see eq . 1 ) and the effective mass of the electron is always much smaller than that of the hole in these ii - vi semiconductors . for example , @xmath0 and @xmath1 in cds are 0.18 and 0.53 , respectively . in the spirit of ema , one can attempt to describe the shifts in the conduction and valence band edges , as @xmath47 where @xmath48 is the variation in the band edge with diameter @xmath11 ; @xmath49 for tvb and @xmath50 for bcb . comparing with the ema ( eq . 1 )
, one expects the fitting parameter @xmath51 to be inversely proportional to the electron ( for bcb , @xmath50 ) or hole ( for tvb , @xmath49 ) effective mass and @xmath52 to equal 2 .
we have fitted the shifts in bcb and tvb as a function of @xmath11 with eq . 4 by varying the parameters @xmath51 and @xmath52 within a least - squared - error approach ;
the resulting best fits are shown in fig .
[ cdsbandedge ] by the solid lines overlapping the calculated data points .
we find that the fits are reasonable , though not very good , in most cases .
more importantly , these fits suggest a gross deviation from the ema predictions ; for example , the variations in tvb and bcb shown in fig .
[ cdsbandedge ] are far from the ema - like @xmath53 dependence and instead the best exponent for @xmath11 is in the range of 1.13 - 1.27 for tvb and 0.65 - 1.05 for bcb , as shown in table ii .
[ all ] shows the variation of the shift in the band gap ( @xmath46 ) for the a@xmath29b@xmath15 semiconductor nanocrystals with a = zn , cd and b = s , se and te as a function of the nanocrystal size .
@xmath46 is calculated in the present model after subtracting the coulomb term ( third term of eq .
1 ) from the calculated difference between the tvb and the bcb to account for the excitonic binding energy , since the experimental data obtained from the uv - absorption include the contribution from the excitonic binding energy . the solid line passing through the calculated data points ( small solid circles ) is the best fit to the calculations .
the best fit is obtained by using simple exponential functions relating @xmath46 to the diameter of the nanocrystallites as @xmath54 while this expression is entirely phenomenological , it has the correct limiting behavior at large @xmath11 .
the advantage of such a best fit is that the @xmath46 for any given system can be readily calculated for any size of the nanocrystallites with the knowledge of the parameter values @xmath55 , @xmath56 , @xmath57 and @xmath58 , which are tabulated in table iii for all the systems investigated here . for comparison
, we also show in the same panels the results obtained from the @xmath7 nearest neighbor tb model ( dashed line ) @xcite and the results from the ema equation ( dotted line ) @xcite .
experimental results available in the literature are also plotted as scattered points with different symbols for comparison with the calculated results .
@xcite there is a plethora of experimental data for zns , cds , cdse and cdte and we see that the present approach provides a better description of the experimental data in all these cases .
the case of znse , where the experimental results are limited , also exhibits good agreement between the experiment and the theory . in the case of znte ,
the present as well as the earlier calculations give almost similar descriptions ; unfortunately , the experimental data are limited and there are large uncertainties in the data , so it is difficult to compare the experimental results with our calculations . for most of the cases , the @xmath12 model with the next nearest neighbor interactions is in better agreement with the experiments compared to the nearest - neighbor - only @xmath7 model .
this is due to the fact that the sole @xmath8 orbital does not account well for the unoccupied states .
these can only be described by the inclusion of the empty anionic @xmath11 orbitals and the anion - anion interactions which are of significance in the description of the bulk electronic structure .
we have calculated the electronic structure as a function of the nanocrystallite size for a@xmath29b@xmath15 semiconductors with a = zn and cd and b = s , se and te , using the tight binding method with the @xmath12 orbital basis set including the a - b and b - b interactions .
it is shown that the shift in the top of the valence band as well as that in the bottom of the conduction band are different from the predictions based on the effective mass approximation , not only in quantitative terms , but also qualitatively .
the calculated variations in the band gaps over a wide range of sizes are compared with all experimental data published so far in the literature .
this comparison shows a very good agreement in every case , suggesting the reliability and the predictive ability of the present approach .
the authors thank o. k. andersen and o. jepsen for the lmto codes and p. mahadevan for providing unpublished @xmath59 results for @xmath30 16 cds nanocrystals .
we acknowledge financial support from the department of science and technology , government of india .
99 c. b. murray , c. r. kagan , m. g. bawendi , ann .
science * 30 * , 545 ( 2000 ) .
a. p. alivisatos , j. phys . chem . * 100 * , 13226 ( 1996 ) . c. b. murray , d. j. norris , m. g. bawendi , j. am .
soc . * 115 * , 8706 ( 1993 ) .
t. vossmeyer , l. katsikas , m. giersig , i. g. popovic , k. diesner , a. chemseddine , a. eychmuller and h. weller , j. phys .
chem . * 98 * , 7665 ( 1994 ) .
j. nanda , s. sapra , d. d. sarma , n. chandrasekharan and g. hodes , chem .
mater . * 12 * , 1018 ( 2000 ) .
o. i. micic , j. r. sprague , c. j. curtis , k. m. jones , a. j. nozik , j. phys .
98 * , 4966 ( 1994 ) .
a. a. guzelian , u. banin , a. v. kadavanich , x. peng , a. p. alivisatos , appl .
phys . lett . *
69 * , 1432 ( 1996 ) .
a. a. guzelian , j. e. bowen katari , a. v. kadavanich , u. banin , k. hamad , e. juban , a. p. alivisatos , r. h. wolters , c. c. arnold and j. r. heath , j. phys
. chem . * 100 * , 7212 ( 1996 ) .
s. v. gapoenko , _ optical properties of semiconductor nanocrystals .
_ cambridge , uk : cambridge univ .
press , 1998 .
u. woggon , _ optical properties of semiconductor quantum dots .
_ berlin : springer - verlag , 1997 . m. a. kastner , phys . today * 46 * , 24 ( 1993 ) .
a. p. alivisatos , science * 271 * , 933 ( 1996 ) .
l. e. brus , j. phys .
solids * 59 * , 459 ( 1998 ) .
l. e. brus , j. phys .
chem . * 90 * , 2555 ( 1986 ) .
b. o. dabbousi , j. rodriguez - viejo , f. v. mikulec , j. r. heine , h. mattoussi , r. ober , k. f. jensen , m. g. bawendi , j. phys .
b * 101 * , 9463 ( 1997 ) .
l. efros and a. l. efros , sov .
* 16 * , 772 ( 1982 ) .
l. e. brus , j. chem . phys .
* 80 * , 4403 ( 1984 ) .
y. kayamma , solid state commun . * 59 * , 405 ( 1986 ) .
h. m. shmidt and h. weller , chem .
* 129 * , 615 ( 1986 ) .
s. v. nair , s. sinha and k. c. rustagi , phys .
b * 35 * , 4098 ( 1987 ) .
y. kayanuma and h. momiji , phys .
b * 41 * , 10261 ( 1990 ) .
g. t. einevoll , phys .
b * 45 * , 3410 ( 1992 ) .
s. v. nair , l. m. ramaniah and k. c. rustagi , phys .
b * 45 * , 5969 ( 1992 ) .
m. v. rama krishna and r. a. friesner , phys .
lett . * 67 * , 629 ( 1991 ) . l. w. wang and a. zunger , phys .
b * 51 * , 17398 ( 1995 ) . h. fu and a. zunger , phys . rev .
b * 55 * , 1642 ( 1997 ) . p. e. lippens and m. lannoo , phys .
b * 39 * , 10935 ( 1989 )
. k. leung , s. pokrant and k. b. whaley , phys .
b * 57 * , 12291 ( 1998 ) .
j. perez - conde and a. k. bhattacharjee , phys . rev .
b * 63 * , 245318 ( 2001 ) .
s. sapra , r. viswanatha and d. d. sarma , j. phys .
phys . * 36 * , 1595 ( 2003 ) .
n. a. hill and k. b. whaley , j. chem .
phys . * 99 * , 3707 ( 1993 ) .
n. a. hill and k. b. whaley , j. chem .
phys . * 100 * , 2831 ( 1994 ) .
s. sapra , d. d. sarma , s. sanvito and n. a. hill , nanolett .
* 2 * , 605 ( 2002 ) .
g. allan , y. m. niquet and c. delerue , appl .
77 * , 639 ( 2000 ) .
s. sapra , n. shanthi and d. d. sarma , phys .
b * 66 * , 205202 ( 2002 ) .
p. vogl , h. p. hjalmarson and j. d. dow , j. phys .
solids * 44 * , 365 ( 1983 ) .
p. mahadevan , unpublished results .
e. dagotto , rev .
phys . * 66 * , 763 ( 1994 ) .
a. d. yoffe , adv . phys . * 42 * , 173 ( 1993 ) .
for example , see the references listed for the experimental data points in fig . [ all ] . v.
l. colvin , a. p. alivisatos and j. g. tobin , phys .
lett . * 66 * , 2786 ( 1991 ) .
j. l@xmath60ning , j. rockenberger , s. eisebitt , j. -e .
rubensson , a. karl , a. kornowski , h. weller and w. eberhardt , solid state commun . * 112 * , 5 ( 1999 ) .
t. van buuren , l. n. dinh , l. l. chase , w. j. siekhaus and l. j. terminello , phys .
* 80 * , 3803 ( 1998 ) .
j. nanda , ph.d .
thesis , indian institute of science , bangalore ( 2000 ) .
p. e. lippens and m. lannoo , mat .
science and engg .
b9 485 ( 1991 ) .
the values for effective masses and dielectric constant were taken from _ data in science and technology : semiconductors other than group iv elements and iii - v compounds _ ed .
o. madelung , springer - verlag , berlin heidelberg , 1992 . y. nakaoka and y. nosaka , langmuir * 13 * , 708 ( 1997 ) .
r. rossetti , r. hull , j. m. gibson and l. e. brus , j. chem .
phys . * 82 * , 552 ( 1985 ) .
h. inoue , n. ichiroku , t. torimoto , t. sakata , h. mori and h. yoneyama , langmuir * 10 * , 4517 ( 1994 ) .
s. yanagida , t. yoshiya , t. shiragami and c. pac , j. phys . chem . * 94 * , 3104 ( 1990 ) . f. t. quinlan , j. kuther , w. tremel , w. knoll , s. risbud and p. stroeve , langmuir * 16 * , 4049 ( 2000 ) .
m. a. hines and p. guyot - sionnest j. phys .
b * 102 * , 3655 ( 1998 ) .
jun , c .- s .
choi and j. cheon , chem .
101 , ( 2001 ) .
y. wang and n. herron , phys .
b * 42 * , 7253 ( 1990 ) .
t. torimoto , h. kontani , y. shibutani , s. kuwabata , t. sakata , h. mori and h. yoneyama , j. phys .
b * 105 * , 6838 ( 2001 ) .
j. nanda , b. a. kuruvilla and d. d. sarma , phys .
b * 59 * , 7473 ( 1999 ) . b. o. dabbousi , c. b. murray , m. f. rubner and m. g. bawendi chem
* 6 * , 216 ( 1994 ) .
a. l. rogach , a. kornowski , m. gao , a. eychmuller and h. weller , j. phys .
b * 103 * , 3065 ( 1999 ) .
s. gorer and g. hodes j. phys .
, * 98 * , 5338 ( 1994 ) . y. mastai and g. hodes , j. phys
b * 101 * , 2685 ( 1997 ) .
y. masumoto and k. sonobe , phys .
b * 56 * , 9734 ( 1997 ) .
h. arizpe - chrez , r. ramrez - bon , f. j. espinoza - betrn , o. zelaya - angel , j. l. marn and r. riera , j. phys .
solids * 61 * , 511 ( 2000 ) . | in order to provide a quantitatively accurate description of the band gap variation with sizes in various ii - vi semiconductor nanocrystals , we make use of the recently reported tight - binding parametrization of the corresponding bulk systems . using the same tight - binding scheme and parameters
, we calculate the electronic structure of ii - vi nanocrystals in real space with sizes ranging between 5 and 80 in diameter .
a comparison with available experimental results from the literature shows an excellent agreement over the entire range of sizes . |
the evolution of the quark - gluon matter produced in ultrarelativistic heavy ion collisions is apparently well described by hydrodynamics with a very small viscosity . given its derivation from an expansion of the energy - momentum tensor in gradients of the velocity field , one expects hydrodynamics to be applicable only when the transverse and longitudinal pressures are sufficiently close , and if the system is close enough to local thermal equilibrium .
an outstanding problem in the application of quantum chromodynamics to the description of the early stages of these collisions is to justify that these conditions are indeed met at a short enough time to apply hydrodynamics .
the main qcd tool for performing these studies is the color glass condensate ( cgc ) framework @xcite , an effective theory designed to describe the highly occupied gluon states encountered in the wave functions of high energy nuclei , and their interactions in a collision . in the cgc ,
the fast partons ( @xmath0 for a nucleus moving in the @xmath1 direction ) are described as static ( thanks to time dilation of all the internal dynamics in a high energy nucleus ) color sources @xmath2 on the light - cone , with a probability @xmath3 $ ] .
the gluons below the cutoff ( @xmath4 ) are described as ordinary gauge fields , coupled eikonally to the fast sources . when this effective description is applied to a collision
, the current @xmath5 is the sum of two terms , corresponding respectively to each of the colliding nuclei .
an important issue in applications of the cgc to the calculation of observable quantities is the dependence on the cutoffs that are introduced to separate the fast and slow partons .
this cutoff is not a physical quantity and should cancel in observables .
however , since it is the upper bound in loop integrals , higher order corrections usually produce logarithms of the cutoff .
it is well known that in electron
nucleus collisions , these logarithms can be absorbed by letting the distribution @xmath3 $ ] depend on the scale @xmath6 according to the jimwlk equation @xcite .
for the cgc to provide a consistent description of nucleus nucleus collisions , one needs to prove that the same distributions @xmath3 $ ] s are sufficient to absorb all the logarithms that arise in these collisions , and that this works for a sufficiently large class of observables .
the first step in this proof is to show that at leading order ( lo ) in the strong coupling @xmath7 , all the inclusive observables ( i.e. those that do not restrict the content of the final state , such as particle spectra , or the expectation value of some local operators in the final state ) can be expressed in terms of the retarded solution of the classical yang - mills equations that vanishes in the remote past ( @xmath8 ) @xcite .
for instance , in terms of this classical field @xmath9 , the gluon spectra at lo read @xmath10 the next step is to compute the next to leading order ( nlo ) correction to these observables . in @xcite
, we have shown that for an inclusive observable @xmath11 , the lo and nlo contributions are formally related by @xmath12_\u \big[{\colorb { \bs a}_\k^*}\,{\mathbbm t}\big]_\v + \int\limits_{_{\u } } \big[{\colorb{\bs\alpha}}\,{\mathbbm t}\big]_\u \bigg]\ ; { \colorb{\cal o}}_{_{\rm lo}}\ ; .
\label{eq : lo - nlo}\ ] ] in this formula , @xmath13 is the operator that generates shifts of the classical field at some initial time ( and the integrals over @xmath14 are integrals over the spatial coordinates at this time ) .
the quantities @xmath15 are small perturbations to the classical field , evaluated at the same time .
the formula is true for any initial time , but some choices simplify the problem by making @xmath15 calculable analytically .
then , the main result in proving the factorization of the logarithms of the cutoffs is @xmath16_\u \big[{\colorb { \bs a}_\k^*}\,{\mathbbm t}\big]_\v + \int\limits_{_{\u } } \big[{\colorb{\bs\alpha}}\,{\mathbbm t}\big]_\u = \log\left(\lambda^+\right)\,{\colorb{\cal h}_1 } + \log\left(\lambda^-\right)\,{\colorb{\cal h}_2 } + \mbox{terms w / o logs}\ ; , \ ] ] where @xmath17 are the jimwlk hamiltonians of the two nuclei .
this formula shows that the logarithms of the two cutoffs do not mix , since they are multiplied by operators that act on a single nucleus . from this formula , it is easy to show that one can absorb all the leading logarithms in two distributions @xmath18 that evolve with their respective jimwlk equation , @xmath19 \ ; { \colora w_1\big[\rho_{_1}\big]}\ ; { \colorb w_2\big[\rho_{_2}\big ] } \ ; { { \colorb{\cal o}}_{_{\rm lo}}[\rho_1,\rho_2]}\ ; .\ ] ] it is the universality of the distributions @xmath20 ( i.e. the fact that the same distributions enter in different types of reactions involving the same nuclei ) that gives its predictive power to the cgc .
at lo , the classical color field produced in a nucleus - nucleus collisions corresponds to longitudinal chromo - electric and chromo - magnetic fields ( figure [ fig : tubes ] - left ) that form _ flux tubes _ whose transverse size is of the order of the inverse saturation momentum .
the energy momentum tensor of this field configuration has a negative longitudinal pressure , and is thus quite far from what can be reasonably described by hydrodynamics .
moreover , it has been known for some time that the classical solutions of yang - mills equations have instabilities : small rapidity dependent perturbations to the boost invariant classical field encountered at lo grow exponentially in time , and eventually become as large as the classical field itself @xcite .
such instabilities occur in loop corrections and lead to secular divergences in observables .
these unstable contributions can be tracked by an improvement of the power counting , where one assigns an exponential of time to each operator @xmath13 .
then one sees that the terms with the strongest divergences ( one exponential factor per power of the coupling @xmath21
see figure [ fig : tubes ] , right , for examples of 2-loop graphs that differ in the strength of their secular divergences ) are obtained by exponentiating the operator in eq .
( [ eq : lo - nlo ] ) .
this exponentiation performs a resummation of a selected subset of all the higher loop corrections .
when applied to the energy - momentum tensor , this resummation reads @xmath22_\u [ { \colorc{\bs a}_\k^*}{\mathbbm t}]_\v}_{{\colorc{\cal g}(\u,\v ) } } + \int\limits_{_{\u } } [ { \colorb{\bs\alpha}}{\mathbbm t}]_\u \bigg]\ ; { t^{\mu\nu}_{_{\rm lo}}}[{\cal a}_{\rm init } ] \nonumber\\ & = & \int[d{\colord\chi}]\ , \exp\bigg[-\frac{1}{2 } \int\limits_{_{\u,\v } } { \colord\chi(\u)}{\colorc{\cal g}^{-1}(\u,\v)}{\colord\chi(\v)}\bigg]\ ; t^{\mu\nu}_{_{\rm lo}}[{\cal a}_{\rm init}+{\colord\chi}+{\colorb\alpha}]\ ; .\end{aligned}\ ] ] the second line is an identity , based on the fact that the operator @xmath13 is the generator of the shifts of the initial classical field @xmath23 .
it shows that this resummation amounts to a gaussian average over fluctuating initial conditions , centered on the classical one . in order to test the effect of this resummation in a simpler setting
, we have studied a scalar field theory with a @xmath24 interaction .
this theory has also unstable classical solutions , due to parametric resonance .
moreover , in four space - time dimensions , this theory is scale invariant at the classical level , as is yang - mills theory .
we have first considered this theory in the case of a fixed volume @xcite .
first of all , one can check explicitly the presence of secular divergences in fixed loop order computations of the energy - momentum tensor ( see figure [ fig : resum ] , left ) . on the right plot of this figure
, we also see that the behavior is quite different after the resummation : not only has the secular divergence been cured , but one also sees that the pressure relaxes to the equilibrium value @xmath25 . in this computation , the initial classical field
was set so that the occupation number is large in a single momentum mode , and zero elsewhere .
rapidly , a population of particles builds up in the neighboring modes . at late times , the distribution has evolved into a classical equilibrium .
interestingly , the best fit requires a positive chemical potential .
this indicates that chemical equilibration has not yet happened at these times , suggesting that the number changing processes are considerably slower than the elastic ones .
moreover , this chemical potential is almost equal to the mass of the quasi - particles , and sees a marked excess of particles in the zero mode , both indicative of bose - einstein condensation ( bec ) . by following the occupation of the zero mode over longer times ( figure [ fig : fk ] , right )
, one sees that indeed this bec is only transient and disappears eventually when inelastic processes equilibrate the particle number . whether a similar phenomenon can occur for gauge fields is still an open issue @xcite .
the previous scalar model can be studied also for a longitudinally expanding system , where the classical background field is boost invariant @xcite . in the left panel of figure [ fig : iso ]
, one first sees that the trace of the pressure tensor relaxes towards the energy density .
moreover , the time dependence of the energy density changes from @xmath26 to @xmath27 , indicating a qualitative change from a very small longitudinal pressure to a longitudinal pressure equal to @xmath28 .
this is confirmed in the right panel of figure [ fig : iso ] : after a very rapid initial decrease , due to the redshifting of the longitudinal momenta , the longitudinal pressure starts growing again due to the instabilities , to eventually become very close to the transverse pressure .
the spectrum of fluctuations appropriate for gauge fields over a boost invariant background has been worked out in @xcite , paving the way for the same simulation in yang - mills theory .
* acknowledgements * f.g .
and t.e . are supported by the agence nationale de la recherche project # 11-bs04 - 015 - 01 .
r.v.s work is supported by the us department of energy under doe contract no.de-ac02-98ch10886 and by an ldrd grant from brookhaven science associates .
, lectures given at cargese summer school on qcd perspectives on hot and dense matter , cargese , france , 6 - 18 aug 2001 , hep - ph/0202270 ; e. iancu , r. venugopalan , quark gluon plasma 3 , eds .
r.c . hwa and x.n .
wang , world scientific , hep - ph/0303204 ; f. gelis , e. iancu , j. jalilian - marian , r. venugopalan , ann . rev .
sci . * 60 * , 463 ( 2010 ) . | we report recent results on the role of instabilities in the isotropization and thermalization of a longitudinally expanding system of quantum fields . heavy ion collisions , color glass condensate , thermalization , bose - einstein condensation |
a _ sandwich beam _ is an engineering model for a three - layer beam consisting of stiff outer face plates and a more compliant inner core layer .
sandwich beam models found in the literature include the models of mead and markus @xcite , rao and nakra ( rn ) @xcite , yan and dowell @xcite and others .
the rn model assumes continuous , piecewise linear displacements through the cross - sections , with the kirchhoff hypothesis imposed on the face plates .
transverse , longitudinal and rotational inertial forces are included in the modeling . in @xcite
several possible multilayer generalizations of the basic three layer sandwich beam structure are derived and analyzed ( in the form of multilayer plates ) . in this paper
we consider a multilayer generalization of the rn model described in @xcite .
the model consists of @xmath0 alternating stiff and complaint ( core ) layers , with stiff layers on the outside .
the stiff layers assume the kirchhoff hypothesis , while the compliant layers admit shear .
the equations of motion for the associated beam model can be written : @xmath1 where @xmath2 primes denote differentiation with respect to the spatial variable @xmath3 and dots denote differentiation with respect to time @xmath4 in the above , @xmath5 represents the transverse displacement , @xmath6 denotes the shear angle in the @xmath7 layer , @xmath8^{\rm t},$ ] @xmath9 denote the longitudinal displacement along the center of the @xmath7 layer , and @xmath10^{\rm t}.$ ] throughout this paper we use the convention that quantities relating to the stiff layers have odd indices @xmath11 and quantities relating to the even layers have even indices @xmath12 .
in addition , @xmath13 are positive physical constants , and @xmath14 where @xmath15 denote the density , thickness , young s modulus , and shear modulus of the @xmath7 layer , respectively .
the vector @xmath16 is defined as @xmath17 where @xmath18 and @xmath19 are the @xmath20 matrices @xmath21 and @xmath22 and @xmath23 denote column vectors with all entries of 1 in @xmath24 and @xmath25 respectively .
the aim of this paper is to prove that the uniform exponential stability of the rn system with standard boundary damping applied at one end point .
consider ( [ main ] ) with the following boundary conditions @xmath26 and the initial conditions @xmath27 where @xmath28 and @xmath29 denote constant positive feedback gains . throughout the paper , we assume @xmath30 boundary controllability of ( [ main ] ) has been studied in several papers . for the three layer case , in @xcite the multiplier method was used to prove exact controllability with a control for each equation applied at an end point .
the moment method was used in @xcite to obtain boundary controllability for the multilayer case , but with the condition that wave speeds of the layers be distinct .
the same approach was used to prove simultaneous controllability ( i.e. , with one boundary control instead of three ) for the three layer case in @xcite . in @xcite exact boundary controllability of the general multilayer system
was proved for a variety of boundary conditions : clamped , hinged , clamped - hinged , and hinged - clamped .
the results in @xcite improve earlier results in that there are no restrictions on the wave speeds or the size of @xmath31 and moreover , exact controllability is proved in the optimal time ( determined by characteristics ) . in @xcite ,
@xcite exact controllability results for the multilayer rn plate system analogous to ( [ main ] ) with locally distributed control in a neighborhood of a portion of the boundary were obtained by the method of carleman estimates .
stability results for layered beam systems closely related to ( [ main ] ) subject to internal damping proportional to rate - of - shear in one or more layers have been studied in several papers ; @xcite , @xcite , @xcite , @xcite . in particular , the approach used in @xcite was successfully applied in the dissertation @xcite to obtain uniform exponential stability results the system ( [ main ] ) with rate - of shear damping included in the compliant layers . concerning boundary feedback stabilization of layered beam models , spectral methods ( based on the riesz basis property )
are applied in @xcite to prove exponential stability results for a laminated beam model in @xcite .
a similar approach is used in @xcite for the mead - markus model described in @xcite .
there are also several results concerning the boundary feedback stabilization of a single rayleigh beam equation e.g. , @xcite , where a uniform exponential decay result is obtained for a rayleigh beam by means of a compact perturbation argument and @xcite where the riesz basis approach is used to obtain a similar stabilization result .
some related uniform stabilization results for the kirchhoff plate are proved in @xcite .
our main result is the following .
[ finaltheorem ] assume ( [ assump ] ) .
then the semigroup generated by @xmath32 is exponentially stable in @xmath33 , i.e. , @xmath34 @xmath35 such that @xmath36 moreover , @xmath37 in the above , @xmath38 @xmath39 and @xmath33 are defined in ( [ semigroupfor1 ] ) , ( [ energy ] ) , and ( [ semigroupdom ] ) respectively .
our methodology in this paper is a combination of techniques used in @xcite , @xcite , and @xcite .
the decoupled system ( i.e. , ( [ main ] ) , with @xmath40 ) consists of a rayleigh beam equation and @xmath41 wave equations .
we prove that the decoupled system has a riesz basis of eigenfunctions and obtain explicit asymptotic estimates on the eigenvalues .
in particular , the eigenvalues of the decoupled system asymptotically lie along a finite number of vertical lines in the left half plane .
we are able to prove that the family of eigenfunctions and generalized eigenfunctions of the decoupled system form a riesz basis and consequently ( see @xcite ) , the _ spectrum determined growth condition _ holds .
this allows us to prove the exponential stability of the decoupled closed - loop system ( see theorem [ exp - ray ] ) .
we mention that the exponential stabilization for the portion of the uncoupled system corresponding to the wave equations is well - known results , e.g. , @xcite , @xcite , @xcite , @xcite .
furthermore a number of results are known for stabilization of the rayleigh beam , e.g. @xcite , @xcite , @xcite , however none of these results are applicable to the rayleigh beam with clamped - hinged boundary conditions which we consider .
therefore we include a detailed proof of the exponential stability for the rayleigh beam .
next , we prove that the system ( [ main])-([initial ] ) has a compact resolvent and is a compact perturbation of the decoupled system .
therefore , exponential stability of ( [ main])-([initial ] ) follows from a perturbation theorem due to triggiani @xcite once it is shown that the semigroup generated by @xmath42 is strongly stable ( see theorem [ stronglystable ] ) . proving the strong stability involves use of dissipativity of the semigroup together with a nontrivial unique continuation argument that is proved in @xcite in application to the associated boundary control problem .
our paper is organized as the following . in section [ wp ] , we give a semigroup formulation of ( [ main])-([initial ] ) .
we prove that the semigroup is a @xmath43semigroup of contractions on an appropriate hilbert space . in section [ decoupled ] , we first prove that the generalized functions corresponding to the single rayleigh beam equation with the feedback applied to the moment forms a riesz basis . then , we show that the decoupled system , i.e. @xmath40 in ( [ main ] ) , has the riesz basis property . finally , we show that the decoupled system is exponentially stable . in section
[ coupled - stab ] , we prove that the system ( [ main ] ) is a compact perturbation of the decoupled system .
finally , we prove our main stabilization result in theorem [ finaltheorem ] .
let @xmath44 let also @xmath45 . from the lax - milgram theorem , @xmath46 is an isomorphism . then ( [ main])-([initial ] ) can be formulated as @xmath47 where @xmath48
let @xmath49 where @xmath50 and @xmath51 may be scalar or vector valued .
define the bilinear forms @xmath52 and @xmath53 by @xmath54 the natural energy of the beam is given by @xmath55 where @xmath56 are the quadratic forms that agree with @xmath57 on the diagonal .
we define the hilbert space @xmath58 by @xmath59 with the energy inner product @xmath60 where @xmath61 * characterization of the domain of @xmath62 * let @xmath63 such that @xmath64 @xmath65 and assume that boundary conditions ( [ bdry1 ] ) hold . also let @xmath66 .
a calculation shows that @xmath67 in the above , @xmath68 hence , using the definition of @xmath52 in ( [ semigroupfor1 ] ) , we find that the following identity holds : @xmath69 we use ( [ eq21 ] ) as a basis for the variational definition of the space @xmath70 more precisely , @xmath71 note that the operator @xmath72 is densely defined .
[ surjective]@xmath73 is surjective , i.e. @xmath74 _ proof .
_ let @xmath75 for given @xmath76 we want to prove the solvability of the system @xmath77 for @xmath78 this is equivalent to prove the solvability of the following system in @xmath79 @xmath80 let @xmath81 if we multiply ( [ oz1 ] ) by @xmath82 and dot product ( [ oz2 ] ) by @xmath83 then integrate by parts and apply ( [ rc1 ] ) , ( [ rangecon ] ) we obtain @xmath84 the bilinear forms @xmath52 and @xmath53 are symmetric , bounded and coercive on @xmath85 and @xmath86 respectively .
moreover , the right hand side of ( [ rangecon3 ] ) is a bounded linear form on @xmath85 .
therefore , by lax - milgram theorem , there exists a unique pair @xmath87 satisfying ( [ rangecon3 ] ) .
therefore , ( [ rangecon ] ) uniquely determines pair @xmath88 + the last step of our proof is to show that @xmath89 i.e. @xmath90 satisfies ( [ eq21 ] ) .
now assume that @xmath91 with @xmath92 then it follows that @xmath93 holds for all @xmath92 therefore in @xmath94 we have @xmath95 if we substitute ( [ rangecon5 ] ) in ( [ rangecon3 ] ) by setting @xmath96 and @xmath97
we obtain that @xmath98 satisfies ( [ eq21 ] ) .
this together with ( [ rangecon5 ] ) implies that @xmath99 . @xmath100
[ disspppp]the infinitesimal generator @xmath32 is dissipative on @xmath101 and it satisfies @xmath102 for all @xmath103 _ proof .
_ by an easy calculation we have the following @xmath104 therefore ( [ dissp ] ) follows .
@xmath100 [ egigen0 ] the point spectrum of @xmath32 does not contain @xmath105 i.e. the following eigenvalue problem @xmath106 with the boundary conditions @xmath107 has only the trivial solution . _ proof .
_ let @xmath108 be defined on the domain @xmath109 where @xmath110 then @xmath111 is a densely defined , self - adjoint , positive definite , and unbounded operator on @xmath112 and therefore @xmath113 exists and is a bounded operator defined on all of @xmath114 where @xmath115 now define the operator @xmath116 then @xmath117 extends to a continuous and self - adjoint operator on @xmath118 and @xmath119 to show this , let @xmath120 so that @xmath121 and @xmath122 then @xmath123 .
\end{aligned}\ ] ] by applying @xmath113 to both sides of ( [ eq9 ] ) , we get ( [ eq90 ] ) .
+ now we show that @xmath117 is non - positive on @xmath124 let @xmath125 so that @xmath126 and @xmath127 then @xmath128 where @xmath129 is an operator defined on the domain @xmath130 and @xmath131 now , we are in the position of solving ( [ eq ] ) . multiplying the second equation in ( [ eq ] ) by @xmath132 and using the first and the third equations in ( [ eq ] ) yields @xmath133 but since @xmath117 is non - positive , the operator @xmath134 is a positive operator ( using ( [ bc ] ) ) .
this implies that @xmath135 therefore @xmath136 by ( [ eq ] ) .
we have the following theorem for the well - posedness of the cauchy problem ( [ main])-([initial ] ) .
[ eigens]@xmath137 is the infinitesimal generator of a @xmath43semigroup of contractions .
therefore for every @xmath138 @xmath139 solves ( [ main])-([initial ] ) , and we have @xmath140 ; \mc{d}(\mc{a})\right)\cap c^1\left([0,t ] ; \mc{h}\right).$ ] moreover , the spectrum @xmath141 of @xmath32 has all isolated eigenvalues .
_ @xmath32 is an m - dissipative operator by lemmata [ surjective ] and [ disspppp ] .
therefore , @xmath137 is the infinitesimal generator of a @xmath43semigroup of contraction by lmer - phillips theorem @xcite . by using the fact that @xmath142 is densely defined and compact in @xmath101 and @xmath143 by lemma [ egigen0 ] ,
@xmath144 is compact at @xmath145 thus compact for all @xmath146 hence the spectrum of @xmath32 has all isolated eigenvalues .
in this section , we prove the exponential stability of the decoupled system : [ decoupled ] @xmath148 with initial and boundary conditions @xmath149 let @xmath150 then the semigroup corresponding to ( [ maindec ] ) is given by @xmath151 where @xmath152 define the bilinear forms @xmath153 and @xmath154 by @xmath155 where @xmath156 are the quadratic forms that agree with @xmath157 on the diagonal .
the corresponding energy inner product on @xmath58 is given by @xmath158 [ eigensss ] the operator @xmath159 defined in ( [ semigroupfordec1 ] ) is the infinitesimal generator of a @xmath43semigroup of contractions .
therefore for every @xmath160 , @xmath161 solves ( [ maindec])-([mainbdconds ] ) and @xmath162 ; \mc{h}\right).$ ] moreover , the spectrum of @xmath163 consists of isolated eigenvalues . _ proof . _ note that @xmath164 where @xmath165 is defined by ( [ eq21 ] ) .
the proof of the theorem [ eigens ] remains valid when @xmath166 , and hence theorem [ eigensss ] follows .
@xmath100 now we find the adjoint operator @xmath167 which is needed in the proof of lemma [ lemomega ] .
[ skew - adjoint]the infinitesimal generator @xmath163 satisfies @xmath168 where @xmath169 denotes the dependence of @xmath163 on the feedback gains @xmath170 . _ proof .
_ let @xmath171^{\rm t}\in \mc{d}(\mc{a}_d),$ ] @xmath172^{\rm t}\in \left(\mc{h } \cap c^{\infty}(\omega)\right).$ ] then , @xmath173 and @xmath174 satisfy the following boundary conditions @xmath175 a calculation using ( [ bcs ] ) shows that @xmath176 this implies that @xmath177 on the space @xmath178 .
it follows from the lemma [ surjective ] that @xmath179 has no larger closed extension and hence @xmath180 .
@xmath100 in this section , we prove the riesz basis property for the eigenfunctions of the rayleigh beam equation with boundary feedback .
a similar analysis applies to the wave equations with boundary feedback of the form ( [ mainbdconds ] ) ( see theorem [ oth ] ) .
the eigenvalue problem corresponding to rayleigh beam in ( [ maindec ] ) is given as the following @xmath181 now let @xmath182 then the solution of ( [ eprob ] ) is @xmath183 where @xmath184 by using the first three boundary conditions @xmath185 and @xmath186 for ( [ gensol ] ) we get @xmath187\\ \label{uofx } & & + is_0 \gamma\left[\theta_0 \sin{\sqrt{\theta_0}l}\sinh{\sqrt{\xi_0}x}-\xi_0\sinh{\sqrt{\xi_0}l } \sin{\sqrt{\theta_0}x}\right].\end{aligned}\ ] ] by using the last boundary condition @xmath188 we obtain the characteristic equation that @xmath189 satisfies @xmath190 since we have @xmath191 by ( [ eq31 ] ) , we find that @xmath192 multiplying ( [ eq43 ] ) by @xmath193 and eventually using ( [ siamtheta ] ) and ( [ siamxi ] ) yields @xmath194 solving ( [ char3 ] ) is equivalent to solving @xmath195 the following theorem characterizes the eigenvalues of ( [ eprob ] ) .
[ expresofevalues ] assume ( [ assump ] ) .
the eigenvalues @xmath196 of ( [ eprob ] ) for sufficiently large @xmath197 consist of complex conjugate pairs @xmath198 and @xmath199 with asymptotic form @xmath200 as @xmath201 where @xmath202 _ proof .
_ first , note that @xmath203 are the solutions of ( [ eq46 ] ) when the right hand side of the equation is zero .
we claim that @xmath204 solve ( [ eq46 ] ) for all sufficiently large @xmath205 without loss of generality , we only consider the case @xmath206 let @xmath207 and @xmath208 now consider the circle @xmath209 then on the disc @xmath210 we have @xmath211 by a simple calculation one easily obtains the following for sufficiently large @xmath212 @xmath213 where we have used ( [ eq43 ] ) , ( [ eq46 ] ) and ( [ eq52 ] ) .
therefore , by rouch s theorem , @xmath214 has a unique zero in the ball @xmath215 for sufficiently large @xmath216 that is , there exists a unique solution of the equation ( [ eq46 ] ) in @xmath217 this proves our claim and hence @xmath218 as @xmath201 by ( [ siamtheta ] ) .
@xmath100 [ asymptots ] assume ( [ assump ] ) holds .
the eigenfunctions @xmath219 corresponding to the eigenvalues @xmath220 of ( [ eprob ] ) have the following asymptotic expressions : @xmath221 which are asymptotically normalized in @xmath222 .
_ first , we find estimates for @xmath223 and @xmath224 application of ( [ siamtheta]),([siamxi ] ) and theorem [ expresofevalues ] shows that for any @xmath225 $ ] we have @xmath226 therefore the characteristic equation ( [ char3 ] ) takes the form @xmath227 now we use ( [ uofx ] ) , ( [ siamtheta ] ) , ( [ siamxi ] ) , ( [ eqq21 ] ) and ( [ eq65 ] ) to get @xmath228 @xmath229 where @xmath230 in the following we verify that @xmath231 we have two cases .
let @xmath206 by ( [ siamtheta ] ) we have @xmath232 when @xmath197 is odd , it is clear that the complex number @xmath233 has a nonzero ( but constant ) real part .
therefore @xmath231 when @xmath197 is even , we must have @xmath234.\ ] ] but @xmath235 otherwise by ( [ siamxi ] ) and ( [ eq65 ] ) , we have @xmath236 and by taking the square of both sides and after cancelations we get a contradiction : @xmath237 now let @xmath238 then @xmath239 but since the last expression above has a nonzero ( but constant ) real part , we have @xmath231 this proves the first part of the theorem .
now if we set @xmath240 then the second part of the theorem follows that eigenvectors are asymptotically normalized , i.e. @xmath241 @xmath100 [ lemomega ] assume ( [ assump ] ) .
the generalized eigenfunctions @xmath242 of ( [ eprob ] ) corresponding to the eigenvalues @xmath243 are @xmath244linearly independent in @xmath245 moreover , @xmath246 is @xmath247linearly independent in @xmath248 _ proof .
_ we first prove that the eigenfunctions @xmath249 of the adjoint eigenvalue problem ( see ( [ eadjprob ] ) below ) , are biorthogonal to the eigenfunctions @xmath250 of ( [ eprob ] ) . by using lemma [ skew - adjoint ]
, we consider the following adjoint eigenvalue problem : @xmath251 this is exactly the same boundary value problem as ( [ eprob ] ) .
therefore @xmath252 the only difference is the expression of the eigenfunctions of ( [ eadjprob ] ) given by @xmath253 it is possible to check that @xmath254 if @xmath255 if @xmath256 the inner product @xmath257 is uniformly bounded ( from ( [ eigenfunc1 ] ) each term has a uniform asymptotic bound ) .
hence a uniform bound for ( [ 21 ] ) exists .
therefore , @xmath250 is @xmath244linearly independent in @xmath258 this proves the first part of lemma [ lemomega ] . to show that @xmath259 is @xmath247linearly independent in @xmath260 we readjust the inner product ( [ inner ] ) for @xmath222 with an equivalent one as the following : @xmath261 define the map @xmath262 by @xmath263 it is clear that @xmath264 is an isomorphism : @xmath222 to @xmath248 hence @xmath244linearly independence is preserved
this proves lemma [ lemomega ] .
[ rieszbasisprop ] assume ( [ assump ] ) .
the generalized eigenfunctions @xmath250 of ( [ eprob ] ) forms a riesz basis in @xmath258 _ proof .
_ it is known that both @xmath265 and @xmath266 are orthonormal bases in @xmath267 therefore , it is easy to see that @xmath268 is also an orthonormal basis in @xmath269 now let @xmath270 where @xmath271 @xmath272 is a bounded linear operator from @xmath273 to @xmath273 since @xmath274 for @xmath275 we have @xmath276 therefore @xmath277 forms a riesz basis in @xmath248 now let @xmath278 where @xmath279 are the eigenvectors corresponding to the eigenvalues @xmath280 since @xmath281 is @xmath244 linearly independent in @xmath282 by lemma [ lemomega ] , and @xmath283 i.e. @xmath284 it follows from bari s theorem @xcite that @xmath285 is a riesz basis on @xmath248 hence @xmath286 forms a riesz basis in @xmath258 @xmath100 the following theorem can be obtained by the same procedure . [ oth]assume ( [ assump ] ) . consider @xmath287 then , the eigenfunctions of ( [ nothing1 ] ) @xmath288 corresponding to the branches of eigenvalues @xmath289 forms a riesz basis in @xmath290 where + @xmath291 @xmath292 @xmath293 for all @xmath294 and @xmath295 [ exp - ray ] assume ( [ assump ] ) .
then the semigroup generated by @xmath163 is exponentially stable on @xmath33 , i.e. , @xmath296 such that @xmath297 where @xmath298 _ proof .
_ the riesz basis property for the rayleigh beam equation ( theorem [ rieszbasisprop ] ) together with riesz basis property for the system of wave equations ( theorem [ oth ] ) imply that the eigenfunctions @xmath299 of the operator @xmath163 form a riesz basis in @xmath300 hence , as is well known , the growth bound for the associated semigroup is determined by spectrum of the generator .
we know from theorems [ expresofevalues ] and [ oth ] that the eigenvalues @xmath301 of @xmath302 have the expressions @xmath303 as @xmath201 and @xmath293 for all @xmath205 furthermore since @xmath304 is dissipative , all eigenvalues have non - positive real parts . hence ,
if we show that there are no eigenvalues on the imaginary axis , then the theorem is proved . for the wave equations ,
this is trivial to show , and is well - known .
for the boundary conditions we have , for the rayleigh beam , the possibility of imaginary eigenvalues lead to the following overdetermined eigensystem @xmath305 this system was shown in @xcite to have only the trivial solution .
therefore @xmath306 is an exponentially stable semigroup on @xmath307 and ( [ exp ] ) holds .
in this section , we show that one boundary feedback for each equation is enough to obtain the uniform stabilization of the multilayer rn beam . first , we will consider the decomposition @xmath308 of the semigroup generator of the original problem ( [ semigroupfor1 ] ) where @xmath163 is the semigroup generator of the decoupled system and it is defined by ( [ semigroupfordec1 ] ) , and the operator @xmath309 is the coupling between the layers defined as the following when @xmath313 we have @xmath314 and @xmath315 and therefore @xmath316 since @xmath317 is an isomorphism , the last terms in ( [ opb ] ) satisfy @xmath318 which are compactly embeddded in @xmath319 and @xmath320 respectively . hence the operator @xmath321 is compact in @xmath58 .
@xmath100 _ proof .
_ we know that our system ( [ main])-([initial ] ) is dissipative by ( [ dissp ] ) . if we can show that there are no eigenvalues on the imaginary axis , or in other words , the set @xmath322 has only the trivial solution , i.e. @xmath323 then by la salle s invariance principle , the system ( [ main])-([initial ] ) is strongly stable . since we have eliminated the possibility of a zero eigenvalue in lemma [ egigen0 ] , ( [ eq81 ] ) corresponds to @xmath324 and @xmath325 where @xmath326 therefore , proving the strong stability of the ( [ main])-([initial ] ) reduces to showing that the following eigenvalue problem @xmath327 with initial and overdetermined boundary conditions @xmath328 has only the trivial solution , i.e. @xmath329 this same overdetermined system came up in proving observability for the corresponding boundary control problem in @xcite , where the uniqueness of the zero solution was proved using a multiplier type argument .
now we prove our main theorem for the exponential stability of the solutions ( [ main])-([initial ] ) : _ proof of theorem [ finaltheorem ] .
_ we know that @xmath330 the semigroup @xmath331 is strongly stable on @xmath33 by theorem [ stronglystable ] and the operator @xmath332 is a compact in @xmath33 by lemma [ compact ] .
therefore , since the semigroup generated by @xmath333 is uniformly exponentially stable in @xmath33 then the semigroup @xmath334 is uniformly exponentially stable in @xmath33 by e.g. , the perturbation theorem of triggiani @xcite . @xmath100
( mr2679641 ) [ 10.3934/dcdsb.2010.14.1279 ] a. a. allen and s. w. hansen , _ _ analyticity and optimal damping for a multilayer mead - markus sandwich beam , _ _ _ discrete contin .
b ( 4 ) , _ * 14 * ( 2010 ) , 12791292 .
( mr1758723 ) [ 10.1142/s0218202500000306 ] s. w. hansen and i. lasiecka , _ _ analyticity , hyperbolicity and uniform stability of semigroups arising in models of composite beams , _ _ _ math .
models meth .
, _ * 10 * ( 2000 ) , 555580 .
( mr2833259 ) [ 10.3934/mcrf.2011.1.189 ] s. w. hansen and o. y. imanuvilov , _ _ exact controllability of a multilayer rao - nakra plate with free boundary conditions , _ _ _ math .
control relat .
fields , _ * 1 * ( 2011 ) , 189230 . ( mr2859867 ) [ 10.1051/cocv/2010040 ] s. w. hansen and o. y. imanuvilov ,
_ _ exact controllability of a multilayer rao - nakra plate with clamped boundary conditions , _ _ _ esaim control optim .
var . , _ * 17 * ( 2011 ) , 11011132 .
[ 10.1109/cdc.2005.1582645 ] s. w. hansen and r. rajaram , _ _ simultaneous boundary control of a rao - nakra sandwich beam , _ _ _ proc .
44th ieee conference on decision and control and the european control conference , _
( 2005 ) , 31463151 .
( mr1122306 ) [ 10.1016/0022 - 0396(91)90022 - 2 ] i. lasiecka and r. triggiani , _ _ exact controllability and uniform stabilization of kirchhoff plates with boundary controls only in @xmath335 _ _ _ j. differential equations , _ * 93 * ( 1991 ) , 62101 .
( mr876804 ) [ 10.1016/0022 - 0396(87)90025 - 8 ] i. lasiecka and r. triggiani , _ _ uniform exponential decay of wave equations in a bounded region with @xmath336-feedback control in the dirichlet boundary conditions , _ _ _ j. differential equations , _ * 66 * ( 1987 ) , 340390 .
( mr953013 ) [ 10.1090/s0002 - 9939 - 1989 - 0953013 - 0 ] r. triggiani , _ _ lack of uniform stabilization for noncontractive semigroups under compact perturbation , _ _ _ proc .
, _ * 105 * ( 1989 ) , 375383 . (
mr2193496 ) [ 10.1137/040610003 ] j. m. wang , g. q. xu and s. p. yung , _ _ exponential stabilization of laminated beams with structural damping and boundary feedback controls , _ _ _ siam j. cont .
, _ * 44 * ( 2005 ) , 15751597 .
( mr2192066 ) [ 10.1051/cocv:2005030 ] j. m. wang , b. z. guo and b. chentouf , _ _ boundary feedback stabilization of a three - layer sandwich beam : riesz basis approach , _ _ _ esaim control optim .
var . , _ * 12 * ( 2006 ) , 1234 . | we consider the problem of boundary feedback stabilization of a multilayer rao - nakra sandwich beam .
we show that the eigenfunctions of the decoupled system form a riesz basis .
this allows us to deduce that the decoupled system is exponentially stable .
since the coupling terms are compact , the exponential stability of the coupled system follows from the strong stability of the coupled system , which is proved using a unique continuation result for the overdetermined homogenous system in the case of zero feedback .
a. zkan zer scott w. hansen |
smoothed particle hydrodynamics ( sph ) has become a popular tool for studying astrophysical flows since its introduction in the mid-1970s by @xcite and @xcite . instead of solving the equations of hydrodynamics on a mesh
, sph operates by associating fluid elements with particles that characterize the flow .
local properties are estimated by `` smoothing '' the attributes of particles ( neighbors ) near a given point .
typically , the particles carry a fixed mass ( unless converted to another form , such as collisionless stars ) , and the smoothing is done in a spherically symmetric manner using an interpolation function known as the smoothing kernel . because it is gridless , sph is naturally spatially adaptive if the scale of the smoothing , set by the smoothing length @xmath4 , varies as the particles move with the flow .
if individual particle timesteps are employed , the scheme will also be temporally adaptive .
sph has been used with success especially in investigations of the formation and evolution of galaxies and large scale structure , where spatial and temporal adaptivity are essential because of the large range in scales present .
along with the successes has come an understanding of various limitations with sph .
examples include : an artificial clumping instability for particular choices of the smoothing kernel @xcite ; a lack of energy or entropy conservation for various forms of the equations of motion when the smoothing lengths vary spatially @xcite ; and a poor handling of instabilities at shearing interfaces owing to artificial surface tension effects @xcite .
many of the problems have already been overcome through a variety of improvements to the original formulation of sph , such as the introduction of generalized smoothing kernels @xcite , fully conservative equations of motion based on variational principles @xcite , and various methods to eliminate artificial surface tension effects @xcite
. however , some issues remain unresolved , which has led to numerous misconceptions within the community of users of sph .
for example , because densities in sph are calculated using smoothed estimates , rather than by solving the mass continuity equation , the flow of the fluid on small scales is not described correctly , as emphasized by @xcite .
one claimed advantage of sph over grid based methods is that the previous history of fluid elements can be determined straightforwardly because individual particles retain their identity over time . with mesh codes , this can be accomplished only by introducing tracer particles into the flow so that the evolution of the density field can be reconstructed @xcite .
in fact , this `` advantage '' of sph owes to the fact that this method does not solve the equations of motion correctly on the smoothing scale . here , we focus on another issue with sph that has received little attention in the literature , but for which confusion abounds : the numerical convergence of this technique .
a commonly used approach for making sph spatially adaptive is to allow the smoothing length of each particle to decrease or increase depending on the local density of particles .
one possible choice is to scale the smoothing lengths in proportion to the mean interparticle separation according to @xmath11 @xcite .
a benefit to this scaling is that the number of particles used to compute smoothed estimates ( particles neighboring a given point ) , @xmath5 , is roughly constant , optimizing the efficiency of the method from one region to another .
but , this choice is somewhat arbitrary and leads to difficulties that are not widely recognized . in sph ,
local quantities are estimated by smoothing continuous fluid variables with an interpolation kernel .
the resulting convolutions are evaluated numerically by approximating the integrals with discrete sums .
these sums are calculated over @xmath5 fluid elements ( particles ) near a given point . for the scaling
noted above @xmath12 constant , and is typically limited to relatively small values @xmath13 to maximize spatial resolution .
however , the approximation of the convolution integrals by discrete sums entails an error that depends directly on @xmath5 and not @xmath3 or @xmath4 .
therefore , as the resolution is improved by increasing @xmath3 and reducing @xmath4 , numerical convergence is not possible because the error in the discrete estimates does not vanish if @xmath5 is held constant .
systematic studies of the convergence properties of sph are lacking , but evidence that this source of error exists is present in the literature .
for example , in his review of sph , @xcite performed a series of tests to empirically determine the convergence rate of sph .
the behavior we suggest can be seen clearly in the gresho vortex problem , as in fig .
6 of @xcite . here
, various runs were performed at improving resolution by increasing the total number of particles , @xmath3 , and reducing the smoothing lengths , @xmath4 , but holding the number of neighbors , @xmath5 fixed .
the solid curves in this figure compare the error in each case to a known analytic solution . for small particle numbers , increasing @xmath3 decreases the error , but eventually the error saturates and plateaus at value that depends in detail on the choice of artificial viscosity .
however , if the runs with varying @xmath3 are compared to the highest resolution versions ( dashed curves ) , the error shows a monotonically decreasing trend with @xmath3 , for all choices of the artificial viscosity .
this behavior can be explained only by the presence of a residual source of error that does not depend directly on @xmath3 and @xmath4 . in what follows ,
we show that this error arises in the discrete sums used to approximate the local smoothing convolutions and that it depends directly only on the number of neighbors in these sums , @xmath5 . for a random distribution of particles
, this residual error essentially arises from shot noise in the discrete estimates and is expected to scale as @xmath14 .
however , as emphasized by @xcite , sph particles are usually not randomly distributed but are instead arranged in a `` quasi - regular '' pattern because of local forces between them .
an analysis of the error in this situation suggests that the noise should scale with @xmath5 as @xmath15 , where @xmath16 is the number of dimensions .
( note that to this order of accuracy , the error estimate vanishes for @xmath17 , meaning that 1-dimensional tests can not be used to gauge the numerical convergence of sph for @xmath18 , contrary to common belief . ) using simple numerical tests , we verify that the expected @xmath19 behavior is a reasonable characterization of the discreteness error for multi - dimensional simulations with sph .
therefore , the discreteness error , error@xmath20 , can be bounded by the estimates @xmath21 error@xmath20 @xmath22 . using this scaling , we suggest an optimal compromise between accuracy and efficiency by requiring that this error decline at the same rate as the error associated with the smoothing procedure itself , which directly involves only the smoothing lengths , @xmath4 . in detail
, the outcome depends on the form of the smoothing kernel , but for typical choices we find that the number of neighbors in the smoothed estimates should increase with the total particle number as @xmath7 . in this case
, the smoothing lengths would scale as @xmath23 , rather than with the average separation between particles as @xmath24 , in which case @xmath4 represents the geometric mean of the size of the system and the average interparticle separation .
our empirical tests verify that numerical convergence can indeed be obtained with sph , if the neighbor number grows systematically with improved resolution . of course
, this comes with the consequence of increased computational expense relative to holding @xmath5 fixed , and implies an unfavorable cpu scaling of @xmath25 compared to grid codes which are typically @xmath8 . this should not be surprising , since sph is a monte carlo - like algorithm and it is well - known in other contexts that , while being flexible and general , monte carlo methods typically have unfavorable convergence properties relative to other more specialized techniques .
ultimately , applications with sph must therefore choose between accuracy and computational efficiency .
what is clear , however , is that if sph is used with constant @xmath5 , numerical convergence is not possible .
we emphasize that this issue with the convergence of sph is well - known in some disciplines ( e.g. @xcite ) .
however , the specific problem we analyze here has been virtually ignored in discussions of the use of sph in applications to cosmology and galaxy formation .
in sph , densities are estimated from the particles by smoothing .
this involves convolving a continuous field quantity , @xmath26 , with a smoothing function @xmath27 through @xmath28 where @xmath29 and @xmath30 denote the true field and its smoothed version .
the smoothing length @xmath4 represents a characteristic width of the kernel over which the desired quantity is spread .
the smoothing kernel is normalized to unity and asymptotically approaches a dirac @xmath31-function when @xmath1 .
this can be achieved with a sufficiently large number of particles to describe the continuous system as @xmath0 .
assuming that the volume within the smoothing kernel is sufficiently sampled with these discrete points , we can further approximate the above integral by discrete summations .
the volume " element @xmath32 is estimated from @xmath33 where @xmath34 and @xmath35 are the mass and density of particle @xmath36 and we have @xmath37 with this approximation , we immediately obtain the density estimate in its discrete form @xmath38 as used in sph according to : @xmath39 where the sum is over all the particles within the volume element centered at @xmath40 . a smoothing function with compact support
is usually used to limit the volume " to finite extent in order to minimize computation time .
equation now operates on a finite number @xmath5 of particles near a given point . in order to satisfy the consistency of this step , which is to approach the continuous limit with such finite summations ,
the condition @xmath2 is required . moreover , in applications with finite @xmath3 and @xmath5 , there is a lack of self - consistency based on the above approach .
if we would represent a scalar field @xmath29 with a constant value , say @xmath41 , the following condition @xmath42 is generally not satisfied with the density field estimated from .
this lack of self - consistency is most extreme where large density gradients are present , e.g. between two phases of flow @xcite or close to a boundary @xcite . in other words ,
the volume estimate in sph from equation with a finite @xmath5 is not an accurate partitioning of space . in order to satisfy the condition of consistency and self - consistency ,
the following conditions @xmath43 indeed should be met if no other fixes such as normalization are applied .
next we discuss the consistency of the euler equations that sph actually solves .
the original euler equations in lagrangian form describe the evolution of density , momentum , and internal energy according to : @xmath44 @xmath45 @xmath46 where @xmath47 , @xmath48 and @xmath49 represent the density , velocity , and internal energy per unit mass .
the equation of state @xmath50 is a closure equation for the above system , where @xmath51 is the adiabatic index of the gas . following the derivation in @xcite , the equations of motion
can be discretized as @xmath52 @xmath53 @xmath54 where the factors @xmath55 and @xmath56 depend on derivatives of the density with respect to the smoothing lengths .
in practice instead of @xmath49 , we can also use a variable @xmath57 from @xmath58 to solve the energy equation as in @xcite .
these two approaches are equivalent in principle .
@xcite have calculated the errors in the continuity and momentum equations with the above sph formulation assuming a finite @xmath5 ( ignoring the factors @xmath55 , which spoils the fully conservative nature of the equations ) as @xmath59 @xmath60 where @xmath61 , @xmath62 are matrices close to the identity matrix @xmath63 and @xmath64 is a non - vanishing error vector .
@xmath61 , @xmath62 and @xmath64 are all determined by the particle distribution within smoothing length @xmath4 ( see @xcite for detailed definitions ) . as a result
, these errors will remain at the same level while we increase @xmath3 if @xmath5 is fixed .
however , a consistent scheme would require all the error terms in the above two equations go down as @xmath65 .
similar conclusions can also be drawn for the energy equation .
based on the above discussions of the density estimate and the discretized euler equations in sph , the condition @xmath66 , @xmath67 and @xmath68 has to be met in order to have a _ consistent and self - consistent _ scheme .
moreover , reducing @xmath4 will lead to finer time steps according to the cfl condition .
if the magnitude of the error from a fixed @xmath5 remains constant , the overall error will build up more rapidly in a high resolution run than a lower resolution one .
the error in sph from the smoothing procedure ( independent of particle distribution ) can be approximated to lowest order by : @xmath69 for the usual b - splines @xcite and the wendland functions @xcite , this is a second order term where @xmath70 = 2 .
higher order precision can be achieved by constructing different classes of smoothing kernels . however , not all are appropriate since smoothed estimates of positive definite quantities can take on negative values .
the error from discretizing the integral convolutions in the smoothing procedure depends on whether or not particles are ordered in space . optimistically ,
if the particles are quasi - ordered and not randomly distributed , it has been conjectured that the discretization error can be approximated by @xcite @xmath71 this estimate is based on the study of the complexity associated with the low discrepancy sequence in numerical integration by @xcite .
this indicates that the actual number of particles within the volume of each sph particle should match the _ expected _ number of particles .
clearly , this is highly dependent on the randomness in the actual distribution of sph particles .
the pressure force between sph particles always tends to regulate the distribution of the sph neighbors into an isotropic distribution .
however , this restoring " pressure force does not push the particles into an exact desired distribution since the number of neighbors is finite .
hence , the velocity field exhibits noise on small scales .
such velocity noise will further induce fluctuations in the density field .
in realistic applications , especially with highly turbulent flows , it is questionable whether the pressure forces can effectively regulate sph particles into a quasi - ordered configuration . on the other hand ,
if the distribution of particles is _ truly _ random , the convergence of sph should follow ( slow ) monte - carlo behavior : @xmath72 we further parameterize the dependence of discretization error on the number of neighbors from these two extreme situations as @xmath73 where @xmath74 . from the perspective of convergence
, we require that as the total number of sph particles is made larger , we will be able to simulate finer scale structure .
the smoothing error is already consistently reduced by making @xmath4 smaller . at the same time
, we must also use more neighbors in the discrete sums to combat the discretization error , which will eventually dominate if no action is taken .
we can write these conditions as @xmath75 if we parameterize the dependence of smoothing length @xmath4 on @xmath3 written by @xmath76 then the relation between @xmath5 and @xmath3 is just @xmath77 the requirements of convergence impose the follow conditions on the dependence of @xmath78 on @xmath3 : @xmath79 and @xmath80 for any case @xmath81 , there is a power - law dependence of @xmath5 on @xmath3 . in order to balance the smoothing and discretization errors ,
i.e. @xmath82 , we have @xmath83 .
the dependence of of @xmath4 and @xmath5 can be expressed as , @xmath84 for commonly used smoothing kernel forms , we have @xmath85 .
recall for random distributions we have @xmath86 and for quasi - ordered distributions @xmath87 , so @xmath78 for these two extremes is @xmath88 and @xmath89 .
the dependence of @xmath5 on @xmath3 is thus between @xmath90 $ ] .
this suggests a simple intermediate choice is given by : @xmath91 in detail , other choices would follow if higher - order smoothing kernels are used , which would require an even stronger scaling of @xmath5 with @xmath3 .
this can be readily seen from the expression of @xmath5 in equation [ eq : ngbscaling ] , where the power - law index @xmath92 approaches 1 for sufficiently large @xmath70 if @xmath51 fixed .
for example we have @xmath93 for the smoothing kernel functions constructed by @xcite , and then @xmath5 should vary between @xmath94 $ ] .
this indicates that the regularity in the particle distribution , which poses a strong limit for numerical interpolation , has to be improved to be much better than a quasi - ordered configuration if higher order smoothing kernels are employed .
we note that an earlier suggestion of increasing @xmath5 to reduce the error was proposed by @xcite from 1-d error analysis of sph .
on the surface , their analysis on the arbitrary spaced particles is similar to our result of quasi - order distribution .
the difference here is that the discretization error in sph in multi - dimensions is much worse than the _ truncation error _ inferred from the second euler - maclaurin formula in 1-d given by @xcite .
a direct application with the scaling in equation [ eq : ngbscaling ] will not work with classical sph codes .
if more than @xmath95 neighbors are used in the discrete estimates , the sph particles will quickly form close pairs , as demonstrated in @xcite and @xcite .
such a configuration actually makes the discretization errors worse , which is opposite to our purpose . for the kernel functions widely used in sph codes , the function @xmath96 tends to flatten when two particles are close together .
consequently , the net repulsive force also diminishes between close pairs .
when a large number of neighbors @xmath5 is used , particles will be less sensitive to small perturbations within the smoothing kernel .
in reality , the density estimate error fluctuates around the true value as a function of @xmath5 as shown in @xcite .
the whole system thus favors a distribution of particles with close pairs which minimizes the total energy .
hence the entire system would not achieve the resolution one aims for .
this clumping instability is the primary reason that the convergence study with variable @xmath5 is not possible in sph simulations @xcite with conventional smoothing kernels . with the cubic spline ,
sometimes additional repulsive forces are added in the inner smoothing length in the equations of motion to fight against the pairing instability , as in @xcite .
@xcite instead proposed a centrally peaked kernel to mitigate against clumping .
@xcite have shown that wendland functions are free from the clumping instability for large @xmath5 when used to perform smoothing in sph .
the desirable property of the wendland functions is that they are smooth and have non - negative fourier transforms in a region with compact support .
similarly , @xcite has shown that the sinc family of kernels is able to avoid particle clumping . in what follows
, we adopt the wendland @xmath97 function , defined on @xmath98 $ ] as @xmath99 in the p - gadget-3 code ( an updated version of gadget-2 @xcite ) to study the convergence rate of sph .
we also update the viscosity switch in p - gadget-3 using the method by @xcite to effectively reduce the artificial viscosity in shear flows while maintaining a good shock capturing capability .
this effectively reduces the viscosity in the flows producing results closer to the _ inviscid _ case .
this switch also greatly enhances the accuracy of sph in the subsonic regime especially where large shear is present .
the time integration accuracy is also improved following the suggestions by @xcite and @xcite . to verify the absence of the clumping instability with our new code , we plot the distribution of distances to the closest neighbor particle from a realistic situation . in figure
[ fig : pair_instability ] , we compare the performance of the wendland @xmath97 with the original cubic spline for the gresho vortex test @xcite .
the details of the set - up can be found in the following sections . in the upper panel of figure [ fig : pair_instability ] , we show the distribution of the distance to the nearest neighbor for each sph particle in a subdomain of the simulation at @xmath100 with the cubic spline .
this test involves strong shearing motions from a constant rotating velocity field . at @xmath100 ,
the vortex has completed almost a full rotation .
the distance to the nearest neighbor is further normalized to the mean interparticle separation in the initial conditions . with @xmath101
, the distribution of distances shows that most of the particles are still well - separated from one other .
this situation quickly worsens as soon as we use @xmath102 , which is the critical number for the cubic spline reported by @xcite .
a significant portion of the particles now have their closest neighbor within a factor of @xmath103 of the initial separation . for an even higher @xmath104 ,
most of the particles have formed close pairs as indicated by the peak near @xmath105 . as a comparison
, the lower panel of figure [ fig : pair_instability ] shows that close particle pairs are not present with the wendland @xmath97 as we vary @xmath5 from 100 to 200 ( equivalent to @xmath10655 with cubic spline ; see @xcite ) to 500 .
this test confirms the good performance of sph against the pairing instability with the wendland function and gives us confidence that we can use this code to perform further tests to study the convergence rate with varying @xmath5 .
we use a randomly distributed particle set to test the density estimate in sph .
@xmath107 particles are distributed randomly within a 3-d box with unit length .
we use the density estimate routine in sph with varying @xmath5 from 40 to 3200 to derive the density field accordingly . in figure
[ fig : random_data_set ] , we plot the histogram of the density estimated at the position of each particle for varying @xmath5 .
the variance within the density field is largest for small @xmath5 values , which are shown in the red colors .
as the number of neighbors increases , the distribution slowly approaches a gaussian like distribution peaked at @xmath108 .
this behavior shows the trade - off between variance and bias in the sph density estimate .
larger @xmath5 hence larger @xmath4 will increasingly smear out fluctuations on short scales .
however this does not show the actual accuracy of the sph density estimate .
in fact , the _ precise _ density field in this set - up is difficult to give but the behavior of the density field can be seen from a pure statistical argument . to be more specific
, the number of particles falling within a fixed volume associated with each particle follows a _ poisson _ distribution .
the expected number of such occurrences goes linearly with larger volume ( larger @xmath5 ) . as a result
, the standard deviation of the density distribution follows a @xmath109 trend .
so should the trend of the standard deviation of the density distribution given by sph if the density estimate is accurate .
however , this is not strictly true as we show in the lower panel of figure [ fig : random_data_set ] , where the black crosses are the measured standard deviation of the density distribution with varying @xmath5 .
there is a significant spread that is much greater for small @xmath5 than the guiding line indicating @xmath109 .
the deviation is already significant for @xmath110 , which is the recommended number by @xcite based on a shock tube test . in the upper panel of figure [ fig : random_data_set ]
, we can also see there is a long tail for small @xmath5 values .
this suggests there is an overestimate of density where we suspect the self - contribution of each particle to the density estimates plays an exaggerated role . in order to verity this ,
we carry out density estimates with self - contribution excluded for the same configuration of particles .
the standard deviation of the density distribution as a function of @xmath5 is shown in the red filled circles in the lower panel of figure [ fig : random_data_set ] .
this brings the relation much closer to the expected @xmath109 line where the deviation at small @xmath5 is greatly reduced due to the subtraction of the self - contribution term . excluding the self - contribution is not the usual practice in modern sph codes .
@xcite actually advocated to use equation [ eq : sphdensity ] with self - contribution excluded to give a better density calculation based on a similar comparison in 2-d .
@xcite however argues that the distribution of sph particles is essentially different from a random distribution .
if particles are distributed randomly , individual particles do not care about the locations of other particles . in sph , due to the repulsive pressure force , each sph particle is trying to establish a zone where other sph particles can not easily reside .
however , as we see in figure [ fig : pair_instability ] , sph forms close pairs so that the argument by @xcite is no longer valid . note that it is not uncommon to see a value above @xmath102 in the literature with the cubic spline .
the sph density estimate will then give a biased result for irregularly distributed particles .
this comparison actually favors the use of a large @xmath5 where the distribution is highly disordered .
although the above experiment is based on a situation where the particle distribution is truly random , the realistic situation is between such a truly random distribution and a quasi - regular one .
however , quasi - regular distributions are hard to achieve in multi - dimensional flows because of shear .
c c a random distribution of particles , as discussed in the above section , is an extreme case for sph simulations .
we relax these particles according to the procedure described in @xcite to allow them to evolve into a glass configuration . given an @xmath111 repulsive force , the particles settle in a equilibrium distribution which is quasi force - free and homogeneous in density .
this mimics the other extreme case where sph particles are well regulated by pressure forces .
nevertheless , the outcome of this relaxation is not perfect where tiny fluctuations in the density field are still present because of the discrete nature of the system .
we now measure the sph density estimate on this glass configuration for different @xmath5 and estimate the standard deviation of the density for each @xmath5 .
as the upper panel in figure [ fig : glass_data_set ] shows , the density distribution is much narrower than that for a random set of points . for sufficiently high @xmath5 ,
the distribution approaches a dirac-@xmath31 distribution .
the convergence is also much faster than in figure [ fig : random_data_set ] .
in fact , we observe a @xmath112 trend for the glass configuration .
for this set up , there is no need to exclude the self - contribution term , which is consistent with @xcite .
this measured @xmath112 rate is quite interesting in itself .
@xcite conjectured that the discretization error with sph behaves as @xmath113 .
this means that the randomness in the distribution of sph particles is closer to a low discrepancy sequence rather than to a truly random one .
our experiment shows this @xmath112 ( neglecting the @xmath114 ) rate is certainly appropriate for a glass configuration , which is carefully relaxed to an energy minimum state .
however , this rate may not hold for a general applications .
thus , these two examples support the dependence of the density estimate error on @xmath5 as in the previous analysis .
namely , @xmath109 for truly random data points and @xmath112 for a distribution with good order .
realistic applications will fall between these two extremes and could be more biased towards the @xmath109 case .
c c we have noted that a constant scalar field is not exactly represented by the sph method . based on our discussion of the self - consistency of the sph density estimation , we can further measure the magnitude of this inconsistency and its dependence on the number of neighbors @xmath5 .
define a parameter @xmath115 as @xmath116 which represents the sph estimate of a constant scalar field @xmath117 .
this parameter @xmath115 should follow a peaked distribution around 1 with some errors .
we compute this parameter with a simple extra routine after the density calculation in the code since this parameter requires the sph density field as input . in figure
[ fig : volume_error ] , we plot the distributions of the calculated values of @xmath115 for randomly distributed particles with several different @xmath5 values .
the distributions of @xmath115 are indeed rather broad around 1 .
similarly as the errors in the density estimate , the standard deviation of the distribution is consistently larger for smaller @xmath5 values .
some extreme values of q with small @xmath5 are even on the order of unity .
the peaks of the distribution with small @xmath5 are actually slightly below 1 , which indicates some bias towards a density overestimate .
this is consistent with the above discussion of the self - contribution to the density overestimate at small @xmath5 values . for this random distribution of particles , increasing @xmath5 will lead to a more accurate volume estimate by reducing the inconsistency in the density estimate .
for example , the spread of the distribution of @xmath115 for @xmath118 is greatly reduced compared to @xmath119 .
as we repeat this experiment with even more neighbors , the distribution of @xmath115 approaches a dirac-@xmath31 function eventually with the error in the volume estimate reduced to zero . in this limit ,
the self - consistency of sph is finally restored . however , as in the lower panel of figure [ fig : volume_error ] , such an improvement depends slowly on the number of number @xmath5 at a rate @xmath120 .
we also consider the cubic spline in this test , which is included in the lower panel of figure [ fig : volume_error ] .
the dependence of q on @xmath5 for a cubic spline follows the same @xmath109 trend .
the cubic spline shows a slightly lower magnitude for the same @xmath5 compared with the wendland @xmath121 function .
this is because the wendland @xmath121 is slightly more centrally peaked than the cubic spline . to have the same effective resolution length ,
a higher number of @xmath5 is required for wendland @xmath121 ( see the table of such scaling in @xcite ) .
the behavior of @xmath115 with respect of @xmath5 is consistent with the two smoothing functions .
we then repeat this calculation with the particles in a glass configuration .
the results of the wendland @xmath121 function and cubic splines are presented with the solid symbols in figure [ fig : volume_error ] .
as we see from the previous section , the relaxed particle distribution reduces the errors in the density estimate .
the inconsistency in the volume estimate , as measured in @xmath115 , declines as a function of @xmath122 in this case .
this is true for both the wendland @xmath121 function and cubic spline . in practice ,
when we use sph , the true distribution is unknown .
so the error in the density estimate is difficult to quantify .
we suggest a simple procedure to measure q in simulations to give us an estimate of the overall quality of the density estimate .
based on the comparison between the @xmath115 value for randomly distributed particles and particles in a glass configuration , a typical number @xmath123 with cubic spline as often adopted in the literature leads to a volume estimate error between 1@xmath124 and 10@xmath124 from the true value .
however , a glass configuration is a rather idealized and we suspect that sph particles will not maintain such good order in most applications .
this error in the volume estimate , which is self - inconsistency in the sph density estimate , also has dynamical consequences as the volume estimate , @xmath125 is used for the derivation of the equations of motion .
in addition to the error terms derived by @xcite , an error on the order of a percent level is already present from the density estimate step itself .
the quality of the particle distribution is crucial to minimizing this error .
however , once the particles deviate from a regular distribution , the errors in the equations of motion will pose another challenge for the particles to return back to an ordered configuration .
one way to impose a small error in the volume estimate is to use a large @xmath5 .
increasing @xmath5 can regulate this error with a dependence between @xmath109 and @xmath112 .
in this section , we perform dynamical tests with sph for several problems to study the convergence rates in each situation .
the evolution in multi - dimensional flows is more complex than in 1-d and , as noted in 1 , one - dimensional tests can not be used to judge convergence of sph in multi - dimensions .
the test problems are chosen so analytic behaviors are known in advance ( @xcite ) .
also the test problems consider both smooth flows and highly turbulent flows .
these test problems are purely _ hydro _ as no further compilations from other factors such as gravity are present .
we also do not consider problems with shocks for our tests as the presence of discontinues will reduce the code to being at best first order accurate .
the first problem is the gresho vortex test @xcite .
we quantify the @xmath126 velocity error with sph and its convergence rate as a function of the total number of sph particles .
this test involves a differentially rotating vortex with uniform density in centrifugal balance with pressure and azimuth velocity specified according to @xmath127 and @xmath128 the value of @xmath129 is taken to be @xmath130 following @xcite and @xcite , as in the original paper of @xcite .
@xcite and @xcite also investigate the behavior by varying this constant background pressure @xmath131 .
the velocity noise will decline for decreased @xmath131 such that the system is more stable .
this test problem is suitable for examining the noise in the velocity field since the density field is uniform . the vortex should be time independent having the initial analytical structure .
@xcite however found that this test poses a challenge to sph since the vortex quickly breaks up with gadget .
@xcite , @xcite , and @xcite have tested their sph formulation with the same problem and found improvements with a better parameterization of artificial viscosity . the convergence rate reported by @xcite and
@xcite is similar to the one given by @xcite , @xmath132 , while @xcite give a rate of @xmath133 .
this discrepancy is due to a modified version of the equations of motion by @xcite .
the set - up of this problem has three sharp transition points which is problematic for sph .
this test is also very sensitive to the particle noise which will can induce a false triggering of artificial viscosity .
we initialize the particles on an n@xmath134n@xmath13416 lattice .
the number of particles vertically assures that we can extend @xmath5 to several thousand for each particle without overlap . the @xmath126 velocity error as a function of the number of effective 1-d resolution elements @xmath3
is shown in figure [ fig : gresho_convergence ] .
the black squares are the simulations with @xmath135 and the blue filled circles are the ones with @xmath110 . for the red diamonds , we run the simulation with increasing @xmath5 as @xmath136 starting from @xmath135 at the lowest resolution @xmath137 .
the convergence rates are as @xmath138 , @xmath132 and @xmath139 respectively .
the former two are consistent with @xcite , @xcite and @xcite .
also , a larger number of neighbors gives a slightly faster convergence rate in this test ( @xmath138 vs. @xmath132 ) .
for @xmath135 and @xmath140 , the @xmath141 velocity error at the highest resolution at @xmath142 is actually _ above _ the convergence rate .
this is what is expected based on the argument for consistency in the sph method .
a fixed @xmath5 may work well for a range of different resolutions until the resolution is ultimately so high that the errors from the density estimate , from the discretized continuity equation and momentum equation , are not consistently reduced according to the corresponding resolution length @xmath4 .
the result with varying @xmath5 , on the other hand , shows a much faster convergence rate as @xmath139 with the highest resolution at @xmath143 with @xmath144 .
this validates the arguments presented earlier and shows the result given by sph does _ converge _ to the proper solution and the convergence rate is improved in this test problem as we increase @xmath5 when @xmath65 .
such a convergence rate is actually close to the rate @xmath145 measured with moving mesh code arepo @xcite for the same problem . .
convergence rate of the velocity field in the gresho vortex test as a function of @xmath146 with sph .
blue filled circles are the results with a fixed number of neighbors @xmath110 while the black filled squares are the ones with a smaller number of neighbors @xmath147 .
the convergence rate shows a @xmath132 behavior for the former and @xmath138 for the latter .
the red squares are with increased @xmath5 as a function of total number of sph particles @xmath3 .
the convergence rate for this run scales almost as @xmath139 .
such a convergence rate is close to @xmath133 for the grid - based code in @xcite . ]
we have also included two series of the gresho vortex test with the cubic spline as in the green symbols in figure [ fig : gresho_convergence ] .
the performance with the cubic spline is disappointing in this test . with @xmath148 ,
the @xmath126 error reaches to a minimum value at an intermediate resolution at @xmath149 . for @xmath150 , where the clumping instability is reducing the effective resolution ,
the minimum @xmath126 error is achieved at @xmath151 . in terms of the absolute error
, the highest resolution actually fails to give the best results .
this behavior is also observed in the gresho vortex test and the kh instability test in @xcite .
this is another vortex problem described by @xcite and employed by @xcite and @xcite .
it also has a time - invariant analytic solution . unlike the previous test ,
the density and velocity profiles are both smooth .
this problem will test the convergence rate of the solver for smooth flows .
we generate the initial conditions in a box of size @xmath152 ^ 2 $ ] with periodic boundary conditions .
similar to the previous test , several additional identical layers are stacked along the @xmath153 direction to get a thin slab configuration .
the velocity field is set up as @xmath154 @xmath155 the distribution of the density field and the internal energy per unit mass are calculated according to the following temperature distribution @xmath156 as @xmath157 and @xmath158 .
consequently , the entropy @xmath159 is a constant within the domain . as in @xcite
, we then measure the @xmath160 error in the density field at @xmath161 obtained with our code with respect to the analytical density field for different resolutions .
the result if shown in figure [ fig : yee_convergence ] where the blue filled circles are obtained with @xmath110 while the red diamonds are with varying @xmath5 .
the dashed line is a @xmath162 trend with the dashed - dotted line taken from @xcite measured with the arepo code .
the overall magnitude of the error with sph is also close to the one for the arepo simulation in @xcite .
in fact , the result for @xmath110 with sph shows a second order convergence for the first four resolutions , but for the two highest resolution runs the @xmath160 density error starts to reach a plateau . beginning with @xmath140 at @xmath163 ,
we re - simulate the two high resolution runs with @xmath164 . as a result , the overall error is reduced and is now consistent with the expected rate of @xmath162 . the above result based on the isentropic vortex problem shows that sph _ can _ be a second - order accurate scheme as long as the other errors do not dominate over the smoothing error " introduced from the density estimate .
again , the condition @xmath165 and @xmath166 needs to be satisfied .
as for the fixed @xmath5 run , starting from @xmath167 the other errors rather than the smoothing error " become dominant and start to degrade the performance of sph . in order to see the effects of the discretization error , we have repeated this experiment with smaller @xmath5 and also using the cubic spline .
for the two series of @xmath168 and @xmath169 , as in green symbols in figure [ fig : yee_convergence ] , the convergence rate systematically slows down from @xmath170 to @xmath171 from a second order convergence rate . the convergence rate with the cubic spine , where @xmath148 , is the slowest one .
moreover , it shows signs of turning over at the highest resolution with the cubic spline . .
convergence rate of the isentropic vortex test as a function of @xmath146 with sph code .
the error is computed from the @xmath160 norm of the density field at @xmath161 when the vortex has completed a full rotation .
blue filled circles are the result for a fixed number of neighbors @xmath110 .
the convergence rate exhibits a @xmath162 behavior except for the two high resolution runs .
the red filled diamonds are the two high resolution runs for increasing @xmath5 and now the error is consistent with the expected @xmath162 trend . ] in this section , we move to a more complex problem involving the chaotic feature of multi - dimensional flows with sph . in particular , we simulate subsonic turbulence in the gas ( which is weakly compressible ) . following @xcite and @xcite , we employ a periodic box with unit length filled with an isothermal gas @xmath87 with unit mean gas density and unit sound speed .
the initial positions of the particles are set on a cubic lattice .
the stirring force field is set up in fourier space and we only inject energy between k = 6.28 and k = 12.56 .
the amplitude of these modes follows a -5/3 law as : @xmath172 , while the phases are calculated for an ornstein@xmath173uhlenbeck process @xmath174 such that a smoothly varying driving field with finite correlation time is obtained .
the parameters in the ornstein@xmath173uhlenbeck process are identical to the ones adopted by @xcite and @xcite as well as @xcite .
a helmholtz decomposition in k - space is applied before the fourier transformation to yield a pure solenoidal component in the driving force field . at every time step , before the hydro force calculation , this driving routine is called to compute the acceleration .
the magnitude of the driving force is adjusted such that the @xmath175 velocity is at @xmath176 once the system reaches a quasi - equilibrium state . besides the routine to calculate the driving force , we have also included a function to measure the velocity power spectrum on the fly .
the turbulent velocity power spectrum is calculated whenever a new snapshot is written .
we use the
nearest neighbor sampling " method for our estimate .
finally the 1d power spectrum is obtained by averaging the velocity power with each component in fixed @xmath177 bins .
we use the time - averaged velocity power spectra between @xmath178 and @xmath179 for the following comparison .
we emphasize that our sph code is able to give improved performance in this test over the previous studies of @xcite and @xcite .
a snapshot of the velocity field is shown in figure [ fig : subsonic_high_res ] in a high resolution test with @xmath180 particles .
our sph code is able to resolve finer structures compared to the other two studies noted .
the velocity field is qualitatively closer to the one simulated with grid based codes .
this is a result of the use of the artificial viscosity switch proposed by @xcite and the wendland @xmath121 smoothing kernel .
four different resolutions , @xmath107 , @xmath181 , @xmath182 and @xmath183 , are used for a resolution study .
we actually fix the smoothing length @xmath4 for all the simulations based on @xmath110 for the @xmath107 resolution run and subsequently scale @xmath5 for the other three cases .
this is done to directly test the effect of varying @xmath5 on the hydrodynamics by controlling the same resolution length @xmath4 so that the smoothing error is fixed .
the projected velocity maps of these four resolutions are shown in figure [ fig : turbulence_resolution ] .
the velocity fields are similar to one another between these four simulations . in figure
[ fig : turbulence_convergence ] , we compare the velocity power spectrum for these four resolutions .
we multiply the power spectrum by @xmath184 so that the inertial range described by @xmath185 shows up as a horizontal line . the power spectrum shows good agreement with the @xmath186 law on large scales .
the velocity power goes down significantly below that once the wave number @xmath187 is above 40 .
however , there is significant power contained on the smallest scales .
the spatial extent where neighbor particles are searched corresponds to @xmath188 , which is the range where the turn - over is observed .
this turn - over can be seen as the result of dissipation of velocity noise on the resolution scale @xcite . improvements with varying @xmath5 in this test problems can be seen .
the effect of varying @xmath5 on the velocity power spectrum is visible on both large and small scales . with more neighbors , the velocity field on the kernel scale ( smoothing length @xmath4 )
is better regulated as the power is reduced by an order of magnitude .
this can be also seen in that the turn - over point for higher resolution increasingly shifts to a higher @xmath187 value . on large scales where the modes are well resolved
, more neighbors also help to bring the power spectrum into better agreement with the @xmath186 law .
this can be verified by comparing the red and purple lines for @xmath183 and @xmath107 in figure [ fig : turbulence_convergence ] .
the variance on the power spectrum between @xmath189 and @xmath190 is reduced for the latter case .
the velocity power contained within @xmath190 and @xmath188 is also significantly higher in the @xmath183 run with @xmath191 than @xmath107 with @xmath110 .
the result is that there is increasingly less power in the noisy motion on the kernel scale but more power in a coherent fashion .
the conclusion is that @xmath5 does play an important role in this subsonic turbulence test as suggested by @xcite .
this also agrees with the consistency requirement for sph in order to recover the continuous limit that the discretization errors " also need to be reduced with @xmath192 .
figure [ fig : volume_error_turbulence ] shows the value of @xmath115 , as defined above , in the turbulence tests at @xmath193 .
we have also over - plotted the trend of q from the static distribution of random particles and the glass configuration in the previous section .
this allows us to use @xmath115 , which describes the deviation from an exact partition of unity , to access the degree of disorder in realistic applications of sph .
in fact , the distribution of @xmath115 for these six snapshots are between the two static cases .
the trend of @xmath115 with respect to @xmath5 , however , is quite shallow among these dynamical cases , indicating a @xmath109 trend .
sph is not able to re - order the particles to a distribution with the degree of order characterized by the glass configuration .
the degree of disorder could be much higher in post - shock flows frequently encountered in astrophysical simulations than what we have found in these weakly compressible turbulence flows with sph since the density variation is much higher in the former . .
a high resolution subsonic turbulence test with @xmath180 particles showing the magnitude of the velocity field at @xmath194 for a thin slice in the box .
a cross comparison between this figure and the ones in @xcite and @xcite at the same resolution shows that our sph code is able to resolve the finest structures among all three .
the velocity field is qualitatively closer to the one simulated with a grid based code .
this is a result of the use of the artificial viscosity switch proposed by @xcite and the wendland @xmath121 smoothing kernel .
@xmath195 is used in this example . ]
ll ll the velocity power spectrum in a driven subsonic turbulence test at different resolutions .
the velocity power is multiplied by @xmath184 , where @xmath187 is the wave number , such that the inertial range will show up as a horizontal line .
different total numbers of sph particles are used while the number of neighbors are set up such that the smoothing length " @xmath4 is the same for all the simulations . even with the same nominal
h - resolution " , the @xmath183 run shows an improvement over the @xmath107 run , especially for the location of the point of turn - over from the minimum on small scales and the increased velocity power on large scales . ] the distribution of @xmath196 for several snapshots at the same @xmath197 in the subsonic turbulence tests in order to access the degree of disorder in realistic sph applications .
the trend of q from the static random particles and the glass configuration are also over - plotted here .
the distribution of @xmath115 for these six snapshots is between the two static cases .
the trend of @xmath115 with respect to @xmath5 , however , is quite shallow in these dynamical cases .
these points roughly follow a @xmath109 trend .
sph is not able to reorder the particles to a distribution with the degree of order characterized by a glass configuration . ]
the main findings of our study are as follows : * we argue that the requirements of consistency and self - consistency within the sph framework require that @xmath198 and @xmath199 need to be satisfied to obtain the true continuum behavior of a flow .
we summarize the errors from the smoothing step and from the discretization step in the sph density estimate and propose a power - law scaling between the desired @xmath5 given the total number of sph particles @xmath3 in order to have a consistent and convergent scheme .
the dependence @xmath200 $ ] is derived based on a balance between these two types of errors for commonly used smoothing kernels .
* we verify the error dependence on @xmath5 for the discretization error with a quasi - ordered glass configuration and a truly random one .
the range for the power - law index between @xmath5 and @xmath3 is obtained based on these two extreme conditions .
the discretization error in the density estimate decreases roughly as @xmath201 for the glass configuration and as @xmath109 for the random distribution .
* we propose a simple method to calculate a parameter @xmath115 , which measures the deviation from an exact partition of unity in sph .
the error in the volume estimate in sph is further examined for a glass configuration and a random one . in agreement with the behavior of the density estimate error
, this quantity also shows the same dependence on @xmath5 in these two situations . in realistic applications ,
the error introduced with the density estimate step itself is on the order of several percent for the @xmath5 value used in most published work . *
we use the wendland @xmath121 function as a new smoothing kernel to perform dynamical tests with a varying @xmath5 to avoid the clumping instability .
we confirm that particle pairs do not form even with a large value of @xmath5 .
the artificial viscosity by @xcite also greatly improves the performance of our sph code . as a consequence ,
the results from our study are not influenced by the numerical artifacts from these two aspects as in some of the previous studies . * a fixed @xmath5 , which indicates an inconsistent scheme , gives a slow convergence rate as reported by previous studies in the two vortex problems .
as shown in the figures , the convergence rate levels off and eventually turns over for sufficiently fine resolution . also the error obtained with the cubic spline is larger than for the wendland @xmath121 function .
the cubic spline actually exaggerates the discretization error as the wendland @xmath121 function is much smoother . * for smooth flows , as seen in the isentropic vortex problem
, varying @xmath5 according to the proposed power - law dependence can give second - order convergence .
the same improvement is seen in the gresho vortex test , where several discontinuities are included .
the convergence rate @xmath202 with varying @xmath5 is closer to the rate @xmath203 measured with the second - order accurate moving mesh arepo code for the same problem . * for highly turbulent flows
, the velocity noise poses another challenge in addition to the errors in the density estimate for the sph particles to retain an ordered configuration .
though our code is able to give a better result in the subsonic turbulence test compared with previous studies , the velocity noise on the kernel scale is always present .
this velocity noise can be reduced by an order of magnitude with a larger @xmath5 , but the inertial range resolved with sph code only shows a slight improvement . *
we measure the randomness of sph particles in the above subsonic turbulence tests using the calculation of @xmath115 .
this further verifies our assumptions when deriving the dependence of @xmath5 on @xmath3 .
the distribution of sph particles in this example is indeed between a truly random configuration and a quasi - ordered case .
this approach can be used in general to quantify the randomness in sph in applications .
unfortunately , sph is not able to rearrange the particles into a highly ordered configuration characterized by a low discrepancy sequence . *
the usual practice of using a fixed @xmath5 indicates that the same level of noise will be present even if higher resolution is used .
this error can be stated as a
zeroth order " term which is _ independent of resolution_. moreover , the errors introduced can be roughly approximated by a gaussian distribution around the true value provided @xmath5 is large enough so that the self - contribution term is not significant even for highly random distributions .
we emphasize that the inconsistency and the errors associated with particle disorder is not a specific feature to a particular sph code but is rather a generic problem .
in fact , the set of equations in gadget are fully conservative since they are derived from the euler - lagrange equations ( see @xcite , @xcite and @xcite ) .
to our knowledge , this approach has not been fully adopted by the community yet .
the conservation of mass , linear momentum , angular momentum and total energy has been long recognized to be the unique advantage over other techniques . indeed , its conservative nature , its robustness and its simple form have made it a popular tool for astrophysical modeling and in other disciplines .
however , the errors associated with the finite summation @xmath204 , usually termed as noise " in the sph literature , and its impact on the convergence of sph has not been well understood so far . as we discussed earlier , such an error naturally and inevitably emerges from the density estimate itself and also from the momentum equation when evolving the system . obviously , one promising direction to improve the particle method is to use a more accurate volume estimate .
it is possible to derive a numerical scheme which can reproduce a polynomial up to the @xmath205 order polynomial exactly following the procedure by @xcite . in practice
, it is sufficient to eliminate the zero - th order error by renormalization such that the corrected scheme is second - order accurate by symmetry .
recently , @xcite proposed a new class of gridless lagrangian methods which use finite - element godunov schemes .
these gridless methods dramatically reduce the low - order errors and numerical viscosity in sph method and significantly improve accuracy and convergence .
they show better performance in resolving fluid - mixing instabilities , shocks and subsonic turbulence , which is critical in a wide range of physical and astrophysical problems .
we caution against equating numerical precision with physical accuracy of a simulation because both physics and numerics play roles in the modeling of a complex system .
for example , we compared cosmological hydrodynamical simulations of a milky - way like galaxy using both the improved gadget code ( presented in this paper ) and the gridless gizmo code by @xcite with the same initial conditions , physics prescriptions and resolutions , and found that both simulations produced similar galaxy properties such as disk mass , size and kinematics , although gizmo resolves the spiral structures better .
these results suggest that the robustness of astrophysical simulations depends not only on the numerical algorithms , but also on the physical processes , as highlighted also in a number of recent papers on galaxy simulations ( e.g. , @xcite ) .
we thank volker springel , phil hopkins , mark vogelsberger , jim stone , andreas bauer and daniel price for valuable discussions on this work .
we further thank paul torrey for his contributions to the experimental setups in the early stage and suggestions throughout the project .
lh acknowledges support from nasa grant nnx12ac67 g and nsf grant ast-1312095 , yl acknowledges support from nsf grants ast-0965694 , ast-1009867 and ast-1412719 .
we acknowledge the institute for cyberscience at the pennsylvania state university for providing computational resources and services that have contributed to the research results reported in this paper .
the institute for gravitation and the cosmos is supported by the eberly college of science and the office of the senior vice president for research at the pennsylvania state university . ,
o. , moore , b. , stadel , j. , potter , d. , miniati , f. , read , j. , mayer , l. , gawryszczak , a. , kravtsov , a. , nordlund , . ,
pearce , f. , quilis , v. , rudd , d. , springel , v. , stone , j. , tasker , e. , teyssier , r. , wadsley , j. , & walder , r. 2007 , , 380 , 963 , a. c. , fryxell , b. , plewa , t. , rosner , r. , dursi , l. j. , weirs , v. g. , dupont , t. , robey , h. f. , kane , j. o. , remington , b. a. , drake , r. p. , dimonte , g. , zingale , m. , timmes , f. x. , olson , k. , ricker , p. , macneice , p. , & tufo , h. m. 2002 , , 143 , 201 , c. , wadepuhl , m. , parry , o. h. , navarro , j. f. , jenkins , a. , springel , v. , teyssier , r. , carlson , e. , couchman , h. m. p. , crain , r. a. , dalla vecchia , c. , frenk , c. s. , kobayashi , c. , monaco , p. , murante , g. , okamoto , t. , quinn , t. , schaye , j. , stinson , g. s. , theuns , t. , wadsley , j. , white , s. d. m. , & woods , r. 2012 , , 423 , 1726 , j. , crain , r. a. , bower , r. g. , furlong , m. , schaller , m. , theuns , t. , dalla vecchia , c. , frenk , c. s. , mccarthy , i. g. , helly , j. c. , jenkins , a. , rosas - guevara , y. m. , white , s. d. m. , baes , m. , booth , c. m. , camps , p. , navarro , j. f. , qu , y. , rahmati , a. , sawala , t. , thomas , p. a. , & trayford , j. 2014 , arxiv e - prints:1407.7040 | we study the convergence properties of smoothed particle hydrodynamics ( sph ) using numerical tests and simple analytic considerations .
our analysis shows that formal numerical convergence is possible in sph only in the joint limit @xmath0 , @xmath1 , and @xmath2 , where @xmath3 is the total number of particles , @xmath4 is the smoothing length , and @xmath5 is the number of neighbor particles within the smoothing volume used to compute smoothed estimates .
previous work has generally assumed that the conditions @xmath0 and @xmath1 are sufficient to achieve convergence , while holding @xmath5 fixed . we demonstrate that if @xmath5 is held fixed as the resolution is increased , there will be a residual source of error that does not vanish as @xmath6 and @xmath1 .
formal numerical convergence in sph is possible only if @xmath5 is increased systematically as the resolution is improved . using analytic arguments
, we derive an optimal compromise scaling for @xmath5 by requiring that this source of error balance that present in the smoothing procedure . for typical choices of the smoothing kernel , we find @xmath7 .
this means that if sph is to be used as a numerically convergent method , the required computational cost does not scale with particle number as @xmath8 , but rather as @xmath9 , where @xmath10 , with a weak dependence on the form of the smoothing kernel . |
ultra - luminous infra - red galaxies ( ulirgs ) , identified by iras and defined by their high far - infrared luminosities ( @xmath1810@xmath19 l@xmath20 ) , have spectral energy distributions dominated by the 8 - 1000-@xmath21 m mid - far ir emission ( e.g. , soifer et al . 1987 ; clements et al . 1996a ; sanders & mirabel 1996 ) and show luminosities and space densities in the local universe similar to those of qsos ( soifer 1986 ; soifer , houck & neugebauer 1987 ) . the excess of far - ir emission in these objects is generally thought to indicate the presence of large amounts of hot dust , but the mechanism for heating the dust remains controversial . intense nuclear starbursts and heavily - shrouded active galactic nuclei ( agn ) have both been advocated and a combination of both energy sources currently seems likely , with starbursts dominating at l@xmath22@xmath2310@xmath24 l@xmath16 and the agn - dominated fraction increasing to 35%@xmath2550% above this luminosity ( genzel et al .
1998 ; veilleux , sanders & kim 1999 ) .
ulirgs are almost always found in strongly interacting systems or advanced mergers ( clements et al .
1996a , b ; clements & baker 1996 ) , during which large - scale torques on the intersteller medium and loss of angular momentum drives gas towards the central few kilo - parsecs of the system , where it is thought to result in intense starburst and/or agn activity ( e.g. , barnes & hernquist 1991 ; mihos & herquist 1994 , 1996 ; norman & scoville 1988 ) . in this scenario , an ( obscured ) agn is switched on as the galaxy merger reaches an advanced state , and is then progressively unveiled as the circumnuclear starburst processes ( and expels ) the surrounding gas .
obscuration is high in the mid - ir emitting regions of ulirgs , with model - dependent extinctions ranging from a@xmath26@xmath05@xmath251000 ( genzel et al .
however , progression from starburst- to agn - powered emission as the merger advances is not evident in present studies ( e.g. , genzel et al .
if ulirgs do contain obscured progenitors of future optically - bright quasars ( e.g. sanders et al .
1988 ) they may offer a valuable opportunity to study an important stage in the evolution of quasars ( and powerful radio galaxies ) in the local universe . alternatively , highly obscured agn might not be common in the ulirg population since strong agn activity , once triggered , might quickly destroy the obscuring screen , producing an easily detectable agn ( e.g. lutz , veilleux & genzel 1999 ) . in this context , the prototypical ulirg arp 220 ( ic4553/4 , ugc09913 ) , a double - nucleus system , is an extremely interesting object . in the diagnostic diagrams of genzel et al .
( 1998 , their fig .
5 ) it appears as one of the most completely starburst - powered ulirg in their sample , despite the fact that it is in an advanced merger state . in these final stages of merging , peak gas densities ( and related activity ) are thought to occur , when the merging galaxies are within a @xmath0kpc of one another and are in the process of coalescing ( mihos & hernquist 1996 ) ; high angular resolution observations of the central kiloparsec are therefore vital to determine the gas dynamics in this important region .
this paper presents the first subarcsecond - resolution , spatialling - resolved spectral imaging of the distribution and kinematics of neutral hydrogen ( ) in the central 900 pc of arp 200 , improving on previous observations by more than a factor of ten in angular resolution .
hi is a valuable tracer of galactic structure and dynamics over a wide range of size scales and is a valuable tool for probing probe the gaseous conditions in the highly - obscured , compact nuclear regions of ulirgs which are often unaccessible at optical wavelengths .
indeed ulirgs contain large amounts of neutral hydrogen ( e.g. @xmath18 10@xmath27@xmath2510@xmath28 atoms - mirabel 1982 ) , and recent vla synthesis maps of _ emission _ in objects such as arp 220 ( hibbard , vacca , & yun 2000 ) , have revealed massive gaseous tidal features ( m@xmath29 @xmath18 10@xmath30 m@xmath16 ) extending over several tens of kpc beyond the optical emission and demonstrating the dramatic redistribution of material during galaxy merger .
unfortunately , such studies of in _ emission _ are limited by surface brightness sensitivity to angular resolutions of @xmath186@xmath31@xmath2510@xmath31 with current instruments . instead , _ absorption _ measurements can be performed to angular resolutions as high as a few milliarseconds in the presence of a bright radio continuum background source ( e.g. peck & taylor 1998 ; carilli & taylor 2000 ) ; subarcsecond resolutions achievable with merlin probe the distribution and kinematics of , in absorption , on scales of @xmath010@xmath25100 pc in nearby galaxies ( e.g. mundell et al .
1995 ; cole et al .
1998 ) .
here we present new , 022 ( 81-pc ) resolution , @xmath221-cm absorption observations of the central @xmath0900 pc of arp 220 with merlin .
the observations and data analysis are described in sect .
[ sectionobservations ] and the results in sect .
[ sectionresults ] .
we then compare the derived masses with those obtained from the molecular gas observations , and discuss the present geometry of the system and the history of the merger in view of these new observations and the available data .
we adopt a distance of @xmath0 77 mpc for arp 220 , at which 1 correspond to @xmath0 370 pc .
the central region of arp 220 was observed at @xmath221 cm , on 26th january , 1996 , with the multi - element radio linked interferometer ( merlin ) , which at this time consisted of 8 telescopes ( including the 76-m lovell telescope ) .
the maximum baseline length of the array is 217 km , corresponding to @xmath0 1 m@xmath2 at @xmath221 cm and yielding an effective spatial resolution of 022 .
dual polarizations were recorded for an 8-mhz bandwidth in spectral line mode , with 64 channels per polarization , resulting in a channel width of 125 khz ( 26.3 ) . as the first channel is reserved for on - line information ,
the centre of the band lies at channel 31.5 , corresponding to an optical heliocentric velocity of 5496.2 .
total observing time on arp 220 was @xmath07 hours , with 2 minutes observations of the phase calibrator 1551 + 239 interleaved every 5 minutes throughout the observations .
absolute flux calibration was determined from observations of 3c286 , assuming a flux density of 14.731 jy at 1420.4 mhz ( baars et al .
0552 + 398 was used to determine bandpass corrections for the spectral line data .
after preliminary amplitude calibration and gain - elevation corrections were applied , using local merlin software , the data were transferred to the nrao astronomical image processing system ( aips - van moorsel , kemball & greisen 1996 ) for all subsequent calibration , editing and imaging .
standard calibration and data editing ( greisen & murphy 1998 ) were performed on the ` pseudo - continuum ' dataset , which is formed by averaging the central 75% ( 6 mhz ) of the band ( analogous to the vla ` channel 0 ' dataset ) .
the resulting map of the phase calibrator was used to guide additional data editing and self - calibration ; the derived phase and gain corrections were applied to the arp 220 pseudo - continuum data , from which a preliminary image was produced and used in further data editing and phase self - calibration .
the total corrections and data flags were then applied to the spectral line data and bandpass calibration was performed . due to good bandpass stability , no significant improvement to the continuum images was achieved by iteratively performing self - calibation and bandpass calibrations .
finally , the telescopes were reweighted according to their gains .
the continuum contribution ( formed from the line - free channels 5 - 15 and 50 - 60 ) was subtracted in the @xmath32 plane using aips task uvlin .
the corresponding ` true ' continuum dataset was formed from the same line - free channels in the @xmath32 plane using avspc .
these two datasets were then fourier transformed with uniform weighting ( pixel size 45 mas ) , deconvolved and added together to form the absorption cube ( 512 @xmath3 512 @xmath3 63 ) with the correct continuum level adjacent to the line .
a circular restoring beam of size 022 was used for both datasets and the r.m.s .
noise levels were @xmath0 0.1 and 0.5 per channel for the continuum image and spectral line cube , respectively .
the commonly - used phase - referencing technique of interspersing observations of the target source with regular and frequent observations of a nearby calibrator with known position allows the absolute position of the target source to be determined ; the accuracy of the calibrator position is therefore important .
our phase calibrator , 1551 + 239 , was originally identified in the calibrator survey of patnaik et al .
( patnaik et al . 1992 ; browne et al . 1998 ) and lies in ` region 2 ' or declination range 20@xmath2535 ( wilkinson et al . 1998 ) .
the typical positional uncertainty for the survey is 50 mas but ` region 2 ' suffered from larger systematic errors than other regions of the survey .
the origin of these errors is unknown and the positional uncertainty for 1551 + 239 , which was omitted from the final published survey , might be as large as 100 mas , predominantly in declination ( thomasson , private communication ) .
positions derived from previous vla images at 5 ghz ( baan & haschick ( 1995 ) , 15 and 22 ghz ( norris , 1988 ) also show a range of values , differing primarily in declination by @xmath0100@xmath25200 mas .
consequently , the positional uncertainty for arp 220 is no better than 100 mas .
nevertheless , the derived position for the western nucleus , @xmath33 = 15@xmath3432@xmath354689142s , @xmath36 = 2340080215 , compares well with positions derived from co observations by ssy99 , @xmath37 = 15@xmath3432@xmath354688s , @xmath38 = 2340080 ( @xmath1101 ) , and 5-ghz merlin observations by baan ( private communication ) , @xmath37 = 15@xmath3432@xmath354689053s , @xmath38 = 2340080509 .
it should be noted that although the positional uncertainties limit very accurate absolute registration with observations from other instruments , the merlin results presented here are not otherwise affected since arp 220 is strong enough for self - calibration to be performed .
examination of the observed absorption line profiles revealed that not all are well approximated by a single gaussian profile , due to the presence of additional narrow components and/or wings overimposed on the main , relatively broad component .
simple , automatic moment analysis is therefore difficult and may result in corrupted velocity fields , where a single gaussian component does not represent the centroid velocity accurately .
we therefore performed a multiple gaussian profile fitting of the absorption spectra using a modified version of the fit / spec software ( rousset 1992 ) , originally developed to fit optical emission lines .
the fitting was performed in two stages .
first a single gaussian profile was fitted automatically to each absorption spectrum and residuals were examined to detect regions where additional components where likely to be present ( i.e. regions with residuals higher than three times the @xmath39 = 0.5 noise level ) - these regions coincided with each nucleus . for each nucleus , the spectral characteristics of the additional components were then determined by fitting to individual spectra with the most contrast .
next , a multi - component gaussian fit to the line profiles was performed , using one relatively broad gaussian component ( the ` main ' component ) and a set of narrower components with fixed characteristics ( see table 1 ) .
this second step was successful for the western nucleus which exhibited the strongest , most distinct , additional components , but failed for the eastern nucleus where the additional components were poorly contrasted ( peak amplitude typically lower than 6@xmath39 ) .
therefore , the moment maps presented in this paper correspond to a single gaussian component fit to spectra from off - nuclear regions and the eastern nucleus , and a multiple gaussian component fit to spectra from the western nucleus .
errors on the centroid velocity in the eastern nucleus due to the use of a single gaussian profile are typically lower than 10 .
the merlin image of the @xmath221-cm ( 1.4-ghz ) radio continuum emission from the nuclear region of arp 220 is shown in figure 1 .
the two nuclei ( baan et al .
1987 ; norris 1988 ; baan & haschick 1995 , bh95 hereafter ) are clearly resolved and are embedded in more complex extended emission .
the nuclei are labelled e and w in figure1 ( labelled a and b in the 5-ghz image of bh95 ) .
their peak and integrated fluxes are [email protected] & [email protected] mjy ( e ) and [email protected] & [email protected] mjy ( w ) , respectively .
the total integrated flux density of the source is @xmath0 [email protected] mjy .
a large - area box flux yields a slightly higher integrated flux , indicating the presence of structures extended on scales larger than 1 - 2 .
the presence of extended emission is also suggested by the higher integrated flux densities of 302 mjy ( white & becker , 1992 ) and 312 mjy ( condon & dressel , 1978 ) measured with single dish telescopes at 1.4 ghz and 2.4 ghz respectively .
gaussian fitting to the merlin image reveals that the eastern nucleus is well resolved ( as can be seen in figure 1 ) , with a deconvolved size ( fwhm ) of 440 @xmath3 345 mas ( i.e. @xmath0160 @xmath3 130 pc ) in pa = 550 @xmath11 07 .
the western nucleus is only marginally resolved with a deconvolved size ( fwhm ) of 245 @xmath3 190 mas ( i.e. 90 @xmath3 70 pc ) in pa = 96.3 @xmath11 07 .
the derived position for the western nucleus is @xmath37 = 15@xmath3432@xmath354689142s , @xmath38 = 2340080215 which corresponds well with observations at other frequencies ( see section [ sectionobservationspositions ] for full discussion ) .
in addition to emission from the two nuclei , a tongue of @xmath221-cm continuum emission is seen to the south - east of the western nucleus ( figure 1 ) and corresponds to emission previously identified by bh95 ( component c in their paper ) at 5 ghz .
similar weak extensions are seen around the eastern nucleus . finally , a broad spur of continuum emission ( labelled t in figure 1 ) is seen north - west of the western nucleus .
it extends over 08 in pa @xmath4050 .
this weak component is detected over @xmath0 11 beam areas , with a peak flux density of @xmath0 1.6 and an integrated flux density of 6.6 .
there is a hint for the presence of a similar component in the 5-ghz naturally weighted image of bh95 ( their fig . 1 , upper panel ) , but their angular resolution ( 043 ) is insufficient to deblend it from the emission of the western nucleus .
figure 2 shows the merlin spectral channel images of absorption detected across the central @xmath0600 pc of arp 220 .
comparison of the channel images with the continuum image , which is displayed to the same scale in the top left panel , shows that absorption is clearly detected against each continuum nucleus , as well as against the bridge of continuum that connects the two nuclei .
figure 3 shows some representative absorption spectra , taken from eight locations across the continuum structure ; many of the absorption lines are asymmetric with some showing additional velocity components ( e.g. location ` er ' ) .
the ability to image neutral gas in absorption depends on both the structure of the background continuum emission and the distribution of the foreground absorbing gas ; the neutral gas in arp 220 is spatially extended and so the absorption is well resolved at this angular resolution , in particular against the extended continuum emission of the eastern nucleus .
the velocity range over which absorption is detected ranges from @xmath0 5160 to 5725 and compares well with the width of the single dish absorption line at 10% of the peak absorption ( mirabel & sanders 1988 ) .
however , this is lower than the full width at zero intensity ( fwzi ) of the single dish profile , @xmath0 640 ( mirabel & sanders 1988 ) .
if weak absorption is present against weak , extended radio continuum emission , this absorption would not be detectable in the merlin observations , and this may explain the additional 50- width at the red and blue extremes of the single dish profile .
hi absorption is not detected against the continuum component c , to the south - east of the western nucleus ( figure 3 ) , to a limiting 3@xmath39 absorption depth of 1.5 .
assuming a constant column density across c , this implies a maximum column density @xmath0 1.7 @xmath3 10@xmath4 t@xmath8(k ) , where t@xmath5(k ) is the spin temperature for which the exact value is unkown but is found to be typically @xmath010@xmath41@xmath2510@xmath42 k in the galaxy ( heiles & kulkarni 1988 ) .
the weakness of continuum component c prevents the determination of stringent upper limits for , but the relatively blue colour of the infrared component coincident with c ( scoville et al . 1998 ) provides support for a low column density in this region .
very weak absorption may be present against the spur of weak , extended continuum emission ( t ) to the north - west of the western nucleus , but the continuum emission is weak in this region ( peak flux density @xmath0 1.6 ) , preventing a reliable determination of column density ; a 3@xmath39 detection would require a column density @xmath18 1.3 @xmath3 10@xmath43 t@xmath8(k ) .
close inspection of the spectral channel images in figure 2 , over a velocity range @xmath05500 @xmath25 5300 , suggests a possible weak absorption component extended in the same pa as the continuum extension ( @xmath4050 ) , but more sensitive observations would be required to confirm its detection and measure the column density .
figure 4 shows the two - dimensional column density distribution and velocity field of the main kinematic component , as derived from multi - component gaussian profile fitting of the spectra ( see sect .
[ sectionobservationsanalysis ] ) .
characteristics of the main and additional kinematic components in the eight spectra displayed in figure 3 are listed in table 1 and examples of gaussian fitting to components in the eastern and western disks are shown in figure 5 and figure 6 , respectively .
the two nuclei are spatially well separated at this resolution ( as shown in figure 2 ) and a velocity gradient is clearly visible across each nucleus . across the eastern nucleus
, the velocity gradient is 1010 @xmath11 20 kpc@xmath12 in pa @xmath0 55 , while the marginally resolved western nucleus shows a gradient of 830 @xmath11 20 kpc@xmath12 in pa @xmath0270 .
the neutral hydrogen column densities across the two nuclei are high , lying in the range 8@xmath310@xmath4 t@xmath8(k)@xmath6 @xmath6@xmath44 t@xmath8(k ) , with the higher column densities generally being found in the eastern nucleus .
the corresponding optical extinctions , if column density and optical extinction in arp 220 follow the same relation as that seen in the galaxy ( staveley - smith & davies 1987 ) , are 0.052 t@xmath8(k)@xmath6 a@xmath10 @xmath6 0.155 t@xmath8(k ) mag and are detailed in tab . [ tableresultsfit ] .
the spin temperature is unknown , but for the minimum spin temperature t@xmath9@xmath0100 - 200 k , these extinction values compare well with those derived by scoville et al .
( 1998 ) from high resolution infrared _ hst _ nicmos images at 1.1 , 1.6 and 2.22 @xmath21 m and confirm that the radio continuum nuclei lie in regions of relatively high extinction .
an estimate of the column density of molecular hydrogen can be inferred from the co observations of ssy99 , who measured integrated co fluxes s@xmath45 = 120 jy ( e ) and 187 jy ( w ) . assuming an excitation temperature of 40 k , a co : h@xmath46 abundance x(co ) = 2.7 @xmath3 10@xmath47 ( lacy et al . 1994 ) and optically thin emission , we infer h@xmath46 column densities n@[email protected] @xmath3 10@xmath28 ( e ) and 4.3 @xmath3 10@xmath28 ( w ) .
these values are comparable to the mean column densities = 1.6 @xmath3 10@xmath43 t@xmath8 ( e ) and 1.3 @xmath3 10@xmath43 t@xmath5 ( w ) for t@xmath9@xmath0100@xmath25200 k and result in total hydrogen column densities n(h ) @xmath04.4 @xmath3 10@xmath28 ( e ) and 5.6 @xmath3 10@xmath28 ( w ) ; these columns densities are likely to be lower limits due to the uncertainty in the values of t@xmath9 ( heiles & kulkarni 1988 ) and x(co ) ( frerking , langer & wilson 1982 ) , excitation conditions and optical depth . indeed , the nature of the ir emission in arp 220 , agn or starburst - driven , has long been controversial and the amount of obscuration present is important .
high frequency observations have failed to find a hard x - ray or soft @xmath49-ray component consistent with an agn , suggesting that the column density required to obscure any hidden agn must be @xmath5010@xmath51 ( rieke 1988 ; dermer et al .
1997 ) . alternatively ,
if @xmath6 10@xmath52 , more than 80% of the total ir luminosity must have a non - agn origin . even if the spin temperature is as high as t@xmath8 @xmath0 10@xmath53 k , as might be expected in the vicinity of strong radio continuum emission ,
the peak measured column density is @xmath02.4 @xmath310@xmath52 , suggesting that either a large additional quantity of ionized and molecular gas is required to provide the necessary obscuration towards any agn component or the ir emission is dominated by a starburst .
the kinematics of the main absorption component are consistent with two counterrotating disks associated with the two nuclei .
we therefore confirm the results from the lower spatial resolution ( 05 ) co observations of sakamoto et al .
in particular , the velocity field of the disk in the more compact western nucleus is clearly apparent in our higher resolution ( 022 ) observations .
the kinematic major axis for this western nuclear disk ( @xmath0 270 ) is consistent with the value inferred from the co observations ( 263 , ssy99 , their tab . 1 ) but is perpendicular to the north - south velocity gradient inferred from the formaldehyde emission observations of bh95 .
this discrepancy can be explained if the gaseous component responsible of the formaldehyde emission is distinct from the western disk gas .
the location of the peak of formaldehyde emission , @xmath0 01 south of the western radio continuum peak ( bh95 ) , is coincident with the position of the group of luminous radio supernovae detected by smith et al .
this suggests that the source of the pumping photons is the intense starburst occuring in the western disk and therefore belongs to it .
however , the masing gas can be located far away from the continuum source .
baan , gsten & haschick ( 1986 ) give _ upper limits _ on the distance between the continuum source and the masing gas in arp 220 ranging from 250 pc to 1.1 kpc .
given these upper limits , the masing gas could be molecular gas located in front of the western nucleus and unrelated to the western nuclear disk or a warp in the eastern / large - scale molecular disk . indeed , the horizontal isovelocity contours of co(2 - 1 ) emission seen to the south of the western nucleus ( ssy99 - fig . 3 , upper right panel ; downes & solomon 1998 - fig . 18 , upper right panel ) would support this interpretation for a non - coplanar component of molecular gas .
using mean column densities = 1.6 @xmath3 10@xmath43 t@xmath8 and 1.3 @xmath3 10@xmath43 t@xmath8 for the eastern and western nuclei respectively and semi - major axes 167 pc @xmath3 118 pc ( e ) and 102 pc @xmath3 65 pc ( w ) , corresponding to the radius at which the absorption line strength is greater than 3@xmath39 , we infer masses m@xmath29 = 7.5 @xmath3 10@xmath42 t@xmath8 m@xmath16 ( e ) and 2.0 @xmath3 10@xmath42 t@xmath5 m@xmath16 ( w ) .
these masses are ( [email protected])@xmath3 10@xmath42 t@xmath54 times lower than the mass of molecular gas in the disks derived from the co observations ( @xmath0 10@xmath14 m@xmath16 ; downes & solomon 1998 , ds98 hereafter ; ssy99 ) ; given the uncertainty in the co mass determination , the and co masses would be comparable for high t@xmath9 @xmath010@xmath42 k. assuming the neutral gas is distributed in two circularly rotating disks of radius 167 pc ( e ) and 102 pc ( w ) , the observed velocity gradients ( see sect . [ sectionresultsabsorption ] ) imply dynamical masses
m@xmath13 = 1.1 @xmath3 10@xmath14 ( sin@xmath15 ) m@xmath16 ( e ) and 1.7 @xmath3 10@xmath17 ( sin@xmath15 ) m@xmath16 ( w ) .
the -inferred dynamical mass for e is slightly lower than values derived from co observations ( 2.4 and @xmath18 1.9 @xmath3 10@xmath14 ( sin@xmath15 ) m@xmath16 for ds98 and ssy99 , respectively ) and indeed the co emission ( ds98 , ssy99 ) out to a radius of @xmath003 ( 111 pc ) , although weak and not fully resolved , appears to show a steeper velocity gradient than that of the .
for the western nucleus , our dynamical mass is very similar to that inferred by ds98 ( 1.7 @xmath3 10@xmath17 sin@xmath15 m@xmath16 ) but 10 times lower than that inferred by ssy99 ( @xmath18 1.5 @xmath3 10@xmath14 sin@xmath15 m@xmath16 ) for a similar disk radius .
the reason for this difference is not clear and higher angular resolution co observations might determine whether the co is located deeper in the nuclear potential well than the hi .
the fact that the two nuclei have counterrotating gas disks , indicates that the merger in arp 220 involved a prograde - retrograde encounter of two gas - rich progenitor disk galaxies .
numerical simulations of such mergers can be found in mihos & hernquist ( 1996 , mh96 hereafter ) .
in particular , these simulations show that the prograde and retrograde disks will behave differently during the merger and may also experience different star formation episodes ( mh96 - sect .
4.2 ) . in arp 220
, qualitative comparison between the observations and the models of mh96 suggests that the eastern nucleus , which rotates in the same sense as the 1-kpc co disk ( scoville et al . 1991 ; scoville , yun & bryant 1997 , syb97 hereafter ; ds98 ) and 100-kpc hi disk ( hibbard , vacca & yun 2000 ,
hvy2000 hereafter ) in which it is embedded , represents the retrograde progenitor while the western nucleus and north - western optical plume originate from the prograde progenitor ( hvy2000 hereafter ) . initially , during the first close passage , the retrograde disk ( and corresponding halo )
sweeps up and accretes a significant fraction of the gas of the prograde disk ( 30% in the prograde - retrograde _ planar _ encounter simulations of mh96 ) . in arp 220 , the eastern nuclear disk as well as the kiloparsec - scale co disk ( scoville et al .
1991 ; syb97 ; ds98 ) , which comprise a large fraction of the mass of the system , are probably the result of this process .
thus , given the similarities between the pa and kinematics of these two overlaping disks , it is likely that the two constitute a single ( warped ) disk , and so we use the term ` eastern disk ' in the following discusision to refer to the association of these two disks . the second argument in favor of the eastern nucleus being the retrograde one , comes from the morphology of the western nucleus . in the mh96 simulations and after the first close passage , the prograde nucleus has lost up to 50% of its mass , either accreted by the retrograde disk or launched into tidal tails .
this is qualitatively consistent with the observed properties of the western disk , which is more compact , less massive than its companion and is likely the origin of the north - western optical plume ( hvy2000 ) .
if the western nucleus is indeed the core of the prograde galaxy , this has important implications for the current geometry of the system
. based on the good agreement between the mean velocity of the western disk and the line of sight velocity of gas in the large - scale co disk at this position ( syb97 ) , previous studies have assumed that the western disk was orbiting in the plane of the ` large - scale ' co disk ( e.g. see fig . 7 in syb97 or fig . 5 in ssy99 ) .
in this context , the deprojected locations of the two nuclei were derived from their mean line of sight velocities , and the western nucleus was placed closer along our line of sight than the eastern nuclear disk ( e.g. see fig .
24 in ds98 ) .
however , for the western nucleus to be the prograde nucleus its orbit must pass _ behind _ the eastern nucleus , so the location of the western nucleus on its orbital path is therefore important .
the absence of an absorption signature from eastern - disk gas in the spectra of the western nucleus indicates that the western disk lies above the eastern disk , while the mean blueshifted velocity of the western disk with respect to the eastern disk ( @xmath0 @xmath25170 , ssy99 ) , permits only the far - sided half of the orbit ( see figure 7 ) ; combining these two constraints results in the western disk lying on the orbit quadrant above the eastern disk , on its final inward journey towards collision with the eastern disk ( see figure 7 ) .
the presence of disturbed in the bridge ( b ) connecting the two nuclei further supports this merger scenario with the eastern edge of the western disk merging with the main disk . continuing our comparison with the models of mh96 and assuming that the eastern nucleus is the retrograde nucleus
, we can investigate how and when two important starburst events in arp 220 fit in such a scenario .
the first event is that which gave birth to the galactic superwind and the associated kiloparsec scale bubbles observed in arp 220 ( heckman , armus & miley 1987 1990 , ham90 hereafter ; heckman et al .
1996 , hdew96 hereafter ) . given the kinematic expansion time of the bubbles ( 3 @xmath3 10@xmath55 yr , hdew96 , from ham90 ) and the time needed by a ( continuous ) starburst to deposit significant mechanical energy into its environment ( @xmath23 10@xmath55 yr , fig . 56 in leitherer & heckman 1995 , lh95 hereafter ) , the starburst powering this outflow must have started a 10@xmath55 to 10@xmath17 years ago ( additional time may have been necessary for the superwind to break out from the gaseous disk ) .
as the orientation of the outflow is roughly perpendicular to the major axis of the eastern disk , it is likely that this starburst took place in this disk .
assuming that the arp 220 merger is currently in the stage just before the final fusion of the two gas rich systems ( i.e. t @xmath0 58 - 60 in fig . 12 in mh96 )
this corresponds 1.3 @xmath3 10@xmath55 yr ( see mh96 - sect .
2.1 ) . ] to t @xmath0 50 - 56 in the simulations .
in the simulations , the retrograde disk does indeed experience a continuous period of star formation ( which is stronger than in the prograde disk , mh96 ) during these intermediate times ( fig . 13 in mh96 ) .
note that , assuming that all the mechanical energy deposited by the starburst is injected into the wind , the mechanical luminosity of the superwind ( @xmath0 10@xmath56 erg s@xmath12 , hdew96 ) corresponds to a star formation rate of @xmath0 10 - 100 m@xmath16 yr@xmath12 ( lh96 - fig .
56 for a continuous star formation rate ) .
these values are lower than the recent star formation rate estimated in arp 220 ( @xmath0 240 m@xmath16 yr@xmath12 , anantharamaiah et al . 2000 ) , but are consistent with the models which predict a sharp increase in the star formation rate in the final stage of the merger compared to the intermediate times ( mh96 - fig .
14 , second row ) .
the second starburst event , more recent and probably more violent ( consistent with the models of mh96 described above ) , is the one revealed by the large number of luminous radio supernovae detected in the western nucleus ( smith et al .
1998 , slld98 hereafter ) .
the rate of type ii supernovea inferred from these observations is 1 - 4 yr@xmath12 and corresponds to a star formation rate ranging from 50 to 800 m@xmath16 yr@xmath12 ( lh95 - fig . 6 for a continuous star formation rate ) .
this is consistent with the value of 160 m@xmath16 yr@xmath12 that can be inferred line emission in their observations come from the western nucleus .
if we scale the total star formation rate ( 240 m@xmath16 yr@xmath12 ) accordingly , this gives a star formation rate for the western nucleus alone of @xmath0 160 m@xmath16 yr@xmath12 .
] for the western nucleus from the observations of anantharamaiah et al .
one striking feature of the distribution is the luminous radio supernovae ( slld98 ) .
figure 8 shows the positions of the radio supernovae and oh megamaser emission features with respect to the merlin 1.4-ghz continuum emission .
the fitted positions of the e and w nuclei as derived from the merlin 1.4-ghz continuum image ( sect .
[ sectionresultscontinuum ] ) and from similar merlin observations made at 5 ghz ( baan , private communication ) are also shown .
there is a clear offset between the rsn and the nuclear positions , particularly for the western nucleus where the supernovae are located @xmath001 @xmath25 015 _ south _ of the western nucleus and consequently , the compact central feature of the w1 oh megamaser emission ( lonsdale et al . 1998a 1998b ) coincides with the position of the western nucleus . in a simplistic model of star formation occurring in an edge on disk
, the rsn distribution would be expected to coincide with the nucleus . however ,
if star formation is occurring throughout the disk and the nearside of the disk is tilted slightly to the south , as suggested by nicmos observations of arp 220 ( scoville et al .
1998 ) , rsn in the near half of the disk would be more easily visible .
in particular a moderate column density of ionized gas associated with the bulge - like component suggested by scoville et al .
( 1998 ) and the longer path length to rsn on the far side of the disk would result in rsn radio emission being free - free absorbed .
emission measures 1.1@xmath310@xmath55@xmath6 em @xmath62.4@xmath310@xmath55 @xmath57 pc , corresponding to optical depths 1.2@xmath6 @xmath58 @xmath62.6 , would be required to free - free absorb emission from rsn with the same flux density at 1.67 ghz as those observed by slld98 , assuming a uniform screen of ionized absorbing gas with an electron temperature t@xmath59=10@xmath53 k. such emission measures compare well with those inferred from radio recombination line observations ( anantharamaiah et al .
2000 ) which , however , suggest more complex models for the ionized gas than a single density uniform slab .
alternatively , the rsn positional offset might be a consequence of positional uncertainty of the merlin observations ( see sect .
[ sectionobservationspositions ] ) ; future observations are required to confirm the registration of merlin and vlba images .
we have presented the highest angular resolution observations to date of the neutral gas distribution and kinematics and the @xmath221-cm continuum emission in the central @xmath0900 pc of the advanced merger system , arp 220 .
spatially resolved absorption was detected against the morphologically complex and extended @xmath221-cm radio continuum emission ; the two nuclei are spatially well separated at this resolution and a velocity gradient is clearly visible across each nucleus consistent with the being in two counterrotating disks of neutral hydrogen , with a small bridge of gas connecting the two .
the column densities across the two nuclei are high , lying in the range 8@xmath310@xmath4 t@xmath5(k)@xmath6 @xmath6@xmath7 t@xmath8(k ) ( t@xmath9 is spin temperature ) and corresponding to optical extinctions of 0.052 t@xmath8(k)@xmath6 a@xmath10 @xmath6 0.155 t@xmath5(k ) mag , with the higher column densities being found in the eastern nucleus .
velocity gradients of 1010 @xmath11 20 kpc@xmath12 in pa @xmath0 55 and 830 @xmath11 20 kpc@xmath12 in pa @xmath0270 across the eastern and western nuclei respectively imply corresponding dynamical masses m@xmath13 = 1.1 @xmath3 10@xmath14 ( sin@xmath15 ) m@xmath16 ( e ) and 1.7 @xmath3
10@xmath17 ( sin@xmath15 ) m@xmath16 ( w ) , assuming the neutral gas is distributed in two thin , circularly rotating disks .
we propose a new merger geometry in which the two nuclei are in the final stages of merging , similar to conclusions drawn from co studies ( ssy99 ) , but instead of being coplanar with the main co disk ( in which the eastern nucleus is embedded ) , we suggest that the western nucleus lies above it and will soon complete its final merger with the main disk .
the bridge of seen connecting the two nuclei further supports this scenario .
we suggest that the collection of radio supernovae ( rsn ) , detected in vlba studies , in the more compact western nucleus represent the second burst of star formation associated with this final merger stage and that free - free absorption due to ionised gas in the bulge - like component can account for the observed rsn distribution .
we thank toby moore , nick scoville , john porter , andrew wilson , sylvain veilleux and lee mundy for useful discussions and peter thomasson for help with the merlin data .
we are grateful to willem baan for kindly providing merlin 5-ghz - derived positions prior to publication and the anonymous referee for useful comments .
cgm acknowledges financial support from the royal society .
this research was partially supported by nsf grant ast9527289 to the university of maryland .
merlin is a u.k .
national facility operated by the university of manchester on behalf of the particle physics and astronomy research council .
this research has made use of nasa s astrophysics data system abstract service ( ads ) and the nasa / ipac extragalactic database ( ned ) , which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration .
lccccccc & b & er & en & eb & wb & wn & wr + @xmath60 ( arcsec ) & @xmath250.405 & @xmath251.125 & @xmath250.945 & @xmath250.675 & @xmath250.270 & 0 & @xmath610.225 + @xmath62 ( arcsec ) & @xmath250.090 & @xmath610.045 & @xmath250.135 & @xmath250.360 & 0 & 0 & 0 + + f@xmath63 ( ) & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] + n@xmath64/t@xmath8 ( 10@xmath43 @xmath65 ) & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] + a@xmath66/t@xmath8 ( 10@xmath67mag ) & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] + velocity ( ) & 5387@xmath1120 & 5591@xmath115 & 5486@xmath1119 & 5354@xmath115 & 5364@xmath1113 & 5401@xmath112 & 5448@xmath1112 + fwhm ( ) & 390@xmath1170 & 169@xmath1112 & 161@xmath115 & 120@xmath1112 & 300@xmath1140 & 238@xmath115 & 300@xmath1140 + + f@xmath63 ( ) & & & [email protected] & & [email protected] & [email protected] & [email protected] + + n@xmath64/t@xmath8 ( 10@xmath43 @xmath65 ) & & & [email protected] & & [email protected] & [email protected] & [email protected] + a@xmath66/t@xmath8 ( 10@xmath67mag ) & & & [email protected] & & [email protected] & [email protected] & [email protected] + velocity ( ) & & & 5478@xmath112 & & 5336@xmath1118 & 5404@xmath113 & 5449@xmath1116 + fwhm ( ) & & & 132@xmath115 & & 317@xmath1150 & 270@xmath117 & 340@xmath1150 + + n@xmath64/t@xmath8 ( 10@xmath43 @xmath65 ) & & & [email protected] & & @xmath23 0.1 & [email protected] & @xmath23 0.08 + a@xmath66/t@xmath8 ( 10@xmath67mag ) & & & [email protected] & & & [email protected] & + velocity ( ) & & & 5618@xmath116 & & 5326@xmath68 & 5326@xmath68 & 5326@xmath68 + fwhm ( ) & & & 80@xmath1112 & & 50@xmath68 & 50@xmath68 & 50@xmath68 + n@xmath64/t@xmath8 ( 10@xmath43 @xmath65 ) & & & & & [email protected] & [email protected] & [email protected] + a@xmath66/t@xmath8 ( 10@xmath67mag ) & & & & & [email protected] & [email protected] & [email protected] + velocity ( ) & & & & & 5428@xmath68 & 5428@xmath68 & 5428@xmath68 + fwhm ( ) & & & & & 50@xmath68 & 50@xmath68 & 50@xmath68 + + | we present new , high angular resolution ( @xmath00@xmath122 ) merlin observations of neutral hydrogen ( ) absorption and @xmath221-cm radio continuum emission across the central @xmath0900 parsecs of the ultraluminous infrared galaxy , arp220 .
spatially resolved absorption is detected against the morphologically complex and extended @xmath221-cm radio continuum emission , consistent with two counterrotating disks of neutral hydrogen , with a small bridge of gas connecting the two .
column densities across the two nuclei are high , lying in the range 8@xmath310@xmath4 t@xmath5(k)@xmath6 @xmath6@xmath7 t@xmath8(k ) ( t@xmath9 is spin temperature ) and corresponding to optical extinctions of 0.052 t@xmath8(k)@xmath6 a@xmath10 @xmath6 0.155 t@xmath5(k ) mag , with the higher column densities in the eastern nucleus .
velocity gradients are clearly visible across each nucleus , reaching 1010 @xmath11 20 kpc@xmath12 in pa @xmath0 55 and 830 @xmath11 20 kpc@xmath12 in pa @xmath0270 for eastern and western nuclei respectively .
these gradients imply dynamical masses m@xmath13 = 1.1 @xmath3 10@xmath14 ( sin@xmath15 ) m@xmath16 ( e ) and 1.7 @xmath3 10@xmath17 ( sin@xmath15 ) m@xmath16 ( w ) , assuming the neutral gas is distributed in two thin circularly rotating disks .
we propose a merger model in which the two nuclei represent the galaxy cores which have survived the initial encounter and are now in the final stages of merging , similar to conclusions drawn from previous co studies ( sakamoto , scoville & yun 1999 - ssy99 hereafter ) .
however , we suggest that instead of being coplanar with the main co disk ( in which the eastern nucleus is embedded ) , the western nucleus lies above it and , as suggested by bridge of connecting the two nuclei , will soon complete its final merger with the main disk .
we suggest that the collection of radio supernovae ( rsn ) detected in vlba studies in the more compact western nucleus represent the second burst of star formation associated with this final merger stage and that free - free absorption due to ionised gas in the bulge - like component can account for the observed rsn distribution . |
any feasible electronic application of a polymer material implies that the material can conduct an electric current via some kind of charge transport mechanism .
velocity of charge carriers ( electrons or holes ) may be an important parameter per se effectively determining operating characteristics of the device or it could influence other important parameters such as carrier injection or recombination efficiency .
some degree of structural disorder is present in most polymers currently used in real life applications , so they naturally belong to the general category of disordered organic materials . at the same time
disordered low molecular weight organic materials often demonstrate similar general features of charge transport ; for this reason in this article various properties of charge transport in disordered organic materials are mostly discussed without specific emphasis on polymeric nature of a material . for many years
the study of charge carrier transport in disordered organic materials was mostly motivated by hope to use them in various electronic and optoelectronic devices . at the same time
this problem is an interesting and sometimes even intriguing area of modern condensed matter physics .
good review of the area may be found in recent monographs @xcite .
we consider here only materials which are insulators under usual conditions and attain short - lived conductivity after injection of charge carriers under the action of laser pulse or electric discharge and completely leave aside such highly conducting materials as doped polyacetylene , polyanyline etc .
typical materials include low molecular weight organic glasses , polymers doped with aromatic organic molecules ( usually amines or hydrozones ) , polysilanes , polyconjugated polymers , and other compounds .
this diverse variety demonstrates a lot of common transport features unambiguously indicating a common transport mechanism .
transport properties of materials are usually characterized by carrier drift mobility @xmath0 and its dependence on applied electric field @xmath1 , temperature @xmath2 , and other relevant parameters .
the most direct method of the mobility measurement is the time - of - flight method ( tof ) where carriers are generated at the vicinity of one electrode of the plane capacitor ( organic material fills the space between electrodes ) and then drift to the opposite electrode under the action of applied uniform electric field @xmath1 . analyzing the shape of the photocurrent transient
, one can calculate the transit time @xmath3 ( the moment when current begins to drop , thus indicating arrival of carriers to the opposite electrode ) and then the mobility @xmath4 where @xmath5 is thickness of the transport layer , @xmath6 is the average velocity of charge carriers , and @xmath7 is the applied voltage .
dominant majority of experimental data cited in this article was obtained by the tof method .
let us briefly summarize major experimental findings offering important clues on the nature of charge transport in such materials.@xcite 1 .
molecularly - doped polymers offer a natural possibility to study dependence of the mobility on dopant concentration .
it was found that @xmath8 where @xmath9 is the average distance between dopant molecules .
this relation suggests that the transport occurs as a series of hops between localized states ( transport sites ) originating , in this particular case , at the dopant molecules .
however , in some exceptional cases the radius @xmath10 of the wave function of a transport site is unreasonably large @xcite .
mobility increases exponentially with the increase of temperature .
there is no consensus on the form of this dependence : some authors favor a simple activation dependence @xmath11 while others argue that the dependence @xmath12 is more accurate .
it is generally accepted that an exponentially strong temperature dependence may be most naturally explained by suggesting that the major factor governing charge transport in disordered organic materials is an energetic disorder : random locations and orientations of molecules in the bulk of the material produce random fluctuations of site energies @xmath13 . then the dependence @xmath14 should be connected with the density of states @xmath15 .
there is no reason for @xmath15 to maintain the same form in different materials and this may be the reason for different forms of @xmath14 .
the most surprising fact is the mobility field dependence . indeed , for carrier hopping between spatially localized sites we should expect @xmath16 just estimating a typical drop in the difference of carrier energy between two transport sites separated by the typical intersite distance @xmath9 . yet instead of eq [ mu(e)_naive ] an almost ubiquitous poole - frenkel ( pf ) dependence emerges in the materials @xmath17 leading to much stronger field dependence of the mobility in weak field .
uncertainty in the form of dependence @xmath14 is mostly originated from the difficulty to measure the quasi - equilibrium ( nondispersive ) mobility in a wide temperature range . in striking contrast to this limitation , dependence ( [ mu(e)_pf ] )
was measured for some materials in really wide field range ( an unrivaled example is ref . with @xmath1 varying from @xmath18 to @xmath19 v / cm ) . for a long time
an explanation of the pf dependence has been considered as a major problem for the charge transport theory . since the early paper of gill , who suggested an empirical formula @xmath20 , \label{gill}\ ] ] based on the study of charge carrier transport in polyvinylcarbazole,@xcite it was very tempting to attribute the field dependence of @xmath0 to the influence of charged traps . in the case of trap - controlled charge transport with charged particles
( having charge of the opposite sign to carriers ) serving as traps , an applied electric field @xmath1 leads to the decrease of activation energy of the carrier escape from a trap . in this case
@xmath21 ( here @xmath22 is a dielectric constant ) and calculated values of @xmath23 are usually close to the measured ones by the order of magnitude , though discrepancies by a factor of 23 are typical . yet this classical pf mechanism attained a lot of criticism because there is no evidence for charged traps to be a common constituent in very different classes of disordered organic materials @xcite .
a first really successful transport model was the gaussian disorder model ( gdm ) @xcite .
in the gdm transport sites are arranged on a regular lattice and are assigned site energies @xmath13 drawn independently from a gaussian distribution having variance @xmath24 .
this model explained many features of charge transport in disordered organic materials , among them the temperature dependence of mobility , transition from the quasi - equilibrium transport to the dispersive - like with temperature decrease , and others features .
still the explanation of the pf field dependence remained an unsolved problem because the gdm can reproduce this dependence in the limited field range only , not significantly wider than @xmath25 v / cm , in weaker field dependence @xmath26 has the usual form ( [ mu(e)_naive ] ) .
moreover , the only reason for the approximate pf dependence in the gdm is the miller - abrahams hopping rate @xcite .
there is no reliable evidence that this particular rate is the universal hopping rate in a wide variety of disordered materials .
for this reason it was highly desirable to retain all useful features of the gdm and yet incorporate some new feature permitting to explain the pf field dependence .
a history of the gdm modification that eventually lead to the successful explanation of the pf dependence ( [ mu(e)_pf ] ) began with the study of dipolar contribution to the total energetic disorder in organic materials .
computer simulation of the distribution of energies for a carrier moving through a simple cubic lattice with sites occupied by randomly oriented dipoles ( thus , the energetic disorder in this case originates from the electrostatic charge - dipole interaction ) was carried out by dieckmann , bssler , and borsenberger @xcite ; they found that @xmath15 has the gaussian form for high concentration of dipoles and goes to the lorentzian one for low dipole concentration .
they calculated the energy variance @xmath27 and found that typical values of @xmath28 are 0.050.1 ev which means that the dipolar contribution should be a significant part of the total @xmath29 ( according to the gdm , @xcite @xmath30 ev is needed to fit experimental temperature dependence of @xmath0 ) .
later it was found that for this model of dipolar glass ( dg ) function @xmath15 could be calculated analytically @xcite . for a simple cubic lattice @xmath31 (
here @xmath32 is a lattice scale , @xmath33 is a dipole moment , and @xmath34 is a fraction of sites occupied by dipoles ) and the distribution of @xmath35 , while indeed having the gaussiam form for @xmath36 , has the long tail @xmath37 for @xmath38 . in @xmath39 sample of the simple cubic lattice model of dipolar glass .
black and white spheres represent sites with positive and negative values of @xmath13 , correspondingly , while the radius of a sphere is proportional to the absolute value of @xmath13 .
sites with small absolute values of @xmath13 ( less than @xmath40 ) are not shown for the sake of clarity .
, width=288 ] the most significant property of the dg model is a strong spatial correlation in the distribution of @xmath41 and the energy correlation function decays very slowly with distance @xcite @xmath42 ( here angular brackets denote statistical average ) . in a correlated gaussiam medium the conditional probability @xmath43 which is the probability density to have energy @xmath35 at the site @xmath44
if the site energy at the reference site @xmath45 is @xmath46 has the form @xcite @xmath47 , \hskip10pt \sigma^2_\delta=\sigma^2_{\rm d}-\frac{c^2(\vec{r}-\vec{r}_0)}{\sigma^2_{\rm d}}. \label{cond}\ ] ] keeping in mind that @xmath48 , we see that equation [ cond ] expresses the tendency of sites that are spatially close to have close energies , in striking contrast to the gdm case ( see fig.[fig1 ] ) . ) . here
@xmath49 is the number of sites in a cluster , @xmath50 is the number of clusters and @xmath51 is the total number of sites ; cluster is defined as a connected set of sites having energies greater than some ( arbitrary ) threshold energy ( @xmath52 in this particular case ) .
, width=288 ] another manifestation of the correlated nature of dg is an exact result for the dependence of the distribution of average domain energy @xmath53 on domain size @xmath54 .
for a gaussian random medium without any spatial correlation the distribution has the gaussian form with variance @xmath55 while for spherical domains in dg @xcite @xmath56 where @xmath9 is the radius of the domain .
the difference between ( [ domain_nc ] ) and ( [ domain_dip ] ) suggests that large domains are much more common in dipolar matrix and , indeed , relative number of large clusters in dg is greater by many orders of magnitude in comparison with the uncorrelated medium ( see fig .
[ fig_c]).@xcite at the same time gartstein and conwell suggested that a correlated nature of the energy landscape should significantly affect charge transport in disordered medium , effectively increasing the typical scale of the hop and , thus , enhancing mobility field dependence .
@xcite the next major step was achieved by dunlap , parris , and kenkre , who found that for charge carrier hopping in 1d random gaussian energy landscape @xcite @xmath57 \right ) } , \hskip10pt \beta = 1/k_{\rm b}t . \label{1d}\ ] ] for the case of strong dipolar disorder @xmath58 equation [ 1d ] gives @xmath59 , \label{1dres}\ ] ] and , in general , if @xmath60 then @xmath61 . at last
, computer simulation confirmed that the dipolar correlation leads to the pf mobility field dependence for 3d carrier transport @xcite ( see fig .
[ 3ddip ] ) and on the basis of extended simulation an empirical relation was suggested @xcite @xmath62\sqrt{eae/\sigma_{\rm d}}\right ] .
\label{3dres}\ ] ] equations [ 1dres ] and [ 3dres ] have essentially the same dependence on @xmath1 and @xmath2 and only numerical coefficients are different . thus , the simplified 1d model could serve for a quick estimation of the mobility field dependence in disordered materials .
( from top curve downward ) .
the lowest curve is the mobility for the gdm for @xmath63 . if @xmath64 ev and @xmath65 , then @xmath66 for @xmath67 v / cm.,width=288 ] good qualitative description of charge transport in a correlated medium is presented in ref . .
the principal difference between a correlated random medium and an uncorrelated one is that in correlated medium deepest valleys of the energetic landscape ( having maximal escape time for a carrier at @xmath68 ) are also the widest .
for this reason carrier escape time for such valleys decreases drastically for @xmath69 .
for every magnitude of @xmath1 a critical size @xmath70 exists such that valleys of this size are the most efficient for carrier capture .
simple calculation gives @xmath71 and @xmath9 in eq [ mu(e)_naive ] should be substituted for @xmath70 .
@xcite this is the physical reason for the pf dependence .
an attracting feature of the dg model is absence of any significant free parameter : the energy scale ( i.e. @xmath29 ) may be calculated using ( [ sd ] ) provided @xmath33 , @xmath32 , and @xmath34 are known .
still , numerical coefficients in eq [ 3dres ] are certainly not constants ( for example , they weakly depend on @xmath10 @xcite ) , so equation [ 3dres ] should be considered rather as a guide and not as a strict relation for @xmath72 in any particular polar material .
another important feature of the dg model is that , unlike the gdm , the form of the mobility field dependence for moderate field does not depend on the particular kind of hopping rate : for @xmath73 mobility dependence retains its pf form ( even the slope is the same @xcite ) for any nonpathological hopping rate .
it is worth to mention , though , that in ( partially ) orientationally ordered dipolar matrices correlation properties of the distribution of @xmath13 may be very different . in such matrices transport properties
may have nothing in common with the usual pf picture .
a vast collection of experimental data indicates that the model of dipolar glass can not serve as a universal explanation for the pf mobility field dependence in disordered organic materials . if the reason for this dependence is the spatial correlation of the dipolar type ( [ c_dip ] ) , then it is absolutely impossible to explain the existence of the pf dependence in nonpolar materials , where it was routinely observed @xcite .
more close study of experimental data reveals , however , that in such materials the experimentally tested field range is not too broad ( not significantly greater than one order of magnitude or even less ) and , moreover , sometimes clearly visible deviations from straight lines can be observed when @xmath74 is plotted against @xmath75 @xcite . in some cases these deviations
lead to the upward convexity of the mobility curve , while in others they lead to the downward convexity .
quite formally , this behavior may be described by @xmath76 with @xmath77 being in some cases smaller than 0.5 and in some cases greater than 0.5 .
this observation hints that , quite possibly , in weakly polar organic materials the real mobility field dependence is not a true pf dependence , but rather a quasi - pf dependence with @xmath78 .
this quasi - pf dependence can successfully imitate true pf dependence in not so wide field range . according to the result of 1d model @xcite ,
the necessary condition for validity of relation ( [ gener ] ) is the algebraic behavior of correlation function @xmath79 .
one possible reason for such behavior is the contribution from randomly oriented quadrupoles .
quite frequently , transport molecules contain highly polar groups having significant dipole moments .
these groups may be arranged in such a way that the total dipole moment of the molecule is close to zero , but the quadrupole moment of the molecule is large enough .
for example , for the particular case of two identical groups having dipole moment @xmath33 , oriented in opposite directions and separated by distance @xmath80 , the total dipole moment of the molecule is exactly zero , but the quadrupole moment @xmath81 is nonzero . in close analogy with the dg model
@xcite we may consider the model of quadrupolar glass ( qg ) , its simplest incarnation being a simple cubic lattice with sites occupied by randomly oriented quadrupoles @xcite .
calculation of the correlation function @xmath82 for the qg model gives @xcite @xmath83 where @xmath34 is the fraction of lattice sites , occupied by quadrupoles . if @xmath84d , @xmath85 , @xmath86 , @xmath87 , then for the totally filled lattice @xmath88 ev . in 1d approximation
the carrier mobility in the qg model for the case of strong disorder @xmath89 has the form @xmath90 \label{eq11}\ ] ] this particular field dependence can successfully imitate the pf dependence in not too wide field range @xmath91 .
moreover , there is another factor that can bring the quadrupole field dependence even closer to the pf dependence .
this factor is an effect of dispersive transport : in dispersive regime @xmath92 .
@xcite if plotted as @xmath74 vs @xmath93 , the dispersive component bends the straight line , making the mobility curve convex upward , thus pushing field dependence more close to the pf one ( contribution of the dispersive component is evident , for example , for photocurrent transient in the data of ref . even at high temperature ) .
detailed computer simulation of the 3d charge transport in the qg model has not been carried out yet , but preliminary data unambiguously support the principal result of 1d model for the mobility field dependence @xmath94 @xcite .
analogy with the dg model suggests that the functional type of the mobility dependence on @xmath1 and @xmath2 could be well captured by the 1d model , while numerical coefficients should be different .
extensive comparison of the qg model with experimental data is still absent , though preliminary analysis of transport data for nonpolar polysilanes indicates that dependence @xmath95 is certainly not worse ( in terms of statistical correlation coefficient @xmath96 ) than the pf dependence @xcite .
recently a new model was suggested to explain the pf mobility field dependence in nonpolar polyconjugated polymers such as polyphenylene vinylens ( ppvs ) @xcite . according to this model
, the major source of the energetic disorder in ppvs is almost static intramolecular fluctuations of the torsion angle of benzene ring , resulting in the carrier energy fluctuations with the correlation function @xmath97 where @xmath98 is the intramolecular restoring force constant and @xmath99 is the intermolecular restoring force constant .
this function in the limit case @xmath100 gives an appropriate behavior needed to provide the pf dependence . in
densely packed ppvs typical values of @xmath101 should be small , @xcite and according to the 1d model the mobility field dependence has the form @xmath102 where @xmath32 is the short - range cutoff and @xmath103 is the linear electron - vibration coupling .
field and temperature dependence of the mobility in the la model is different ( though not very different ) from the prediction of the qg model .
thus , a careful study of the dependence of @xmath0 on @xmath1 and @xmath2 may provide an opportunity to discriminate between these models . at present
, there is no possibility to state which particular model better describes the mobility field dependence in nonpolar materials .
we should note , though , that the model of quadrupolar glass does not relay on any particular property of the nonpolar material ( apart from its quadrupolar nature ) . from this point of view
it is preferable over la model that , obviously , could hardly be applied for the explanation of the properties of various low molecular weight nonpolar organic glasses .
these materials demonstrate the same pf field dependence as ppvs and can be perfectly well described by the qg model .
high intramolecular flexibility is a necessary prerequisite for significant amplitude of the energetic disorder in the la model , yet experimental data unambiguously demonstrate that some nonpolar materials consisting of very rigid molecules may still have significant energetic disorder ( a good example of the material with planar rigid nonpolar molecules having @xmath104 ev is discussed in ref . ) .
it is worth to add that the very observation that served as a first clue for the la model , the significant difference between slopes of the pf dependence for two particular polyconjugated polymers - one is traditional meh - ppv @xcite and the other one ( having very small slope of the mobility curve ) is stiff - chain polyfluorene @xcite - may be reasonably well explained using the qg model .
indeed , the local stiffness of the main polymer chain should by itself decrease the qudrupolar disorder .
in addition , the molecule of polyfluorene contains only carbon and hydrogen atoms , while the molecule of more flexible meh - ppv contains oxygen atoms as well - this means that the qudrupolar moment of the monomer unit for meh - ppv is greater .
hence , both factors acts in the same direction , decreasing the quadrupolar disorder in polyfluorene and weakening the mobility field dependence .
recent papers on charge transport in ppvs offer a limited opportunity to compare qg and los alamos models .
@xcite these papers offer data on the temperature dependence of the pf coefficient @xmath105 ( see eq [ pf ] ) . by definition @xmath106 and for the la model @xmath107 .
although in the qg model coefficient @xmath108 is not a true constant with respect to @xmath1 , its field dependence is very weak @xmath109 and we can approximately treat it as an effective constant .
results of the fit @xmath110 are shown in fig .
[ fig_fit ] . obtained by fitting the temperature dependence of @xmath108 for different ppvs on @xmath77 in @xmath110 .
the best fit corresponds to the minimum of @xmath111 .
curves for different ppvs from ref . are marked a , b , c , and d ; curve a represents data from ref .
( the same ppv as for the curve a , but different thickness of the transport layer).,width=288 ] for most ppvs the best @xmath77 is close to 1.5 and , thus , more close to 1.25 ( qg model ) than to 1 ( la model ) . a possible reason for the discrepancy with the qg model may be the specific technique used in refs . where mobility was calculated from the current - voltage characteristics in the space - charge limited conduction regime assuming the pf mobility field dependence
thus , the mobility field dependence was not calculated at all , only the coefficient @xmath108 was optimized to fit experimental current - voltage curves .
the direct study of the temperature dependence of the corresponding true qg coefficient @xmath112 obtained from tof experiments should be a better way to test the qg model . in a typical case the total energetic disorder in a material
has several contributions of different nature @xmath113 ( say , dipolar and quadrupolar components ) .
in such situation the correlation function ( assuming statistical independence of individual components ) is a sum of terms having different dependence on distance .
the resulting mobility dependence in the 1d model has the form @xmath114\right ) , \label{mu_tot}\ ] ] which can not be written in the form of an explicit simple formula for nontrivial functions @xmath115 . moreover
, mobility field dependence could present a clear indication of the existence of several contributions to ( [ utot ] ) only when mobility is measured in an exceptionally wide field range .
careful analysis shows that calculation of individual dipolar and quadrupolar contributions to the total disorder could be performed only in the case when mobility field dependence is measured in the field range spanning at least three orders of magnitude @xcite .
this is hardly possible in real experiments .
most organic materials contain impurities serving as traps for charge carriers ( i.e. such impurities have energy levels laying deeper than levels of majority of transport sites ) .
typically , energies of trap sites have no spatial correlation and produce an additional spatially uncorrelated disorder .
important question is : to what extent general properties of charge transport in correlated medium are insensitive to the presence of traps ( later we will limit our consideration to the the particular case of dipolar medium ) ?
this problem has been extensively studied in the 1d approximation and by means of 3d computer simulation @xcite . again ,
major results are the same for both approaches and may be summarized as follows : 1 .
addition of traps leads to the decrease of the carrier drift mobility ; 2 .
pf dependence in weak fields remains unchanged without respect to trap depth @xmath116 and concentration @xmath34 ; 3 . in stronger fields
a linear field dependence @xmath117 emerges .
these conclusions seem to be in striking contradiction with recent experimental observations for transport of holes in doubly doped polymer layers .
@xcite molecules of one dopant , added in small concentration , and possessing significantly lower ionization potential , served as traps for charge carriers . in these studies
it was found that for shallow traps the pf dependence remains mainly untouched @xcite , while for deep traps a linear field dependence was observed in the whole field range . @xcite the most puzzling experimental result is the unusual dependence of the mobility on trap concentration @xmath34 @xcite @xmath118 with @xmath119 instead of expected dependence with @xmath120 for trap - controlled transport .
again , this result does not agree with the theoretical dependence where @xmath120 .
@xcite where @xmath121 is the time of intersection of asymptotes to the plateau and trailing edge of the transient ; and 2 ) @xmath122 where @xmath123 is the time for photocurrent to decay to the half of its plateau value.,width=288 ] quite unexpected solution for all these puzzles was suggested in ref . .
surprisingly , the problem is not connected with any fault of the theory or inaccuracy of modern experimental technique but rather with the interpretation of the experimental data , namely with the particular way of the mobility calculation from the photocurrent transient temporal dependence . in tof experiment , mobility
is usually calculated by two methods ( fig .
[ fig_mu ] ) .
the first method is the method of choice for most experimental papers .
unfortunately , this particular method is too sensitive to the variation of the shape of the transient with variation of @xmath1 ( or @xmath34 ) and tends to overestimate the contribution of fast under - relaxed carriers .
@xcite the second method produces mobility that is much more close to the strict definition of the mobility as @xmath124 where @xmath125 is an average carrier velocity .
the sensitivity of the first method is so significant that even in the case when good current plateau with @xmath126 const is observed in the whole range of variation of the parameter ( e.g. , @xmath1 ) , there is a possibility to obtain principally different mobility field dependences for @xmath127 and @xmath128 ( see fig .
[ fig_x ] ) . , @xmath129 , @xmath130 , and @xmath131 lattice planes ( equivalent to transport layer thickness of 30 - 40 @xmath0 m ) for different methods of mobility calculation : @xmath132 - diamonds , @xmath133 - triangles , and @xmath134 - squares .
here @xmath34 is a fraction of sites occupied by traps.,width=288 ] note , that for this particular case in the whole field range the transient is more or less nondispersive and has a well - defined plateau ( fig . [ fig_x2 ] ) .
we can conclude that in some cases mobility , calculated by the first method , effectively masquerades a real field dependence of the true mobility .
the same reason explains an unusual dependence of the mobility on trap concentration .
@xcite : 0.021 ( thin line ) and 1.70 ( thick line ) , correspondingly , with time axis rescaled to make values of @xmath123 equal for both transients .
all other parameters are the same with fig .
[ fig_x ] .
note the different shapes of transients.,width=288 ] the second method produces much more reliable results and should be used instead of the first one .
this conclusion is in sharp contradiction with the common belief that field dependences of @xmath135 and @xmath133 are essentially the same ( experimental evidence for the difference may be found in literature , @xcite but it is usually ignored ) .
the use of @xmath133 is especially important in situations where we have to choose among not very distinctive alternatives , such as qg or pf mobility field dependence in nonpolar materials .
original pf mechanism was for a long time considered unacceptable for explanation of the mobility field dependence in disordered organic materials due to absence of charged traps ( arguments provided in ref . suggested that trap density is certainly less than @xmath136 @xmath137 ) . yet in a recent paper the pf model was revitalized by rackovsky and scher , @xcite who argued that a very low density of coulomb traps is enough to produce an essentially nondispersive pf charge transport ( they estimated that @xmath138-@xmath136 @xmath139 should be a sufficient density for transport layer with the thickness of 10 @xmath0 m ) .
because such low density can easily avoid detection , they suggested that the pf model still can be considered as a possible candidate for explanation of charge transport in disordered organic materials .
calculation of the pf effect in ref . was carried out for the usual case of an isolated coulomb trap only .
yet charged traps produce strongly spatially correlated energy landscape , and the hopping charge motion in such landscape usually has many features that can not be captured by the approximation taking into account interaction of charge carrier with the isolated force center .
early indication that charge transport in charged medium may significantly differ from predictions of the model of isolated coulomb trap was presented by dunlap and novikov @xcite . in recent papers
@xcite this problem has been considered in more detail for a special case of the medium with equal concentration of randomly located positive and negative traps without using the isolated trap approximation .
@xcite the result is in drastic contrast with the approximation of isolated coulomb trap .
for example , at @xmath140 ( here @xmath77 is trap density ) a transition from mobile to immobile carriers occurs and average carrier velocity @xmath141 for @xmath142 in the infinite medium , so in finite transport layers the only possible regime is a dispersive transport with mobility depending on thickness .
@xcite by accident , mobility field dependence for @xmath142 retains the pf form , but its temperature dependence and , more important , concentration dependence differ drastically from the prediction of the isolated trap model . @xcite for small concentration of traps @xmath143^{1/2 } , \label{pf_true}\ ] ] @xmath144 here the lattice version of the model is considered assuming that fraction of sites occupied by charges is small @xmath145 but @xmath146
. experimental test of the mobility dependence on the concentration of charged traps should be the best test of the result ( [ pf_true ] ) . by
now there are no experimental data on this dependence ( actually , no experimental data at all on charge transport in disordered organic materials containing controllable concentration of static charges ; quite possibly , such data are very difficult to obtain ) .
nonetheless , the mobility temperature dependence ( [ pf_true ] ) seems to be too strong to describe existing experimental data .
for this reason we believe that the model of charged traps can not be suggested as a serious candidate for explanation of charge transport in disordered organic materials .
all variety of disordered organic materials could be subdivided into different classes on the basis of their spatial correlation properties in the distribution of energies of transport sites .
materials from different classes have different transport properties though differences are not very pronounced . in order to discriminate between different classes of the mobility field and temperature dependence , analysis of experimental data
should be carried out with utmost care . in this respect
an early attempt to characterize all materials in a unified manner ( the gaussian disorder model ) is unjustified while we can now understand , why it was so successful .
to some extent charge transport theory is ahead of experiment : major predictions ( mostly concerning charge transport in nonpolar materials ) still lack experimental test .
for example , the reasonable question is : which model , quadrupolar glass or los alamos , is more suitable to describe charge transport in nonpolar ppvs ?
a careful and purposeful experimental study of the transport properties of nonpolar materials could resolve this problem .
* acknowledgements * i am greatly indebted to a.v .
vannikov , p.m. borsenberger , d.h .
dunlap , p.e .
parris , v.m .
kenkre , h. bssler , a. kadaschuk , and z. soos for numerous discussions .
i am especially indebted to s. nespurek for providing me with data on transport properties of polysilanes
. partial financial support of the international science and technology center grant 2207 and the russian fund for basic research grant 02 - 03 - 33052 is acknowledged .
this work was partially supported by the university of new mexico and utilized the unm - alliance loslobos supercluster at the albuquerque high performance computing center .
20 pope , m. ; swenberg , c. e. electronic processes in organic crystals and polymers ; oxford university press : new york , 1999 .
borsenberger p.m. ; weiss , d.s .
organic photoreceptors for imaging systems ; marcel dekker : new york , 1997 .
borsenberger , p.m. ; magin , e.h . ; van der auweraer , m. ; de schryver , f.c .
phys status solidi a 1993 , 140 , 9 - 47 .
bssler , h. phys status solidi b 1993 , 175 , 15 - 56 .
schein , l.b . ; borsenberger , p.m. chem phys 1993 , 177 , 773 - 781 .
schein , l.b . ; peled , a. ; glatz , d. j appl phys 1989 , 66 , 686 - 689 .
gill , w.g .
j appl phys 1972 , 43 , 5033 - 5040 .
novikov , s.v . ; dunlap , d.h . ; kenkre , v.m . ; parris , p.e . ; vannikov , a.v .
phys rev lett 1998 , 81 , 4472 - 4475 .
dieckmann , a. ; bssler , h. ; borsenberger , p.m. j chem phys 1993 , 99 , 8136 - 8141 .
novikov , s.v . ; vannikov , a.v .
jetp 1994 , 79 , 482 - 485 .
novikov , s.v .
; vannikov , a.v .
j phys chem 1995 , 99 , 14573 - 14576 .
novikov , s.v .
phys status solidi b 2000 , 218 , 43 - 46 .
novikov , s.v . ; vannikov , a.v . ,
mol crystals and liquid crystals 1998 , 324 , 177 - 182 .
gartstein , yu.n . ; conwell , e.m .
chem phys lett 1995 , 245 , 351 - 358 .
dunlap , d.h . ; parris , p.e . ; kenkre , v.m .
phys rev lett 1996 , 77 , 542 - 545 .
novikov , s.v . ; vannikov , a.v .
spie proc 1996 , 2850 , 130 - 138 .
novikov , s.v . ; vannikov , a.v .
synth met 1997 , 85 , 1167 - 1168 .
dunlap , d.h . ; kenkre , v.m . ; parris , p.e .
j imag sci technol 1999 , 43 , 437 - 443 . parris , p.e .
j chem phys 1998 , 108 , 218 - 226 .
novikov , s.v . ; vannikov , a.v .
spie proc 1997 , 3144 , 100 - 109 .
borsenberger , p.m. ; gruenbaum , w.t . ; magin , e.h .
physica b 1996 , 228 , 226 - 232 .
borsenberger , p.m. ; gruenbaum , w.t . ; magin , e.h .
phys status solidi b 1995 , 190 , 555 - 563 .
sinicropi , j.a . ; cowdery - corvan , j.r . ; magin , e.h . ; borsenberger , p.m. chem phys 1997 , 218 , 331 - 339 .
heun , s. ; borsenberger , p.m. chem phys 1995 , 200 , 245 - 255 .
novikov , s.v .
; dunlap , d.h . ;
kenkre , v.m .
, spie proc 1998 , 3471 , 181 - 191 .
scher , h. ; montroll , e.w .
phys rev b 1975 , 12 , 2455 - 2477 .
novikov , s.v . ; vannikov , a.v . , mol crystals and liquid crystals 2001 , 361 , 89 - 94 .
nespurek , s. private communication .
; smith , d.l . ; saxena , a. ; martin , r.l . ; bishop , a.r .
phys rev lett 2000 , 84 , 721 - 724 .
yu , z.g . ;
smith , d.l .
; saxena , a. ; martin , r.l . ; bishop , a.r .
phys rev b 2001 , 63 , 085202 .
campbell , i.h . ; smith , d.l . ; neef , c.j . ; ferraris , j.p .
appl phys lett 1999 , 74 , 2809 - 2811 .
redecker , m. ; bradley , d.d.c . ; inbasekaran , m. ; woo , e.p .
appl phys lett 1998 , 73 , 1565 - 1567 .
blom , p.w.m .
; de jong m.j.m .
; van munster m.g .
phys rev b 1997 , 55 , r656-r659 .
martens , h.c.f . ; blom , p.w.m . ;
h. f. m. schoo , h.f.m .
phys rev b 2000 , 61 , 7489 - 7493 .
novikov , s.v .
proceedings of nip15 : international conference on digital printing technologies ; society for imaging science and technology : orlando , fl , 1999 ; 683 - 686 .
veres , j. ; juhasz , c. phil mag b 1997 , 75 , 377 - 387 .
wolf , u. ; bssler , h. ; borsenberger , p.m. ; gruenbaum , w.t .
chem phys 1997 , 222 , 259 - 267 .
lin , l.b . ; borsenberger , p.m. ; magin , e.h . ; visser , s.a . ; gruenbaum , w.t .
spie proc 1998 , 3471 , 174 - 180 .
novikov , s.v .
j imag sci technol 1999 , 43 , 444 - 449 .
rackovsky , s. ; scher , h. j chem phys 1999 , 111 , 3668 - 3674 .
dunlap , d.h . ; novikov , s.v .
spie proc 1997 , 3144 , 80 - 91 .
novikov , s.v .
proceedings of nip17 : international conference on digital printing technologies ; society for imaging science and technology : fort loderdale , fl , 2001 ; 764 - 767 .
novikov , s.v .
phys status solidi b 2003 , in press .
the case of charged medium with unequal concentrations of positive and negative traps is unusual because in this medium energetic disorder is spatially inhomogeneous and the nondispersive charge transport is not possible at all . | general properties of charge carrier transport in disordered organic materials are discussed .
spatial correlation between energies of transport sites determines the form of the drift mobility field dependence .
particular kind of spatial correlation in a disordered material depends on its nature .
mobility field dependences have to be different in polar and nonpolar materials .
different methods of mobility calculation from the shape of photocurrent transient are analyzed .
a widely used method is very sensitive to the variation of the shape of the transient and sometimes produces results that effectively masquerade the true dependence of the mobility on electric field or trap concentration .
arguments in favor of the better , more reliable method are suggested .
charge transport in materials containing charged traps is considered without using the isolated trap approximation and this leads to qualitatively different results .
they indicate that the effect of charged traps can hardly be responsible for experimentally observed transport properties of disordered organic materials .
* keywords : * charge transport ; correlated disorder ; computer simulations * charge carrier transport in disordered polymers * * s. v. novikov * a.n .
frumkin institute of electrochemistry , leninsky prosp .
31 , moscow 119071 , russia |
the bl lac object mkn 501 is very close ( @xmath2 , ulrich et al .
@xcite ) and has been studied extensively at all wavelengths . together with its sister object
mkn 421 , it was among the two first bl lac objects with known radio ( colla et al .
@xcite ) , x - ray ( schwartz et al .
@xcite ) and tev gamma - ray ( quinn et al .
@xcite , bradbury et al .
@xcite ) counterparts .
recently , it was also marginally detected at photon energies @xmath3 mev by the egret instrument on board cgro ( kataoka et al .
@xcite ) . during 1997
, the object was found to be in an extreme high state with a tev flux on average 20 times higher than in 1996 ( breslin et al .
the source exhibited strong variability on timescales of days with a possible quasi - periodically varying component with a timescale of about 25 days ( kranich et al .
@xcite ) . to complete the list of reasons for excitement for the observers , mkn 501 reached in some of its flares fluxes of more than 10@xmath4 @xmath5s@xmath6 ( above 1.5 tev ) - the most intense tev emission ever measured so far from any astronomical object . however , the shortest observed variability timescale ( 5 hours , aharonian et al .
@xcite ) was significantly longer than that observed for mkn 421 ( gaidos et al .
@xcite ) .
all available tev observatories monitored the event for several months ( samuelson et al .
@xcite , aharonian et al .
@xcite,@xcite , hayashida et al .
@xcite , djannati - ata et al .
the most complete dataset was produced by the hegra cherenkov telescope `` 1 '' ( ct1 ) ( aharonian et al . ,
@xcite ) which was even able to observe mkn 501 under the presence of moonlight , however with reduced sensitivity , thereby filling many gaps in the lightcurve .
this telescope also obtained the confirmatory observations in 1996 . also from hegra come probably the most accurate spectral measurements in the tev regime .
they were carried out by the hegra system of ( at the time 4 ) cherenkov telescopes ( cts ) and are largely concurrent with the ct1 measurements , however with less time coverage ( aharonian et al .
@xcite ) .
the origin of the tev @xmath1-ray emission and the reasons for its variability are still essentially unknown .
the most popular models explain the tev emission as near - infrared to uv photons which have been upscattered via the inverse compton effect by very high energy electrons which are known from radio , optical and x - ray observations to be present in the jets of bl lac objects .
possible sources of the seed photons for compton scattering are the synchrotron radiation produced within the jet by the same population of relativistic electrons ( synchrotron self - compton model , ssc ; marscher & gear @xcite ; maraschi , celotti & ghisellini @xcite ; bloom & marscher @xcite ) , or radiation from outside the jet ( external inverse compton model , eic )
. this external radiation could be the quasi - thermal radiation field of an accretion disk surrounding a supermassive black hole which is generally believed to power the relativistic jets .
the accretion disk radiation can enter the jet either directly ( dermer , schlickeiser & mastichiadis @xcite , dermer & schlickeiser @xcite ) or after being rescattered by circumnuclear material ( sikora , begelman & rees @xcite , blandford & levinson @xcite , dermer , sturner & schlickeiser @xcite ) .
it is also possible that synchrotron radiation produced within the jet and reflected by circumnuclear debris is the dominant source of soft photons during flares ( ghisellini & madau @xcite , bednarek @xcite ) , although it has been shown that this process is unlikely to be efficient in the case of bl lac objects ( bttcher & dermer @xcite ) .
as an alternative , mannheim ( @xcite ) has suggested a hadronic model in which protons are the primarily accelerated particles in the jet and the @xmath1-ray emission is produced by secondary pions and electron - positron pairs produced in photopion and photopair production interactions of the ultrarelativistic protons in the jet with external radiation .
the time - averaged broadband spectrum of mkn 501 has been fitted using this model by mannheim et al .
( @xcite , @xcite ) . however , the attempts to explain the short and intermediate - term variability of blazars with hadronic jet models have only just started ( rachen & mannheim @xcite ) . for this reason
, we concentrate on leptonic jet models in this paper since the variability time scales predicted by these models are in good agreement with the observed intraday variability of blazars . recently , fossati et al .
( @xcite ) have compared the broadband spectra of different types of @xmath1-ray emitting agn and suggested a continuous sequence fsrqs ( flat spectrum radio quasars ) @xmath7 lbls ( low - frequency peaked bl lac objects ) @xmath7 hbls ( high - frequency peaked bl lac objects ) , characterized by decreasing bolometric luminosity , decreasing dominance of the total energy output in @xmath1-rays compared to the emission at lower frequencies and a shift of the peak frequencies of the synchrotron and the @xmath1-ray component towards higher frequencies .
recent modelling efforts of various blazar - type agn have revealed that this sequence is consistent with a decreasing importance of external radiation as a source of soft photons for compton scattering in the jet ( ghisellini et al .
this suggests that the extreme hbls like mkn 501 or mkn 421 can be well fitted with strongly ssc - dominated jet models , as was shown , e. g. , by mastichiadis & kirk ( @xcite ) and pian et al .
( @xcite ) . from the ssc model one expects a strong correlation between the x - ray and the tev emission . and
indeed , by comparing the daily tev measurements with daily averages from the all sky monitor ( asm ) on board the rossi x - ray timing explorer ( rxte ) , a significant correlation with a most probable time - lag of 0 days ( no time - lag ) was found ( aharonian et al .
@xcite ) . furthermore
, a clear overall x - ray high state was visible in the 1997 asm measurements which coincided with the tev high state ( see also figure [ fig - lightcurve ] ) . and ,
as discovered by bepposax , the x - ray peak of the spectral energy distibution ( sed ) had shifted from 10 - 20 kev in 1996 to 100 - 200 kev in 1997 ( pian et al .
@xcite ) , consistent with the assumption that the x - ray and tev-@xmath1-ray flares are produced by a more powerful electron acceleration , shifting the high - energy cutoff of the electron distribution to higher energies and hardening the x - ray spectrum .
this observation was , however , only made during one flare in 1997 . in this paper
we give evidence that the hardening took place during all flares in 1997 .
an alternative explanation for the synchrotron spectral changes , in terms of a steadily - emitting helical - jet model , has been presented by villata & raiteri ( @xcite ) . apart from the short - term variability on timescales of hours to days , there is also a longer - term variability in blazars which has so far been investigated mainly in the optical regime ( see e.g. katajainen et al .
( @xcite ) and references therein ) and recently also in x - rays ( e.g. mc hardy @xcite ) .
this variability shows remarkable amplitudes ( e.g. 4.7 mag for mkn 421 in the optical ) and in the case of oj 287 there is even evidence for periodicity . in this respect mkn 501
is not yet very well explored .
mkn 501 has been observed in the tev regime since its discovery in 1995 .
the source showed low emission close to the sensitivity limits until the onset of the 1997 high state .
since the duty cycle of the tev observatories is only about 10 % , this does not prove the absence of strong short flares prior to this high state , but the increased average intensity of the source can be described as an increase in flaring probability from 1996 to 1997 by at least an order of magnitude .
this description is especially appropriate since even during the high state , the source returns to quiescent ( comparable with 1996 ) levels of emission for periods of up to a few days .
these transitions are seen irrespective of the presence of short , strong flares on time - scales of a day .
the last observations in 1996 were made in august and found mkn 501 still quiescent while the first observations in 1997 were made in february and found the source already flaring .
the transition from low to high flaring probability obviously took place on timescales shorter than half a year .
the correlated x - ray data from the rxte asm confirm this and show in addition that the change has been a smooth process over several months ( see e.g. figure 8 in aharonian et al .
@xcite ) . in this article
we explore the multi - wavelength variability of mkn 501 over medium timescales , and try to relate its behaviour to the physical parameters of a leptonic jet model . for this purpose
we construct weekly seds using the hegra ct1 flux data and hegra ct system spectral data together with data from longer wavelengths , namely radio , optical , soft x - ray and hard x - rays .
see table [ tab - instruments ] for an enumeration of the instruments and energy ranges .
this paper may represent an important step toward understanding the origin of strong changes in the flaring probability of mkn 501 and blazars in general , but our data is clearly not sufficient to give a final answer to this question . all data except that from hegra and the rxte asm are published here for the first time .
the batse data are especially valuable since they confine the intensity at the x - ray peak of the sed .
the observations used to fit the multi - wavelength spectrum cover mostly the synchrotron part of the sed . only the tev data explore what is believed to be the inverse compton emission , although osse and egret observations cover a few days in 1996 and 1997 . as a guide line for our fitting procedure we take into account
the highest ever observed egret flux from mkn 501 , @xmath8 photons @xmath5 s@xmath6 with a photon spectral index @xmath9 , measured during 1996 mar 25 - 28 ( kataoka et al .
@xcite ) , as an upper limit . in the definition of time - bins for the multi - wavelength dataset , we start by subdividing the hegra ct1 lightcurve and bin all other data accordingly .
the data from hegra are binned in time from march through october 1997 ( mjd 50514 - mjd 50708 ) resulting in 28 equidistant time bins .
this weekly temporal resolution is an order of magnitude larger that the longest observed tev intraday variability timescale of 15 h ( aharonian et al .
@xcite ) and is believed appropriate , given the nature of the available data and the medium scale variability timescale which we have chosen to explore .
however , the hegra points are not spread as uniformly over time as are the batse and rxte data .
the `` center of gravity in observation time '' ( defined as the weighted mean of the observation time of the daily points each weighted by the duration of the individual observation ) from hegra would therefore in general not coincide with that from batse and rxte . in order to compensate for this
, we calculate the center of gravity in observation - time for each weekly hegra timebin and use these as time - bin-_centers _ for the other data .
the edges of these bins are then defined by the average of two adjacent bin - centers .
figure [ fig - lightcurve ] shows the lightcurves from each instrument which went into this analysis .
the weekly points after the binning are given in table [ tab - data ] .
the first column in the table gives the bin - centers as described above .
the following subsections describe the data used to compile the table .
the optical observations were performed using the telescopes and filters listed in table [ tab - instruments ] .
all the observations were made with ccd - cameras .
all ccd - images were treated the normal way with flat field and bias corrections .
the magnitudes were measured using either daophot in iraf ( not and tuorla data ; for more details see katajainen et al .
@xcite ) , or with the robin - procedure developed in torino ( e.g. villata et al .
@xcite ) , or using the automated reduction routine developed in perugia ( e.g. tosti et al .
all magnitudes were measured using 10 arcsecond aperture .
the use of the same aperture is important because mkn 501 has a large host galaxy ( e.g. nilsson et al .
@xcite ) . for the calibration we used the calibration sequence given by fiorucci & tosti ( @xcite ) and villata et al .
( @xcite ) .
for some of the weekly observing periods , no optical data were available .
however , the optical flux is extremely important for constraining the spectral index of the synchrotron spectra in the optical x - ray range , which , in turn , is essential for constraining the spectral index of the electron spectrum . as can be seen in table [ tab - data ] , the optical flux exhibits only moderate variability on the time scales considered here .
thus , for fitting purposes we assume that the optical flux during those viewing periods for which no optical observations were available , was within the range of optical fluxes observed during the whole campaign which yields @xmath10 erg @xmath5 s@xmath6 .
the radio observations were made with the metshovi radio telescope at 22 ghz as part of the ongoing quasar monitoring program , which started in 1980 .
currently about 85 sources , mainly northern flat spectrum quasars , should be observed monthly at 22 and 37 ghz .
the total number of observations is now over 40000 . for details of the observing strategy and reductions , as well as
the data until 1995.5 see tersranta et al .
( @xcite ) .
the integration time of each point was essentially the same , so single points were formed from the data points in the individual timebins by simple averaging .
since the beginning of 1996 , the all - sky monitor ( asm ) of the rossi x - ray timing explorer ( rxte ) satellite has been observing mkn 501 in the 2 - 12 kev energy band .
the data used in this analysis are taken from the publicly available asm data products provided by the asm / rxte teams at mit and at the rxte sof and gof at nasa s gsfc .
these measurements are given by the authors ( see e.g. levine et al .
@xcite ) as rates @xmath11 in units of counts per second in three energy bins : 1.3 - 3.0 kev , 3.0 - 5.0 kev , and 5.0 - 12.1 kev . from these
we calculate the total rate @xmath12 by summing the three bins . for mkn 501
we obtain by averaging over the period mjd 50510 - 50710 @xmath13 in order to assess the x - ray spectrum and flux of mkn 501 , we use the asm data for the crab nebula which is publicly available from the same source .
the average rate of the crab nebula during the period mjd 50510 50710 is @xmath14 hence , in soft x - rays , mkn 501 is even during this high state a significantly weaker source than the crab nebula . in the energy range 2 kev to 60 kev
, the spectrum of the crab nebula is a stable power law .
the spectral index of the differential photon spectrum is known with an accuracy of 1% to be @xmath15 ( e.g. toor & seward @xcite , pravdo & serlemitsos @xcite or pravdo , angelini & harding @xcite ) .
the emission measured by the rxte asm is the sum of the steady nebula and the pulsed crab pulsar emission .
the latter has on average a harder spectrum than the nebula .
below 12.1 kev , however , the pulsed fraction of this emission is @xmath16 % .
we can therefore use the knowledge of the crab nebula spectrum to derive an approximate relative calibration factor @xmath17 for the rates @xmath18 and @xmath19 using @xmath20 where @xmath21 kev , @xmath22 kev , and @xmath23 kev are the energy bin edges of the second and third energy bin .
we ignore the first energy bin since it is strongly influenced by interstellar x - ray absorption which is dependent on the column density and hence varies between sources ( remillard @xcite ) .
averaging over the available crab data from the observation period under discussion ( mjd 50510 - 50710 ) we obtain @xmath24 .
there is no indication of a variability of the value of @xmath17 ( see figure [ fig - k ] ) .
hence we can assume that it is also valid for the mkn 501 observations of the same detector . in order take into account that there is a pulsed component with a harder spectrum ,
we add an additional error of 8 % to the error of @xmath25 and obtain thus @xmath26 for mkn 501 we calculate the spectral index @xmath27 of the differential photon spectrum using the equation @xmath28 and varying @xmath27 until the two sides of the equation are equal to an accuracy better than 0.1 % .
the error @xmath29 of this index , we estimate from the approximate formula @xmath30 which leads to @xmath31 where @xmath32 , @xmath33 , and @xmath34 are the errors of the corresponding quantities . the time resolved values for @xmath35 are shown in figure [ fig - alphas ] .
the same method applied to the crab nebula data ( however with @xmath36 ) yields , as expected , on average the spectral index we have put in .
also this is shown in figure [ fig - alphas ] . from a constant fit to the values in this figure , we obtain @xmath37 and @xmath38 thus we find that during the 1997 high state , mkn 501 had on average a significantly harder soft x - ray spectrum than the crab nebula .
the spectral variability of mkn 501 is below the asm s sensitivity .
the distribution of the weekly values is consistent with a constant value ( reduced @xmath39 = 0.93 ) and so is that for the crab nebula ( reduced @xmath39 = 0.95 ) . in order to calculate the @xmath40 values in erg @xmath5 s@xmath6
, we use the knowledge of the flux of the crab nebula . here
we face the problem that there is still a disagreement of up to 25 % between the measured normalization constants of the crab spectrum from different experiments although there is perfect agreement in the spectral index .
the discrepancy seems to stem from not very well understood systematic differences between the detectors ( see the discussion in pravdo , angelini & harding @xcite ) .
we use the two most extreme recent measurements of the differential photon flux of the crab nebula together with the crab pulsar ( pulse - averaged ) and take their average as our normalization and their difference as the systematic error of this quantity . from pravdo & serlemitsos ( @xcite )
we get at 5.2 kev a flux of ( 0.236 @xmath41 0.006 ) photons @xmath5s@xmath6kev@xmath6 , while from pravdo , angelini & harding ( @xcite ) we get ( 0.302 @xmath41 0.001 ) photons @xmath5s@xmath6kev@xmath6 .
the average of these values corresponds to a differential energy flux of @xmath42 where the systematic error is the difference between the averaged values divided by 2 .
the statistical errors of the two measurements are negligible .
the energy flux of any other source with asm rates @xmath43 and @xmath25 and differential photon spectral index @xmath27 is then calculated by @xmath44 & = & ( r_2 + r_3)[\mathrm{s}^{-1 } ] \cdot \frac{\displaystyle f_{crab}(5.2 \mathrm{kev})[\mathrm{erg}\,\mathrm{cm}^{-2}\mathrm{s}^{-1}\mathrm{kev}^{-1 } ] } { \displaystyle ( r_{2,crab}+r_{3,crab})[\mathrm{s}^{-1 } ] } \\ & & \cdot 5.2^{1 + \alpha_{crab } - \alpha } \cdot \frac{\displaystyle 1 - \alpha}{\displaystyle 1 - \alpha_{crab } } \cdot \frac{\displaystyle e_4^{1 - \alpha_{crab } } - e_2^{1 - \alpha_{crab } } } { \displaystyle e_4^{1 - \alpha } - e_2^{2 - \alpha } } \\ & = & r[\mathrm{s}^{-1 } ] \cdot \frac{\displaystyle 5.2^{1 - \alpha } \cdot ( 1 - \alpha ) } { \displaystyle 12.1^{1 - \alpha } - 3.0^{1 -\alpha } } \cdot 3.13 \times 10^{-10 } \\
\end{array}\ ] ] where we correct for the difference in the spectra of crab nebula and the source in question by making the ansatz that the measured differential rates have the same ratio as the differential fluxes .
furthermore we have used the result @xmath45 s@xmath6 obtained from the dataset under discussion . inserting the spectral index of mkn 501
gives : @xmath46 \cdot 2.40 \times 10^{-10 } \mathrm{erg\,cm}^{-2 } \mathrm{s}^{-1}\ ] ] the error of this flux value is determined by propagating all errors of the quantities involved which gives @xmath47 where the term outside the square - root stems from the systematic error of the crab flux normalization .
the hard x - ray fluxes were measured using the burst and transient source experiment ( batse ) onboard the cgro satellite .
although batse is an uncollimated detector , accurate point source fluxes can be measured using the earth occultation method described in harmon et al .
( @xcite ) .
daily sensitivities are 100 mcrab and over integrations of years , sources as weak as 3 mcrab can be detected .
several months are usually needed to obtain a statistically significant flux from mkn 501 with batse , but intense daily flares can be seen , and the overall high state of 1997 allowed useful measurements in each weekly interval even outside the flares . the source is measured only when it sets and rises from behind the earth . since two such occultations occur per spacecraft orbit ( roughly every 90 minutes ) , up to 32 independent flux measurements can be made per day
. there exists , however , a wide variation in this number because of passage of the spacecraft through the south atlantic anomaly , telemetry gaps from loss of tdrss contact , and other events , which occur randomly relative to the mkn 501 steps . for the data shown here between 61 and 211 measurements went into an individual weekly point .
each measurement lasts about 8 seconds giving a duty cycle of about 0.2% . in a 7 day time bin during the period discussed in this paper ,
batse observes of the order of @xmath48 photons from the source ( more during the flares ) .
the fluxes are integrated between 20 and 200 kev , and are calculated by folding the measured counts through the batse detector response assuming a differential source powerlaw spectrum of index -2.0 .
the -2.0 spectrum is the best fit to the flare measured on mjd 50550 - 51 ( between 20 and 1000 kev ) .
spectra for other time intervals were also calculated and were consistent with -2.0 .
uncertainties of 10 % in the spectral index during flare times , larger outside , make spectral variability difficult to assess for this source , so that the index of -2.0 was used for each weekly interval .
fluxes were calculated for smaller energy bands but the single 20 - 200 kev ( median energy 36.4 kev ) point for each interval is the most useful outside intense flares .
seven of the weekly averages are not statistically significant , and one period shows a @xmath49 deficit .
these eight points are inconsistent with a zero - level flux at the 99 % confidence level .
the seven low but positive points are inconsistent with a zero - flux level at the 98 - 99 % confidence level and are well fit by a constant flux of @xmath50 erg @xmath5 s@xmath6 , implying that a flux is present which is below the sensitivity of the batse instrument in weekly integrations .
an analysis of 30 blank fields on the sky shows that with 100 weekly integrations for each field one can not distinguish ( using an f - test ) between a zero - flux level and the weighted mean of the weekly averages for any of the 30 fields .
this implies that systematic effects are not responsible for the excess seen in the low points and that a flux is indeed present .
for this reason these low but positive values have been included as detections in the multiwavelength fits .
as described above , we use the data from the ct1 lightcurve of mkn 501 ( integral flux above 1.5 tev ) published by aharonian et al .
( @xcite ) .
these data are available in daily points based on observations of between 0.5 h and 5 h duration each .
we group these points according to our weekly time - bins and calculate an average flux from the up to seven values weighting each daily point by its observation time .
apart from giving daily flux measurements , the hegra papers aharonian et al .
( @xcite ) , ( @xcite ) , and ( @xcite ) also determine the average spectral shape in the range between 0.5 and @xmath51 25 tev with high accuracy .
they find @xmath52 where @xmath53 @xmath5 s@xmath6 tev@xmath6 , @xmath54 , and @xmath55 tev .
the first error given is the statistical , the second the systematic error .
furthermore , they find that there is no spectral variability up to their sensitivity of @xmath56 on all relevant timescales . in order to include this important spectral information in our model fit
, we make the assumption that there is indeed no spectral variability and extrapolate the points measured at 1.5 tev by ct1 up to 10 tev and down to 0.8 tev . from the average integral flux values @xmath57 in units of photons @xmath5s@xmath6 we obtain the @xmath40 values at photon energy @xmath58 in tev using @xmath59 = 1.6022 \cdot f_{1.5 } \ , { e^{(2 - \alpha ) } \cdot \exp(-e / e_0 ) \over \int\limits_{1.5}^{\infty } d\epsilon \
> \epsilon^{-\alpha } \ , \exp(-\epsilon / e_0)}.\ ] ] the extrapolation uses the measured spectral shape ( equation [ equ - spec ] ) and fully propagates all statistical errors ( error of the ct1 point , error of the spectral index @xmath27 and the error of the cut - off energy @xmath60 ) to form the error of the extrapolated point . the systematic errors of the spectral shape are expected to influence all measured tev spectra in the same way since they are caused by the 15 % uncertainty in the absolute energy calibration of the cherenkov telescopes .
thus , for the purpose of a spectral variability study , the statistical errors are those which actually determine the uncertainty in the differences of the spectral shape between different measurements .
for this reason we propagate only the statistical errors in the extrapolation . in this way
, the columns 6 and 8 of table [ tab - data ] are obtained . in the model
fit ( section [ sec - fitting ] ) we take into account the uncertainty in the energy scale by introducing errors of @xmath61 % along the energy axis .
the energies to which we extrapolate are chosen as a compromise of maximum distance to 1.5 tev and minimum systematic errors .
the latter are increasing up to several 10% towards both ends of the range for which the spectrum has been measured ( see aharonian et al .
@xcite , figure 9 ) but are still small at 0.8 and 10 tev .
the correlation between the three tev points which we introduce into the model fit by performing the described extrapolation is not problematic since we do not plan to calculate absolute @xmath39 values in the fits for proving that the model describes the data better than another . instead
, the fits serve the aim to study the time - dependent behaviour of the model parameters .
the extrapolated points are only a means to take into account the available spectral information .
to each weekly sed we fit the blazar jet model described in detail by bttcher et al .
( @xcite ) .
the model assumes that isolated components ( blobs ) of relativistic pair plasma , which are assumed to be spherical in the co - moving frame , are injected instantaneously into the jet , and follows the self - consistent evolution of the particle and radiation spectra as the blob moves outward along the jet , taking into account all relevant radiation , cooling , and absorption mechanisms : synchrotron radiation , synchrotron self - absorption , synchrotron self - compton scattering , external compton scattering of direct accretion disk radiation , @xmath1-@xmath1 absorption and pair production intrinsic to the source .
the magnetic field is assumed to decay along the jet as @xmath62 , where @xmath63 is the distance from the center of the agn .
the emerging , time - averaged spectra are corrected for @xmath0 absorption by the intergalactic infrared background radiation using the lower model spectrum given by malkan & stecker ( @xcite ) .
since we are interested in weekly averages , the emission from single blobs is time - integrated over the jet evolution and subsequently re - converted into a flux by dividing the fluence by an average repetition time @xmath64 of blob ejection events .
we assume that a fraction @xmath65 of the jet is filled with relativistic pair plasma .
the filling factor is given by @xmath66 , where @xmath67 is the blob radius in the co - moving frame and @xmath68 is the bulk lorentz factor of the blobs . even if the filling factor is close to unity , it is a reasonable approximation to assume that the blobs do not interact with each other because the synchrotron and ssc radiation produced within the jet are isotropic in the comoving frame so that most of the radiation escapes to the sides without interaction with the rest of the jet . as mentioned in the introduction , extreme hbls like mkn 501 or mkn 421
are generally well described by a pure ssc model .
a simple analytic estimate shows that the observed tev @xmath1-ray spectrum can not plausibly be produced by comptonization of radiation from an accretion disk around the putative supermassive black hole in the center of mrk 501 : the spectrum emitted by an optically thick , geometrically thin accretion disk is reasonably well approximated by a blackbody spectrum whose temperature , for a black hole mass of @xmath69 , yields an average photon energy of the disk radiation of @xmath70 .
if external comptonization is to be efficient in competition with the synchrotron self - compton process , the blob has to be rather close to the accretion disk , or a significant fraction of the disk photons has to be rescattered into the jet by surrounding material , so that the bulk of disk photons enters the blob from the side and is blue shifted by a factor of @xmath71 into the blob rest frame .
thus , due to the strong reduction of the klein - nishina cross section for @xmath72 ( where @xmath73 is the photon energy in the comoving frame ) , no significant radiative output at ( observer s frame ) energies @xmath74 ( where @xmath75 is the doppler factor ) , corresponding to @xmath76 gev , will be produced by external comptonization , ruling out this process to explain the observed high - energy spectrum extending to tev energies .
if the high - energy spectrum in the @xmath77 gev regime is produced by the ssc process , then the level of ssc radiation at 1 gev may be estimated by @xmath78 where @xmath79 tev is the energy of the @xmath40 peak in the high - energy part of the spectrum and @xmath80 is the spectral index of the injected electron spectrum . inserting a typical value of @xmath81 and @xmath82 ,
this yields @xmath83 , which is already close to the maximum ever observed of @xmath84 . in reality , the ssc spectrum is not a straight power - law below @xmath85 , but shows a gradual turnover so that the actual ssc flux at 1 gev is substantially higher than the above estimate , leaving little room for an additional eic contribution . for our fitting procedure , we are thus using a simplified version of our jet simulation code , in which the photon output ( but not the electron cooling ) from comptonization of accretion disk radiation is neglected .
these simulations properly account for the self - consistent cooling of the electron population and @xmath0 absorption and pair production in the blob . in the simulations , the exact , angle - averaged synchrotron spectrum of an isotropic electron population as given by crusius & schlickeiser ( @xcite )
is used .
[ ssc_fits ] shows two examples of fits to low and high flux levels of mkn 501 .
the respective fit parameters are given in the figure captions .
inverse - compton cooling on accretion - disk photons may be important close to the base of the jet even if the resulting photon spectra do not contribute significantly to the time - averaged emission and is thus still included in our simulations . due to the non - linear nature of the model system , the fit results can not be described in a simple , linear way as a function of the model parameters .
we therefore construct a three - dimensional mesh of simulations in parameter space , with the electron density @xmath86 , the high - energy cutoff @xmath87 and the spectral index @xmath80 of the injected electron spectrum as parameters which are free to vary on the grid points .
we calculate our parameter grid varying @xmath80 in steps of 0.025 between 1.6 and 2.8 , the total electron number density @xmath86 in steps of 5 @xmath88 between 10 and 120 @xmath88 , and @xmath87 for values of @xmath89 , @xmath90 , @xmath91 , @xmath92 , @xmath93 , @xmath94 , @xmath95 , @xmath96 , and @xmath97 .
we constrain the range of @xmath87 values to @xmath98 because of the kinematic limit and because around this energy , electron cooling due to triplet pair production on the highest - energy synchrotron photons , which is ignored in our simulations , becomes dominant over compton scattering ( mastichiadis et al .
@xcite , anguelov et al .
we find that our simulated spectra are only very weakly dependent on the actual value of @xmath87 .
a change of @xmath87 by a factor 3 typically results in an increase of the reduced @xmath39 of only 0.3 such that the above - mentioned restriction of @xmath87 values has only minor impact on our results .
in fact @xmath87 can be regarded as constant and of the order of @xmath93 .
we point out that the instantaneous synchrotron spectra of individual blobs at the time of injection , which might correspond to short - term x - ray and tev flares , have their synchrotron peak at @xmath99 hz if @xmath100 , in agreement with the shift of the synchrotron peak into the hard x - ray regime during extreme flaring activity ( e. g. , pian et al .
@xcite ) . all other parameters ( in particular , the magnetic field at the particle injection site , @xmath101 g , the low - energy cutoff of the electron spectrum @xmath102 , and the doppler factor , @xmath103 ) are fixed to values allowing good fits to the observed weekly averaged seds using our simulation code .
an estimate for the required parameters can be found on the basis of the location of the synchrotron and ssc peaks of the observed broadband spectra as described below .
although we are assuming the injection of a single power - law distribution of ultrarelativistic electrons into the jet , the time - averaged radiation spectrum will be reasonably well approximated by the one produced by a broken power - law distribution of electrons with spectral index @xmath80 below the break energy @xmath104 , and @xmath105 above the break energy .
@xmath104 may be computed by setting the synchrotron cooling time scale equal to the dynamical time scale of jet evolution ( magnetic field decay ) , which yields @xmath106 where @xmath107 , the height of the injection / acceleration site above the accretion disk is @xmath108 pc , and @xmath109 .
thus , our model calculations will produce a time - averaged synchrotron break at @xmath110 where @xmath111 is an appropriate average of the magnetic field ( in units of @xmath112 g ) over the jet evolution .
for the purpose of these estimates , we neglect factors of @xmath113 for mkn 501 . the location of the peak of the ssc component will be strongly influenced by klein - nishina effects and will thus depend on the actual shape of the synchrotron spectrum , which , in turn , depends on the electron spectral index @xmath80 . considering these effects , tavecchio et al .
( @xcite ) find @xmath114 where @xmath115 , @xmath116 , and @xmath117 .
\label{g_alpha}\ ] ] for @xmath118 , this yields @xmath119 , corresponding to @xmath120 combining eqs .
[ nu_sy ] and [ e_ssc ] and using the average observed @xmath121 and @xmath122 , we find @xmath123 for @xmath118 ( see also tavecchio et al .
similarly , we may use eq .
( 22 ) of tavecchio et al .
( @xcite ) to estimate @xmath124^{(1 - \alpha_1)/2 } \ , \sqrt { 2 f(\alpha_1 , \alpha_2 ) \over c^3 } \
> { \left ( \nu l_{\nu } \right)_{sy } \over t_{var } \ , \sqrt { \left ( \nu l_{\nu } \right)_{ssc}}}\ ] ] where @xmath125 .
using @xmath118 , @xmath126 erg / s , @xmath127 erg / s , and @xmath128 h ( aharonian et al .
@xcite ) , we find @xmath129 combining this with eq . [ bd_1 ] , we have @xmath130 and @xmath131 . these numbers are consistent with the limits found by bednarek & protheroe @xcite , but are slightly outside the allowed region of parameter space as found by tavecchio et al .
( @xcite ) and kataoka et al .
( @xcite ) .
this is because in those papers the broadband spectrum is either characterized by quantities pertaining to individual outbursts or to a long - term quiescent state .
those parameters are not representative of the weekly averages investigated in this paper .
having constructed the three - dimensional mesh of simulations , we compare all weekly seds with the simulated spectra and find the simulation with the smallest @xmath39 .
results of this procedure are described in the next section .
the ssc model for mkn 501 predicts a very strong correlation between the emission at the synchrotron peak ( in soft hard x - rays ) and at the inverse - compton peak ( close to tev @xmath1-ray energies ) .
we have derived an analytic estimate for the expected correlation , for variations of several input parameters . in the following discussion ,
unprimed quantities are measured in the co - moving frame , while a superscript @xmath132 refers to quantities measured in the observer s frame .
we assume that the time - averaged ( cooling ) electron spectrum can be described by a broken power - law , @xmath133 where the injection spectral index @xmath134 , and @xmath104 is the break energy of the spectrum , determined by eq .
[ gamma_b ] .
the normalization is given by @xmath135 .
we are using a @xmath136 approximation for the synchrotron spectrum : @xmath137 where @xmath138 is the dimensionless photon energy and @xmath139 is the characteristic synchrotron energy radiated by an electron of lorentz factor @xmath140 . normalizing the synchrotron luminosity to @xmath141 we have @xmath142 neglecting compton scattering events in the klein - nishina regime , @xmath143 , we may approximate the ssc spectrum by @xmath144 where @xmath145 is the heaviside function .
the evaluation of this expressions is straightforward .
observed fluxes in the asm , batse , and hegra energy ranges are calculated integrating the doppler boosted synchrotron and ssc spectra , @xmath146 , over the energy ranges @xmath147 , @xmath148 , and @xmath149 . in fig .
[ fig_ssc_var ] , we plot trajectories in the @xmath150 and @xmath151 planes of these solutions , varying individual model parameter separately while all others are held constant at values representative of states of moderate x - ray and high - energy @xmath1-ray fluxes .
a variation of the electron density obviously yields a relation @xmath152 since the synchrotron flux depends linearly , the ssc flux quadratically on @xmath86 .
note that this dependence may be altered due to an increasing @xmath0 absorption opacity intrinsic to the source , which is not included in the analytical estimate ( [ lssc ] ) used to compute the hegra flux .
a variation of the electron injection spectral index @xmath80 results in a relation which may be approximated by @xmath153 and @xmath154 , i. e. the dependence is weaker than quadratic .
a variation of the magnetic field strength leads to more complex flux variations due to the back - reaction on the break lorentz factor @xmath104 as a result of radiative cooling .
for relatively strong magnetic fields
( @xmath155 g ) , the x - ray and tev @xmath1-ray fluxes become anti - correlated .
finally , if the variability of this source were dominated by a variation of the bulk lorentz factor @xmath68 , the x - ray and high - energy @xmath1-ray fluxes would be expected to be approximately linearly correlated ( the back - reaction on @xmath104 leads to a slight deviation from a strictly linear correlation ) , as long as the observer is located within the @xmath156 beaming cone of the jet . if @xmath68 increases beyond @xmath157 , both the x - ray and tev @xmath1-ray fluxes start to decrease with increasing @xmath68 . the same quasi - linear correlation would be expected if the variability were due to a bending jet , i. e. a variation of @xmath158 .
the empirical correlation of the tev and the x - ray emission of mkn 501 in 1997 has already been studied extensively using the ct1/ct2 and the cts data from hegra and the soft x - ray data from rxte asm : aharonian et al .
( @xcite ) find the correlation coefficient for the daily averages to be @xmath159 .
this maximum correlation is found for zero time - lag .
we examine the correlation between the weekly rxte , batse and hegra points from table [ tab - data ] .
[ sy_ssc_corr ] shows the observed correlation between the hegra and batse measurements , fitted with a second - order polynomial as well as with a power - law with index 1.4 , which is the theoretical prediction if the variability is caused solely by variations of the electron injection spectral index @xmath80 .
both fits give acceptable values for the reduced @xmath39 ( 1.1 and 1.65 respectively ) . still , due to the large error bars
, we can not confidently distinguish between these and similar correlations on the basis of the currently available data . in any case , there is no indication for a super - quadratic dependence between the x - ray and tev fluxes , which would be inconsistent with a pure ssc model , unless there is a persistent quiescent level of emission above which the observed flaring behaviour is superimposed .
figure [ fig - hegra - rxte ] shows the correlation between the weekly tev and soft - x - ray points .
the linear correlation coefficient is @xmath160 , nearly the same as found by aharonian et al .
( @xcite ) who compare the same data on a daily basis .
the constant term of the linear fit is still consistent with zero .
however , the reduced @xmath39 of 4.7 is too large for a good fit .
this is also the case for a fitted power - law with index 1.6 which gives @xmath161 .
the systematic differences in time - coverage which are not taken into account in the determination of the error bars may be responsible for this .
still , the large linear correlation coefficient suggests that the correlation is nearly linear .
figure [ fig - batse - rxte ] shows the correlation between the weekly hard - x - ray and soft - x - ray points . in this case
, the time coverage is the same for both instruments , only the duty cycles are different .
the correlation coefficient is 0.53 corresponding to a 0.5 % chance probability for a linear correlation .
the constant term of the linear fit is very well consistent with zero .
this figure also illustrates the difference in the dynamical ranges of the variability in the soft and the hard x - ray band . at batse energies , which are believed to be near the high - energy end of the synchrotron spectrum
, the variability amplitude is about 50 % larger than at rxte energies .
this fits nicely into the scheme that the strongest variability takes place at the high - energy ends of both spectral components .
each weekly sed is compared to a three - dimensional mesh of @xmath162 simulations ( 48 different values of @xmath80 , 22 values of @xmath86 , and 9 values of @xmath87 ) , selecting the simulated broadband spectrum with the smallest @xmath39 .
the resulting best - fit parameters are listed in table [ fit_parameters ] , along with the resulting @xmath39 divided by the number of data points .
since we do not have a continuous sequence in parameter space , the actual number of degrees of freedom is questionable so that we use the above quantity to assess the quality of the fit .
our best - fit parameters for each individual weekly averaged broadband spectrum are listed in tab .
[ fit_parameters ] . with few exceptions ( mjd 50549.0 , 50576.9 , 50618.8 , 50696.1 ) , all fits resulted in acceptable @xmath39 values .
we find a correlation between the electron injection spectral index @xmath80 and the x - ray and high - energy @xmath1-ray fluxes , while there is no obvious correlation with the total electron density and/or the high - energy cut - off @xmath87 of the injected electron spectrum .
there also appears to be a weak anti - correlation between the jet filling factor @xmath163 and the x - ray and hegra fluxes .
this could indicate that during states of relatively low activity , the fluxes are dominated by a quasi - steady component from a continuous jet , while in high - activity state the emission is dominated by more isolated , eruptive events .
however , this latter correlation is much less pronounced than the correlation with the electron spectral index and will need to be tested on the basis of future , more sensitive observations . in fig .
[ par_corr ] the temporal variation of the best - fit values of @xmath80 and @xmath164 are compared to the variations of the soft and hard x - ray fluxes and the 1.5 tev flux .
the correlation between the hard x - ray and tev @xmath1-ray fluxes with the injection spectral index is illustrated in figure [ fig - p ] .
furthermore , we show in figure [ fig - peakpos ] the correlation between @xmath80 and the positions of the peaks in the synchrotron and the inverse compton component of the sed .
the peak positions are not fit parameters but were determined by finding the local maxima in the weekly seds .
we find that , in our model , there is a strong correlation between the peak positions and @xmath80 such that these parameters can be regarded as nearly identical .
however , while the peak positions are directly observable , the electron spectral index is the more fundamental quantity .
our results indicate that medium - timescale high activity states in x - rays and high - energy @xmath1-rays are consistent with a hardening of the electron spectrum injected at the base of the jet .
as pointed out in the previous subsection , a pure ssc model in which the tev @xmath1-ray flux is strongly influenced by klein - nishina effects , predicts that the hegra and both the soft and hard x - ray fluxes should roughly be correlated by power - laws @xmath165 with @xmath166 in high flux - level states , in which the contribution of a possible quasi - stationary radiation component is small .
the data available for this study are consistent with this but do not allow a clear distinction between different variability mechanisms , and future observations with increased sensitivity , in particular at multi - gev to tev energies , are needed in order to test this prediction .
we have presented broadband spectra of the extreme hbl mkn 501 during its high state in 1997 , including radio , optical , soft and hard x - ray , and tev @xmath1-ray observations . in this study we concentrated on the medium - timescale variability , using
weekly averaged seds .
we confirmed the strong correlation between the tev @xmath1-ray flux and the hard x - ray flux .
this correlation was found to be non - linear and could be fitted with a second - order polynomial , in agreement with the expectation of an ssc dominated leptonic jet model , if the flux variations are related to fluctuations of the electron density in the jet and/or the spectral index of the electron spectrum at the time of injection into the jet .
the weekly averaged seds were fitted with a leptonic jet model , strongly dominated by the ssc process . with a few exceptions , this model yielded acceptable fits to the observed broadband spectra .
the observed spectral variability of mkn 501 could be explained mainly by variations of the electron spectral index .
no clear correlation between the maximum electron energy and the hard x - ray and tev @xmath1-ray fluxes on the 1-week timescale was found , in contrast to the short - term variability of mkn 501 .
pian et al .
( @xcite ) have shown that the intraday variability of this object is most probably related to an increase of @xmath87 , leading to pronounced flares in hard x - rays , most probably on the synchrotron cooling timescale which is most likely @xmath167 a few hours and thus much shorter than the 1-week timescale considered in this paper .
our result indicates that such synchrotron flares are isolated events and are at most weakly correlated to the activity of the source on the 1-week timescale .
our result that the flaring behaviour on intermediate timescales is consistent with a hardening of the electron spectrum is in contrast to the flaring characteristics observed in quasars .
recently , mukherjee et al .
( @xcite ) have investigated all available broadband data on the very luminous flat - spectrum radio quasar ( fsrq ) pks 0528 + 134 , and found that its flaring behaviour is consistent with an increasing contribution of the external inverse - compton component during flares , possibly related to an increase in the bulk lorentz factor .
the fits to the seds of pks 0528 + 134 required that the average energy of relativistic electrons in the jet shifts towards lower values during flares , in contrast to the results found for mkn 501 .
as pointed out by bttcher ( @xcite ) , this implies that the synchrotron peak is expected to shift towards lower frequencies during flares of fsrqs , while mkn 501 and mkn 421 show clear evidence for a shift of the synchrotron peak to higher frequencies .
we point out that in the present study the magnetic field along the jet and the bulk lorentz factor of individual blobs were fixed , so that we can not confidently rule out variations of the doppler factor , accompanied by appropriate changes of the electron injection spectrum , as the flaring mechanism for mkn 501 .
however , the very moderate variability at optical frequencies , as observed in mkn 501 , leads us to consider this flaring mechanism less likely in this object since it would require a peculiar conspiracy between the doppler factor and the electron spectrum to keep the optical flux at an approximately constant level .
the work of dp is supported by the spanish cicyt grant sb97-b12601316 .
the work of mb was supported by nasa grant nag 5 - 4055 ( until aug . 1999 ) and by chandra postdoctoral fellowship grant number pf 9 - 10007 , awarded by the chandra x - ray center , which is operated by the smithsonian astrophysical observatory for nasa under contract nas 8 - 39073 .
the rxte asm data has been obtained through the high energy astrophysics science archive research center online service provided by the nasa / goddard space flight center .
we thank c.d .
dermer for valuable comments on the manuscript and j.m .
holeczek for providing a fit routine .
.[tab - data ] the weekly @xmath40 datapoints from observations of mkn 501 used for the model fits presented in this paper .
all entries are in units of @xmath189 erg @xmath5 s@xmath6 .
entries exactly equal to 0 are those where no data are available .
this is mainly radio data .
upper limits are indicated by a `` @xmath190 '' symbol and are calculated with 90% confidence level .
the fluxes at 0.8 tev and 10 tev are extrapolated from the flux measurement at 1.5 tev . [ cols="<,<,<,<,>,>,>,>,>",options="header " , ] rcl + @xmath87 & = & electron spectrum high - energy cutoff + @xmath80 & = & electron spectral index ,
@xmath191 + @xmath86 [ @xmath88 ] & = & electron density + @xmath193 [ s ] & = & blob ejection events repetition time scale ( normalization ) + + @xmath194 & = & 0.03 pc ( injection height of blob ) + @xmath195 & = & @xmath196 ( mass of central black hole ) + @xmath197 & = & @xmath198 ergs@xmath6 ( isotropic accretion disk luminosity ) + @xmath199 & = & @xmath200 cm ( blob radius in the comoving frame ) + @xmath201 & = & 200 ( low - energy cutoff of electron sp . ) + @xmath68 & = & 25 ( bulk lorentz factor ) + @xmath75 & = & 30 ( doppler factor ) + @xmath202 & = & 3336 s ( contracted blob crossing time ) + @xmath203 & = & 0.05 g ( magnetic field ) + @xmath204 & = & 75 kms@xmath6mpc@xmath6 ( hubble constant ) + @xmath205 & = & 0.5 ( deceleration parameter ) + @xmath206 & = & 0.034 ( cosmological redshift ) + @xmath207 & = & 133 mpc ( luminosity distance ) + | during the observation period 1997 , the nearby blazar mkn 501 showed extremely strong emission and high variability .
we examine multiwavelength aspects of this event using radio , optical , soft and hard x - ray and tev data .
we concentrate on the medium - timescale variability of the broadband spectra , averaged over weekly intervals .
we confirm the previously found correlation between soft and hard x - ray emission and the emission at tev energies , while the source shows only minor variability at radio and optical wavelengths .
the non - linear correlation between hard x - ray and tev fluxes is consistent with a simple analytic estimate based on an ssc model in which klein - nishina effects are important for the highest - energy electrons in the jet , and flux variations are caused by variations of the electron density and/or the spectral index of the electron injection spectrum .
the time - averaged spectra are fitted with a synchrotron self - compton ( ssc ) dominated leptonic jet model , using the full klein - nishina cross section and following the self - consistent evolution of relativistic particles along the jet , accounting for @xmath0 absorption and pair production within the source as well as due to the intergalactic infrared background radiation . the contribution from external inverse - compton scattering
is tightly constrained by the low maximum egret flux and found to be negligible at tev energies .
we find that high levels of the x - ray and tev fluxes can be explained by a hardening of the energy spectra of electrons injected at the base of the jet , in remarkable contrast to the trend found for @xmath1-ray flares of the flat - spectrum radio quasar pks 0528 + 134 .
_ submitted to the astrophysical journal _ |
the physics program at a high energy linear @xmath1 collider , such as the ilc @xcite or clic @xcite , relies significantly on the jet flavour tagging capability of its detectors . distinguishing @xmath2 , @xmath3 and light quark jets as well as @xmath4 leptons is of prime importance both for extracting signals of new particles , which may couple preferentially to heavy fermions , and for analysing new sectors of particle physics , such as the higgs mechanism .
the identification of heavy quarks and leptons largely relies on their long lifetimes , which result in secondary and tertiary vertices separated from the beam interaction point by distances which range from few hundred microns to several millimetres , or even centimetres at multi - tev collision energies . in order to efficiently identify these secondary vertices ,
it is essential to achieve excellent resolution on the extrapolation of charged particle tracks to their production point , @xmath5 .
the requirement for a 0.5 tev - 1.0 tev linear collider has been identified to be @xmath6 ( @xmath0 m ) , where @xmath7 is the momentum component in the transverse plane , measured in gev .
these asymptotic and multiple scattering terms are respectively 3 and 8 times better than the design performance of the lhc detectors and 1.5 and 3 times better than the best extrapolation resolution ever achieved by a collider experiment detector , that of the vxd3 vertex detector in sld at the stanford linear collider @xcite .
achieving these performances requires the development of novel sensor technologies .
monolithic silicon pixellated sensors , back - thinned to 20 - 50 @xmath0 m have emerged as possibly the most appealing solution and a significant r&d effort is addressing various technologies and architectures in the framework of the world - wide linear collider studies .
small pixel size , fast readout and limited power dissipation , to enable airflow cooling and remove the material burden of active cooling , are some of the main issues being addressed .
this r&d is guided by the physics requirements and detailed simulation studies of tracking performances .
it is essential to validate these simulations with data collected from prototype trackers made of thin pixel sensors with a geometry resembling that envisaged for the ilc , under realistic operating conditions and hit occupancy levels . in this paper
we report the first results of the t966 beam test experiment at fnal , which aims at a detailed study of tracking and vertexing capability of a prototype tracker based on thin cmos monolithic si pixel sensors .
the t966 beam test experiment took data at the meson test beam facility ( mtbf ) at fnal in the summer 2007 and deployed a beam telescope made of four layers of thinned cmos pixel sensors . a prototype thin pixel telescope ( tppt-1 )
had been successfully operated on the 1.5 gev bts electron beam - line at the lbnl advanced light source @xcite .
the t966 beam telescope has better mechanical positioning of the detector layers compared to the tppt-1 condiguration and also includes remotely controlled stages for positioning a detector under test 20 mm downstream from the telescope and inserting a copper target @xmath8 30 mm upstream .
the telescope consists of four detector layers , spaced 15 mm apart .
this provides a geometry quite similar to that envisaged for the ilc vertex tracker , which has five to six sensitive layers spaced by @xmath8 11 mm , located at radii from 15 mm to 60 mm @xcite . in particular , by extrapolating tracks reconstructed from hits on layers 2 , 3 and 4 onto layer 1 , we can study the extrapolation resolution for particle tracks over the same lever arm foreseen at the ilc , where the first layer of the vertex tracker would be positioned @xmath8 15 mm away from the beam interaction point .
the telescope uses the mimosa-5 chip , developed by iphc , strasbourg ( france ) @xcite .
this chip , fabricated in the ams 0.6 @xmath0 m process , features a large active area of [email protected] @xmath10 and more than 1 million pixels , arranged in four independent sectors .
the epitaxial layer is 14 @xmath0 m thick and the pixel size 17@xmath917 @xmath0m@xmath11 .
detectors have been thinned by aptek industries @xcite to ( 50@xmath127 ) @xmath0 m using a proprietary grinding process .
each detector is glued to a pc board using a precision mounting jig which ensures a position accuracy of the detector corners w.r.t .
the boards of @xmath8 20 @xmath0 m .
the board has a cut - out area corresponding to the sensor sensitive region to reduce multiple scattering .
we verify the flatness of the thin chips after gluing them on the pc board using an optical survey system which has a resolution of @xmath8 1 @xmath0 m . we measure the deviation between the height of 100 points on the chip surface and that of a plane interpolated through them .
this has a gaussian width of 7 @xmath0 m .
the pc boards are mounted on the readout board and held in place by precision mechanics which ensures their relative alignment .
we verify the stability by measuring fiducial reference marks on the four pc boards before and after mechanically disturbing the setup . from these measurements we estimate that the sensitive area of each layer is stable within @xmath13 20 @xmath0 m in the @xmath14 coordinate , along the beam axis , and @xmath13 10 @xmath0 m in the @xmath15 and @xmath16 coordinates , transverse to it . the telescope and its readout electronics are installed in an optical enclosure .
the system is aligned on the beam - line using a laser beam and is operated at a constant temperature of + 20@xmath17c , by flowing cold air inside the optical enclosure .
the temperature is monitored by a thermocouple installed between layer 2 and 3 of the telescope .
one sector of each chip , corresponding to a 510@xmath9512 pixel array , is read out through a custom fpga - driven acquisition board .
the trigger is provided by two finger scintillators , s1 and s2 , mounted just in front of the optical enclosure , defining a fiducial area of @xmath81 @xmath10 around the sector of the mimosa chip that is readout .
the s1-s2 coincidence is gated on the 4 s - long spill extraction signal .
the readout cycle consists of a reset followed by the readout of two subsequent frames of the 510@xmath9512 pixel sector .
the chip is read out at 3.125 mhz , which corresponds to 84 ms for reading a full sector .
four 14 bits , 40 msample / s adcs simultaneously read the four sensors , while an array of digital buffers drives all the required clocks and synchronisation signals . the fpga has been programmed to generate the clock pattern and collect the sampled data from the adcs .
a 32 bit wide bus connects the fpga to a digital acquisition board installed on a control pc . due to the slow readout and data transfer ,
the telescope acquires 7 to 10 events / spill .
data are processed on - line by a labview - based program , which performs correlated double sampling and pedestal subtraction , noise computation and cluster identification .
to reduce the amount of data written to disk , only the addresses and pulse heights of the pixels in a fixed matrix around the centre of a cluster candidate having a seed pixel with signal - to - noise ratio above 4 are recorded .
data are converted offline in the lcio format @xcite for the subsequent analysis .
the average pixel noise has been measured to be ( 85 @xmath12 16 ) @xmath18 of equivalent noise charge ( enc ) , which is due both to the noise of the readout electronics and to leakage current from operating the detector at + 20@xmath17c .
a test performed in the lab shows that the noise drops to @xmath8 50 @xmath18 of enc , when operating the detector below + 5@xmath17c . during the 2007 data taking ,
the 120 gev proton beam was extracted from the main injector and attenuated through a pinhole collimator .
the average beam size was measured to be about 4 mm and 9 mm in the horizontal and vertical coordinate , respectively , using two sets of multi - wire chambers located upstream from the telescope .
the average beam divergence , measured from the slope of tracks reconstructed in the telescope , is 0.52 mrad .
the beam intensity was varied during data taking and resulted in an average hit density ranging from 0.07 hits mm@xmath19 to 0.70 hits mm@xmath19 , with local densities of up to 5 hits mm@xmath19 .
for comparison , in the core of hadronic jets from @xmath1 collisions , the average hit density is expected to be in the range 0.2 - 1.0 hits mm@xmath19 for @xmath20=0.5 tev and 0.5 - 2.5 hits mm@xmath19 for 3.0 tev @xcite .
the machine - induced pair background contributes @xmath80.05 hits mm@xmath19 bunch crossing@xmath21 at 15 mm radius at 0.5 tev @xcite and @xmath80.005 hits mm@xmath19 bunch crossing@xmath21 at 30 mm radius at 3.0 tev @xcite for solenoidal fields of 4 t and 5 t , respectively .
local densities up to 5 hits mm@xmath19 are quite comparable to those expected in a linear collider vertex tracker . at the same time our redundancy is lower than that provided by a five - layered vertex tracker and a main tracker , which makes the overall occupancy conditions quite realistic .
the offline analysis is performed using a set of dedicated processors developed for the marlin framework @xcite .
individual processors are used for clustering and hit reconstruction , pattern recognition and track fit and for vertexing .
an additional processor provides signal simulation of the pixels .
the t966 setup is modelled by a simulation program based on the geant-4 package @xcite , which generates particle points of impact and energy deposits in the sensitive detector planes and accounts for the material surrounding the beam - line and the detectors .
simulated data are stored in lcio format and used as input to the cmos pixel simulation processor , pixelsim , implemented in marlin @xcite .
hits are reconstructed from the recorded pixel pulse heights as follows : each detector is scanned for pixels with pulse height over a given signal - to - noise ( s / n ) threshold .
these are designated as cluster ` seeds ' .
seeds are then sorted according to their pulse height values and the surrounding , neighbouring pixels are tested for addition to the cluster .
the s / n threshold is 5 for the seed pixel and 2.5 for the neighbour pixels .
the neighbour search is performed in a 5@xmath95 matrix .
clusters are not allowed to overlap , i.e. pixels already associated to one cluster are not considered for populating another cluster around a different seed .
finally , we require that clusters are not discontinuous , i.e. pixels associated to a cluster can not be interleaved by any pixel below the neighbour threshold .
pixels with noise exceeding two times the average sector noise or giving a pulse height above the seed threshold in more than 10 % of the events of a given run are flagged as noisy and removed offline for the whole run .
columns of pixels with more than 25 pixels having pulse height above the seed threshold in one event are also flagged as noisy and removed for that single event .
selected clusters have an average pixel multiplicity of 2.35 and average signal - to - noise ratio of 10.3 .
these results are in agreement with those predicted by simulation , as shown in figure [ fig : clusters ] .
simulation predicts single point resolutions , @xmath22 , of ( [email protected] ) @xmath0 m , ( [email protected] ) @xmath0 m and ( [email protected] ) @xmath0 m for cluster s / n values of 10 , 15 and 20 .
this last value agrees with the resolution of 1.7 @xmath0 m , reported in @xcite , obtained by operating the detector cold .
the track reconstruction efficiency is estimated using simulation where the noise of each individual layer is tuned on the data .
particle densities comparable to those registered in the data are simulated and pixels are masked to reproduce the effect of noisy pixels and columns .
simulation reproduces well the number of reconstructed particles and the number of associated hits . due to the limited pixel s
/ n performance achieved in our 2007 data taking , the estimated track reconstruction efficiency , after applying the full set of selection criteria as in real data , is 0.69 .
the relative position of the four layers of the telescope is obtained by using the result of an optical survey and a track - based alignment procedure to align the telescope planes with respect to a common reference system .
the origin of this system coincides with the centre of the detector chip in the first layer and the @xmath15 and @xmath16 directions are aligned along the pixel rows and columns in their readout direction . for each telescope layer
we consider six degrees of freedom : the three offsets along the @xmath15 , @xmath16 and @xmath14 axes and the three rotations around them .
these parameters are determined from a @xmath23 fit which minimises the distance between the hit positions measured on each plane and those predicted by the extrapolation from the other planes .
the optical survey provides with the starting positions .
the track - based alignment procedure aligns the detector planes in pairs .
we start with aligning the second telescope plane with respect to the first .
the third plane is then aligned on the extrapolated positions determined from the track segments fitted with the first two planes .
finally , the fourth sensor is aligned with respect to the tracks obtained from a linear fit between the first three telescope planes .
after this initial procedure , any given plane can be re - aligned on the extrapolations of the tracks fitted with the other three planes .
quality cuts are applied on the track slopes and fit @xmath23 . due to a @xmath8 5 - 9@xmath17c difference in temperature between the day , when the telescope is operated , and the night ,
when its power is switched off , the procedure is repeated for each day of data taking , in order to account for possible distortions due to thermal effects .
changes in the relative position of the layers of up to 6 - 7 @xmath0 m are observed , as shown in figure [ fig : align ] .
this result is consistent with the measured stability of the mechanical mounting .
a sub - micron accuracy on the relative position of any pairs of layers is obtained with samples of @xmath13 250 fitted tracks .
these results are relevant for estimating the amount of data needed for aligning a single ladder in a full vertex tracker at a collider .
alignment parameters are stored in a conditions database using the linear collider conditions data ( lccd ) toolkit @xcite , a c++ based lcio framework .
lccd offers the advantage of full database functionality while being straightforward and easy to implement .
conditions data consisting of a set of alignment parameters for each day of data taking are stored in an lcio file and are accessed by a marlin module at run time .
particle tracks are fitted using the reconstructed hits with a simple straight track model .
the pattern recognition starts from hits on a given detector layer and looks for sets of aligned hits .
hit pairs are used as track seeds provided the slope does not exceed 0.0015 .
additional hits are tested if their residual w.r.t the track extrapolation does not exceed 35 @xmath0 m .
hit bundles corresponding to the @xmath23 minimum for a given hit on the starting layer are used to define a track candidate provided that @xmath24 25 and the slope does not exceed 0.0010 .
the slope cut removes lower momentum particles originating from interactions and scattering upstream in the beamline as well as badly reconstructed particle tracks resulting from pattern recognition failures .
the procedure is repeated using all planes .
ambiguities in hit association are then resolved based on the number of hits in a track candidate and its fit @xmath23 .
tracks are fitted and associated hits are flagged .
the remaining unassociated hits are used for a second pass pattern recognition , where three- and two - hit tracks are reconstructed from the hit combinations giving the minimum @xmath23 provided they do not exceed 25 or the minimum slope , respectively . the procedure is tested on simulated events with the same track density as the data .
the fraction of correctly associated hits is [email protected] for tracks with at least three hits .
residuals are computed from the distance between the track extrapolation on a test detector plane and the closest hit , which is not included in the track fit . only tracks having an associated hit on the layer closest to the extrapolation plane are considered .
first we determine the detector single point resolution by including the hits on all four telescope layers in the track fit . using the fact that the layers are equally spaced
, we can extract the single point resolution @xmath22 from the width of the measured residual distribution , corrected by a geometrical factor , under the assumption that @xmath22 is the same for all layers .
we measure a gaussian width of ( [email protected] ) @xmath0 m and ( [email protected] ) @xmath0 m using hits from clusters with an average s / n ratio of 10 and 15 , respectively . using the correction factor of 1.81 for our geometry
, we obtain an estimate of @xmath22 = ( [email protected] ) @xmath0 m and ( [email protected] ) @xmath0 m for cluster hits with average signal - to - noise ratio of 10 and 15 , respectively .
this result is in good agreement with the prediction of our cmos sensor simulation , as discussed above .
then we extrapolate the tracks on the second layer and study the distribution of the residuals between the position of extrapolated and that of the closest hit , which is not included in the track fit .
this distribution has a gaussian width of ( [email protected] ) @xmath0 m for data and ( [email protected] ) @xmath0 m for simulation , for an average cluster s / n ratio of 10.3 . finally , we study the residuals obtained by extrapolating the track 15 mm upstream onto the first telescope layer , which resembles the extrapolation of a particle track back to its production point .
this has a gaussian width of ( [email protected] ) @xmath0 m while simulation predicts ( [email protected] ) @xmath0 m as shown in figure [ fig : residuals ] . in order to measure the dependence of these residuals on the s / n ratio of the hit clusters , we vary the threshold on the seed pixel s / n ratio of all associated hits and
compute the r.m.s .
width of the residual distribution on the first layer as a function of the average s / n ratio of the hits on a track .
results are shown in figure [ fig : residuals_sn ] for data and simulation .
we also evaluate the effect of hit density on the non - gaussian tails of the residual distribution .
the fraction @xmath25 of tracks with a residual on the first layer , which is more than 2.5 @xmath26 away from zero , is studied as a function of the minimum distance , @xmath27 , between the hits associated to the track and the closest , not associated hit on any of the layers 2 , 3 and 4 .
the most probable value of this distance in data is 0.19 mm . as a comparison , in the core of a @xmath2 jet from @xmath28 , @xmath29 at 0.5 tev , disregarding the machine - induced pair background , the most probable value of @xmath30 is 0.40 mm .
the fraction @xmath25 is 0.040 , for data where the distance @xmath27 is larger than 100 @xmath0 m , and increases to 0.050 and 0.085 , when @xmath27 drops below 100 @xmath0 m and 75 @xmath0 m , respectively .
this trend is consistent with that predicted by simulation and gives a measure of the rate of outliers in the pattern recognition in presence of large hit occupancy .
the extrapolation resolution on the first layer can be extracted from the measured width of the residual distribution by subtracting in quadrature the detector resolution , taken from simulation .
this gives an extrapolation resolution of @xmath5=([email protected] ) @xmath0 m and ( [email protected] ) @xmath0 m for cluster s / n values of 10 and 15 , respectively . in an earlier paper ,
we reported the measurement of the extrapolation resolution for 1.5 gev electrons , where we obtained @xmath5=([email protected] ) @xmath0 m with the tppt-1 telescope @xcite .
these results are now compared in figure [ fig : extrap ] to the required track extrapolation resolution for @xmath1 collisions at 0.5 tev at the ilc .
these data reaches the target performance required for the ilc , using a standalone si pixel tracker of similar geometry to that proposed for an ilc detector , in a high density track environment and rather realistic operating conditions with airflow cooling .
a set of data has been taken with a 4 mm thick cu target inserted on the beam - line , @xmath8 30 mm upstream of the first beam telescope plane .
these data are used to collect inclined tracks , from protons scattered in the target , for alignment purposes as well as for a first study of vertex reconstruction performance with a thin pixel tracker telescope in 120 gev @xmath31-cu interactions . for this study
we relax the requirement on the minimum s / n ratio of the seed pixel in the cluster to 4 and we raise the maximum track slope to 0.0085 .
vertices are fitted using a port to the marlin framework of the vt @xcite kalman filter vertex fitter , developed for hera - b and lhcb , according to the following procedure .
first a seed vertex is searched for .
this is defined as a vertex made of a pair of tracks with at least three associated hits , including one on the first layer , and having a slope in excess of 0.00075 .
these cuts suppress combinatorics due to either fake tracks or primary protons not interacting with the target and have been optimised using fully simulated and reconstructed p - cu interaction events .
seeds with a fit @xmath23 below 20 are kept .
additional tracks are tested for association with this seed vertex .
tracks are retained if their @xmath23 contribution does not exceed 10 and the vertex position along @xmath14 does not change by more than five times the estimated uncertainty .
finally , unassociated hits are tested for their compatibility with this vertex .
a track search is performed using the seed vertex as a track point .
tracks reconstructed with at least two associated hits , @xmath24 15 and slope below 0.0100 are accepted .
the vertex is then refitted using all associated tracks .
vertices with @xmath32 2.5 and at least one track with slope in excess of 0.010 are kept .
the distribution of the reconstructed vertex positions along the beam axis , @xmath14 , shows a clear accumulation at the target location .
an event with a reconstructed six - prong vertex in the target is shown in figure [ fig : evtdisplay ] .
the analysis has been applied to both simulated events , where a comparable number of protons per event collide against a copper target as in the actual data , and to data taken without the target in front of the telescope .
these data are used as a control sample to estimate the background from combinatorics .
the data control sample has a flat distribution of the @xmath14 position of the reconstructed vertices , consistent with the sidebands of the data target sample ( see figure [ fig : vertex ] ) .
c + + + the number of tracks associated to these vertices in data agrees with that predicted by simulation for real vertices ( see figure [ fig : vertex ] ) .
the average multiplicity of tracks at the vertex is [email protected] in background - subtracted data and [email protected] in signal simulation .
finally , we study the distribution of the resolution on the vertex position along the particle line of flight .
the estimation of the resolution is validated using simulated events .
the @xmath14 pull distribution , i.e. the difference between the simulated and reconstructed @xmath14 positions normalised to their estimated uncertainty , has a gaussian width of [email protected] .
the average resolution in data is ( 260@xmath1210 ) @xmath0 m for vertices @xmath832 mm away from the first sensitive layer .
this can be compared to the resolution obtained for @xmath33 decay vertices in simulated @xmath2 jets at the ilc , which are @xmath815 mm away from the first sensitive layer .
simulated vertices from @xmath33 mesons with energies in the range 100 gev-150 gev , re - weighted by the associated track multiplicity to match that of the reconstructed vertices in our data , have a resolution of ( 170@xmath1220 ) @xmath0 m .
the analysis of the data collected in the first run of the t966 beam test experiment at fnal mtbf with a beam telescope based on thin monolithic cmos pixel sensors has provided information on tracking and vertexing capabilities in rather realistic conditions . the resolution for extrapolating tracks 15 mm upstream of the first telescope plane
is measured to be 4.2 @xmath0 m to 4.9 @xmath0 m , depending on the s / n of the hit clusters used for reconstructing the tracks .
vertices from 120 gev p - cu interactions have been reconstructed with a resolution of ( 260@xmath1210 ) @xmath0 m along the flight direction .
this work was supported by the director , office of science , of the u.s .
department of energy under contract no.de-ac02-05ch11231 .
we are indebted to e. ramberg and the main injector staff for their assistance and the excellent performance of the machine and to fnal for hospitality .
we gratefully acknowledge the contribution of i. childres , k. deck and m. marins de castro souza to the data taking and data quality checking .
f. gaede , t. behnke , n. graf and t. johnson , in the _ proc . of 2003 conf . for computing in high - energy and nuclear physics _
( chep 03 ) , la jolla , california , 24 - 28 mar 2003 , pp tukt001 , [ arxiv : physics/0306114 ] . m. battaglia , nucl . instrum
. meth .
a * 530 * 33 ( 2004 ) .
k. buesser , _ in the proceedings of 2005 international linear collider workshop ( lcws 2005 ) , stanford , california , 18 - 22 mar 2005 , pp 1116_. f. gaede , nucl .
instrum .
a * 559 * ( 2006 ) 177 .
s. agostinelli _ et al .
_ , nucl .
instrum .
a * 506 * ( 2003 ) 250 .
m. battaglia , nucl .
instrum .
a * 572 * ( 2007 ) 274 .
t. lohse , report desy 95 - 103 . | we present results of a study of charged particle track and vertex reconstruction with a beam telescope made of four layers of 50 @xmath0m - thin cmos monolithic pixel sensors using the 120 gev protons at the fnal meson test beam facility .
we compare our results to the performance requirements of a future @xmath1 linear collider in terms of particle track extrapolation and vertex reconstruction accuracies . , , , , , , , , , monolithic pixel sensor ; particle track and vertex reconstruction |
over the last 15 years , it has become possible to observe molecular emission in nearby galaxies with sufficient resolution and sensitivity to distinguish individual giant molecular clouds ( gmcs ) .
the immediate goal of such studies is to determine whether ( and how ) the gmcs in other galaxies differ from those seen in the solar neighborhood .
the most common method used to address this question has been to use molecular - line tracers of h@xmath1 , in particular @xmath2co(@xmath3 ) , to compare the macroscopic properties ( size , line width , and luminosity ) of gmcs in other galaxies to those of milky way gmcs .
unfortunately , a wide variety of methods have been used to reduce data from spectral line data cubes into macroscopic gmc properties . as a result , many of the differences between gmc populations found in the literature can be attributed , at least partially , to observational artifacts or methodological differences .
it is therefore difficult to assess what the real differences between gmc populations are based on the reported data in the literature . for gmcs
that are either marginally resolved or marginally detected , observational biases can be severe .
figure [ resbias ] shows the variation of the measured spatial size and line width with the resolution for a model cloud .
typical galactic gmcs have sizes of a few 10s of parsecs , comparable to the spatial resolution of many data sets used to study extragalactic gmcs ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
figure [ resbias ] shows that when the size of the beam is comparable to the size of the object , the measured size is much higher than the true size of the object .
millimeter spectrometers and correlators often have excellent frequency resolution , so the spectral resolution bias is usually less important for gmc studies , but it can become substantial when data are binned to increase signal - to - noise .
figure [ extrapmoms ] shows that the measured spatial size , line width , and flux of a real gmc in m 33 ( eprb1 from * ? ? ?
* ) are all strong functions of the sensitivity of the data .
we discuss another major source of bias , the method by which emission is decomposed into gmcs , in the appendix .
to place these biases in the the context of real molecular cloud studies , figure [ distcomp ] shows the peak flux and angular size of typical gmcs ( those of * ? ? ?
* ) as a function of distance .
the sensitivities and resolutions of a representative sample of molecular cloud studies have been indicated as horizontal lines in these plots .
distances to commonly observed objects have also been labeled .
figure [ distcomp ] demonstrates that most studies of extragalactic gmcs are conducted where the clouds of interest are only marginally resolved and are found at low sensitivity .
even future observations of gmcs in the virgo cluster using the atacama large millimeter array ( alma ) will be affected by resolution and sensitivity biases . in this paper
, we examine the effects of biases stemming from finite spatial resolution , spectral resolution , and sensitivity in molecular - line observations of gmcs .
we recommend data analysis methods to produce a standardized set of observed cloud properties that account for these biases .
most of the methods used in this paper have been adopted piecemeal and _ ad hoc _ in previous studies . here
, we endeavor to justify our choice of methodologies and to synthesize various author s techniques for approaching the problems of molecular cloud data analysis . in section [ moments ] , we describe a standardized method to measure three basic properties of an emission distribution size , line width , and flux while accounting for the sensitivity and resolution of a data set . in section
[ physical ] , we discuss how these measurements can be transformed into physical quantities radius , line width , luminosity , and implied mass . finally , we consider the effects of using interferometers to derive cloud properties in section [ interf ] .
the results in these three sections are applicable to all observations of molecular clouds .
in contrast , the methods used for decomposing emission into molecular clouds vary widely , and there is little basis for favoring one method over another in all cases . hence , we defer a presentation of our decomposition algorithm to the appendix of the paper , leaving only a brief discussion of the decomposition problem in section [ decompsect ] .
we conclude the paper by exhibiting several examples of the application of our standardized methods to previously observed data and making recommendations for future observations ( section [ applications ] ) .
the methods described in this paper require a computer program to apply .
a documented software version of the decomposition and measurement algorithms is available from the authors as an idl package .
this section describes how to derive the spatial size , line width , and flux from a region of emission within a spectral line map ( a `` data cube '' ) while accounting for the finite sensitivity and resolution of the data set .
we use moment methods ( e.g. * ? ? ?
* ) , which make full use of position and intensity information without assuming a functional form for the cloud .
they are therefore robust to pathologies in the data _ within _ the cloud .
moments are , however , sensitive to the inclusion of false emission ( noise ) at the edge of a cloud . including noise has the effect of artificially increasing the values of the moment .
therefore the methods outlined here should be employed in conjunction with careful signal identification so that the calculations include as little noise as possible .
we assume throughout this section that the algorithm is being applied to a distribution of real emission that we label a `` cloud '' ( we discuss signal identification and decomposition in the appendix ) .
this subsection describes how to apply moment methods to derive the size , line width , and flux from a distribution of emission ( a `` cloud '' ) within a position - position - velocity data cube .
the data cube consists of a number of pixels that have sizes of @xmath4 , @xmath5 , and @xmath6 in the two spatial dimensions and the velocity dimension , respectively .
the @xmath7th pixel in the data cube has positions @xmath8 and @xmath9 , velocity @xmath10 , and brightness temperature @xmath11 .
we assume that the cloud is contiguous and bordered by an isosurface in brightness temperature of value @xmath12 , so that all of the pixels in the cloud have @xmath13 and the pixels outside the cloud have @xmath14 or are separated from the cloud by emission with @xmath14 .
we begin by rotating the spatial axes so that the @xmath15 and @xmath16 axes align with the major and minor axis of the cloud , respectively .
we determine the orientation of the major axis using principal component analysis .
we find the eigenvectors of the intensity - weighted covariance matrix for the cloud , @xmath17 \mbox { .}\ ] ] in the equations above the sum @xmath18 runs over all pixels within the cloud and @xmath19 and @xmath20 are the intensity weighted mean positions within the cloud ( defined below ) .
we define the new @xmath15 axis to lie along the eigenvector with the largest eigenvalue the major axis of the cloud .
the @xmath16 axis lies perpendicular to the @xmath15 axis along the minor axis of the cloud . in the discussion below
, @xmath15 refers to position along the major axis and @xmath16 refers to position along the minor axis .
rotating the axes in this manner yields information about the axis ratio of the cloud and allows a more careful deconvolution .
this method for determining the position angle of molecular clouds has also been adopted by @xcite . to measure of the size of the cloud , we compute the geometric mean of the second spatial moments along the major and minor axis .
this is @xmath21 , the root - mean - squared ( rms ) spatial size : @xmath22 where @xmath23 and @xmath24 are the rms sizes ( second moments ) of the intensity distribution along the two spatial dimensions .
we adopt this particular functional form since it has been used in previous observational studies @xcite and explored in depth by @xcite with respect to inclination , aspect ratio , and virialization .
we calculate @xmath23 and @xmath25 by : @xmath26 ^ 2}{\sum_{i}^{cloud } t_i } } , \mbox { where } \\
\bar{x } ( t_{edge } ) & = & \frac{\sum_{i}^{cloud } t_i x_i}{\sum_{i}^{cloud } t_i } \mbox{~and } \\
\sigma_{min } ( t_{edge } ) & = & \sqrt{\frac{\sum_{i}^{cloud } t_i \left [ y_i - \bar{y } ( t_{edge } ) \right]^2}{\sum_{i}^{cloud } t_i } } \mbox { , where } \\
\bar{y } ( t_{edge } ) & = & \frac{\sum_{i}^{cloud } t_i y_i}{\sum_{i}^{cloud } t_i } \mbox{.}\end{aligned}\ ] ] in the equations above the sum @xmath27 runs over all pixels within the cloud .
we have written each of the moments as a function of @xmath12 because changing the isosurface that defines the boundary of the cloud ( @xmath12 ) will change the set of pixels included in the sum and therefore the values of the moments .
note that @xmath21 is not the rms distance ( @xmath28 ) from the center of the cloud .
rather it is the analogous to the rms size of the cloud along an arbitrarily chosen axis .
thus , @xmath29 for a perfectly round cloud , while the rms distance from the center for such a distribution is larger , @xmath30 .
also note that @xmath31 is the axis ratio of the cloud and will be @xmath32 for round clouds and @xmath33 for elongated or filamentary clouds .
we calculate the velocity dispersion , @xmath34 in the same manner as the size : @xmath35 ^ 2}{\sum_{i}^{cloud } t_i } } \mbox { , where } \\
\bar{v } ( t_{edge } ) & = & \frac{\sum_{i}^{cloud } t_i v_i}{\sum_{i}^{cloud } t_i } \mbox { .}\end{aligned}\ ] ] the sums again run over all emission in the cloud . for a gaussian line profile , such as that found for most clouds , the full - width half - maximum ( fwhm ) line width ,
@xmath36 will be related to @xmath37 by @xmath38 finally , we calculate the flux of the cloud , @xmath39 using the zeroth moment : @xmath40 if @xmath4 and @xmath5 are in units of arcseconds , @xmath6 in km s@xmath41 , and @xmath11 in k , then the resulting flux will have units of k km s@xmath41 arcsecond@xmath42 . the sensitivity of a dataset influences the cloud properties derived from that data , a fact that we have emphasized in the previous section by explicitly writing the moments as functions of @xmath12 , the cloud boundary ( usually set by the signal - to - noise ratio of the data ) .
figure [ extrapmoms ] shows the variation of spatial size , line width , and flux as a function of sensitivity for a bright cloud in m33 .
the data for this cloud shows a substantial _ sensitivity bias _ ; all of the derived properties are strong functions of the boundary isosurface ( @xmath12 ) . in order to compare data sets with different sensitivities ,
one must account for this bias . in this section
we describe a method to do this by extrapolating the measured properties of a cloud@xmath23 , @xmath24 , @xmath37 , and @xmath43to those we would expect to measure for a cloud within a boundary isosurface of @xmath44 kelvin ( i.e. , perfect sensitivity ) .
we estimate the values of the moments at @xmath45 k by extrapolating from higher values of @xmath12 . this technique was originally suggested for inferring total cloud areas by @xcite and applied to molecular cloud properties by @xcite .
we calculate each of the moments for a sample of boundary temperatures , @xmath12 , ranging from near the peak temperature of the cloud to the lowest boundaries allowed by the data .
thus , we measure the variations of the moments as a function of the boundary temperature , @xmath12 , within the cloud ( this is how we constructed the plot shown in figure [ extrapmoms ] ) .
below , we assume that we have measured each of the four moments for values of @xmath12 ranging from @xmath46 ( the minimum allowed by the data , that is the sensitivity limit ) to @xmath47 ( near the peak temperature of the cloud ) .
we estimate the value of the moments at @xmath44 k by performing a weighted , linear least - squares fit to the measured moments . as an example , we consider @xmath23 .
the data are modeled as @xmath48 and the fit determines the extrapolated moment , @xmath49 . for the fit ,
each pair of data @xmath50 is assigned a weight proportional to the number of data in the cloud with @xmath51 , so that measurements of the moment using more data are weighted more heavily .
practically , this means that points to the left in figure [ extrapmoms ] have higher weights than those to the right .
we use this linear extrapolation for @xmath52 , @xmath53 , and @xmath54 , but we find that a quadratic extrapolation ( including a @xmath55 term ) gives better results for the zeroth moment , @xmath56 ( though we revert to a linear extrapolation when the extrapolated flux is lower than the measured flux ) .
we plot the flux of a gaussian cloud as a function of @xmath12 in figure [ fluxfig ] for an uncorrected zeroth moment and the linear and quadratic extrapolations to illustrate the appropriateness of the quadratic fit to the zeroth moment . at very low sensitivities ( signal - to - noise ratios near unity ) ,
the quadratic extrapolation is very noisy , but for signal - to - noise ratios of @xmath57 or better it does a dramatically better job of recovering the true flux of the cloud ( @xmath58 ) than either the linear extrapolation or no extrapolation .
figure [ extrapmoms ] also shows these extrapolations for a bright cloud in m 33 .
the result of this extrapolation is a set of four moments @xmath59 , @xmath60 , @xmath61 , and @xmath62 that correspond to those we would expect to measure given infinite sensitivity .
the values of these moments should be directly comparable even among datasets with different sensitivities ( values of @xmath46 ) .
note that diffuse emission surrounding a gmc may confuse this method .
if one data set is measured with sensitivity sufficient to detect diffuse emission surrounding a gmc , while another lacks the sensitivity to do so then this approach may not be sufficient to correct for the sensitivity bias
. this problem may be particularly acute when comparing galactic gmcs observed with very good sensitivity to extragalactic clouds with worse signal - to - noise ratios .
interferometric data `` resolves out '' emission significantly more extended than the synthesized beam ( see [ interf ] ) , representing another bias against detecting diffuse emission .
@xcite , @xcite , @xcite , and @xcite find evidence for diffuse emission surrounding gmcs in the milky way and the local group galaxies m 31 , m 33 , and ic 10 , respectively .
any astronomical data set represents the convolution of the intensity of the source with the profile of the instrument used to observe it .
care must therefore be taken in measuring sizes and line widths when the extent of the intensity distribution is comparable to the instrumental profile . in a typical spectral line data cube
two profiles are important : the spatial beam and the width of a velocity channel . in this section ,
we describe simple corrections to account for the effects of finite spatial and spectral resolution .
we `` deconvolve '' the spatial beam from the measured cloud size by subtracting the rms beam size , @xmath63 , from the extrapolated spatial moments , @xmath64 and @xmath65 , in quadrature an approach that is exact for gaussians .
the deconvolved second moment is given by : @xmath66^{1/2 } ~ [ \sigma_{min}^2 \left(0~\mathrm { k } \right ) - \sigma_{beam}^2 ] ^{1/2}}~\mbox { , } \ ] ] where @xmath67 and @xmath68 are extrapolated to the 0 kelvin isosurface as described in [ extrapsect ] .
this extrapolation is necessary to make this deconvolution valid : subtracting the full @xmath63 from the spatial moment measured for only part of the cloud will lead to an overcorrection and thus to an underestimate of the cloud size .
measuring the spatial size along the minor axis is also necessary to ensure that the cloud is indeed resolved in all dimensions .
this is an advantage of the choice of axes ( major / minor ) described above . with sufficient signal - to - noise , this method of deconvolution
provides a robust measurement of cloud size even for marginally resolved clouds .
instrumental resolution also affects the measured line width .
spectrometers measure the average intensity across a channel , rather than sampling the intensity at the center ( nominal frequency ) of that channel .
when the width of the spectral line under consideration is comparable to the bandwidth of a single channel , the line strength varies significantly across an individual channel . in this case
, the average value may differ substantially from the value at line center .
the output of the spectrometer is thus a convolution of the true spectral profile with the profile of an individual channel .
we account for this potential bias towards higher line widths by a simple deconvolution of the channel width from the extrapolated second moment : @xmath69 where @xmath70 is the second moment of the cloud in the @xmath71 dimension extrapolated to 0 kelvin as described in [ extrapsect ] and @xmath72 is the width of a velocity resolution element . although the channel profiles are usually square in shape and not gaussian , we simplify the deconvolution by approximating the channel shape as a gaussian with integrated area equal to that of a square channel with width @xmath72 .
for such a gaussian , @xmath73 .
we use extrapolated moments to measure gmc properties rather than employing an established method from the literature . in this section , we justify our choice by comparing several methods of measuring gmc properties .
we focus on the performance of these methods at marginal resolution and low signal - to - noise , conditions typical of extragalactic gmc observations . determining the radius of a cloud is particularly difficult because gmcs are often asymmetrical with poorly defined boundaries .
several authors have devised methods to return a single characteristic size for complicated emission distributions .
the intensity - weighted second moments in the spatial directions have been used in many studies ( e.g. * ? ? ?
* ) , but are sensitive to both noise and convolution effects . the other commonly used method ( e.g. * ? ? ?
* ; * ? ? ?
* ) is to infer the radius based on the area of the cloud : @xmath74 here @xmath75 is the area of a point source that has been convolved with a beam and measured with the same signal - to - noise as the emission in the map .
finally , @xcite and @xcite adopted the size of the cloud as the mean of the deconvolved fwhms of the emission distribution along two perpendicular directions .
we compare these three methods to the extrapolated moment method presented above across a range of resolutions and sensitivities .
we measure the size of a galactic gmc using each method after convolving it to a desired resolution and adding noise to produce a particular signal - to - noise ratio . for the data , we use the @xmath2co data from the rosette molecular cloud @xcite , which we clip at 2@xmath76 ( the rms noise in the original data set ) and integrate in the velocity dimension to produce a map of integrated intensity . for a range of sensitivities and resolutions , we convolve this map with a gaussian beam and add noise .
we measure the size of the cloud in 100 realizations of the noise for each such \{resolution , sensitivity } pair using ( a ) the extrapolated moment method , ( b ) the moment of the data without extrapolation , ( c ) the area method and ( d ) the fwhm method .
wherever possible we corrected for the effects of beam convolution and signal - to - noise for each of the methods .
the results of the analysis are shown in figure [ radplot ] .
figure [ radplot ] shows the recovered radius as a function of resolution and sensitivity , with the `` true '' radius defined as that measured at very high sensitivity and very good resolution ( i.e. in the original data , the top right corner of each plot ) .
the hashed region of parameter space shows the range of parameters over which each algorithm recovers a radius within 10% of the true value .
the extrapolated moment method has the largest hashed region and so is remarkably robust , recovering the true radius over a large range of sensitivities and resolutions . only at low sensitivity ( @xmath77 ) and marginal resolution ( @xmath78 ) ,
does the derived radii depart systematically from the true radius .
notably , the extrapolated second moment performs quite well at signal - to - noise ratios from 5 to 10 ( in the integrated intensity map ) , values typical of extragalactic co data sets .
by contrast , the uncorrected moment method ( panel b in figure [ radplot ] ) underestimates the size at low signal - to - noise since ( by construction ) the uncorrected moment does not account for emission below the noise level .
similarly , the area method ( panel c ) shows systematic variation at both low signal - to - noise ( where emission drops below the noise level ) and low resolution ( where the convolved area of the emission distribution grows disproportionately because of the filamentary nature of the cloud ) .
finally , the fwhm method ( panel d ) shows large systematic variations since it depends only on the location of the fwhm contour and not on the remainder of the emission distribution .
the region of systematic underestimation at low signal - to - noise but reasonable resolution shows the effects of missing the diffuse emission mentioned in [ extrapsect ] .
the rosette includes more co emission at low intensities than the extrapolated moment predicts from the high intensity data . as a result , when that diffuse emission is not included in the measurement ,
the algorithm underestimates the true radius of the gmc .
this effect is seen panels ( a ) and ( b ) the extrapolated and uncorrected second moments and is more pronounced in the uncorrected second moment , panel ( b ) .
we perform a similar experiment on recovering the line width of an emission line in noisy data .
we measure the recovered line width of a gaussian line of known width using three methods ( a ) the extrapolated moment method ( b ) an uncorrected second moment and ( c ) a gaussian fit to the line . for a range of signal - to - noise levels ( @xmath79 ) and channel widths @xmath72 we measure the recovered line width relative to the known line width .
the mean values of the recovered line for 1000 realizations of the noise are plotted in figure [ dvplot ] .
the extrapolated moment does not show the systematic variation with signal - to - noise seen in the uncorrected moment .
the extrapolated moment is nearly as robust a measure as the gaussian fit for a perfectly gaussian line and will prove superior if the line is not gaussian .
robust recovery of the line width using any method requires @xmath80 .
the formal uncertainty associated with each moment measurement is quite small .
cloud identification and extrapolation represent larger sources of uncertainty , but their effects are difficult to assess formally .
we use bootstrapping methods to estimate the uncertainties in our measurements of cloud properties .
the bootstrapping method determines errors by generating several trial clouds from the original cloud data .
a trial cloud is generated by considering the cloud to be a collection of data @xmath81 for @xmath82 , the number of points in the cloud .
the data are sampled for @xmath83 random values of @xmath7 , allowing for @xmath7 to be repeated .
the properties of the cloud are measured for each trial cloud .
we estimate the uncertainty from the variance of the cloud properties derived from these resampled and remeasured data sets .
the final uncertainty in each property is the standard deviation of the bootstrapped values scaled up by the square root of the oversampling rate .
the oversampling rate , which is usually equal to the number of pixels per beam , accounts for the fact that not all of the data in each cloud are independent . for many interferometric data sets
this is an important effect , since these data can have @xmath84 or more pixels per beam .
we compare the uncertainties produced by the bootstrapping to those derived from repeatedly adding noise to and then reanalyzing a data set .
we use the bright cloud in m 33 shown in figure [ extrapmoms ] .
we conduct 100 realizations of the data plus random noise .
the resulting uncertainties in @xmath52 , @xmath53 , @xmath54 , and the flux are @xmath85 , @xmath86 , @xmath85 , and @xmath85 . repeatedly bootstrapping the same data set ( adjusted to have the same final noise level ) yields average uncertainties of @xmath87 , @xmath87 , @xmath88 , and @xmath89 .
the bootstrapping estimates are higher for this bright cloud because they reflect both the formal uncertainty and the robustness of the result to the removal of a given piece of data . in the low signal - to - noise regime ,
the values for the two methods converge as noise dominates the uncertainty derived from bootstrapping for example , performing the same experiment in a dimmer m 33 cloud with @xmath90 the luminosity of the bright cloud and comparable noise , bootstrapping yields errors of @xmath91 , @xmath92 , @xmath93 , and @xmath94 in the four moments while repeated realizations produces scatters of @xmath95 , @xmath96 , @xmath97 , and @xmath98 .
the bootstrapping method produces a robust , believable estimate of the uncertainty in the measurement of the properties of a particular , defined cloud .
it does not account for uncertainties in the assignment of emission to a cloud either as a result of noise or choice of algorithm .
these uncertainties are more systematic than random in nature and may be best assessed by analyzing the emission distribution using several methods .
the bootstrapping estimate may be treated as an accurate estimate of the uncertainties in the results _ given _ that one adopts the methods presented in this paper .
in this section , we outline how to use the measured size , line width , and flux to calculate several physical quantities of interest : the effective spherical radius , the virial mass , and the luminous mass . throughout this section
we assume that clouds can be described as self - gravitating spheres with density profiles @xmath99 and negligible support from magnetic fields or confinement by external pressure .
we assume below that the data consists of observations of the transition , in units of brightness temperature , but the method is readily adaptable to analogous data sets .
we define a factor @xmath100 that relates the one - dimensional rms size , @xmath101 , to the radius of a spherical cloud @xmath102 : @xmath103 .
it is possible to derive an estimate for @xmath100 based on spherical cloud of radius @xmath102 with a density profile @xmath104 . in this model ,
@xmath105 for a cloud with @xmath106 , @xmath107 , somewhat higher than the empirical correction of @xmath108 derived by solomon et al .
the difference arises , in part , from using @xmath2co as a density tracer .
since @xmath2co emission saturates in dense regions and vanishes from low density regions , the apparent density profile in @xmath2co is shallower than the true density profile .
hence , an appropriate value of @xmath100 likely falls in the range between @xmath108 and @xmath109 . for tracers like @xmath110co with higher critical densities ,
a different value of @xmath100 may be appropriate . since comparison to this `` anchoring '' data set
may be more important than adopting a self - consistent but grossly oversimplified model for a cloud , we recommend the solomon et al .
( 1987 ) definition of the cloud radius , @xmath111 .
note , that adjusting the definition of the radius renders the virial mass formula we present below inexact .
we convert the cloud properties @xmath112 , @xmath113 , and @xmath114 to the physical quantities @xmath102 , @xmath115 , and @xmath116 . for a cloud at a distance of @xmath117 ( in parsecs ) , the physical radius will be @xmath118 = \frac{r ( 0~\mathrm{k } ) [ \mathrm{arcsec}]}{3600 } \times \frac{\pi}{180 } \times d [ \mathrm{pc}]~\mathrm{,}\ ] ] the fwhm line width will be @xmath119 and the luminosity of the cloud , @xmath116 , will be @xmath120 & = & f_{\mathrm{co}}(0~\mathrm{k } ) [ \mathrm{k~km~ s}^{-1}~\mathrm{arcsec}^2 ] \\
\nonumber & \times & ( d[\mathrm{pc}])^2\\ & \times & \left(\frac{\pi}{180\cdot 3600}\right)^2.\end{aligned}\ ] ] a particular co luminosity , @xmath116 , implies a mass of molecular gas , @xmath121 , of @xmath122 & = & \frac{x_{\mathrm{co}}}{2 \times 10^{20 } [ \mathrm{cm}^{-2}/(\mathrm{k~km~s}^{-1 } ) ] } \times 4.4~l_{\mathrm{co}}\\ & \equiv & 4.4~x_2~l_{\mathrm{co}}\end{aligned}\ ] ] where @xmath123 is the assumed co - to - h@xmath1 conversion factor .
this calculation includes a factor of 1.36 ( by mass ) to account for the presence of helium . including helium is important to facilitate comparison with the virial mass , which should reflect all of the gravitating mass in the cloud .
we have adopted a fiducial value of the co - to - h@xmath1 conversion factor of @xmath124 and express changes relative to this value in terms of the parameter @xmath125 .
we compute the virial masses under the assumption that each cloud is spherical and virialized with a density profile described by a truncated power law of the form @xmath104 with no magnetic support or pressure confinement . as with the spherical radius correction ,
the exact density profile of the cloud will affect the correct form of the virial theorem mass .
for @xmath106 , the virial mass is given by the formula ( solomon et al .
1987 ) : @xmath126 and more generally by @xmath127 where @xmath115 is the fwhm velocity line width in , @xmath102 is the radius in pc , and the cloud has a density profile of @xmath128 .
clouds exhibit a range of non - spherical geometries and may be supported by magnetic fields or confined by pressure . therefore , the studying the virial parameter may be more useful than the virial mass itself .
the virial parameter is a constant of order unity that characterizes deviations from the virial theorem applied to a non - magnetic cloud with no external pressure and constant density .
following @xcite , we define the virial parameter as @xmath129 larger - than - unity virial parameters can result from pressure confinement , while @xmath130 may result from significant magnetic support .
incorrect values of the co - to - h@xmath1 conversion factor may skew the result in either direction .
finding @xmath131 means that the clouds are gravitationally bound in the absence of significant magnetic support .
millimeter - wave interferometers are required to resolve even the most massive molecular clouds in galaxies beyond the magellanic clouds ( see figure [ distcomp ] ) .
unfortunately , interferometers are not sensitive to spatial frequencies outside the limited region of the @xmath0 plane that they sample .
practically , this means interferometers do not measure the total flux from the emission distribution ; and structures are resolved out , usually on large angular scales that correspond to small separations in the @xmath0 plane .
ideally , interferometer observations are combined with single - dish observations that supply the missing information . in practice
such observations are conducted infrequently and the unknown total flux and short - spacing information is estimated using deconvolution algorithms such as clean or maximum entropy ( see * ? ? ?
* and references therein ) .
@xcite simulate the results of using only an interferometer to observe galactic gmcs as if these well - studied clouds were located in m31 .
they find that interferometers experience significant ( 50% ) flux loss for their simulated observation , primarily from extended emission around clouds .
however , they find that the flux loss does not change the size and line width of the cloud .
@xcite examine the recovery of large - scale flux distributions from interferometer measurements in more depth and explore the effectiveness of deconvolution algorithms at low signal - to - noise .
they find that deconvolution algorithms recover flux nonlinearly at low sensitivities , finding much less flux at low sensitivities than one would expect .
since much of the data on extragalactic gmcs have low signal - to - noise , this may represent an important bias .
we assess the effects of interferometric biases on the methods presented here by extending the method of @xcite .
we use @xmath132 observations of three galactic gmcs : the orion molecular complex @xcite , the rosette molecular cloud @xcite , and an excerpt from the outer galaxy survey of @xcite which contains the molecular clouds associated with the w3/w4/w5 regions .
we simulate observing these three molecular complexes in m31 ( @xmath133 kpc ) with the bima interferometer .
we fourier transform each plane of each data set into the @xmath0 domain and resample the data along the @xmath0 tracks that would be sampled by bima observations of the data provided the gmcs were in m31 .
the @xmath0 coverage reflects typical observing strategies for extragalactic clouds , such as interleaving observations of the source with calibrators and other sources .
we add thermal noise and phase noise to the @xmath0 data , including a phase noise component with a magnitude that depends on the length of the baseline .
we adjust the scale of the thermal and phase noise to produce the desired peak signal - to - noise ( we find our results depend only weakly on whether the noise is thermal or phase ) . we invert the resulting @xmath0 data using the miriad software package @xcite producing maps separated by 2 km s@xmath41 , and then we deconvolve the dirty maps using a clean algorithm that terminates at the 2@xmath76 level . for each trial cloud
, we then calculate the cloud properties using the methods of [ moments ] and [ physical ] . for comparison , we compute cloud properties using the same procedure to simulate single - dish observations with signal - to - noise and effective resolution identical to the mock interferometer data .
we generate the mock single - dish observations by sampling the transformed image data for an equal number of @xmath0 points as the interferometer data , but the points are normally distributed in the @xmath0 plane and one point is forced to lie at @xmath134 thus sampling the total power .
the width of the @xmath0 point distribution is chosen to give a beam size similar to that of the mock interferometer data .
random thermal and phase noise is added to these data in the same fashion as for the interferometer data .
then the data are inverted using natural weighting and deconvolved in the exact same fashion as the interferometer data ( though the deconvolution step has little effect ) .
again , we extract cloud properties using the methods of [ moments ] and [ physical ] . with these simulations ,
we compare the cloud properties derived from interferometer data to single - dish data that are equivalent in every other fashion , thereby isolating the effects of interferometers on the derived properties .
the additional biases imposed by limited resolution and sensitivity are discussed separately in [ compare ] . here
we focus on mock bima observations of the three molecular complexes using three antenna configurations : the c array ( extended ) , the d ( compact ) array , and a combination of c and d array ( see * ? ? ?
* for details on the configurations ) .
the synthesized beam sizes for these configurations are @xmath135 , @xmath136 and @xmath137 for the d , c and c+d hybrid array observations respectively , corresponding in turn to 54 , 22 and 33 parsecs at 770 kpc .
we conduct 10 sample observations for each cloud in each array at a range of sensitivities ranging from @xmath138 to @xmath139 .
figure [ interf_props ] shows the properties recovered by mock observations of the orion molecular complex for each array and a range of sensitivities .
the values of each properties are normalized by the value recovered by mock single - dish observations at the same sensitivity and resolution .
thus , the only difference between the four sets of properties ( c , d , c+d , and single dish ) is the @xmath0 coverage of the simulated observations .
we find that the derived properties from mock single - dish observations follow the same behavior as the simulations in [ compare ] .
thus , it is possible to decouple the two sets of biases those arising from marginal resolution and sensitivity and those arising from using interferometers and examine only the latter .
we plot the results for orion because these observations show the most dramatic variation of the three complexes studied , but the results are qualitatively the same for all three data sets .
orion is the most sensitive of the three to spatial filtering because it consists of three gmcs and therefore shows more structure than the other two targets .
based on the results of the mock observations , we make the following comments regarding the use of interferometer data alone in measuring cloud properties .
most of these points can be seen visually in figure [ interf_props ] . 1 .
cloud properties measured from interferometric data are biased .
the degree of bias is affected by the sensitivity of the array as well as the @xmath0 coverage of the observations .
a minimum signal - to - noise of 10 is required for stable recovery of cloud properties .
below this level , errors in cloud properties , can approach 100% for interferometer data .
3 . even for intermediate signal - to - noise values ( @xmath140 )
there are significant systematic effects on the cloud properties .
the most extreme effects are on the luminosity measurement , which can be 40% lower than a single - dish observation .
this effect is much less pronounced for measurements of the line width and the radius , which show @xmath141 variations .
this result , that the radius and line width are relatively robust to the spatial filtering of the interferometer , confirms the qualitative results of @xcite .
the values of derived properties are always _ underestimated _ relative to single - dish observations .
4 . even at high sensitivities ,
the spatial filtering of interferometers affects property recovery at the 10% level .
for example , c - array observations of orion systematically underestimate the radius of the cloud by 10% even for very high signal to noise ratios and d - array observations underestimate the flux of the orion by 5% even at high sensitivity .
5 . for interferometer observations ,
the dynamical mass measurements of gmcs are more robust than the luminosity measurements .
this behavior will bias interpretations of the virial parameter in extragalactic observations .
estimates of the co - to - h@xmath1 conversion factor based on the assumption that gmcs are bound or virialized are likely to _ overestimate _ @xmath123 .
6 . for a given signal - to - noise value , observations with the widest range of @xmath0 coverage provide the most robust measurement of cloud properties .
thus , in achieving a given sensitivity , observers should favor arrays with more antennae or observations made in multiple configurations .
the choice of how to decompose an emission distribution into individual clouds may be the most important source of bias in gmc property measurements .
many different methods have been applied to identify gmcs in blended emission , the most prevalent being decomposition by eye and the application of the gaussclumps @xcite or the clumpfind algorithm @xcite . when comparing gmcs between two data sets
, care must be taken to decompose emission in a consistent way across both data sets , preferably using the same algorithm on both data sets .
furthermore , the physical values of any tuning parameters in the algorithms should be matched where possible so that both algorithms search for peaks in the emission over the same spatial scale ( rather than angular or resolution - units ) or intensity range .
this will avoid , for example , comparing `` clumps '' in a galactic molecular cloud to gmcs in another galaxy .
a more extreme method to ensure accurate comparison is to convolve the higher ( spatial ) resolution data set to the spatial resolution of the other data ( e.g. * ? ? ?
* ) . however , this approach clearly sacrifices accuracy of the derived parameters to allow a more careful comparison between two data sets . in the appendix to this paper
we present a robust , conservative , new decomposition algorithm .
this algorithm is designed to avoid creating spurious clouds from noise and to remain sensitive to non - gaussian structures in the data .
additionally , the parameters of the algorithm are fixed to physical values rather than being determined by the data .
this algorithm is designed explicitly with the goal of decomposing emission into gmcs ( rather than clumps or other structures ) with extragalactic data in mind .
the methods for measurement of cloud properties described above including the sensitivity and resolution corrections are independent of the decomposition algorithm and are important no matter what decomposition algorithm is chosen . in order to avoid confusion between these two separate problems , we choose to describe the decomposition algorithm in the appendix .
we conclude the paper by applying these methods to molecular line data sets that have been previously published . in future studies
, the algorithm will be used to evaluate the differences between gmc populations between galaxies . here , we simply present an analysis designed to demonstrate the method s utility .
we present the median corrections found for a large set of extragalactic ( local group ) observations and a test application of our methods to galactic data .
we use all the methods discussed in the previous sections and the decomposition algorithm discussed in the appendix .
we apply the methods outlined here to an array of data from across the local group and measure the properties of 110 spatially resolved clouds to estimate the typical magnitude of the sensitivity and resolution corrections for extragalactic data .
we use bima data on m 33 and m 31 @xcite ; nanten observations of the lmc @xcite ; ovro observations of ic 10 @xcite ; and a sest map of n83 in the smc @xcite .
table [ corrections ] shows the number of clouds measured in each galaxy along with the median sensitivity and resolution corrections applied to the radius , line width , and flux . for comparison
, we also measure the properties of a number of clouds in the outer galaxy ( quadrant 2 , see below ) from the @xcite co survey of the milky way .
table [ corrections ] includes all spatially resolved clouds with masses ( derived from the co luminosity ) of @xmath142 m@xmath143 or more ( @xmath144 of the @xmath145 extragalactic clouds are above this mass ) . the numbers quoted in table [ corrections ] are `` correction factors , '' @xmath146 table [ corrections ] shows that throughout the local group data the corrections suggested in this paper have magnitudes of a few tens of percent .
we draw several conclusions based on these data : 1 .
resolution effects on the size of clouds tend to be significant we would overestimate cloud sizes by @xmath147% if we did not apply a deconvolution . in the milky way data ,
this effect is much less severe .
the sizes of milky way clouds are measured to within @xmath89 before the resolution correction .
unresolved clouds do not contribute to table [ corrections ] , so if the effects of the resolution bias were completely neglected this would be much larger ( a naive approach would measure these clouds to have the size of the spatial beam ) .
2 . resolution effects on the line width are negligible throughout the local group data .
sensitivity effects are also significant . without a correction for the sensitivity bias , the size , line width , and luminosity of clouds
would all be significantly underestimated .
this sensitivity bias is least severe only about 20 30% for the line width , and most significant ( and variable ) for the luminosity .
sensitivity corrections to the luminosity vary from 20% to more than 100% .
the resolution and sensitivity biases for the size measurement tend to cancel out , so that the completely uncorrected radius measurement is often within 10 20% of the corrected value .
this is a happy coincidence of resolution and sensitivity within the local group , not evidence that sensitivity and resolution corrections are unimportant .
the magnitude of corrections across the local group data are fairly uniform .
this is because gmcs near the resolution limit tend to outnumber higher mass gmcs .
unresolved gmcs are not included in the analysis , so the median cloud through all the data sets appears marginally resolved .
6 . in order to compare extragalactic data to galactic data ( with very good sensitivity and resolution and therefore small corrections )
it is crucial to correct for the sensitivity and resolution biases .
l c c c c c c lmc & 46 & @xmath148 & @xmath149 & @xmath150 & @xmath151 & @xmath152 + m 31 & 28 & @xmath153 & @xmath154 & @xmath155 & @xmath156 & @xmath152 + ic 10 & 17 & @xmath150 & @xmath154 & @xmath157 & @xmath156 & @xmath152 + m 33 & 15 & @xmath148 & @xmath149 & @xmath153 & @xmath156 & @xmath152 + smc & 4 & @xmath158 & @xmath149 & @xmath149 & @xmath156 & @xmath152 + mw & 107 & @xmath158 & @xmath158 & @xmath148 & @xmath152 & @xmath152 + the method described in this paper has been designed with extragalactic data in mind . however , a crucial step in interpreting extragalactic measurements is to make a fair comparison with galactic data . in this section
we report some results of applying our decomposition and measurement algorithms to the survey of the second galactic quadrant by @xcite .
we compare the results of this analysis to the results by @xcite and show that our analysis recovers results that are consistent with theirs .
we decompose and analyze @xmath2co(@xmath3 ) from the second quadrant ( survey 17 in table 1 of * ? ? ?
the data set covers the galactic plane from @xmath159 to @xmath160 with a noise level of @xmath161 k. we measure the distance to the molecular emission using the kinematic distances by adopting a flat rotation curve with @xmath162 km s@xmath41 and @xmath163 kpc .
we omit local emission by discarding all elements of the data cube with a kinematic distance less than 2 kpc as well as all elements in the data cube that are connected by significant emission in position or velocity space to such pixels .
we apply the decomposition algorithm described in the appendix and measured sizes , line widths , and luminosities of gmcs using the methods of [ extrapsect ] and [ beamdcsect ] .
the analysis recovers 431 clouds with resolved angular sizes and line widths located within 10 kpc of the sun .
we include the median sensitivity and resolution corrections for massive ( @xmath164 m@xmath143 ) clouds in table [ corrections ] above .
do the results from our algorithm agree with previous studies of galactic gmcs ? the data set covers the region studied by @xcite using the @xmath165 resolution of the fcrao 14 m .
that data set has a lower sensitivity than the @xcite data , so we apply a rudimentary sensitivity correction ( assuming that the gmcs are gaussian and using their peak temperature and boundary isosurface ) to their results and scale to @xmath166 before comparing gmc properties .
we focus on a comparison of the virial parameter between the two studies a full treatment of the galactic `` larson s laws '' is beyond the scope of this paper .
we draw several conclusions from the comparison : 1 .
figure [ virparams ] shows that the virial parameters derived in our analysis largely agree with those found by @xcite .
below masses of @xmath167 m@xmath143 , both surveys find the same virial parameter in a given mass bin .
2 . above @xmath167 m@xmath143
, our analysis of the @xcite data set may yield a slightly higher virial parameter , on average .
this may be evidence for the diffuse emission mentioned in [ extrapsect ] the higher sensitivity @xcite data may include diffuse emission surrounding the gmcs while the @xcite data may miss this effect .
it may also reflect inadequacies in the simple sensitivity correction we apply to the @xcite data .
resolution effects may also play a role the @xcite data set has @xmath168 times better resolution than the @xcite data and so the lower resolution data may tend to lump unbound clouds together .
the number of clouds in the high mass bins is relatively low , so the discrepancy may not be particularly significant .
3 . we apply our algorithm to a small portion of the ogs data and find our corrections for resolution and sensitivity bias increase the mean virial parameter to be consistent with the results from the @xcite data .
this suggests that the differences in the virial parameters for the high mass clouds in figure [ virparams ] may simply be methodological .
both catalogs of outer galaxy clouds find an inverse relationship between luminous mass and the virial parameter , approximately @xmath169 a relationship that was also observed by @xcite for inner galaxy molecular clouds .
thus we find agreement with the results of previous studies of galactic molecular clouds .
our methods applied to galactic data find the same behavior observed in earlier work and we find agreement among the method applied to several data sets .
we have presented a method for measuring macroscopic gmc properties spatial size , line width , and luminosity from a region of emission in a spectral line data cube .
this method corrects for biases from limited sensitivity and resolution and produces reliable results that are directly comparable among a wide variety of data sets .
we correct for limited sensitivity via an extrapolation to a theoretical 0 kelvin isosurface .
we apply a simple quadratic deconvolution to the extrapolated values to account for resolution biases .
we find that bootstrapping methods yield believable estimates of the uncertainties in the derived parameters .
we present a set of suggestions for transforming the derived properties into physical quantities of interest . in the appendix to this paper
we present a new method for decomposing emission into individual gmcs .
this method is conservative and robust , designed to produce robust results from low resolution / sensitivity data . in this section
we have applied all of these methods to an array of extragalactic ( local group ) and galactic data .
we find that the algorithm reproduces established results for galactic gmcs and that the sensitivity and resolution biases are potentially significant often @xmath170 for even the most recent local group gmc measurements . based on this investigation of the observational biases in measuring molecular cloud properties , we note several important points that should be considered in planning observations of gmcs and interpreting the results .
first , resolution and sensitivity biases can be corrected to @xmath171 error provided the cloud has a modest peak signal - to - noise ( @xmath172 ) and is marginally resolved @xmath173 , where @xmath174 is the full width at half maximum extent of the beam .
given current and future telescope capabilities , even the properties of extragalactic gmcs can be accurately measured . from figure [ distcomp ] , we can see that single dish surveys can accurately study clouds more massive than @xmath175 in the magellanic clouds and careful interferometer observations can recover cloud properties for clouds with mass @xmath176 in m31 and m33 .
however , interferometer observations systematically underestimate molecular cloud properties .
all else being equal , interferometers can underestimate fluxes by 40% and cloud radii by 10% relative to single - dish observations .
line widths are largely unaffected by interferometer observations .
the magnitude of the systematic bias depends on the array configuration and the sensitivity of the observations . in general , wider coverage of the @xmath0 plane produces better property recovery . to minimize bias
, observers should favor observations from several array configurations or from arrays with many elements .
if possible , the interferometer data should be supplemented with single - dish observations .
finally , the decomposition algorithm used to separate emission into physically relevant structures will systematically affect molecular cloud properties .
to date , there is no algorithm that should be favored in all circumstances , but any comparative study of gmc properties should be consistent in the choice of algorithm .
for example , the results of a clumpfind algorithm applied to an extragalactic data set are not directly comparable to a catalog produced by a simple contouring method applied to milky way data . provided the same algorithm is used across multiple data sets , referencing algorithm parameters to a common physical scale will minimize systematic differences .
we will make a software version of the decomposition and measurement algorithm available as an idl package .
further , since methodology can affect the results of gmc studies so strongly , we encourage authors working in the field to make their data available to the community after publication in order to facilitate future rigorous comparisons .
we are extremely grateful to tom dame and the millimeter - wave group at the center for astrophysics for providing both the orion data and the quadrant 2 data used in the paper .
we also thank the nanten group at nagoya , especially yasuo fukui and akiko kawamura , providing us with the lmc co data .
we are grateful to fabian walter for providing us with the ovro ic 10 data .
alberto bolatto , jason wright , jon swift , and leo blitz all offered helpful comments on drafts of the paper .
we thank ronak shah for helping us compare our idl version of gaussclumps to the original implementation of the algorithm .
the informed comments of an anonymous referee greatly improved the paper , particularly in encouraging us to explore the effects of interferometers .
er is grateful for support through the national science foundation astronomy & astrophysics postdoctoral fellows program ( ast-0502605 ) .
this work is partially supported by nsf grant 0228963 to the radio astronomy laboratory at uc berkeley .
12 natexlab#1#1 bertoldi , f. , & mckee , c. f. 1992 , , 395 , 140 blitz , l. 1985 , , 296 , 481 , l. & stark , a. a. 1986 , , 300 , l89 blitz , l. , & thaddeus , p. 1980
, , 241 , 676 bolatto , a. d. , leroy , a. , israel , f. p. , & jackson , j. m. 2003 , , 595 , 167 , c. m. , kerton , c. r. , & pomerleau , c. 2003 , , 144 , 47 , t. m. , elmegreen , b. g. , cohen , r. s. , & thaddeus , p. 1986
, , 305 , 892 dame , t. m. , hartmann , d. , & thaddeus , p. 2001
, , 547 , 792 fukui , y. , et al . 1999 , , 51 , 745 helfer , t. t. , vogel , s. n. , lugten , j. b. , & teuben , p. j. 2002 , , 114 , 350 heyer , m. h. , brunt , c. , snell , r. l. , howe , j. e. , schloerb , f. p. , & carpenter , j. m. 1998 , , 115 , 241 , m. h. , carpenter , j. m. , & snell , r. l. 2001 , , 551 , 852 israel , f. p. , et al .
1993 , , 276 , 25 koda , j. , sawada , t. , hasegawa , t. , & scoville , n. 2005 , , in press . , c. , stutzki , j. , rohrig , r. , & corneliussen , u. 1998 , , 329 , 249 leroy , a. , bolatto , a. , walter , f. , & blitz , l. 2005 , in preparation lin , g. , adiga , u. , olson , k. , guzowski , j. , barnes , c. , & roysam , b. 2003 , cytometry a , 56 , 23 , c. f. & zweibel , e. g. 1992 , , 399 , 551 polk , k. s. , knapp , g. r. , stark , a. a. , & wilson , r. w. 1988 , , 332 , 432 rosolowsky , e. , engargiola , g. , plambeck , r. , & blitz , l. 2003 , , 599 , 258 rosolowsky , e. , , submitted . , d. b. , scoville , n. z. , & solomon , p. m. 1985 , , 289 , 373 sanders , d. b. , clemens , d. p. , scoville , n. z. , & solomon , p. m. 1986 , , 60 , 1 , r. j. , teuben , p. j. , & wright , m. c. h. 1995 , in asp conf . ser .
77 : astronomical data analysis software and systems iv , 433+ sheth , k. , vogel , s. n. , wilson , c. d. , & dame , t. m. 2000 , proceedings 232 .
we - heraeus seminar , 37 , p. m. , rivolo , a. r. , barrett , j. , & yahil , a. 1987 , , 319 , 730 , j. & gsten , r. 1990 , , 356 , 513 scoville , n. z. , yun , m. s. , sanders , d. b. , clemens , d. p. , & waller , w. h. 1987 , , 63 , 821 taylor , c. l. , httemeister , s. , klein , u. , & greve , a. 1999 , , 349 , 424 thompson , a. r. , moran , j. m. , & swenson , g. w. 2001 , interferometry and synthesis in radio astronomy by a. richard thompson , james m. moran , and george w. swenson , jr .
2nd ed .
new york : wiley , c2001.xxiii , 692 p. : ill . ; 25 cm .
`` a wiley - interscience publication . '' includes bibliographical references and indexes .
isbn : 0471254924 , vogel , s. n. , boulanger , f. , & ball , r. 1987 , , 321 , l145 , f. 2005 , in preparation , j. p. , de geus , e. j. , & blitz , l. 1994 , , 428 , 693 wilson , b. a. , dame , t. m. , masheder , m. r. w. , & thaddeus , p. 2005
, , 430 , 523 , c. d. & scoville , n. 1990 , , 363 , 435 wilson , c. d. , & reid , i. n.1991 , , 366 , l11 wilson , c. d. , & rudolph , a. l. 1993 , , 406 , 477 wright , m. c. h. 1998 , bima memorandum series , 69 [ ouralg ]
the choice of what emission to identify as a gmc may be the single largest source of uncertainty in measuring and comparing gmc properties .
a number of methods have been employed over the years , from simple contouring methods ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) to fitting three - dimensional gaussians ( gaussclumps , * ? ? ?
* ; * ? ? ?
* ) , to modified watershed algorithms ( clumpfind and its kin , * ? ? ?
* ; * ? ? ?
in this section we present a new decomposition algorithm that is designed to identify clouds at low sensitivities while avoiding the introduction of a false clouds due to noise .
this algorithm consists of two parts : identifying regions of significant emission in the data set and then assigning this emission to individual `` clouds . ''
we first identify regions of contiguous , significant emission in our position - position - velocity data cube .
we estimate the noise in the data set by measuring the rms intensity , @xmath76 , from a signal - free region of the data cube .
we then construct a high - significance mask .
this mask includes only adjacent channels that both have intensities above @xmath177 .
we expand that mask to include all emission above a lower threshold typically two channels above @xmath178 significance that is connected to the original high significance mask through pixels with @xmath179 significance .
the resulting mask contains most of the significant emission in the data cube . lowering the threshold below two channels at @xmath180 runs the risk of biasing the moment measurements towards high values by including false emission ( noise with positive values ) in the cloud . in this section
, we describe the algorithm used to decompose a region of emission into individual subsections representing the physically distinct entities in the data ( `` clouds '' ) . through the description of the algorithm
, there are several parameters that can be varied to produce changes in the resulting decomposition . in general , we set these parameters using physical prior knowledge of the gmcs we are seeking to catalog .
we discuss the choice of these parameters in the following section . 1 .
_ discard small or low contrast regions : _ if a region is too small for us to measure meaningful properties from it , we discard the region .
we require that each region has an area larger than two beam sizes , so that we can measure its size ; and a velocity width of more than a single channel , so that we can measure its line width .
if the intensity contrast between the peak and the edge of the region is less than a factor of two , we lack the dynamic range needed to correct the sensitivity bias and we therefore discard the region .
if a region is not discarded we proceed to the next step .
[ xform]_rescale the data to reduce the effects of substructure : _ molecular clouds contain significant substructure that confuses the decomposition of these sources .
the substructure is often significantly brighter than bulk of the gas in the cloud .
we rescale the data to reduce the contrast between this substructure and the cloud as a whole .
the data are rescaled using the following transform : @xmath181 & ; ~ t\ge t_{clip}. \\ \end{array } \right.\ ] ] this transform reduces the contrast pixels with @xmath182 while preserving the relative brightness distribution .
the value of @xmath183 is left as a free parameter .
the transformed data are used in the decomposition algorithm .
such brightness transforms are frequently used in the decomposition algorithms used in other fields such as medical imaging ( e.g. * ? ? ?
3 . _ identify independent local maxima : _ we identify potential local maxima , by identifying the elements in the data cube that are larger than all their neighbors .
we consider neighbors to be all data that lie within a box with side length @xmath184 in position and @xmath185 in velocity centered on the local maximum .
these are our `` candidate maxima . ''
the parameters @xmath184 and @xmath186 are free parameters . if we find more than one candidate maximum in a region of emission , we proceed to the next steps to further verify each maximum s independence .
if we find a single candidate maximum then we label the region as a cloud and measure its properties as described in the main paper.[locmax ] 4 .
_ find shared isosurfaces and reject small clouds or those with smooth mergers : _ for each _ pair _ of candidate maxima we calculate the value of the highest intensity isosurface to contain both maxima .
we refer to this highest shared isosurface as the _ merge level_. using this set of highest shared isosurfaces , we calculate three properties of interest : 1 .
the area uniquely associated with each maximum ( i.e. the area above the merge level for that maximum ) .
the antenna temperature interval between the merge level and each maximum , referred to as the _ contrast interval_. [ decimate ] 3 .
the fractional amount by which each of the ( unextrapolated ) moment values changes across each shared isosurface ( i.e. @xmath187 for each maximum across each shared isosurface ) .
we consider the two clouds to merge smoothly across the isosurface only if : 1 .
none of the second moments increase by more than 100% for both maxima .
2 . no two of the second moments increase by more than 50% for both maxima .
the flux increases by less 200% for both maxima .
+ we use these three properties to pare maxima from the region .
we reject maxima associated with small areas ( less than two beam sizes , as for the region above ) and contrast intervals less than @xmath188 , a free parameter . choosing @xmath189
significantly reduces the effects of noise on decomposition ( e.g. * ? ? ?
* ) since noise is associated with low contrast intervals .
finally , when a pair of maxima merge smoothly across a shared isosurface , we keep the higher intensity maximum and discard the lower one .
this is a conservative choice in the decomposition algorithm : unless merging the two kernels significantly alters the properties of one of the clouds associated with the separate kernels , we assume the kernels are not physically distinct .
the effect of removing kernels that merge smoothly from our data set is to reduce the algorithm s sensitivity to substructure within clouds .
we iterate this step until we have a set of maxima associated with the required areas and separated from each other by significant jumps in their properties . 5 .
_ define clouds using shared isosurfaces : _ the surviving maxima each correspond to a `` cloud . ''
that cloud consists of the emission within the lowest intensity isosurface uniquely associated with the cloud .
emission that lies below this isosurface is part of a `` watershed '' shared among clouds and we do not assign it to any cloud . by not considering contested emission , i.e. emission that could be associated with distinct local maxima , we avoid the problem of how to properly assign this emission to local maxima .
measure cloud properties : _ finally , we apply the methods described in 2 and 3 to derive spatial sizes , line widths , and luminosities for each cloud .
the transformed data ( step [ xform ] ) are inverse transformed into the original brightness units for this analysis .
several of the algorithm s parameters are left to the choice of the user . without any prior knowledge of the physical objects in the data , we establish default values for these parameters that will produce a reasonable decomposition based solely on the characteristics of the data .
these defaults are chosen to provide sensitivity to real substructure within the data without contamination by noise and can be regarded as the _ minimum _ appropriate values for these parameters in most cases .
the free parameters in the algorithm are the brightness transform threshold ( @xmath183 , step [ xform ] ) position and velocity window size used in searching for local maxima ( @xmath184 and @xmath185 , step [ locmax ] ) , and the minimum contrast temperature ( @xmath190 , step [ decimate ] ) . without prior knowledge of substructure in the data ,
no brightness transform should be applied ( @xmath191 ) .
the minimum values for the parameters in the search for initial local maxima are set by the resolution of the data : @xmath192 and @xmath193 or else separations between local maxima can not be resolved .
finally , we use @xmath194 to prevent the noise in the data set from being recognized by the algorithm as legitimate structure and becoming the basis for the decomposition .
in the main section of the paper , we developed methods that measured physical properties of molecular emission without observational bias .
ideally , the decomposition algorithm used in the analysis would also be free of observational bias .
a good algorithm would , for example , decompose the same emission into similar structures regardless of the resolution and sensitivity of the observations .
to progress towards this goal , we set the parameters of the decomposition algorithm to have similar values in physical units ( pc , km s@xmath41 , k ) rather than data units ( beam widths , channels , @xmath76 ) . in the molecular ism , which has structure on a range of scales , the choice of the physical values for the algorithm parameters must be motivated by prior knowledge of the objects that we wish to identify . the choice of parameters for identifying gmcs would be different from the choice of parameters for identifying clumps and cores . as a population ,
gmcs have size scales of 10s of parsecs , line widths of several km s@xmath41 , and brightness temperatures of @xmath195 k ( e.g. * ? ? ?
in contrast , the clumpy substructure within clouds has a size scale of 1 pc , line widths of order 1 km s@xmath41 and can have brightness temperatures of 30 k or higher ( e.g. * ? ? ?
we select the parameters of the algorithm to find molecular clouds rather than the clumps within them .
in particular , we set the minimum separation between local maxima as @xmath196 pc and the velocity separation to be @xmath197 km s@xmath41 .
we also fix a contrast interval in antenna temperature of @xmath198 k rather than the data - driven value of @xmath199 . finally , we must account for the presence of bright molecular gas substructure within molecular clouds .
for example , orion a has a few distinct regions separated by @xmath200 pc with antenna temperatures in excess of 15 k @xcite .
these are associated with hot molecular gas around young stars like the trapezium cluster . in general , the typical kinetic temperature of gas in gmcs is @xmath201 k so brightness temperatures in excess of 10 k are usually associated with structure _ within _ molecular clouds not changes from cloud to cloud .
hence , to catalog clouds and not substructure , we must reduce the influence of the high @xmath202 data with the parameter @xmath183 .
we use @xmath203 k to maintain the full sensitivity for the gas separating gmcs ( which typically has @xmath204 k in the @xcite data ) while reducing the influence of brighter gas associated with a single gmc . using @xmath205
k has little influence on extragalactic data where beam deconvolution typically averages out the presence of bright substructure within the clouds .
we summarize our choices of parameters in table [ params ] when the parameters are motivated by the data ( data - based ) or by physical assumptions about the structures being extracted ( gmc physical priors ) .
the resolution or the sensitivity of a data set may be sufficiently poor that the physical parameters are unattainable in the data set ( e.g. @xmath206pc , @xmath207 k ) . in this case , the decomposition must be regarded with caution as it may not be directly comparable to other data sets .
the adopted values are appropriate for decomposing data sets of @xmath2co emission .
the values of @xmath183 and @xmath190 would be different for other tracers . to demonstrate that the proposed algorithm is actually an improvement on existing methods , we analyze trial data sets with known properties using this algorithm and compare the results to those from the gaussclumps and clumpfind algorithms to the trial data .
we use the clfind algorithm implemented in miriad @xcite and our own idl implementation of the @xcite gaussclumps algorithm , which we have compared to the standard algorithm operating on a real data set with satisfactory results .
we first examine how adept the three algorithms are at decomposing a pair of blended clouds .
we construct a series of data cubes with two unresolved model clouds separated in position by a variable distance .
the model clouds each have a peak signal - to - noise ratio of 10 with a gaussian line profile . for each value of the separation
, we decompose 100 data cubes with different realizations of the noise using the three algorithms , using data based choices for the algorithm parameters .
figure [ nclplot ] ( left ) shows the mean number of clouds recovered by each algorithm as a function of the separation ( we count clouds with peak signal - to - noise larger than 5 as `` recovered '' ) . as the trial clouds are moved farther apart , the typical number of clouds detected by each algorithm increases by 1 .
however , only the algorithm presented here consistently recovers a single cloud at low separations .
gaussclumps and clumpfind produce false clouds from the noise .
the jump in the number of clouds detected by the gaussclumps algorithm occurs where the separation equals the resolution , whereas both the current algorithm and clumpfind are able to distinguish the clouds only if their separation is over 1.5 times the resolution .
gaussclumps appears to be able to distinguish tight blends of clouds but is the most susceptible to noise . as a second test of the algorithms
, we compared the number of clouds that the algorithms recover from a single cloud with a circular top - hat brightness profile ( @xmath209 otherwise ) and peak signal - to - noise of 10 .
the size of the cloud is varied with respect to the resolution of the data set and 100 data sets for each value of @xmath210 are decomposed by each algorithm .
the non - gaussian brightness profile confounds all of the algorithms but to varying degrees .
clumpfind detects an increasing number of spurious clouds as the cloud grows , suggesting that the number of false clouds grows with the volume studied . despite the non - gaussian profile
, gaussclumps does surprisingly well with large sources . the current algorithm , however , does the best job of detecting a single source in the presence of noise . for analyzing data with a relatively low signal - to - noise
, we find the decomposition algorithm presented in this appendix should be favored for identifying clouds .
the orion molecular cloud is among the best studied of all gmcs , so it is an good place to compare the results of our methods to those of previous work . for this comparison ,
we use the data and results of the recent , uniform survey of the entire orion - monoceros region by @xcite .
we analyze their final data set using the methods in this paper ( with physical priors for the decomposition algorithm parameters ) and summarize the results in figure [ orion_fig ] , an integrated intensity map of the region . the decomposition results in 81 molecular clouds , most of which are associated with the galactic plane at the top of the figure .
the results of the algorithms decomposition of the region are good : major molecular clouds are identified as single entities in the decomposition .
the algorithm succeeds where other cloud identification schemes would face difficulty .
nearly all of the emission in the data is connected above a single isosurface so that simple contouring methods would identify the entire orion - monoceros region as a single molecular cloud , though more complicated algorithms @xcite may succeed .
clumpfind and gaussclumps would isolate individual peaks and decompose clouds into their substructure as was intended by their design ( clumpfind identifies 617 clumps in the same data ) .
the properties of the clouds agree well with the values published for a human decomposition of the emission .
the results of the analysis are given in table [ orion_props ] .
for comparison , we adopt the distances of @xcite .
we compare the results of the algorithm to their results , after scaling their mass up by 10% to account for a difference in the adopted co - to - h@xmath1 conversion factors . in several cases
the masses agree quite well ( to within 5% , orion a , mon r2 , scissors ) , while other features show @xmath211 differences .
these systematic discrepancies arise from differences in how emission is assigned into structures .
for example , the algorithm only identifies the central region of orion b as a molecular cloud ; it does not include emission near the location of orion east that is nominally part of the orion b cloud because the assignment of this emission is contested with neighboring clouds .
similarly , the algorithm characterizes the northern filament region as five distinct clouds rather than the single large cloud that a human decomposition produced . despite these differences ,
the results of the algorithm are reassuring in most cases , the well - known molecular clouds are identified as single clouds and there is good agreement between their derived physical properties and the results of previous studies .
lcccccc orion a & 480 & 12 . &
@xmath212 & @xmath213 & @xmath214 & @xmath215 + orion b & 500 & 9.1 & @xmath216 & @xmath217 & @xmath218 & @xmath219 + orion east & 120 & 0.013 & @xmath220 & @xmath221 & & + mon r2 & 800 & 12 . & @xmath222 & @xmath223 & @xmath224 & @xmath225 + crossbones & 470 & 1.9 & @xmath226 & @xmath227 & @xmath228 & @xmath229 + northern filament & 390 & 1.9 & @xmath230 & @xmath231 & @xmath232 & @xmath233 + as a final test of the decomposition and analysis algorithm , we apply the algorithm to simulated observations of the orion - monoceros region with beam sizes and signal - to - noise values typical of extragalactic gmc observations .
we simulate observations by convolving the data set to the desired resolution and adding a convolved data cube of noise which is scaled up to give the desired peak signal - to - noise ratio in the data .
we perform this for a peak signal - to - noise ( @xmath234 ) values of 10 and 30 combined with beam sizes of 10 , 20 and 50 pc .
the results of the decompositions are displayed in figure [ decompdemo ] . at coarse resolution and low sensitivity
, the fainter clouds are undetectable .
however , the algorithm identifies each of orion a , orion b , mon r2 and the northern filament in at least one of the trial data sets . at high resolution
( 10 pc ) and peak signal - to - noise ( @xmath235 ) , the algorithm successfully identifies the clouds with properties that are consistent , within the uncertainties determined by the algorithm , with the masses of the clouds in table [ orion_props ] .
the systematic effects of poor resolution manifest themselves for beam sizes @xmath236 pc where the decomposition of blended emission results in @xmath237 variations in the properties of the clouds relative to those found in the original data set for both high and low @xmath234 . for a beam size of 20 pc ,
the clouds are not resolved since this is over twice the size of the clouds along their minor axes .
finally , at 50 pc resolution , mon r2 is not found by the algorithm and high @xmath234 is required to distinguish orion a and orion b. there is a variation of 100% ( 0.3 dex ) in the derived physical parameters for the clouds ; this variation is reflected in the estimates of the uncertainties . as expected based on the 20 pc resolution data ,
the clouds are not spatially resolved for a beam size of 50 pc .
an accurate decomposition of emission into typical galactic gmcs requires a beam size @xmath238 pc though only a modest sensitivity is required : @xmath239 . | we review methods for measuring the sizes , line widths , and luminosities of giant molecular clouds ( gmcs ) in molecular - line data cubes with low resolution and sensitivity .
we find that moment methods are robust and sensitive making full use of both position and intensity information and we recommend a standard method to measure the position angle , major and minor axis sizes , line width , and luminosity using moment methods . without corrections for the effects of beam convolution and sensitivity to gmc properties , the resulting properties may be severely biased .
this is particularly true for extragalactic observations , where resolution and sensitivity effects often bias measured values by 40% or more .
we correct for finite spatial and spectral resolutions with a simple deconvolution and we correct for sensitivity biases by extrapolating properties of a gmc to those we would expect to measure with perfect sensitivity ( i.e. the 0 k isosurface ) .
the resulting method recovers the properties of a gmc to within 10% over a large range of resolutions and sensitivities , provided the clouds are marginally resolved with a peak signal - to - noise ratio greater than 10 .
we note that interferometers systematically underestimate cloud properties , particularly the flux from a cloud .
the degree of bias depends on the sensitivity of the observations and the @xmath0 coverage of the observations . in the appendix to the paper we present a conservative , new decomposition algorithm for identifying gmcs in molecular - line observations
this algorithm treats the data in physical rather than observational units ( i.e. parsecs rather than beams or arcseconds ) , does not produce spurious clouds in the presence of noise , and is sensitive to a range of morphologies . as a result , the output of this decomposition should be directly comparable among disparate data sets . |
the essential physics of the hole - doped high-@xmath14 cuprates can be described as that of a doped mott insulator , which exhibits several phases varying from antiferromagnetic insulator to metallic superconductor .
the elucidation of the basic physics of these various complicated phases by studying the elementary excitations could yield information essential for uncovering the mechanism of high-@xmath14 superconductivity .
experimental techniques that can probe the charge excitations have been dramatically developed in recent years .
in particular , resonant inelastic x - ray scattering ( rixs ) in the hard x - ray regime has attracted much attention due to its ability to probe the momentum dependence of the charge excitations .
this technique was initially applied to nio ; @xcite in that case a large enhancement of the inelastic signal with incident photon energy was observed near the ni k edge .
more recently this technique has been applied to various cuprates including high-@xmath14 materials @xcite and other highly - correlated electron systems .
@xcite a recent rixs study near the cu k edge in undoped la@xmath15cuo@xmath16 ( lco ) has revealed three features in its charge excitation spectrum .
@xcite two features , labeled a and b , appear at energy transfers @xmath17 ev and 3.9 ev , respectively ; these features show the largest intensity enhancement at an incident photon energy of @xmath18 kev .
the other feature , labeled c , appears at @xmath19 ev , and resonates at the higher incident energy of @xmath8 kev .
since the resonance energy of the a and b features corresponds to the absorption feature associated with a well - screened state , @xcite these two have been attributed to the charge - transfer ( ct ) excitations from the o @xmath20 to the cu @xmath21 upper hubbard band . on the other hand ,
the resonance energy of feature c is associated with a poorly screened state , and has been attributed to a molecular orbital ( mo ) excitation that involves a transition between bonding and anti - bonding states of the cu @xmath21 and o @xmath20 orbitals . @xcite recently , there has been a number of rixs studies of doped cuprate superconductors .
@xcite for the la@xmath0sr@xmath1cuo@xmath2 ( lsco ) system , kim _ et al . _
@xcite showed that the two features of the ct excitations in lco are also observed for @xmath22 , but only a single feature is observed at @xmath23 ev for @xmath13 .
in addition , a continuum - like excitation appears for @xmath13 below the charge transfer excitation energy ( @xmath24 ev ) at the zone center ( 0 0 ) and at the zone boundary @xmath10 .
@xcite the zone center result is consistent with optical measurements which show a transfer of the spectral weight from the ct excitation to lower energies .
@xcite for the mo excitation , the excitation energy increases with doping .
@xcite this was attributed to the decreased cu - o distance in doped samples , which leads to an increased energy splitting between the bonding and antibonding molecular - orbitals due to the increased @xmath25 hybridization .
although measurements have been carried out for underdoped and optimally doped samples , the overdoped region has not yet been studied with rixs .
such studies should yield important insights into the nature of the charge excitations observed with the rixs technique .
we note that thus far lsco appears to be the only high-@xmath14 system where large , high quality heavily overdoped crystals are available . in this paper , we report rixs results for overdoped lsco with @xmath26 and 0.30 ; the primary purpose of this study is to elucidate the evolution of the charge excitations in the overdoped region .
we find that the ct and mo excitations in the overdoped compounds exhibit resonantly enhanced intensity over a broader range of incident energies than those of an optimally - doped and undoped ( parent ) compound .
we also observe that the near - edge x - ray absorption spectrum exhibits a feature associated with the well - screened final state , which is well - defined similar to its counterpart in underdoped samples . on the other hand
, we do not find a well - defined peak corresponding to the poorly - screened final state .
profiles measured utilizing a high energy resolution spectrometer configuration indicate that the continuum excitation exists below 3 ev at @xmath11 in addition to the @xmath12 and @xmath10 positions . on top of the continuum intensity , we observed an additional excitation at @xmath27 ev , which exhibits resonantly enhanced intensity around @xmath28 kev .
the paper is organized as follows .
section ii describes the experimental details .
the ct and mo excitations measured with a conventional rixs setup are presented sec .
iii , while a detailed study of the continuum intensity below 3 ev using the high resolution setup is reported in sec .
finally the results are summarized in sec . v.
since the results for the @xmath26 and @xmath4 crystals are quantitatively similar , we do not distinguish between these two samples in our discussion unless there is a special need .
single crystals of lsco with @xmath26 and @xmath4 for the rixs measurements were grown by the traveling - solvent floating - zone ( tsfz ) method in the same manner as those used for neutron - scattering experiments .
@xcite the grown crystal rods have a typical size of 6 mm in diameter and 80 mm in length ; they were cut into pieces 5 mm @xmath29 4 mm @xmath29 1 mm along the tetragonal @xmath30 , @xmath31 and @xmath32-directions , respectively , and annealed at 850 @xmath33c for 12 hours in flowing oxygen to compensate for oxygen deficiencies .
the temperature dependence of the superconducting shielding fraction determined by magnetization measurements shows that the @xmath26 sample has @xmath34 k whereas the @xmath35 sample shows no bulk superconductivity down to 2 k. the @xmath36 surfaces , which the x - ray beams in the rixs experiments are incident upon , were polished using a 1 @xmath37 m grit size .
the rixs experiments were performed at the undulator beam line 9idb at the advanced photon source at argonne national laboratory . for the measurements of the ct and mo excitations , we used the si ( 333 ) reflection from the double bounce main monochromator , together with a spherical ge(733 ) analyzer with 1 m radius of curvature .
this setup gives an instrumental energy resolution of 0.4 ev full - width at half maximum ( fwhm ) .
for the measurement of the continuum intensity below @xmath38 ev , a si(444 ) channel - cut secondary monochromator was used in addition to a si(111 ) primary monochromator . a spherical ge(733 )
analyzer with a 2 m radius of curvature was also utilized .
this spectrometer configuration yielded a net energy resolution of 0.13 ev ( fwhm ) .
more technical details of the finer resolution setup has been reported in ref . .
the energy resolution has been measured by scanning an incoherent background signal along the energy direction . for the finer resolution setup
, we cooled the samples down to 9 k by a closed cycle refrigerator to reduce a possible phonon contribution to the elastic background intensity .
however we found that the elastic peak width along the energy direction does not show clear temperature dependence .
thus , we conclude no phonon effect on the elastic peak . for all measurements , the scattering plane was vertical and the polarization of the incident x - ray was fixed along the @xmath31-axis .
typical scans were performed by counting for 2 minutes at each data point , and the counts were normalized to monitor counts corresponding to 1 s. in our analysis , we have used the following background subtraction and normalization method .
the largest source of background arises from the contribution of the elastic ( @xmath39 ) intensity .
the energy gain ( @xmath40 ) side spectrum is expected to represent purely background intensity , since the detailed balance factor at this temperature and energy scale vanishes .
therefore , the background subtracted intensity can be written as @xmath41 @xmath42 , where @xmath43 is the raw intensity , assuming that the elastic background is symmetric in @xmath44 . unless noted otherwise , we used the intensity of the cu k@xmath45 emission line as a normalization factor , since this emission intensity is expected to be proportional to the sample volume probed . at last , throughout this paper
, we use tetragonal notation with @xmath46 along the cu - o - cu direction , corresponding to @xmath47 @xmath48 .
first , we present the experimental results for the main spectral features , which correspond to the ct and the mo excitations .
figure 1 shows the evolution of the rixs spectra as a function of the incident energy @xmath7 near the cu k edge at the positions @xmath49 corresponding to a position near the zone center @xmath50 and @xmath51 corresponding to the zone boundary position @xmath52 .
unfortunately , the zone center @xmath53 position is contaminated at @xmath54 presumably by the tails of the bragg peaks at ( 3 , 0 , 1 ) and ( 3 , 0 , @xmath55 ) , so that the data have been obtained at positions slightly away from the zone center . however
, a relatively high elastic intensity remains at @xmath56 , producing a broader tail than that at @xmath10 .
nevertheless , a similar incident energy dependence of the rixs spectra is observed at both positions with a ct excitation around 4 ev and the mo excitation around 8 ev , exhibiting resonance enhancements at @xmath57 kev and @xmath58 kev , respectively .
= 5.5 in to study any possible change of the rixs spectra as a function of momentum and doping , the spectral weight has been calculated by simply summing up the background subtracted and normalized intensity between 2 ev and 12 ev .
spectral weights so obtained for the data at @xmath56 and @xmath10 are shown in fig .
2(a ) by circles and squares , respectively .
although the ( near ) zone center data show smaller spectral weights , the @xmath7 dependence appears to be very similar at both positions .
this is more clearly seen in fig .
2(b ) which shows the @xmath7 dependences of the spectral weights normalized to their maximum values .
thus , although the excitation energies are slightly different at different @xmath59 positions as shown in fig . 1 , the resonance behavior is similar . = 3.3 in
we observe that the spectral weight does not show a well - defined peak at @xmath60 kev , which is the mo excitation resonance energy , whereas the spectral weight has a peak at @xmath61 kev , the ct excitation resonance energy .
we also found that the x - ray absorption spectrum shown in fig .
2(c ) , which is measured by monitoring the intensity of cu k@xmath45 line at @xmath62 kev , does not exhibit a well - defined peak associated with the so - called poorly - screened state around @xmath6 kev .
this result should be contrasted with the results in the undoped samples .
figure 2(d ) shows the spectral weight of la@xmath15cuo@xmath16 calculated in the same manner using the data reported in ref . .
it is clearly seen that the spectral weight of la@xmath15cuo@xmath16 has two well - defined peaks at the energies where the ct and the mo excitations resonate .
consistent with this , the absorption spectrum of la@xmath15cuo@xmath16 also has the well - defined peaks for both well - screened and poorly screened states as shown in fig .
@xcite since such a two - peak structure has also been observed for the @xmath13 sample , @xcite the disappearance of the higher energy peak in the absorption spectrum and the broad nature of the rixs spectral weight distribution seem to be characteristic of overdoped metallic samples .
= 3.3 in the dispersions of the ct and the mo excitations have been studied by collecting the rixs spectra at fixed @xmath63 kev and 8.998 kev at several @xmath64-positions between @xmath65 and @xmath66 , and between @xmath65 and @xmath67 .
the data have been fitted by a lorentzian function with a sloping background to determine the excitation energies .
the dispersions so obtained are shown in fig .
the ct excitation appears at 3.7 ev at the zone center and reaches 4.2 ev at @xmath10 and 5 ev at @xmath11 .
this dispersion is qualitatively similar to that observed in the @xmath13 sample .
@xcite the actual profiles , however , show systematic changes with increasing hole concentration .
the left panel of fig .
4 indicates profiles of the ct peak for @xmath68 , @xmath69 ( from refs . and ) and @xmath4 . to compare the ct peak itself ,
the intensity is normalized to the ct peak intensity .
it may be clearly seen that the ct peak shifts its weight to higher energy with increasing doping at all @xmath64 positions .
this is consistent with an increase of the ct gap with doping predicted by tsutsui _
et al . _
@xcite and agrees quantitatively with their numerical calculations .
@xcite the large value of this shift ( @xmath70 ev ) is somewhat surprising , given that the chemical potential shift due to the hole doping observed in photoemission experiments is less than 0.5 ev.@xcite however , the upper hubbard band shifts to higher energy as holes are doped .
@xcite this combined effect of the chemical potential shift and the shift of the unoccupied band seems to contribute to the large overall shift of the ct feature .
it is also interesting to note that the large shift of the ct peak energy in the overdoped regime may imply a large overlap of the o @xmath20 and the lower hubbard cu @xmath21 bands , resulting in the existence of partially doped - holes with cu @xmath21 character in the overdoped regime .
this in turn would suppress magnetic correlations as reported by recent neutron scattering measurements of the overdoped samples .
@xcite however , detailed numerical calculations are necessary to draw any firm conclusions .
= 3.3 in the nature of the mo excitation also appears to differ from that in the @xmath13 sample .
@xcite although the zone boundary has a lower excitation energy than the zone center as is the case for @xmath13 , the dispersion of the overdoped sample is flat in the large @xmath64 region as shown in fig .
3 . as discussed in ref .
, the energy and the dispersion of the mo excitation increases as the cu - o distance decreases , and concomitantly the hybridization between the cu @xmath71 and the o @xmath72 orbitals becomes stronger .
the excitation energy of 7.6 ev at @xmath73 is very close to that of the @xmath13 sample which has the same cu - o distance as the @xmath26 sample , and , moreover , if we define the dispersion as the difference between the energies at @xmath10 and @xmath74 , the overall dispersion of @xmath75 ev of the present sample is similar to that of the @xmath13 sample ( @xmath76 ev ) .
the plots on the left side of fig . 4 show clear evidence for the appearance of a low energy excitation below the ct peak for the doped samples . although the excitation spectrum around 1 ev for the undoped sample decreases to the background level ( intensity at @xmath77 ev ) , the intensity at 1 ev remains well above the background in the doped samples . as reviewed in sec .
1 , kim _ et al . _
@xcite have observed that the continuum intensity at low energy ( @xmath78 ev ) exists at ( 0 0 ) and at @xmath10 , but not at @xmath11 . to test this in the overdoped sample ,
we have compared the rixs spectra for @xmath26 at @xmath10 and @xmath79 .
the top - right panel of fig .
4 shows the as - measured data of rixs spectra taken with @xmath80 kev including the peak around 19 ev due to the cu k@xmath45 fluorescence emission .
the bottom - right panel shows the same spectrum after normalization using the cu k@xmath45 intensity . both spectra exhibit a constant intensity between 1 and 3 ev as shown by the shaded area ,
although the continuum intensity at the @xmath11 position appears to be smaller than that at @xmath10 .
thus , the data are suggestive of the existence of the continuum excitation even at @xmath11 in the overdoped samples .
= 3.3 in = 3.3 in we have also measured the rixs spectra for @xmath35 at 9 k utilizing the high resolution instrumentation ( @xmath81 ev ) which enables us to study the detailed structure of the continuum excitation . figure 5 shows the evolution of the rixs spectra as a function of @xmath7 between 8.992 and 8.994 kev at @xmath10 .
it appears that there is a small peak on top of the continuum intensity at @xmath82 ev resonating in the range of @xmath83 kev .
this may be more clearly seen in the right panel of the figure , where the background subtracted spectra are plotted on a different scale to emphasize the continuum region .
the data below 1 ev have a large ambiguity due to the tail of the elastic peak , while the increase of @xmath84 above 2.5 ev is due to the ct peak .
a peak at 1.5 ev is , nevertheless , clear in the range of @xmath85 kev .
similar features have been found at the @xmath11 position .
figure 6 shows plots analogous to those in fig . 5 for @xmath11 .
in the right panel , the profile at @xmath86 kev exhibits a peak of 0.5 counts / s appearing around @xmath87 ev on top of the continuum intensity of 1 count / s , whereas the data at @xmath88 and @xmath89 kev show only a flat continuum intensity .
as @xmath7 increases to 8.994 kev , the peak appears to become broader and larger , and seems to shift to higher energy .
we have studied the @xmath64-dependence of this excitation around @xmath11 with fixed @xmath7 at 8.993 ev .
figure 7 shows the background - subtracted and normalized rixs intensity @xmath90 at @xmath91 ev along the @xmath92 direction which is equivalent to the trajectory in a reduced momentum space shown by an arrow in the inset .
it is evident that a peak develops at @xmath79 with increasing energy transfer and it reaches an intensity of @xmath93 counts / sec . at 1.8 ev on top of the @xmath64-independent continuum intensity of 1 count / sec ,
consistent with the data for @xmath86 ev in fig .
it is open question what the origin of this excitation is . from a consideration of the excitation energy of this feature ,
a possible origin is the local @xmath94-excitation which has been observed around 1.5 ev .
@xcite however more detailed measurements are necessary to draw definitive conclusions .
we have studied heavily overdoped lsco with @xmath26 and 0.30 using the resonant inelastic x - ray scattering technique near the cu k edge .
we observe several features of the charge - transfer and molecular - orbital excitations in these materials .
first , the resonance behavior in this overdoped sample appears to be broader in incident photon energy than that of the undoped parent material .
furthermore , in the absorption spectra of the overdoped samples , the peak corresponding to the poorly - screened state is damped .
this presumably is related to the metallic nature of the overdoped samples , and the better screening due to mobile charge carriers .
second , we found that the ct peak energy increases with doping @xmath95 .
this agrees with the prediction of an increase of the ct gap energy with increasing @xmath95 and specifically , the numerical calculations of tsutsui _ et al . _
@xcite and tsutsui .
@xcite our study of the low energy excitations has shown that the continuum intensity appears at all momentum transfers for the overdoped samples . from detailed measurements utilizing the high energy resolution of @xmath96 ev
, we observed an additional excitation on the continuum intensity at @xmath11 and @xmath10 , which shows resonantly enhanced intensity at @xmath28 kev .
it will be important to study the dispersion of this feature , as well as to determine whether or not this excitation exists in the samples with lower doping to elucidate the origin of this excitation .
the authors thank d. casa , d. ellis , j. p. hill , k. tsutsui , c. t. venkataraman , and k. yamada for invaluable discussions .
the work at the university of toronto is part of the canadian institute of advanced research and supported by natural science and engineering research council of canada .
the work at bnl is supported by the u. s. department of energy , division of material science , under contract no .
de - ac02 - 98ch10886 .
the aps is supported by the u. s. department of energy , basic energy sciences , office of science , under contract no .
w-31 - 109-eng-38 . | resonant inelastic x - ray scattering ( rixs ) at the copper k absorption edge has been performed for heavily overdoped samples of la@xmath0sr@xmath1cuo@xmath2 with @xmath3 and @xmath4 . we have observed the charge transfer and molecular - orbital excitations which exhibit resonances at incident energies of @xmath5 and @xmath6 kev , respectively . from a comparison with previous results on undoped and optimally - doped samples ,
we determine that the charge - transfer excitation energy increases monotonically as doping increases .
in addition , the @xmath7-dependences of the rixs spectral weight and absorption spectrum exhibit no clear peak at @xmath8 kev in contrast to results in the underdoped samples .
the low - energy ( @xmath9 ev ) continuum excitation intensity has been studied utilizing the high energy resolution of 0.13 ev ( fwhm ) .
a comparison of the rixs profiles at @xmath10 and @xmath11 indicates that the continuum intensity exists even at @xmath11 in the overdoped samples , whereas it has been reported only at @xmath12 and @xmath10 for the @xmath13 sample .
furthermore , we also found an additional excitation on top of the continuum intensity at the @xmath11 and @xmath10 positions . |
it has been proposed long ago ( for recent reviews see @xcite ) that the non - trivial dynamics due to the attractive interaction of pions with quarks ( or nucleons ) allows for the existence of non - uniform phases in the low temperature and high chemical potential regime of the qcd phase diagram . as the sign problem affects lattice qcd at non - zero baryon density , the use of alternative approaches such as low energy models of the nambu jona - lasinio ( njl ) @xcite type , became a useful exploratory tool in the studies of strongly interacting matter . in this
talk we outline some results reported in detail in @xcite , pertaining to the study of the njl model with the t hooft determinant @xcite applied to the case where the light quark chiral condensates assume the shape of the chiral wave .
we work at zero temperature and in the chiral limit for the @xmath0 and @xmath1 sector , while the @xmath2 quark has a non - vanishing current mass .
the chiral wave _ ansatz _ proposed in @xcite , with the corresponding quark orbitals and energy levels , forms a self consistent solution of the euler - lagrange equations of the model in the chiral limit .
they have the form @xmath3 where @xmath4 is the pauli matrix acting in the isospin space , @xmath5 is the quark dynamical mass , @xmath6 denotes the momentum of the quark , and @xmath7 is the wave vector of the condensate modulation .
we choose the @xmath8-axis to coincide with @xmath7 ( note that the @xmath9 branch has a lower energy than @xmath10 ) . for the @xmath2 quark a uniform condensate background is considered . the application of techniques of ref .
@xcite yields the thermodynamic potential of the model in the mean field approximation as @xmath11 where @xmath12 ( @xmath13 ) are twice the quark condensates .
the integrals @xmath14 stem from the fermionic path integral over the quark bilinears which appear after bosonization , while @xmath15 corresponds to the stationary phase contribution to the integration over the auxiliary bosonic fields .
the njl coupling strength is @xmath16 , while @xmath17 is the ozi - violating t hooft determinant coupling . from the value evaluated at the dynamical masses
@xmath5 , a subtraction of its value evaluated at @xmath18 is made @xcite ( denoted by the @xmath19 symbol in eq .
( [ omega ] ) ) .
we use a regularized kernel corresponding to two pauli - villars subtractions in the integrand @xcite , namely @xmath20 .
the dirac and fermi sea contributions , @xmath21 and @xmath22 , can be written explicitly as @xmath23 where @xmath24 .
the @xmath25 notation refers to the subtraction of the same quantity evaluated for @xmath18 and @xmath26 , which is done to set the zero of the potential at a uniform gas of massless quarks .
the superscript @xmath27 in the definition of @xmath28 refers to the energy branch , whereas the subscript refers to the sign in front of the chemical potential in the exponent .
the @xmath29 term is needed for thermodynamic consistency @xcite .
the minimization of the @xmath30 with respect to @xmath5 and @xmath31 has to be done self - consistently with the resolution of the following stationary phase equations ( where @xmath32 stands for the current masses of the quarks ) : @xmath33
in our study we chose the parameters to reproduce a reasonable value for the vacuum dynamical mass of the light quarks ( @xmath34 ) .
the current masses were set to @xmath35 , @xmath36 , the remaining three parameters of the model ( @xmath16 , @xmath17 and @xmath37 ) can then reduced to two : the t hooft coupling strength , @xmath17 , and the dimensionless curvature , @xmath38 . in the chiral limit
@xmath39 is the critical value above which dynamical chiral symmetry breaking occurs as a crossover for @xmath40 and as a first order transition for higher values . with no ozi - violating term ,
the light and strange sectors are decoupled : two independent first order transitions occur for the light and strange sector , with the latter shifted to higher @xmath41 due to the current @xmath2-quark mass ( both occur at values of @xmath41 close to the vacuum dynamical mass @xmath42 ) . at high enough @xmath41 ,
the energetically favourable solution is always a modulated one , with a non - vanishing @xmath31 the non - uniform phase develops .
asymptotically , this solution goes to @xmath43 and becomes degenerate with the trivial one . above @xmath44 ,
a window for energetically favourable finite-@xmath31 solutions appears in the vicinity of the first - order transition encompassing it ; the transition is not to a vanishing condensate , but @xmath45 .
the window ends with the light condensate going continuously to zero . for @xmath46 this
windows extends and merges with the one at higher chemical potentials .
-dependence of @xmath47 , @xmath48 and @xmath31 .
each row corresponds to a different value of the t hooft coupling strength ( @xmath49=\mathrm{gev}^5 $ ] ) .
thicker lines denote the finite-@xmath31 solutions . in the grey regions the value of @xmath31 is undetermined , since @xmath50 . ] turning on flavour mixing couples the @xmath51 and strange sectors . at fixed curvature
, we find several different scenarios depending on the value of @xmath17 , as seen in fig .
[ grafpainelnjlhsu3 ] where we show the @xmath52 case . for @xmath53
a new solution branch appears , with a shark - fin shape for @xmath47 in the vicinity of @xmath54 ( see first row of fig .
[ grafpainelnjlhsu3 ] ) .
there are therefore three intervals of @xmath41 where an energetically favourable finite-@xmath31 solution exists .
they are delimited by two first order transitions and three - crossovers ( the crossovers correspond to the disappearance / emergence of a non - vanishing @xmath47 ) .
the two first order transitions occur slightly below ( for the one taking place near @xmath55 ) and slightly above ( near @xmath56 ) , thus excluding the occurrence of the @xmath26 transitions .
a zoom of the behaviour of the chiral condensates near the transitions can be seen in fig .
[ solhlsnjlhtau14k500zoom ] .
for @xmath57 the first two @xmath41 windows with finite-@xmath31 solutions merge , resulting in the disappearance of the corresponding crossover transitions , as can be seen in the second row of fig .
[ grafpainelnjlhsu3 ] . with @xmath58 the last transition to a vanishing @xmath31 solution
does not occur and is substituted by a first - order transition between the two phases with finite @xmath31 ( third row of fig . [ grafpainelnjlhsu3 ] and fig . [ grafsolnjlhtau14k1000 ] ) .
the dependence on @xmath17 of these critical chemical potentials can be seen in fig .
[ mucritvsk ] .
this work has been supported by the centro de fsica computacional of the university of coimbra , fundao para a cincia e tecnologia , project : cern / fp/116334/2010 and the grant sfrh / bpd/63070/2009/. this research is part of the eu research infrastructure integrating activity study of strongly interacting matter ( hadronphysics3 ) under the 7th framework programme of eu , grant agreement no .
wb acknowledges the support of the polish national science centre , grants dec-2011/01/b / st2/03915 and dec-2012/06/a / st2/00390 .
99 w. broniowski , acta phys.pol.supp .
* 4 * ( 2012 ) 631 .
m. buballa , s. carignano , arxiv:1406.1367 [ hep - ph ] .
y. nambu , phys .
lett . * 4 * ( 1960 ) 380 ; phys . rev . *
122 * , ( 1961 ) 345 ; * 124 * ( 1961 ) 246 ; v. g. vaks and a. i. larkin , sov .
jetp * 13 * ( 1961 ) 192 .
j. moreira et al . , phys .
d * 89 * ( 2014 ) 3 , 036009 .
g. t hooft , phys . rev .
d * 14 * ( 1976 ) 3432 .
f. dautry , e. nyman , nucl .
phys . a * 319 * ( 1979 ) 323 .
a. a. osipov , b. hiller , eur .
j. c * 35 * ( 2004 ) 223 . b. hiller et al . , phys .
d * 81 * ( 2010 ) 116005 .
a. a. osipov , m. k. volkov , sov .
* 41*:3 ( 1985 ) 500 . | the possible existence of non - uniform phases in cold dense quark matter in the light quark sector ( @xmath0 , @xmath1 , @xmath2 ) is addressed within the nambu - jona lasinio model extended to include the flavour - mixing t hooft determinant .
the effect of changes in the coupling strengths of the model is discussed .
it is seen that the inclusion of the strange sector catalyses the appearance of the non - uniform phases , extending the domain for their existence . |
magnetically dominated outflows from stars and accretion discs , are central to the evolution and properties of these objects . because they can efficiently extract rotational energy , and convert it into wind kinetic energy , magnetic outflows are ubiquitous in powering a variety of astrophysical systems .
rotationally powered pulsars are among the best example of relativistic outflows produced by a fast compact rotator .
the high magnetic field at the surface ( @xmath1 g ) and short rotation period ( @xmath2 s ) , induce an electric field at the surface itself which is strong enough to lift electrons and ions and accelerate them to relativistic speeds . these accelerated particles ,
as they move along curved magnetic field lines , produce the pulsed emission that is observed . in the magnetosphere of the neutron star
, these particles produce pairs . if the charge density exceeds the goldreich - julian value then conditions can be modeled in term of ideal rmhd .
the typical lorentz factor of the plasma in the pulsar magnetosphere , within the light cylinder , is @xmath3 .
however from observations of the interaction of the pulsar wind with the environment , and especially the surrounding snr , we know that far away from the source the wind is ultrarelativistic ( @xmath4 ) and weakly magnetized @xcite .
the basic idea is that the wind is magneto - centrifugally accelerated , and that the high lorentz factor is achieved by almost completely converting magnetic energy into kinetic energy . in this sense , pulsars constitute the prototype of relativistic accelerators and the basic physics of magneto - centrifugally driven relativistic winds apply to a wide variety of astrophysical objects from accretion disks to grbs and microquasars . in all of these systems a plasma
is assumed at the base with small outflow velocity ( compared to the asymptotic velocity ) , embedded in a strong magnetic field .
magnetorotation then accelerates the plasma and converts the magnetic energy into relativistic motion .
the study of winds from neutron stars can give an useful insight into the general problem of relativistic outflows .
in the simple 1d radial model for relativistic winds from compact rotators by @xcite , solutions can be parametrized in term of @xmath5 , where @xmath6 is the rotation rate , @xmath7 the magnetic flux and @xmath8 the mass flux .
@xmath9 is the maximum lorentz factor the wind can achieve if all magnetic energy is converted into kinetic energy , and at the base of a pulsar it is @xmath10 .
the results by @xcite show that , due to relativistic effects , the asymptotic lorentz factor tends to @xmath11 , and conversion of magnetic to kinetic energy is minimal .
this is known as the @xmath12 problem @xcite .
the low efficiency of the michel model is due to the assumption of strictly radial flow .
if this assumption is relaxed and the flow is allowed to diverge more than radially from the light cylinder up to the snr , then acceleration is more efficient @xcite .
it is known from the theory of classical winds that the presence of a magnetic field can cause collimation of the flow along the rotation axis , and divergence along the equator , so in principle one might expect magnetic collimation to be also responsible for the acceleration of the wind .
however when the flow speed reaches values close to @xmath0 the coulomb term in the electromagnetic force can not be neglected ( as in non relativistic mhd ) .
the strength of the induced electric field is comparable with the magnetic field and the coulomb term balances the magnetic hoop stress .
solutions of the relativistic grad - shafranov equation for a monopolar field in the trans - fast regime @xcite have shown that acceleration is too slow and the asymptotic results of the michel model apply also in 2d axisymmetric geometry , meaning that the flow remain radial .
all simulations were performed using the code developed by @xcite .
it is a high - order finite difference uct scheme , that solves the equation of relativistic mhd in general curvilinear coordinates .
the scheme is based on an approximate hll solver , that avoids wave decomposition .
the first axisymmetric numerical simulations of the acceleration of a relativistic wind were presented by @xcite , where the case of a zero pressure plasma in a monopolar magnetic field , with uniform injection conditions at the base of the star was investigated .
results show that when @xmath13 exceeds 5 collimation and acceleration of the wind become marginal .
our recent investigation @xcite has focused on the study of the acceleration of relativistic wind in the case the pressure is small compared with the rest mass energy density but not negligible .
it has been proposed that soon after the birth of a proto neutron star ( ns ) , neutrino heating might rise the sound speed in the neutron star atmosphere to value @xmath14 @xcite .
the heating might be able to drive a wind against the gravitational attraction of the ns .
such wind might be accelerated to high @xmath13 by the magnetic field .
if acceleration is efficient , the wind can reaches @xmath15 , enough to trigger a grb .
we have investigated the axisymmetric wind from rotating neutron star in schwartzschield metric .
pressure , density and the radial component of the magnetic field are fixed at the ns surface ( @xmath16 ) .
a 2d spherical @xmath17 grid was used , with 100 uniform cells in @xmath18 , and a logarithmic scale with 200 cells per decade in @xmath19 .
solutions in the 2d case can be parametrized by using the value @xmath9 averaged over all stream lines .
results in the case of a monopolar magnetic field show that for @xmath20 the flow appears to be radial , the terminal @xmath13 is @xmath21 , pointing to inefficient acceleration .
moreover the energy and angular momentum fluxes rapidly approach the force - free value . despite the fact that in the close magnetosphere @xmath22 , asymptotic results agree with the flow structure found by @xcite .
energy flux scales as @xmath23 , and @xmath24 , in agreement with the exact monopole solution .
results suggest that a rapidly rotating magnetar will not be able to simultaneously produce a collimated and relativistic outflow .
relativistic velocities can be reached only for @xmath25 , and in this case most of the energy would be concentrated in the equatorial plane instead that in a jet .
, axes are in units of @xmath16
. left : contours represent magnetic surfaces , colors represent @xmath26 , the dashed line is the alfvenic surface , coincident with the light cylinder , the dotted line is the fast - magnetosonic surface .
right : lorentz factor in the wind . from @xcite ,
title="fig : " ] , axes are in units of @xmath16
. left : contours represent magnetic surfaces , colors represent @xmath26 , the dashed line is the alfvenic surface , coincident with the light cylinder , the dotted line is the fast - magnetosonic surface .
right : lorentz factor in the wind . from @xcite , title="fig : " ] we computed solutions also in the more realistic cases of a dipolar magnetic field at the surface of the ns . due to stability issue in numerical rmhd it was not possible to study cases with @xmath27 .
numerical resistivity on current sheets at the edge of the closed zone , causes the plasma to rotate faster that corotation , while at the equator reconnection takes place and plasmoids are formed . to obtain steady solutions one
is forced to suppress resistivity , thus enhancing the instability of the code . moreover , in the case of a dipole the magnetic and flow surface are not aligned with the grid , and in this case inversion from conservative to primitive variables is not stable for @xmath28 . despite the presence of a closed zone , and the fact that within the light cylinder the field structure is close to a dipole
, we found that solutions can again be parmaterized in term of @xmath9 , now defined as an average done only on open streamlines . in this case
we find that energy and angular momentum losses are the same of a monopolar case with the same @xmath9 . again as soon as @xmath29 solutions approach the force - free limit @xcite , outside the light cylinder , and for @xmath30 the asymptotic structure resembles the exact monopole , with energy flux peaking at the equator and a higher equatorial lorentz factor .
similar results are found also by @xcite .
contrary to expectations @xcite we also find that the location of the y - point is inside the light cylinder .
this is due to the presence of a finite pressure plasma at the base of the star . in this case
the extent of the closed zone is limited by the pressure equilibrium at the y - point .
this implies that even for large @xmath9 the asymptotic wind structure is close to the force - free monopole solution , the spindown time can be shorter , and the total energy losses larger that what is inferred by applying the dipole formula .
, axes are in units of @xmath16
. left : contours represent magnetic surfaces , colors represent @xmath26 , the dashed line is the alfvenic surface , coincident with the light cylinder , the dotted line is the slow - magnetosonic surface .
right : same as the left panel , the dotted line is now the fast - magnetosonic surface .
note the radial structure in the asymptotic region . from @xcite.,title="fig : " ] , axes are in units of @xmath16
. left : contours represent magnetic surfaces , colors represent @xmath26 , the dashed line is the alfvenic surface , coincident with the light cylinder , the dotted line is the slow - magnetosonic surface .
right : same as the left panel , the dotted line is now the fast - magnetosonic surface .
note the radial structure in the asymptotic region . from @xcite.,title="fig : " ]
a way to constraint the properties of the wind at large distances is to study its interaction with the environment and especially with the surrounding snr .
the pulsar wind blows a cavity inside the snr , filled with a hot relativistic pair plasma and magnetic field , known as pulsar wind nebula , which shines in non thermal synchrotron emission from radio to x - rays .
recent x - ray images of the central region in pwne have shown the presence of a well defined axisymmetric structure known as `` jet - torus '' . a main emission torus is observed in the equatorial plane , with the presence of an inner ring of emission ; a perpendicular jet is present .
sometimes an enhancement of emission close to the pulsar location , known as the `` knot '' is also observed .
the observed doppler boosting suggests velocities in the torus @xmath31 and in the jet @xmath32 .
such structure is too complex to be interpreted in term of existing 1d models .
moreover the presence of a jet seems to contradict the result on the wind structure that collimation is not efficient .
recent numerical simulation @xcite has shown that such structure naturally arises as a consequence of an anisotropic energy distribution and sufficiently high magnetic field in the nebula . assuming a split monopole for the wind energy and lorentz factor distribution , as derived from 2d simulation of relativistic winds , and a magnetization higher than @xmath33 , simulated synchrotron maps are able to recover qualitatively all the main features of the jet - torus structure .
the torus and inner ring as well as the knot are explained in term of region where the flow is moving at a relativistic speed toward the observer , while the formation of the jet is associated with collimation by hoop stresses inside the nebula . in all simulations a 2d spherical @xmath17 grid was used , with 100 uniform cells in @xmath18 , and a logarithmic scale with 200 cells per decade in @xmath19 .
a wind with energy distribution typical of a monopole is injected at the base , in an environment representative of snr ejecta .
results suggest that the emission features observed in the nebulae might be compared with simulations to constraint the wind properties .
preliminary results show that a better agreement is achieved if an unmagnetized region in the equatorial plain is assumed .
such unmagnetized region could be due to dissipation of the current sheet @xcite , which in turn could provide a mechanism for accelerating the wind and solving the @xmath9 problem .
recent numerical results have confirmed theoretical predictions about the acceleration of relativistic pulsar winds .
the conversion of magnetic to kinetic energy appears to be minimal , the flow closely resembles the exact monopole solution , there is no evidence for the formation of a collimated energetic jet .
the use of a more realistic dipolar field does not seem to change significantly the wind properties far from the light cylinder . on the contrary
the energy and angular momentum losses seem to depend only on the integrated value of @xmath9 , and for @xmath34 on the amount of open magnetic flux . an alternative way to investigate the wind properties at least at large distances comes from the study of its interaction with the surrounding snr .
modeling of the `` jet - torus '' structure in pwne have shown that such features are due to the details of the nebular flow , which in turn depends on the wind properties .
simulations show that an energy and magnetic field distribution in the wind resembling the monopole solution , can indeed explain the main details of the observations .
arons , j. , 2004 , young neutron stars and their environments , i.a.u .
symposium,218 , 163 begelman , m. c. & li , z .- y . , 1994 , apj , 426 , 269 beskin , v. , kuznetsova , i. & rafikov , r. , 1998 , mnras , 299 , 341 bogovalov , s. , 1999 , a&a , 349 , 1017 bucciantini , n. , del zanna , l. , amato , e. , & volpi , d. 2005 , , 443 , 519 bucciantini , n. , thompson , t. a. , arons , j. et al . , 2006 , to be published in mnras , astro - ph/0602475 contopoulos , j. , kazanas , d. & fendt , c , 1999 , , 511 , 351 del zanna , l. , bucciantini , n. & londrillo , p. 2003
, , 400 , 397 del zanna , l. , amato , e. , & bucciantini , n. 2004 , , 421 , 1063 del zanna , l. , volpi , d. , amato , e. , & bucciantini , n. 2006 , accepted for publication in a&a , astro - ph/0603080 helfand , d. j. , gotthelf , e. v. , & halpern , j. p. 2001
, , 556 , 380 hester , j. j. , et al .
2002 , , 577 , l49 kennel , c. f. & coroniti , f. v. , 1984 , , 283 , 694 kennel , c. f. & coroniti , f. v. , 1984 , , 283 , 710 kirk , j. g. & skjaeraasen , o. , 2003 , , 591 , 366 komissarov , s. s. , 2006 , mnras , 367 , 19 komissarov , s. s. , & lyubarsky , y. e. 2003 , , 344 , l93 komissarov , s. s. , & lyubarsky , y. e. 2004 , , 349 , 779 michel , f. c. , 1969 , apj , 158 , 727 thompson , t. a. , chang , p. , & quataert , e. 2004 , , 611 , 380 weisskopf , m. c. , et al .
2000 , , 536 , l81 | the recent development of numerical schemes for relativistic mhd ( rmhd ) allows us to model the acceleration and outflow properties of winds from compact sources .
theoretical models suggest that acceleration and collimation of the flow are extremely inefficient when the speed is close to @xmath0 , in contrast with many observations .
numerical results for an axisymmetric rotator , both in the case of monopolar and dipolar magnetic field will be presented , suggesting that in ideal rmhd acceleration is indeed inefficient .
i will also mention numerical challenges and stability problems of present simulations .
finally i will discuss the interaction of a pulsar wind with the surrounding snr .
i will present emission maps based on numerical simulations of the flow inside the nebula and how they can be used in comparison with observations to derive informations about the properties of pulsar winds far away from the source . |
hawking famously showed @xcite that black holes emit a stationary thermal flux with temperature @xmath0 which is perceived by the asymptotic observers at late - times .
but that is not all that happens in the system that undergoes a gravitational collapse .
for example , in a typical collapse scenario ( see fig .
[ penrosecol ] ( left ) ) , the usual hawking effect is understood due to the ` peeling ' of modes that travel close to the horizon just before the collapse and escape to infinity with an infinite redshift .
this is depicted by ray number @xmath1 and the corresponding static observer encounters a temperature with the standard tolman shift which becomes unity in the asymptotic limit .
however , we can have a similar null ray numbered @xmath2 here , which would be just inside the horizon and hits the singularity at some finite time . the infallers
( radial or otherwise ) would encounter this situation .
the third possibility is of the ray @xmath3 that is inside the collapsing star all the time and here we would have comoving observers ( the inside spacetime being friedmann for dust collapse ) .
thus , we can have many different observers / frames which _ should _ see the effect of collapse on the quantum vacuum differently and there is a rich structure beneath waiting to be explored . on a related note , a class of observers termed the _
infallers _ have recently sprung a hot debate in the community on what would happen _ at the horizon crossing _
la the hawking effect .
it is usually argued @xcite on the basis of equivalence principle that the infalling observer should see a quantum vacuum state while crossing the horizon and nothing special .
however , this line of thought has been questioned lately by amps @xcite claiming on the quantum - information - theoretic grounds that the infalling observers will encounter a _ firewall _ at the horizon . amidst these speculations which are not completely understood at this time
, we seek to know what happens at least _ semi - classically _ , relying on the well - established grounds .
[ penrosecol ] , point @xmath2 corresponds to the radial infallers while those inside the collapsing body reach point @xmath3 presenting different aspects of the hawking phenomenon . *
* construction of globally defined null coordinates .
any point @xmath4 on the spacetime can be ascribed two coordinates @xmath5 which correspond to one ray coming directly from @xmath6 and the other traced back to @xmath6 after a reflection from the origin . here , we take vaidya null collapse where the thin null shell lies at @xmath7 separating the minkowski interior from the schwarzchild exterior.,title="fig : " ] , point @xmath2 corresponds to the radial infallers while those inside the collapsing body reach point @xmath3 presenting different aspects of the hawking phenomenon . *
* construction of globally defined null coordinates .
any point @xmath4 on the spacetime can be ascribed two coordinates @xmath5 which correspond to one ray coming directly from @xmath6 and the other traced back to @xmath6 after a reflection from the origin . here , we take vaidya null collapse where the thin null shell lies at @xmath7 separating the minkowski interior from the schwarzchild exterior.,title="fig : " ] working on these lines , barbado et al @xcite introduced the notion of an ` effective ' temperature as a measure of the response of an unruh - dewitt detector @xcite moving on a given trajectory .
they found that the effective temperature increases for a detector dropped from infinity on a radially infalling geodesic as the infall progresses and reaches the temperature _ four _ times the hawking temperature at the horizon .
as much as this result is intriguing , it should be noted their analysis does not give any direct link with the detector response nor does it provide any understanding to how thermal or non - thermal the effect is .
another point of concern is the choice of vacuum which they choose to be some new unconventional state in extended schwarzchild geometry .
we might think the peculiar result could be an artefact of this contraption .
but more importantly , the use of both the ingoing and outgoing eddington - finkelstein coordinates together is not suitable if we wish to study the interior of the black hole horizon . with this retrospection of the cards on the table
, we now proceed to bring in a new deck @xcite with the following assumptions and characteristics : ( a ) we introduce a new set of globally defined coordinates which are regular everywhere .
( b ) working with the massless scalar field , the vacuum state in a collapse scenario is a natural and uniquely defined in - vacuum on @xmath6 .
( c ) we define and study a suitably defined time - dependent ` temperature ' function also relating it explicitly to udw response functions by an _ adiabatic expansion _ and finally ( d ) we look at the observables energy density and fluxes built from the renormalized stress energy tensor ( rset ) which explicitly depend on observer s kinematics for an alternate description .
this also confirms the results of the effective temperature approach .
this paper is meant to give a concise description of the formalism and results obtained in the previous works @xcite which relate with the collapse of a thin null shell ( ingoing vaidya geometry ) and to provide a preview of the results ( relating to time - like dust collapse ) which are a work in progress @xcite .
without further ado , we shall present the new procedure and its application to the thin null shell collapse . as remarked earlier , it begins with the introduction of the _ new _ coordinates . *
* coordinatization : * any point @xmath4 on the spacetime manifold in the collapse scenario ( fig
. [ penrosecol ] ( right ) ) can be ascribed two coordinates @xmath5 which correspond to one ray coming directly from @xmath6 and the other traced back to @xmath6 after a reflection from the origin .
this construction is completely general and valid for any type of collapse . for the vaidya collapse where the thin null shell lies at @xmath7 separating the inside minkowski from the schwarzchild exterior ,
this coordinatization can be computed ( by ray tracing ) in terms of the eddington - finkelstein ingoing coordinates to give @xmath8 and @xmath9\quad\quad\textrm{for}\ v\geq0 , \end{array } \right.\end{gathered}\ ] ] where @xmath10 and @xmath11 is the lambert @xmath12-function .
the metric in these coordinates is given ( with @xmath13 ) by , [ nullmetric ] * * field setup : * the minimally coupled massless scalar field in the @xmath14-wave approximation at @xmath4 is given by ~_r , that is , superposition of two spherical waves one impinging directly and one that comes after a reflection from @xmath15 .
this also gives the interpretation to the coordinates @xmath5 as being tied to the eikonal approximation .
* * detector response : * the response of a detector @xcite moving on a general trajectory @xmath16 can be written as ( ) = 2_0^ds e^-is g((),(-s ) ) where the wightmann function , @xmath17 takes the logarithmic form @xcite , [ 2d ] g(x , y)((v_+(x)-v_+(y)-i0)(v_-(x)-v_-(y)-i0 ) ) for @xmath14-waves .
this splits the response into two : @xmath18 decoupling the ingoing and outgoing modes giving _ two temperatures _ ! the response function in its generality is difficult to compute analytically .
however , if @xmath19 is any solution of the differential equation , ( ) = -k ( ) ( ) with @xmath20 as the slowly varying function , then we can consider the adiabatic expansion of the response functional , which gives , @xmath21 = \f{2\pi}{\omega(e^{\omega / t}-1)}\big(\,1\,+\ , \f{\dot{t}(\tau)}{t(\tau)^{2}}\f{\omega}{8\pi t(\tau)}\f{e^{\omega/2t(\tau)}}{\sinh^{2}(\omega/2t(\tau))}+\cdots\big).\end{aligned}\ ] ] that is , all observers , asymptotic or non - asymptotic will perceive a _ quasi - thermal _ spectrum at t_=|| which is related to the ` peeling ' condition .
it becomes proper thermal temperature in the uv limit ( @xmath22 ) or under adiabatic condition , @xmath23 .
* * energy density and fluxes : * for an alternate prescription , we can compute the rset of the quantum field in @xmath24 dimensions from @xmath25 for @xmath14-waves , the @xmath26 dimensional counterpart is usually considered @xcite by multiplying the above expressions with the inverse square distance @xmath27 .
given the velocity of the observer @xmath28 and a vector @xmath29 such that @xmath30 , we can construct two local invariant observables _ energy density _ and _ flux _ as @xmath31 and @xmath32 respectively and which depend explicitly on the observer s kinematics .
these have the advantage over detector response in being local ( not dependent on the past trajectory ) and exact ( no need of adiabatic expansion ) . @ccc@ eff .
temperatures & asymptotic & at horizon crossing + @xmath33 & @xmath34 & @xmath35 + @xmath36 & 0 & @xmath37 +
with the above procedure , the usual hawking effect for an asymptotic observer at rest relative to the hole can be recovered as follows .
for this observer , @xmath38 satisfies -=and subsequently v^-()-v^-(-s)e^-(-s)/2(s/2 ) which recovers _ thermal spectrum _ at @xmath39 [ thermalspectrum ] _
while @xmath40 in this case . for a non - asymptotic static observer at some finite radius @xmath41 from the hole ,
the temperature gets scaled by the redshift factor , @xmath42 .
while many other trajectories can be considered , we shall look at the results for the radially infalling observer characterized by energy per unit rest mass , @xmath43 or the initial radius from where the infall begins , @xmath44 .
the analysis of the effective temperatures here then shows a dependence on @xmath43 and is presented in table [ radialtemp ] ( these results are also seen to hold in the case of time - like collapse@xcite ) .
the non - stationarity of the regime shows in having two temperatures now .
the temperature @xmath33 presents itself with a doppler shift for @xmath45 and the result in the near - horizon limit matches with ref .
@xcite for @xmath46 , i.e. , for the detector dropped in from infinity .
the surprise is that in the near - horizon region @xmath36 ( incoming modes ) dominates over @xmath33 ( outgoing modes ) for @xmath47 ( see fig [ inoutnearh ] left ) which implies a duality between the asymptotic and the near - horizon observer in that the near - horizon observer sees a dominant and high radiation from the _ sky _ @xcite .
however , it is to be noted that the adiabaticity in these cases is @xmath48 and hence the temperatures can only be interpreted in the uv limit . from another angle , we can compute the energy density and fluxes for various observers .
for the static observers in the late - time limit , we have , @xmath49 these expressions reduce to the hawking energy density and flux for the asymptotic observers with @xmath50 and otherwise the temperature is given by the tolman blueshift factor .
the _ onset _ of thermality depends on the location of the observer .
it is earlier for the near - horizon observers and late for those away from the horizon .
the flux diverges at the horizon .
the energy density presents a contrast that in the near - horizon region it reaches a maximum positive value , then decreases and becomes negative ( fig .
[ infalluf ] ) .
this corresponds to negative energy fluxes directed into the black hole which have been discussed before ( see ref .
@xcite ) . in the near horizon limit which shows dependence on energy and the dominance of ingoing mode ( @xmath36 ) over outgoing ( @xmath33 ) for @xmath51 . *
* we have the flux in the near horizon region confirming the effective temperature results.,title="fig : " ] in the near horizon limit which shows dependence on energy and the dominance of ingoing mode ( @xmath36 ) over outgoing ( @xmath33 ) for @xmath51 . *
* we have the flux in the near horizon region confirming the effective temperature results.,title="fig : " ] [ infalluf ] or @xmath52 here ) and diverge only at the eventual singularity .
for the static observers , as we go near the horizon , the flux diverges at the horizon due to infinite acceleration .
the energy density , however , reaches a maximum and then decrease to become negative and diverges negatively at the horizon.,title="fig : " ] or @xmath52 here ) and diverge only at the eventual singularity . for the static observers ,
as we go near the horizon , the flux diverges at the horizon due to infinite acceleration .
the energy density , however , reaches a maximum and then decrease to become negative and diverges negatively at the horizon.,title="fig : " ] for the radially infalling observer , the energy density and flux increases ( see fig .
[ infalluf ] ) as the infall occurs and is _ finite _ and _ regular _ at the horizon for the infall beginning far from the horizon and diverges only at the eventual singularity . at the horizon ,
these observables are given by , @xmath53 which show agreement with the effective temperatures such that for @xmath47 , that is , if the infall occurs starts close to the horizon , a high flux is perceived at the horizon crossing ( see fig .
[ inoutnearh ] right ) . * but why the clicking ?
* we can show by considering artificial black holes which are flat at the horizon , that the effective temperature due to outgoing modes @xmath33 vanishes at the horizon and also the flux contribution corresponding to it .
however , the effect due the ingoing modes remains .
this shows that curvature has something to do with the detector response and the flux .
when the outgoing modes do not feel the curvature at the horizon , we see no effect but the ingoing modes surely encounter some curvature while travelling to the horizon and account for the observed effect . the varying riemann curvature ( as the infall progresses )
gives the clicking at the horizon .
we see that using a new set of tools and procedure unravels some new and surprising results in the age old problem of quantum fields in gravitational collapse . the response of an udw detector can be computed using the adiabatic expansion and
hence we see not one but _ two _ temperatures attributed to ingoing and outgoing modes .
this formalism can be applied to many different trajectories . in particular , we see for the infalling radial observers , in the near - horizon limit , the temperature due to ingoing modes dominates over outgoing modes for _ slowly _ moving detectors .
a look at the energy density and flux confirms this .
so the punchline is that _ if you wish to drop into the black hole , start far off_. also , now we know , at least semi - classically , that there _ is _ clicking at the horizon and we need to dig in further for more insights .
i thank thanu padmanabhan and dawood kothawala for various discussions , matteo smerlak and sumanta chakraborty for the collaborations @xcite and @xcite respectively .
i also thank iucaa and the organisers of ere 2014 for the financial support that allowed me to attend the meeting . | hawking effect was dug out of the gravitationally collapsed site forty years back when it was realised that quantum effects at the horizon could propagate outward to infinity giving rise to a thermal flux at late - times .
however , the situation regarding non - asymptotic observers was never completely clear .
also , recently a debate has sprung in the community as to what would infalling observers perceive while crossing the horizon .
we set out to settle this question and more at least semi - classically in the articles [ arxiv:1304.2858 ] and [ arxiv:1404.0684 ] with a fresh approach .
we introduce a new set of coordinates that are regular everywhere , consider the adiabatic expansion of detector response and its link to the trajectory - dependent ` effective ' temperature / s and also local invariant observables , energy density and flux , built from renormalized stress energy tensor .
this paper is a concise summary of the new procedure and the results obtained thereof . |
very high - energy gamma - rays with energies around tev are powerful tools to investigate the cosmic accelerators .
they arise from either leptonic ( cosmic microwave background ( cmb ) or other soft seed photons up - scattered by accelerated electrons and their bremsstrahlung emission ) or hadronic ( the decay of neutral pions , arising from the collision of high energy protons and interstellar matter ) processes . the strong evidence for electron acceleration up to more than @xmath4 tev has been obtained with x - ray observations of , for example , young supernova remnants ( snrs ) that are believed to be the most probable cosmic - ray accelerators in our galaxy ( e.g. , * ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
on the other hand , we have not yet derived firm evidence for proton acceleration ; although tev gamma - rays have been detected by several of young snrs , such as rx j1713.7@xmath03946 @xcite and rx j0852.0@xmath04622 @xcite , their origin is not yet fully understood ( e.g , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
recently , a survey of the inner part of our galaxy has revealed several new tev @xmath5-ray sources @xcite .
some of them have no counterpart in the x - ray regime .
then , the flux ratio , defined by @xmath6 is more than @xmath4 @xcite . on the other hand , young supernova remnants ( snrs ) and pulsar wind
nebulae have @xmath7 less than @xmath8 .
it is interesting that the tev unidentified sources with large @xmath7 may show evidence for hadron acceleration because the leptonic inverse - compton model requires an unusually small magnetic field strength ( @xmath9 g ) .
so far , several tev gamma - ray objects have been followed up in x - rays , which provide us a lot of hint on the emission process of tev gamma - rays ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
is one of such tev unidentified sources .
it was discovered by the galactic plane survey of high energy stereoscopic system ( h.e.s.s . )
@xcite , and subsequent follow - up observation was performed recently @xcite .
the observed flux from the whole object is estimated as @xmath10 erg s@xmath2@xmath3 the low galactic latitude and the fact that the source is spatially extended strongly argues for a galactic origin of . mainly consists of three bright emission regions ( named region a , b , and c ; * ? ? ?
* ) , and the angular size of the whole object is @xmath11 , which is one of the largest tev unid sources .
has direct positional coincidence with an unidentified egret source ( 3eg j1744@xmath03011 , * ? ? ?
* ) , which implies proton acceleration because the leptonic ic model for the tev emission usually predict very dim gev @xmath5-rays .
@xcite suggested that a molecular cloud coincides with the tev emitting region , although the precise distance to the molecular cloud is unknown .
this region was also observed with _ xmm - newton _ , which gives an upper - limit on the diffuse x - ray emission , @xmath12 erg @xmath3s@xmath2 , from the region a which is the main part of and has a tev energy flux of @xmath13 erg @xmath3s@xmath2 .
then , the tev - to - x - ray flux ratio , @xmath7 , is larger than @xmath8 . in this paper , we report on the deep mapping observations of with x - ray satellite _ suzaku _ @xcite .
section [ sec : obs ] introduces the observations .
our analysis is explained in [ sec : analysis ] , and we discuss our findings in [ sec : discuss ] .
we carried out mosaic observations of hess j1745@xmath0303 with _ suzaku_. the mapping used four pointings as listed in table [ tab : obslog ] .
_ suzaku _ has two active instruments , four x - ray imaging spectrometers ( xis0xis3 ; * ? ? ?
* ) each at the focus of an x - ray telescope ( xrts * ? ? ?
* ) and a separate hard x - ray detector ( hxd ; * ? ? ?
all four spectrometers were active in the first observation , whereas only three could be used subsequently due to a problem with xis 2 .
xis1 is a back - illuminated ( bi ) ccd , whereas the others are front - illuminated ( fi ) .
the xis was operated in the normal full - frame clocking mode .
the spaced - row charge injection @xcite was used in the later three observations ( table [ tab : obslog ] ) . the data reduction and analysis
were made using headas software version 6.4 .
, version 2.2.7.18 of the processed data , and xspec version 11.3.2 .
we filtered out data obtained during passages through the south atlantic anomaly ( saa ) , with elevation angle to the earth s limb below 5@xmath14 , or with elevation angle to the bright earth s limb below 13@xmath14 . in order to avoid the contamination of emission from the bright earth . in this paper , we concentrate on the xis data analysis , because hess j1745@xmath0303 is located near the galactic plane , and the hxd data is contaminated by the galactic ridge x - ray emission ( grxe ; c.f .
* ; * ? ? ?
* ) and bright point sources , which are quite difficult to remove correctly .
we made mosaic intensity maps in the 0.52.0 kev , 2.08.0 kev , and 6.36.5 kev bands of the four pointing observations as described in the following .
the raw mosaic images were made in each energy band . in order to estimate the continuum emission in the narrow iron intensity band , we derived the spectrum over the entire region and fitted 38 kev photons with a power - law plus narrow gaussians which represent emission lines .
the 23 kev band photons were not used in order to avoid the influence of soft discrete emission ( two snrs ; described later ) .
the best - fit photon index was @xmath15 .
we compared the count rates of the power - law component in the 2.08.0 kev and 6.36.5 kev bands , the latter is 1.15% of the former , and normalized the 2.08.0 kev image with the ratio .
the normalized continuum image was subtracted from the 6.36.5 kev image , which is the `` neutral iron intensity image '' .
the non - x - ray background ( nxb ) image for each energy band was generated from the nxb database provided by the suzaku team .
the nxb images were chosen to make the average geomagnetic cu - off rigidity same as that of the observation , and extracted from the same region in the detector . in this procedure ,
xisnxbgen package developed by @xcite was used . after subtracting nxb images from the raw images , we combined the nxb - subtracted images using ximage package .
the exposure maps were made for 0.52.0 and 2.08.0 kev bands with the xissim package @xcite .
for the neutral iron line map , the 2.08.0 kev energy range was used .
the exposure maps were also combined . dividing the nxb - subtracted images with the exposure maps
, we made the mosaic images shown in figure [ fig : images ] .
our observations cover all of the region a of . on the eastern edge of
, we can see some diffuse components in the soft x - ray band , ( see figure [ fig : images](a ) ) .
they coincide with two radio snrs , [email protected] and [email protected] @xcite .
some point sources are also seen , as reported by @xcite .
however , no excess was found within the region .
the hard band image ( figure [ fig : images]b ) also shows no diffuse source . in the neutral iron line map ( figure [ fig : images]c ) , on the other hand
, some excess is shown around region a. we selected source and background regions ( `` src '' and `` bgd13 '' hereafter ) as shown with solid lines in figure [ fig : regions ] , to make the upper - limit of the x - ray emission of hess j1745@xmath0303 region a. the background regions were divided to 3 regions ( bgd13 in figure [ fig : regions ] ) .
two types of background should be considered , nxb and the grxe ( * ? ? ?
* for example ) .
the latter has positional dependence along the galactic longitude .
we thus used bgd1 and 2 regions to make the average longitude of each region similar .
the 38 kev count rates of xis 0 are @xmath16 and @xmath17 cnts s@xmath2arcmin@xmath18 for the source and the background regions , respectively . for making the upper - limit of the x - ray flux in the source region ,
the response and auxiliary files were constructed using xisrmfgen and xissimarfgen @xcite in the headas package , with the assumption that the emission comes from the source region uniformly .
the photons from the background regions were subtracted from the source spectra .
we fit the background - subtracted spectra in the 210 kev band with a power - law model with fixed @xmath19 , which is significantly harder value than the tev emission @xcite , and to derive the 90% upper - limit of 2.1@xmath20 ergs @xmath3s@xmath2 .
the larger photon indices were also used to estimate the systematic errors of the upper - limit and derived the smaller value .
we also derived the surface brightness of the neutral iron emission line in each region .
only the fi ccd data were used since the bi ccd has larger background in the hard x - ray band .
we fitted 58 kev spectra with a power - law component , narrow gaussians for neutral , he - like , and h - like iron lines , and for mn - k , and ni - k lines as nxb emission .
the best - fit models and parameters of the neutral and he - like iron lines in each region are shown in figure [ fig : spectra ] and table [ tab : iron ] . the surface brightness of he - like iron is almost constant between these regions .
we added all background regions to increase statistics and derived the average values of line intensity .
after the procedure , the neutral iron line is significantly stronger in the source region than those in background regions , whereas the he - like iron line is compatible in both regions .
the total excess of the line in these regions is @xmath21 ph s@xmath2@xmath3 . for the spectral analysis of the two snrs
, we used the regions with dashed - lines in figure [ fig : regions ] .
background photons were accumulated from the source - free regions in the same observation .
the background - subtracted spectra are shown in figure [ fig : snrs ] .
the spectra are rather soft and there are emission lines from highly ionized si and s in both spectra .
these facts indicate that these emissions are thermal , which is consistent with the asca results @xcite .
we then fitted these spectra with non - equilibrium ionization collisional plasma model ( nei in xspec ; * ? ? ?
* ) with absorption .
for the absorption model , we used the cross sections of @xcite and assumed solar abundances @xcite .
the xis team reported the gain uncertainty of a few ev ( c.f .
* ) , thus the gain is allowed to vary within a few percent separately .
@xcite reported that the spectrum of [email protected] was well described with the two - temperature model with high - s abundance in the higher temperature component .
so we have added another thermal component in the present analysis .
we treated some abundances in table [ tab : snrs ] as free parameters and fixed the others at 1 solar . the best - fit models and parameters
are shown in figure [ fig : snrs ] and table [ tab : snrs ] .
the results are consistent with the _ asca _ results , but the abundance and the ionization time scale are better determined thanks to the better energy resolution of xis .
we expect a difference in flux , since our observations do not cover the entire remnants .
the absorption column of [email protected] is smaller than sources in the galactic center ( gc ) , thus this snr is likely a foreground source . on the other hand
, [email protected] shows a very large absorption column , which implies that it is in the gc region .
@xcite reported that hess j1745@xmath0303 region a had spatial coincidence with a molecular cloud @xcite .
we suggest that the tev emission is also associated with the neutral iron line emission , although we can not unambiguously conclude that due to the poor statistics . the positional coincidence of the neutral iron line and the molecular cloud reminds us of reflected x - rays originating in the past activity in the gc ( e.g. , * ?
if the molecular cloud is optically thin , the line emission intensity , @xmath22 , should be proportional to the mass of the molecular cloud and the viewing angle from the gc , so that @xmath23 where @xmath24 , @xmath25 , and @xmath26 are the mass of the molecular cloud , the angular size of the molecular cloud viewed from the gc , and the angular separation between the cloud and the gc , respectively @xcite .
we compared the masses of the molecular cloud coincident with source a of hess j1745@xmath0303 and the most famous x - ray reflection nebula , sgr b2 @xcite .
the latter shows neutral iron line emission with an intensity of @xmath27 ph s@xmath2@xmath3 .
table [ tab : mc ] shows the characteristics of these molecular clouds . for the molecular cloud at region a of hess j1745@xmath0303
, we obtain @xmath28 ph s@xmath2@xmath3 on the assumption of time - independent ejected x - ray luminosity at the gc .
this estimated value is compatible with the observed one , @xmath21 ph s@xmath2@xmath3 .
therefore , we conclude that this molecular cloud is a new example of x - ray reflection nebula and that the molecular cloud is located in the gc region .
we believe that hard x - ray observations are the strong tools to judge whether the molecular clouds are in the gc region or not .
we derive a 90% x - ray flux upper - limit for the hess j1745@xmath0303 region a in the 210 kev band of @xmath1 ergs s@xmath2@xmath3 , which is tighter than the _ xmm _ results @xcite .
the flux ratio between 210 kev and 110 tev , @xmath7 , is larger than 4 , which could be larger than tev emitting shell - type snrs and pwne @xcite . in the following ,
we discuss specific models of the emission mechanism to reproduce the observed results of x - ray and tev gamma - rays for the region a of hess j1745@xmath0303 .
in particular , we focus on whether or not the upper limit of nonthermal x - ray emission , which is derived in the present work , gives us significant constraints on the theoretical models .
we consider hadronic and leptonic models separately .
since is near the galactic center , the gev emission may be overlaid with the intense galactic diffuse emission and uncertainty is very large .
hence we neglect it here for simplicity , and take into account only x - ray and tev gamma - ray bands . in hadronic models , tev gamma - rays originate from accelerated protons .
it is assumed that the proton energy spectrum has a form of @xmath29 with an index , @xmath30 , and cut - off energy , @xmath31 . here
, the value of @xmath30 is highly uncertain , so we adopt @xmath32 as a typical value .
the high - energy protons collide with target protons in the ism or molecular cloud , generating neutral ( @xmath33 ) and charged ( @xmath34 ) pions .
the former decays into gamma rays , while the decay of the latter causes secondary electrons and positrons producing bremsstrahlung , inverse compton , and synchrotron emission .
although it may be likely that the region a of is associated with a molecular cloud and target matter density @xmath35 is high , the low-@xmath35 case can not be , at present , excluded .
so , we consider two cases , in which @xmath36 @xmath37 and @xmath38 @xmath37 , independently . the value of @xmath35 in the latter case is adopted according to @xcite . generally speaking , in the former , bremsstrahlung emission is not dominant in the tev gamma - ray band , while it may often be dominant in the latter . at
first we consider the case @xmath36
@xmath37 ( see the left panel of fig . [
fig : spectrum ] ) .
since the value of @xmath35 is small , emissions from secondary electrons is dim in the gamma - ray band , so that the @xmath33-decay emission dominates .
we find that as long as @xmath39 @xmath37 , the bremsstrahlung emission of secondary electrons and positrons is not dominant in the gamma - ray band .
we fix @xmath40 tev in order to fit the observed spectral shape in the tev energy range .
then , as can be seen in the left panel of fig .
[ fig : spectrum ] , derived upper limit on nonthermal x - rays are explained if the magnetic field @xmath41 mg .
this is rather loose limit on the magnetic field . because of the small value of @xmath31 , the maximum energy of secondary electrons and positrons is small , @xmath42 tev , so that the large value of @xmath43 is necessary for bright synchrotron x - rays emitted by them
this comes from the fact that the characteristic frequency of electron / positron synchrotron radiation is given by @xmath44 where @xmath45 is the energy of secondary electrons or positrons . for assumed distance of 8.5 kpc ,
the total energy of accelerated protons is estimated as @xmath46 erg .
next we consider the case @xmath38
@xmath37 ( see the right panel of fig . [
fig : spectrum ] ) .
if the magnetic field strength is small ( @xmath47 g ) , the bremsstrahlung emission from secondary electrons and positrons is comparable to the @xmath33-decay gamma - ray emission .
however , in the strong field cases , the synchrotron cooling effect significantly suppresses the number of @xmath48 , resulting small contribution to the gamma - rays . on the other hand ,
the synchrotron radiation of secondary @xmath48 is brighter for larger @xmath43 .
similar to the case of @xmath36 @xmath37 , the magnetic field is no larger than @xmath49 mg .
@xcite derived the relatively high magnetic field value , 0.20.6 mg , which is measured from the zeeman splitting of maser lines . the molecular cloud associated with the region
a might have the similar magnetic field strength , and then , @xmath50 mg field and the number density of the cloud @xmath51
@xmath37 is roughly consistent with the previous result on other molecular clouds @xcite .
as long as the field strength in the cloud is @xmath52 mg or less , the expected flux of secondary synchrotron radiation in the x - ray band is rather small and below the observed upper limit . for assumed distance of 8.5 kpc ,
the total energy of accelerated protons is estimated as @xmath53 erg .
then , the mean energy density of high - energy protons at region a of is calculated as @xmath54 ev @xmath37 if we assume the volume of the molecular cloud of @xmath55 @xmath56 @xcite , in the leptonic model , tev gamma - rays arise via bremsstrahlung and inverse compton process ( with cmb seed photons ) emissions of primarily accelerated electrons , and x - rays arise from their synchrotron radiation .
again we assume that the electron energy spectrum has a form of @xmath57 with an fixed index , @xmath58 , and the cut - off energy , @xmath59 . as well as the hadronic model , both low- and high-@xmath35 cases
are discussed .
first we consider the case @xmath60
@xmath37 ( see the left panel of fig . [
fig : spectrum2 ] ) . in this case , the tev gamma - ray band is dominated by the ic emission . in order to explain the observed hard tev spectrum , we adopt @xmath61 tev .
then , the relatively weak magnetic field , @xmath62 @xmath63 g , is necessary in order for the flux of synchrotron x - rays to be smaller than the observed upper limit .
we find that for an assumed distance of 8.5 kpc , the total energy of accelerated electrons is estimated as @xmath64 erg , which is very large compared with the typical young snrs ( e.g. , * ? ? ?
next we consider the case @xmath38
@xmath37 ( see the right panel of fig . [
fig : spectrum2 ] ) . in this case , the tev gamma - ray band is dominated by the bremsstrahlung emission . when the magnetic field is larger than @xmath65 g , the synchrotron cooling effect is significant and the electron spectrum is modified enough for the emission spectrum in the tev gamma - ray band to deviate from the observed one . as long as @xmath66 @xmath63 g , we can fit the tev gamma - ray spectrum with bremsstrahlung emission , and the synchrotron x - rays below the observed upper limit . for assumed distance of 8.5 kpc ,
the total energy of accelerated electrons is estimated as @xmath67 erg and @xmath68 erg for @xmath69 @xmath63 g and @xmath70 @xmath63 g , respectively . if we assume the volume of the molecular cloud of @xmath55 @xmath56 @xcite , the mean energy density of high - energy protons at region a of is calculated as @xmath71 ev @xmath37 and @xmath72 ev @xmath37 for @xmath69 @xmath63 g and @xmath70 @xmath63 g , respectively . in the case of @xmath38 @xmath37 ,
the energy - loss time scale , @xmath73 , for tev - emitting electrons via synchrotron and bremsstrahlung radiation is less than @xmath74 yr .
then , the diffusion length scale is estimated as @xmath75 pc , where we assume the bohm type diffusion , @xmath76 .
even if @xmath77 , this length scale is much smaller than the source size of region a of ( @xmath78 pc assuming the distance of 8.5 kpc ) , which implies that the high - energy electrons coming from a nearby accelerator can not penetrate into deep interior of a molecular cloud . on the other hand , when @xmath60 @xmath37 , relatively large energy of accelerated electrons is required .
so , the leptonic models considered above may be unlikely .
however , in order to derive the firm conclusion , we need to discuss further detail with more information on , such as the intrinsic source size , the number density of target matter , more precise spectrum in all wavelength ( from radio to gev tev gamma - rays ) , and so on
. spatial coincidence of a molecular cloud and the tev gamma - rays at region a of hess j1745@xmath0303 suggests that the tev emission is possibly due to accelearated protons interacting the molecular cloud .
the source of high energy protons is still unknown .
the past gc activity might be able to make protons @xcite .
however , it is rather difficult due to the low mass of the molecular cloud in our region .
nearby snrs and pulsars , on the other hand , could be high - energy proton injectors .
[email protected] is an energetic snr and interacting with the molecular cloud @xcite and is located in the gc region .
we confirmed that the distances to this snr and the molecular clouds are similar , but we could not find any signature of the interaction in x - rays .
hidden snrs by the molecular cloud also could be the source of accelerated particles .
the average absorption column of the molecular cloud is @xmath79 @xmath3 , which is enough to hide the soft x - ray emission from the snrs . for further study ,
more detailed radio observations are needed .
[email protected] is an unrelated source with its significantly smaller distance @xcite .
a pulsar b1742@xmath030 has been argued as a possible cosmic ray injector ( e.g. , * ? ? ?
* ) , but the estimated distance ( 2.1 kpc ; * ? ? ?
* ) is significantly smaller than that of .
we thus concluded that the pulsar b1742@xmath030 is also unrelated to .
deep mapping observations of a part of have been carried out with _ suzaku_. the main part of , region a , is entirely covered in the present observation .
we made a tight upper limit of hard x - rays of the region , @xmath1 ergs s@xmath2@xmath3 in the 210 kev band .
the flux ratio between 110 tev and 210 kev is larger than 4 , which could make one of the `` dark particle accelerators '' .
an excess of the neutral iron emission line on this region is found .
this possible excess should be due to the x - ray reflection from a molecular cloud by the past active gc .
this provides strong evidence for the interaction between the molecular cloud system in the gc region . with the small flux in the hard x - ray band and coincidence with a molecular cloud
, the tev emission from could be due to protons encountered the molecular cloud .
we would like to thank the referee for useful comments and suggestions .
the authors also thank h. murakami , a. kawachi , t. oka , and y. fukui , for their useful comments on molecular clouds and x - ray reflection nebulae . this work was supported in part by grant - in - aid for scientific research of the japanese ministry of education , culture , sports , science and technology , no .
19@xmath804014 ( a. bamba ) , no . 18740153 and no
. 19047004 ( r. yamazaki ) , and also supported in part by pparc grant , pp / d000394/1 , eu grant mrtn - ct-2006 - 035863 , the european union through the marie curie research and training network `` universenet '' , mrtn - ct-2006 - 035863 ( k. kohri ) .
aharonian , f. et al .
, 2002 , a&a , 393 , l37 aharonian , f. et al . , 2004a , nature , 432 , 75 aharonian , f. et al . ,
2005a , science , 307 , 1938 aharonian , f. et al .
, 2005b , a&a , 439 , 1013 aharonian , f. et al . , 2005c , a&a , 437 , l7 aharonian , f. et al . ,
2006a , apj , 636 , 777 aharonian , f. , et al .
2006b , , 439 , 695 aharonian , f. , et al . 2008 , , 483 , 509 anders , e. , & grevesse , n. 1989 , , 53 , 197 bamba , a. , yokogawa , j. , sakano , m. , & koyama , k. 2000 , , 52 , 259 bamba , a. , ueno , m. , koyama , k. , yamauchi , s. , 2003a , , 589 , 253 bamba , a. , yamazaki , r. , ueno , m. , koyama , k. , 2003b , , 589 , 827 bamba , a. , yamazaki , r. , yoshida , t. , terasawa , t. , koyama , k. , 2005a , , 621 , 793 bamba , a. , yamazaki , r. , hiraga , j. , 2005b , apj , 632 , 294 bamba , a. et al . , 2007 , , 59 , s209 bitran , m. , alvarez , h. , bronfman , l. , may , j. , & thaddeus , p. 1997
, , 125 , 99 borkowski , k.j . , lyerly , w.j .
, & reynolds , s.p .
2001a , , 548 , 820 crutcher , r. m. 1991 , apj , 520 , 706 ellison , d. c. , 2001 , ptps , 143 , 125 enomoto , r. et al .
2002 , , 416 , 823 hartman , r. c. et al . 1999 ,
apjs , 123 , 79 inui , t. , koyama , k. , matsumoto , h. , & tsuru , t. g. 2008 , arxiv e - prints , 803 , arxiv:0803.1528 ishisaki , y. et al .
2007 , , 59 , s113 katagiri , h. et al .
, 2005 , apj , 619 , l163 koyama , k. , awaki , h. , kunieda , h. , takano , s. , & tawara , y. 1989 , , 339 , 603 koyama , k. , petre , r. , gotthelf , e.v . ,
hwang , u. , matsura , m. , ozaki , m. , & holt s.s .
1995 , , 378 , 255 koyama , k. et al .
1997 , pasj , 49 , l7 koyama , k. et al .
2007 , , 59 , s23 larosa , t. n. , kassim , n. e. , lazio , t. j. w. , & hyman , s. d. 2000 , , 119 , 207 lazendic j. s. , slane p. o. , gaensler b. m. , reynolds s. p. , plucinsky p. p. , hughes j. p. , 2004 ,
apj , 602 , 271 lemiere , a. et al .
2007 , in proc . of 30th int .
cosmic ray conf .
( mexico ) , ( p.42 in arxiv:0710.4057 ) murakami , h. , koyama , k. , & maeda , y. 2001 , , 558 , 687 roberts , m. s. e. , romani , r. w. , & kawai , n. 2001 , , 133 , 451 matsumoto , h. et al .
2007 , , 59 , s199 matsumoto , h. et al .
2008 , , 60 , s163 mitsuda , k. et al .
2007 , , 59 , s1 morrison , r. , & mccammon , d. 1983 , , 270 , 119 nakajima , h. et al .
2008 , , 60 , s1 nobukawa , m. , et al .
2008 , , 60 , 191 pannuti , t. g. , allen , g. e. , houck , j. c. , & sturner , s. j. , 2003 , apj , 593 , 377 robinson , b. yusef - zadeh , f. , & roberts , d. 1996 , bull . am .
soc . , 28 ,
948 serlemitsos , p. et al .
2007 , , 59 , s9 takahashi , t. et al .
2007 , , 59 , s35 tawa , n. , et al .
2008 , , 60 , 11 taylor , j. h. , manchester , r. n. , & lyne , a. g. 1993 , apjs , 88 , 529 tian , w. w. et al .
2007 , apj , 657 , l25 uchida , k. , morris , m. , & yusef - zadeh , f. 1992 , , 104 , 1533 uchiyama , y. , aharonian , f. a. , takahashi , t. , hiraga , j. s. , moriguchi , y. , & fukui , y. , 2005 , in proc . of international symposium on high energy gamma - ray astronomy ,
heidelberg , july 2004 ( astro - ph/0503199 ) uchiyama , y. , aharonian , f. a. , tanaka , t. , takahashi , t. , & maeda , y. 2007 , , 449 , 576 yamauchi , s. , & koyama , k. 1993 , , 404 , 620 yamazaki , r. et al .
2006 , mnras , 371 , 1975 yamazaki , r. et al .
2008 , arxiv:0806.3303 yuasa , t. et al .
2007 , , in press p5pccccc 501010010 & 2006 - 10 - 07 & ( 266.26 , @xmath030.37 ) & 53 & off + 502016010 & 2008 - 03 - 02 & ( 266.22 , @xmath030.11 ) & 73 & on + 502017010 & 2008 - 03 - 06 & ( 266.47 , @xmath030.09 ) & 75 & on + 502018010 & 2008 - 03 - 08 & ( 266.06 , @xmath030.24 ) & 82 & on p15pcccccc neutral iron + line center [ ev ] & 6.42 ( 6.416.43 ) & ( fixed to src1 ) & 6.38 ( 6.366.41 ) & 6.41 ( 6.376.45 ) & 6.40 ( 6.386.42 ) + surface brightness [ @xmath81ph @xmath3s@xmath2arcmin@xmath18 ] & 6.5 ( 5.87.2 ) & 5.9 ( 3.7 - 8.1 ) & 6.6 ( 4.38.9 ) & 2.3 ( 1.23.4 ) & 3.2 ( 2.53.9 ) + he - like iron + line center [ ev ] & 6.68 ( 6.676.69 ) & ( fixed to src1 ) & 6.67 ( 6.666.68 ) & 6.69 ( 6.676.71 ) & 6.68 ( 6.676.69 ) + surface brightness [ @xmath81ph @xmath3s@xmath2arcmin@xmath18 ] & 11 ( 1012 ) & 12 ( 1014 ) & 17 ( 1420 ) & 10 ( 911 ) & 11 ( 1012 ) p10pccc @xmath83 [ kev ] & 0.35 ( 0.320.38 ) & 1.1 ( 0.72.0 ) + mg & 1.4 ( 1.11.6 ) & + si & & 1.9 ( 1.32.9 ) + fe & 0.4 ( 0.20.6 ) & + @xmath84 [ @xmath85 @xmath37s ] & 2.3 ( 1.53.7 ) & + @xmath86 [ kev ] & & 2.0 ( 1.72.5 ) + s & & ( @xmath8740 ) + @xmath88 [ @xmath89 @xmath3 ] & 1.6 ( 1.51.8 ) & 2.3 ( 1.62.5 ) + flux ( 210 kev ) [ ergs s@xmath2@xmath3 ] & @xmath91 & @xmath92 + @xmath93/d.o.f .
& 109.7/70 & 64.0/41 | _ suzaku _ observations of a tev unidentified ( unid ) source , hess j1745@xmath0303 , are presented . a possible excess of neutral iron line emission is discovered , and is likely associated with the main part of hess j1745@xmath0303 , named `` region a '' . it may be an x - ray reflection nebula where the x - rays from previous galactic center ( gc ) activity are reflected by a molecular cloud .
this result further strengthens the assumption that the molecular cloud which is spatially coincident with region a of hess j1745@xmath0303 is located in the gc region .
the tev emission from molecular clouds is reminiscent of the diffuse tev gamma - rays from the gc giant molecular clouds , and it could have the same emission mechanism . with deep exposure mapping observations by _ suzaku _ , a tight upper - limit on the 210 kev continuum diffuse emission from region a is obtained , as @xmath1 ergs s@xmath2@xmath3 .
the flux ratio between 110 tev and 210 kev is larger than 4 .
possible scenarios to reproduce wide - band spectra from kev to tev are examined .
thermal x - rays from nearby two old supernova remnants , [email protected] and [email protected] , are detected , and their emission properties are well determined in the present study with deep exposure . |
in the late 1960s , vlastimil ptk ( 19251999 ) derived the _ method of nondiscrete induction_. this method for estimating the convergence of iterative processes was originally motivated by a quantitative refinement ptk had found for the closed graph theorem from functional analysis in 1966 @xcite .
he published about 15 papers on the method , five of them jointly with potra in the 1980s , and the work on the method culminated with their 1984 joint monograph @xcite . for historical remarks on the development of the method
see ( * ? ? ?
* preface ) or @xcite .
ptk described the general motivation for the method of nondiscrete induction in his paper `` what should be a rate of convergence ? '' , published in 1977 @xcite : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ it seems therefore reasonable to look for another method of estimating the convergence of iterative processes , one which would satisfy the following requirements .
* it should relate quantities which may be measured or estimated during the actual process .
* it should describe accurately in particular the initial stage of the process , not only its asymptotic behaviour since , after all , we are interested in keeping the number of steps necessary to obtain a good estimate as low as possible . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ this seems to be almost too much to ask for . yet , for some iterations the method of nondiscrete induction indeed leads to analytical convergence estimates which satisfy the above requirements .
the essential idea , as we will describe in more detail below , is that in the method of nondiscrete induction the rate of convergence of an iterative process is considered _ a function rather than just a number_. ptk derived convergence results satisfying the above requirements for a handful of examples including newton s method @xcite ( this work was refined in @xcite ) , and an iteration for solving a certain eigenvalue problem @xcite .
for these examples it was shown that the convergence estimates resulting from the method of nondiscrete induction are indeed optimal in the sense that in certain cases they are attained in every step of the iteration .
in addition to the original papers , comprehensive statements of such sharpness results are given in the book of potra and ptk ; see , e.g. , ( * ? ? ?
* proposition 5.10 ) for newton s method and ( * ? ? ? * proposition 7.5 ) for the eigenvalue iteration .
despite these strong theoretical results , it appears that the method of nondiscrete induction never became widely known . even in the literature on newton methods for nonlinear problems
it is often mentioned only marginally ( if at all ) ; see , e.g. , deuflhard s monograph @xcite .
part of the reason for this neglect of the method may be the lack of numerically computed examples in the original publications .
the goals of this paper are : ( 1 ) to motivate and prove a theorem on nondiscrete induction due to potra and ptk that is directly applicable in the analysis of iterative processes , ( 2 ) to explain on two examples ( namely the newton iterations for the matrix polar decomposition and the matrix square root ) how the method of nondiscrete induction can be applied to matrix iterations in numerical linear algebra , and ( 3 ) to demonstrate the method s effectiveness as well as discuss some of its weaknesses in the numerical linear algebra context .
it must be stressed upfront , that most theoretical results in sections [ sec : polar ] and [ sec : appl ] also could be derived using the general theory of newton s method for nonlinear operators in banach spaces as described in ( * ? ? ?
* chapters 2 and 5 ) and in the related papers of potra and ptk , in particular @xcite .
the strategy in this paper is , however , to apply the method of nondiscrete induction without any functional analytic framework and differentiability assumptions directly to the given algebraic matrix iterations .
the paper is organized as follows . in section [
sec : method ] we describe the method of nondiscrete induction and derive the theorem of potra and ptk .
we then apply this theorem in analysis the newton iterations for computing the matrix polar decomposition ( section [ sec : polar ] ) and the matrix square root ( section [ sec : appl ] ) . for both iterations
we prove convergence results and illustrate them numerically . in section [ sec : concl ]
we give concluding remarks and an outlook to further work .
in most of the publications on the method of nondiscrete induction , ptk formulated the method based on his `` induction theorem '' , which is an inclusion result for certain subsets of a metric space ; see , e.g. , @xcite , @xcite , @xcite , @xcite , or the book of potra and ptk ( * ? ? ? * proposition 1.7 ) . instead of the original induction theorem we will in this paper use a result that was stated and proven by potra and ptk as a `` particular case of the induction theorem '' in ( * ? ? ? * proposition 1.9 ) ; also cf . @xcite . unlike the induction theorem ,
this result is formulated directly in terms of an iterative algorithm , and therefore it is easier to apply in our context . in order to fully explain the result and its consequences for the analysis of iterative algorithms we will below give a motivation of the required concepts as well as a complete proof of the assertion without referring to the induction theorem .
let @xmath0 be a complete metric space , where @xmath1 denotes the distance between @xmath2 .
consider a mapping @xmath3 , where @xmath4 denotes the domain of definition of @xmath5 , i.e. , the subset of all @xmath6 for which @xmath7 is well defined .
for each @xmath8 we may then consider the iterative algorithm @xmath9 with iterates given by @xmath10 if all iterates are well defined , i.e. , @xmath11 for all @xmath12 , then the iterative algorithm @xmath9 is called _
it is clear from ( [ eqn : algorithm ] ) that a ( meaningful ) iterative algorithm can converge to some @xmath13 only when the distances between @xmath14 and @xmath15 decrease to zero for @xmath16 .
this requirement , and the rate by which the convergence to zero occurs , are formalized in the method of nondiscrete induction by means of a ( nondiscrete ) family of subsets @xmath17 that depend on a positive real parameter @xmath18 .
the key idea is that for each @xmath19 , where possibly @xmath20 , the set @xmath21 should contain all elements @xmath22 for which the distance between @xmath23 and @xmath24 is at most @xmath18 , i.e. @xmath25 , and then to analyze what happens with the elements of @xmath21 under one application of the mapping @xmath5 .
this means that the method compares the values @xmath26 hopefully , the new distance @xmath27 is ( much ) smaller than the previous distance @xmath28 . using the family of sets this can formally be written as @xmath29 where @xmath30 should be a ( positive ) `` small function '' of @xmath18 that needs to be determined .
this function is called the _ rate of convergence _ of the iterative algorithm . since in the limit
the distances between the iterates should approach zero , the limiting point(s ) @xmath31 of the algorithm are contained in the limit set @xmath32 , which is not to be constructed explicitly but is defined as @xmath33 where @xmath34 denotes the closure of the set @xmath35 .
note that the set @xmath32 possibly is not a subset of @xmath36 .
we need the following formal definition of a rate of convergence .
[ def : rate ] let @xmath37 be a given real interval , where possibly @xmath20 .
let @xmath38 be a function and denote by @xmath39 its @xmath40th iterate , i.e. , @xmath41 and @xmath42 for @xmath43 .
the function @xmath30 is called a rate of convergence on @xmath44 when the corresponding series @xmath45 converges for all @xmath46 .
the reason for the convergence of the series @xmath47 in ( [ eqn : sigma ] ) will become apparent in the proof of the following result of potra and ptk , cf .
* proposition 1.9 ) .
[ thm : potpta ] let @xmath0 be a complete metric space , let @xmath3 be a given mapping , and let @xmath8 . if there exist a rate of convergence @xmath30 on a real interval @xmath37 , with the corresponding function @xmath47 as in ( [ eqn : sigma ] ) , and a family of sets @xmath17 for all @xmath46 , such that * @xmath48 for some @xmath49 , * @xmath25 and @xmath50 for each @xmath46 and @xmath51 , then : 1 .
@xmath9 is a meaningful iterative algorithm and the sequence @xmath52 of its iterates converges to a point @xmath53 .
( in particular , @xmath54 . ) 2 .
the following relations are satisfied for all @xmath12 : + a ) @xmath55 , + b ) @xmath56 , + c ) @xmath57 .
+ the right hand side of the bound in ( 2.c ) is a strictly monotonically decreasing function of @xmath40 . moreover ,
if equality holds in ( 2.c ) for some @xmath58 , then equality holds for all @xmath59 . since @xmath60 , we know that @xmath61 exists
. now ( ii ) implies that @xmath62 we can apply the same reasoning to @xmath63 , which yields the existence of @xmath64 with @xmath65 continuing in this way we obtain a sequence of well defined iterates @xmath66 with @xmath67 this shows items ( 2.a ) and ( 2.b ) .
next observe that , for all @xmath12 and @xmath68 , @xmath69 where we have used ( 2.b ) .
convergence of the series @xmath70 for all @xmath46 implies that the sequence @xmath71 is a cauchy sequence , which by completeness of the space converges to a well defined limit point @xmath31 . from @xmath55 and @xmath72 for @xmath73 , we obtain @xmath53 .
hence we have shown ( 1 ) .
for all @xmath12 and @xmath68 , @xmath74 taking the limit @xmath75 shows that @xmath76 , which proves ( 2.c ) . since @xmath77 and @xmath49
, we have @xmath78 for all @xmath12 , and thus @xmath79 for the proof of the last assertion about the equality in ( 2.c ) we refer to the proof of ( * ? ? ?
* proposition 1.11 ) .
( in that proof it should read `` if ( 10 ) is verified '' instead of `` if ( 9 ) is verified '' . ) from the formulation of theorem [ thm : potpta ] it becomes apparent why ptk considered his method a `` continuous analogue '' of the classical mathematical induction ; also cf . his own explanation in @xcite : in condition ( i )
we require the existence of at least one set @xmath80 that contains the initial value @xmath81 .
this corresponds to the base step in the classical mathematical induction .
condition ( ii ) then considers what happens with the elements of the set @xmath21 under one application of the mapping @xmath5 .
this corresponds to the inductive step . in the classical theory of iterative processes ,
an iteration @xmath15 is called _ convergent of order @xmath82 _ when there exist a constant @xmath83 and a positive integer @xmath84 , such that @xmath85 in particular , @xmath86 and @xmath87 give `` linear '' and quadratic convergence , respectively , where in the first case we require that @xmath88 . as ptk pointed out ,
e.g. , in @xcite , such classical bounds compare quantities that are not available at any finite step of the process , simply because @xmath31 is unknown .
in addition , since @xmath89 is the same fixed number for all @xmath59 and @xmath84 often is large , such bounds in many cases can not give a good description of the initial stages of the iteration ; cf .
the quote from @xcite given in the introduction above .
item ( 2.c ) in theorem [ thm : potpta ] , on the other hand , gives an _ a priori _ bound on the error norm in each step @xmath12 of the iterative algorithm .
moreover , since the convergence rate is a function ( rather than just a number ) , there appears to be a better chance that the bound is tight throughout the iteration .
finally , we note that from the statement of theorem [ thm : potpta ] it is clear that if @xmath90 and @xmath5 is continuous at @xmath31 , then @xmath91 , i.e. , @xmath31 is a fixed point of @xmath5 .
in contrast to classical fixed point results , however , theorem [ thm : potpta ] contains no explicit continuity assumption on @xmath5 . as shown in @xcite
( see also ( * ? ? ?
* section 3 ) ) , the banach fixed point theorem , which assumes lipschitz continuity of @xmath5 , can be easily derived as a corollary of theorem [ thm : potpta ] .
let @xmath92 be nonsingular , and let @xmath93 be a singular value decomposition with @xmath94 , @xmath95 for @xmath96 , and with unitary matrices @xmath97 .
the factorization @xmath98 is called a polar decomposition of @xmath99 . here
@xmath100 is unitary and @xmath101 is hermitian positive definite .
the theory of the matrix polar decomposition and many algorithms for computing this decomposition are described in higham s monograph on matrix functions ( * ? ? ?
* chapter 8) .
one of the algorithms for computing the matrix polar decomposition is newton s iteration ( * ? ? ?
* equation ( 8.17 ) ) : @xmath102 this iteration can be derived by applying newton s method to the matrix equation @xmath103 ; cf .
* problem 8.18 ) . as shown in (
* theorem 8.12 ) , its iterates @xmath104 converge to the unitary polar factor @xmath100 , and they satisfy the convergence bound @xmath105 where @xmath106 denotes any unitarily invariant and submultiplicative matrix norm .
this is a bound of the form ( [ eqn : classical ] ) that shows quadratic convergence of the iteration ( [ eqn : polar1])([eqn : polar2 ] ) .
the derivation of this bound as well as further a posteriori convergence bounds for this iteration in terms of the singular values of @xmath99 can also be found in ( * ? ? ?
* theorem 3.1 ) .
we will now analyze the algebraic matrix iteration ( [ eqn : polar1])([eqn : polar2 ] ) using theorem [ thm : potpta ] . in the notation of the theorem we consider @xmath107 and we write the distance between @xmath108 as @xmath109 , where @xmath106 denotes the spectral norm .
we also consider the mapping @xmath3 defined by @xmath110 we need to determine @xmath36 and define the sets @xmath17 for some interval @xmath37 on which the nondiscrete induction is based .
the value of @xmath111 is not specified yet ; it will be determined when the rate of convergence has been found . obviously , @xmath36 is equal to the set of the invertible matrices in @xmath112 , and hence every @xmath113 must be invertible .
we also require that every @xmath113 satisfies @xmath114 ; cf . the motivation of theorem [ thm : potpta ] and the first assumption in ( ii ) of the theorem .
moreover , for @xmath115 we get @xmath116 .
inductively we see that every @xmath104 that can possibly occur in the iteration ( [ eqn : polar1])([eqn : polar2 ] ) is of the form @xmath117 for some diagonal matrix @xmath35 with real and positive diagonal elements . for the given matrix @xmath99 and each @xmath46 we therefore define @xmath118 in particular , @xmath119 .
now let @xmath113 , so that @xmath120 for some @xmath121 with @xmath122 for @xmath96 .
then @xmath123 , and @xmath124 from ( [ eqn : pp1 ] ) we see that in the limit @xmath125 we obtain @xmath126 , i.e. , the limiting set @xmath32 consists only of the unitary polar factor of @xmath99 . in order to determine a rate of convergence we compute @xmath127 and hence @xmath128 here we have used the definition of @xmath129 from ( [ eqn : pp1 ] ) and that @xmath130 .
note that @xmath131 and @xmath132 for all positive @xmath133 .
the maxima in ( [ eqn : pp1 ] ) and ( [ eqn : pp2 ] ) are attained at the same point @xmath134 . without loss of generality
we can assume that @xmath135 .
elementary computations show that @xmath136 @xmath137 i.e. , both @xmath138 and @xmath139 are strictly monotonically decreasing for @xmath140 and strictly monotonically increasing for @xmath141 .
this means that @xmath142 if @xmath143 resp .
@xmath144 , then obviously the maximum for both functions is attained at @xmath145 resp . @xmath146 . in the remaining case
@xmath147 we will use that @xmath148 and @xmath149 for all @xmath150 , which can be easily shown .
suppose that the maximum in ( [ eqn : pp1 ] ) is attained at @xmath146 , i.e. , @xmath151
. then also @xmath152 , and thus we must have @xmath153 because of the ( strict ) monotonicity of @xmath138 for @xmath141 .
but since @xmath139 is also ( strictly ) monotonically increasing for @xmath141 , we get @xmath154 . an analogous argument applies when the maximum in ( [ eqn : pp1 ] ) is attained at @xmath145
. for @xmath155 there exist an index @xmath156 and a real number @xmath157 such that @xmath158 the previous lemma then implies @xmath159 where we have used that the function @xmath30 introduced in ( [ eqn : omegapp ] ) is ( strictly ) monotonically increasing for @xmath160 .
ptk showed in @xcite ( also cf .
@xcite or ( * ? ? ?
* example 1.4 ) ) that @xmath30 is a rate of convergence on @xmath44 in the sense of definition [ def : rate ] with the corresponding series @xmath47 given by should not be confused with the singular values @xmath161 of @xmath99 .
this slight inconvenience is the price for retaining the original notation of ptk @xcite . ]
@xmath162 he pointed out in @xcite that @xmath163 for large @xmath18 , and @xmath164 for small @xmath18 . thus , the rate of convergence @xmath30 can describe both a linear and a quadratic phase of the convergence ; cf .
the numerical examples in section [ sec : numerpp ] .
we have shown above that @xmath113 implies @xmath165 for @xmath30 defined in ( [ eqn : omegapp ] ) , and hence we have satisfied condition ( ii ) in theorem [ thm : potpta ] . in order to satisfy condition ( i )
we simply set @xmath166 , then @xmath167 is guaranteed ( recall that @xmath168 ) .
theorem [ thm : potpta ] now gives the following result .
[ thm : pp ] let @xmath169 with @xmath170 , @xmath95 for @xmath96 , and unitary matrices @xmath97 .
then the iterative algorithm ( [ eqn : polar1])([eqn : polar2 ] ) is meaningful and its sequence of iterates @xmath171 converges to the unitary polar factor @xmath172 of @xmath99 .
moreover , with @xmath173 , and @xmath30 resp .
@xmath47 defined as in ( [ eqn : omegapp ] ) resp .
( [ eqn : sigmapp ] ) , we have @xmath174 the inequality ( [ eqn : boundpp ] ) is an equality for all @xmath43 .
it remains to show that ( [ eqn : boundpp ] ) is an equality for all @xmath43 . according to the last part of theorem [ thm : potpta ] it suffices to show that @xmath175 . with @xmath115
we get @xmath176 , so that @xmath177 ( recall that @xmath178 for all @xmath46 by definition of the series @xmath47 . ) the fact that ( [ eqn : boundpp ] ) is an equality for all @xmath43 clearly demonstrates the advantage of using functions rather than just numbers for describing the convergence of iterative algorithms .
the equality could be expected when considering that potra and ptk showed that the convergence estimates obtained by the method of nondiscrete induction applied to newton s method for nonlinear operators in banach spaces are sharp for special cases of scalar functions ; see , e.g. , ( * ? ? ?
* proposition 5.10 ) .
one can interpret the iteration ( [ eqn : polar1])([eqn : polar2 ] ) in terms of @xmath179 uncoupled scalar iterations , and can then derive theorem [ thm : pp ] using results in ( * ? ? ?
* chapter 5 ) or the original papers @xcite . we will illustrate the convergence of the iteration ( [ eqn : polar1])([eqn : polar2 ] ) and the bound ( [ eqn : boundpp ] ) on four examples .
all experiments shown in this paper were computed in matlab . for numerical purposes we first computed a singular value decomposition @xmath93 using matlab s command svd and then applied ( [ eqn : polar1])([eqn : polar2 ] ) with @xmath180 .
the error in every step @xmath12 , given by @xmath181 , is shown by the solid lines in figures [ fig:1pp ] and [ fig:2pp ] .
the pluses ( @xmath182 ) in these figures indicate the corresponding values of the right hand side of ( [ eqn : boundpp ] ) . it should be stressed that the right hand side of ( [ eqn : boundpp ] ) is a simple scalar function that only depends on the value @xmath183 .
this function may also be computed by the more explicit expressions given in @xcite or ( * ? ? ?
* chapter 5 ) .
we present numerical results for four different matrices : generated using matlab s command gallery(moler,16 ) , this is a @xmath184 symmetric positive definite matrix with 15 eigenvalues between @xmath185 and @xmath186 , and one small eigenvalue of order @xmath187 . for this matrix @xmath188
; the numerical results are shown in the left part of figure [ fig:1pp ] .
generated using gallery(fiedler,88 ) , this is a @xmath189 symmetric indefinite matrix with 87 negative eigenvalues and 1 positive eigenvalue . for this matrix @xmath190
; the numerical results are shown in the right part of figure [ fig:1pp ] .
the @xmath191 matrix @xmath192 , where @xmath193 is an @xmath194 jordan block with eigenvalue @xmath195 . for this matrix @xmath196 ; the numerical results are shown in the left part of figure [ fig:2pp ] . generated using gallery(frank,12 ) , this is a @xmath197 nonnormal matrix with ill - conditioned real eigenvalues ; the condition number of its eigenvector matrix computed in matlab is @xmath198 . for this matrix @xmath199 ; the numerical results are shown in the right part of figure [ fig:2pp ] .
( solid ) and @xmath200 ( pluses ) for the moler matrix ( left ) and the fielder matrix ( right ) . ]
( solid ) and @xmath200 ( pluses ) for the moler matrix ( left ) and the fielder matrix ( right ) . ]
( solid ) and @xmath200 ( pluses ) for the jordan blocks matrix ( left ) and the frank matrix ( right ) . ]
( solid ) and @xmath200 ( pluses ) for the jordan blocks matrix ( left ) and the frank matrix ( right ) . ] in the examples we observe the typical behavior of newton s method : an initial phase of linear convergence being followed by a phase of quadratic convergence in which the error is reduced very quickly . while the convergence bound ( [ eqn : boundpp ] ) is an equality for all @xmath43 , it may happen that @xmath201 as well as @xmath202 ; see the results for the moler matrix and the frank matrix . as shown in theorem [ thm : potpta ]
, the error is guaranteed to decrease strictly monotonically after the first step . in practice
the iteration can be scaled for stability as well as for accelerating the convergence ; see @xcite or ( * ? ? ?
* section 8.6 ) for details .
we give some brief comments on scaled iterations in section [ sec : concl ] .
let @xmath92 be a given , possibly singular , matrix .
any matrix @xmath203 that satisfies @xmath204 is called a square root of @xmath99 . for the theory of matrix square roots and many algorithms for computing them
we refer to ( * ? ? ?
* chapter 6 ) .
one of the algorithms for computing matrix square roots is newton s iteration ; cf .
* equation ( 6.12 ) ) : @xmath205 similar to the newton iteration for the matrix polar decomposition , the iteration ( [ eqn : newton1])([eqn : newton2 ] ) can be derived by applying newton s method to a matrix equation , here @xmath206 ; see @xcite for details . as shown in ( * ? ? ?
* theorem 6.9 ) , the iterates @xmath104 converge ( under some natural assumptions ) to the principal square root @xmath207 , and they satisfy the convergence bound @xmath208 where @xmath106 denotes any submultiplicative matrix norm .
this is a bound of the classical form ( [ eqn : classical ] ) that shows quadratic convergence .
the approach for analyzing the iteration ( [ eqn : newton1])([eqn : newton2 ] ) using theorem [ thm : potpta ] contains some differences in comparison with the one for the matrix polar decomposition .
these differences will be pointed out in the derivation .
ultimately they will lead to a convergence result giving formally the same bound as in theorem [ thm : pp ] ( with the appropriate modifications ) , but containing a strong assumption on the initial matrix @xmath209 . as in section
[ sec : polar ] we consider @xmath107 and the spectral norm @xmath106 . now the mapping @xmath3 is defined by
@xmath210 we need to determine @xmath36 and define the sets @xmath17 for some interval @xmath37 , where we will again specify @xmath20 in the end .
clearly , @xmath36 is equal to the set of the invertible matrices in @xmath112 , so that each @xmath113 must be invertible .
the first major difference in comparison to the iteration for the polar decomposition , where the iterates converge to a unitary matrix , is that in case of ( [ eqn : newton1])([eqn : newton2 ] ) the conditioning of the iterates may deteriorate during the process .
in fact , when @xmath99 is ( close to ) singular , any of its square roots will be ( close to ) singular as well . as originally suggested by ptk in @xcite ( also cf .
@xcite ) , such deterioration may be dealt with by requiring that each @xmath113 satisfies @xmath211 where @xmath212 is some positive function on @xmath44 that will be specified below .
further requirements for each @xmath113 are that @xmath213 commutes with @xmath99 ( cf .
( [ eqn : newton1 ] ) ) and that @xmath114 , which is the first condition in ( ii ) of theorem [ thm : potpta ] .
we thus define for the given matrix @xmath92 and @xmath214 the set @xmath215 satisfying ( i ) and the second condition in ( ii ) of theorem [ thm : potpta ] will now prove the convergence of newton s method ( [ eqn : newton1])([eqn : newton2 ] ) to a matrix @xmath216 with the error norm in each step of the iteration bounded as in ( 2.c ) of theorem [ thm : potpta ] .
we will have a closer look at the set @xmath32 below .
for the second condition in ( ii ) , let @xmath46 and @xmath113 be given . since @xmath213 commutes with @xmath99 , we see that the matrix @xmath217 commutes with @xmath99 .
we need to show that @xmath218 for a positive function @xmath212 on @xmath44 and a rate of convergence @xmath30 on @xmath44 . using @xmath219 and @xmath220
, we get @xmath221 from which we obtain the condition @xmath222 next , note that @xmath223 can equivalently be written as @xmath224 using this and the fact that @xmath213 and @xmath225 commute , we get @xmath226 consequently , @xmath227 which gives the condition @xmath228 the inequality in ( [ eqn : ineq ] ) represents the second major difference in the derivation compared to the matrix polar decomposition . in ( [ eqn : ineq ] ) we have used the submultiplicativity of the matrix norm , while in ( [ eqn : eq])([eqn : pp3 ] ) no inequalities occur because the matrices are effectively diagonal when considered under the unitarily invariant norm .
the absence of inequalities in the derivation ultimately led to a convergence bound in theorem [ thm : pp ] that is attained in every step @xmath43 . the two ( sufficient ) conditions ( [ eqn : cond1 ] ) and ( [ eqn : cond2 ] ) for @xmath229 form a system of functional inequalities that was ingeniously solved by ptk in @xcite .
here it suffices to say that for any @xmath230 the rate of convergence @xmath30 on @xmath231 and the corresponding function @xmath47 given by @xmath232 ( i.e. ( [ eqn : omegapp ] ) and ( [ eqn : sigmapp ] ) for @xmath233 ) and the function @xmath234 which is positive on @xmath235 , satisfy the conditions ( [ eqn : cond1 ] ) and ( [ eqn : cond2 ] ) .
it remains to satisfy the condition ( i ) in theorem [ thm : potpta ] , i.e. , to show that for some @xmath236 there exists an @xmath237 .
this will determine the parameter @xmath230 , which so far is not specified .
we require that @xmath209 commutes with @xmath99 , and that @xmath238 if we define @xmath239 then the second inequality is automatically satisfied .
some simplifications of the first inequality lead to @xmath240 hence we must have @xmath241 , and @xmath242 is a feasible choice of the parameter in order to satisfy all conditions . theorem
[ thm : potpta ] now shows that under these conditions the iterative algorithm ( [ eqn : newton1])([eqn : newton2 ] ) is meaningful and converges to some @xmath243 .
let us now consider the set @xmath32 .
first note that , for any @xmath244 and @xmath113 , @xmath245 see ( [ eqn : asquare ] ) . taking the limit
@xmath125 we see that each @xmath246 must satisfy @xmath204 .
moreover , each @xmath246 commutes with @xmath99 and satisfies @xmath247 . if @xmath248 , then all matrices in the limiting set @xmath32 are invertible and @xmath249 . if @xmath250 , then a singular limiting matrix is possible and in this case @xmath32 is not a subset of @xmath36 .
this will be discussed further below and numerically illustrated in the next section . in summary
, we have shown the following result .
[ thm : main ] let @xmath92 and let @xmath251 be invertible and commute with @xmath99 . if @xmath252 then the iterative algorithm ( [ eqn : newton1])([eqn : newton2 ] ) is meaningful and its sequence of iterates @xmath171 converges to a matrix @xmath253 that commutes with @xmath99 and satisfies @xmath254 .
moreover , with @xmath30 and @xmath47 as in ( [ eqn : rate ] ) , where @xmath255 as in ( [ eqn : gammadef ] ) , we have @xmath256 if @xmath248 , then @xmath257 is invertible with @xmath258 .
we will now discuss the condition ( [ eqn : condition ] ) .
first note that from @xmath259 it follows that ( [ eqn : condition ] ) is implied by @xmath260 .
it therefore appears that the condition is satisfied whenever @xmath209 commutes with @xmath99 , is well conditioned and sufficiently close ( in norm ) to a square root of @xmath99 .
however , as we will see below , the condition ( [ eqn : condition ] ) seriously limits the applicability of theorem [ thm : main ] .
the fact that the initial matrix @xmath209 has to satisfy some conditions ( unlike in the case of theorem [ thm : pp ] ) is to be expected , since not every matrix @xmath99 has a square root ; cf .
* section 1.5 ) .
if an invertible matrix @xmath209 that commutes with @xmath99 and satisfies ( [ eqn : condition ] ) can be found , then the convergence of the newton iteration guarantees the existence of a square root of @xmath99 .
ptk referred to this as an _ iterative existence proof _
@xcite .
now let @xmath261 be real and symmetric with the real eigenvalues @xmath262 and let @xmath263 for some real @xmath264
. then @xmath209 is invertible and commutes with @xmath99 , and ( [ eqn : condition ] ) becomes @xmath265 this condition can not be satisfied if @xmath266 .
( note that in this case the principal square root of @xmath99 does not exist ( * ? ? ?
* theorem 1.29 ) . ) on the other hand , if @xmath267 and hence @xmath99 is positive semidefinite , then a simple computation shows that each matrix of the form @xmath268 satisfies ( [ eqn : condition ] ) and thus the conditions of theorem [ thm : main ] .
if @xmath269 and hence @xmath99 is singular , then for every @xmath270 equality holds in ( [ eqn : condalpha1 ] ) .
we then get @xmath250 in ( [ eqn : gammadef ] ) , and in this case the rate of convergence is @xmath271 a numerical example with a real , symmetric and singular matrix is shown in the right part of figure [ fig:1 ] .
the following result shows that , despite the inequality in ( [ eqn : ineq ] ) , the bound ( [ eqn : bound ] ) is attained in every step for certain @xmath99 and @xmath209 .
[ cor : exactsq ] if @xmath272 is symmetric with eigenvalues @xmath273 and @xmath209 is as in ( [ eqn : x0 ] ) , then the conditions of theorem [ thm : main ] are satisfied , the sequence of iterates @xmath171 of ( [ eqn : newton1])([eqn : newton2 ] ) converges to the principal square root @xmath207 of @xmath99 , and the bound ( [ eqn : bound ] ) is an equality in each step @xmath12 .
we already know that @xmath99 and @xmath209 satisfy the conditions of theorem [ thm : main ] .
let @xmath274 be an eigendecomposition of @xmath99 with @xmath275 orthogonal and @xmath276 .
then @xmath277 with @xmath278 is the principal square root of @xmath99 . it suffices to show that @xmath279 , then the result follows from the last part of theorem [ thm : potpta ] .
using @xmath263 and @xmath280 we get @xmath281 and thus @xmath282 where we have used that @xmath283 and @xmath284 .
similar to theorem [ thm : pp ] , the sharpness result in the above corollary could also be derived using results of potra and ptk ( e.g. , ( * ? ? ?
* proposition 5.10 ) ) , when one interprets the iteration ( [ eqn : newton1])([eqn : newton2 ] ) for a real and symmetric positive semidefinite matrix @xmath99 and with @xmath263 in terms of @xmath179 uncoupled scalar iterations . finally , we note that for a real and symmetric positive definite matrix @xmath99 with eigenvalues @xmath285 the condition ( [ eqn : condition ] ) with @xmath286 and a real @xmath264 becomes @xmath287 an elementary computation shows that this condition can not be satisfied for any real @xmath264 when @xmath288 . if @xmath289 , then @xmath290 for some @xmath291 , and @xmath292 with @xmath293 is a feasible choice . obviously , the conditions on @xmath99 and @xmath294 for the applicability of theorem [ thm : main ] are in this case much more restrictive than for initial matrices of the form @xmath263
. we will now illustrate theorem [ thm : main ] on some numerical examples . as originally shown by higham @xcite , the newton iteration for the matrix square root in the form ( [ eqn : newton1])([eqn : newton2 ] )
is numerically unstable ; see also ( * ? ? ?
* section 6.4 ) . for our numerical experiments
we have used the mathematically equivalent `` in iteration '' ( cf .
* equation ( 6.20 ) ) ) , originally derived by iannazzo @xcite : @xmath295 this variant is numerically stable and has a limiting accuracy on the order of the unit roundoff ( * ? ? ?
* table 6.2 ) . in practice
one uses scalings for accelerating the convergence ( * ? ? ?
* section 6.5 ) ; for brief comments on scalings see section [ sec : concl ] .
we use three initial matrices of the form @xmath296 this choice of @xmath297 is motivated by the argument leading to ( [ eqn : x0 ] ) , although this argument only applies to real symmetric matrices .
we point out that for none of our example matrices @xmath99 an initial matrix of the form @xmath298 satisfies the condition ( [ eqn : condition ] ) , although the newton method always converges when started with them ( except for the singular matrix defined below , where @xmath298 can not be used ) .
this demonstrates the severe restriction of the condition ( [ eqn : condition ] ) for the general applicability of theorem [ thm : main ] . for the given @xmath209 we first run the iteration ( [ eqn : in1])([eqn : in2 ] ) until it converges to a matrix @xmath257 that is an approximate square root of @xmath99 .
typically in our examples @xmath299 .
we then run the iteration again and compute the error @xmath300 in each step @xmath12 .
these errors are shown by the solid lines in figures [ fig:1][fig:3 ] .
the corresponding a priori values @xmath200 of the upper bound ( [ eqn : bound ] ) are shown by the pluses @xmath301 in figures [ fig:1][fig:2 ] .
we present numerical results for the moler matrix ( left part of figure [ fig:1 ] ) , the jordan blocks matrix ( left part of figure [ fig:2 ] ) and the frank matrix ( figure [ fig:3 ] ) as defined in section [ sec : numerpp ] .
we do not use the symmetric indefinite fiedler matrix from section [ sec : numerpp ] since the newton iteration ( [ eqn : newton1])([eqn : newton2 ] ) for this matrix does not converge when started with @xmath302 .
instead , we present results for the following matrices : the @xmath303 diagonal matrix with eigenvalues @xmath304 .
the results for this matrix are shown in the right part of figure [ fig:1 ] .
the @xmath191 matrix @xmath305 , i.e. , the previous jordan blocks matrix with the eigenvalue 1.5 replaced by 1.0 .
the results for this matrix are shown in the right part of figure [ fig:2 ] .
for the moler matrix and the singular matrix we have @xmath306 for @xmath307 . hence the bound ( [ eqn : bound ] )
is sharp by corollary [ cor : exactsq ] , which is illustrated by the numerical results . while for the moler matrix we observe the typical behavior of newton s method , the convergence for the singular matrix is linear throughout , which is explained by ( [ eqn : singular ] ) .
the bound ( [ eqn : bound ] ) is very tight for the jordan blocks matrix , but for the modified jordan blocks matrix the actual convergence acceleration in the quadratic phase is larger than the one predicted by the bound .
finally , for the frank matrix none of the initial matrices @xmath302 with @xmath308 ( here @xmath309 ) satisfies the condition ( [ eqn : condition ] ) , and hence no convergence bound is given by theorem [ thm : main ] .
the constants in condition ( [ eqn : condition ] ) have the following values : [ cols="^,>,^",options="header " , ] although theorem [ thm : main ] is not applicable , the newton iteration ( eventually ) converges , as shown in figure [ fig:3 ] .
the figure also shows that the convergence is characterized by an initial phase of complete stagnation .
it is a challenging task to determine a rate of convergence that can closely describe such behavior .
( solid ) and @xmath200 ( pluses ) for the moler matrix ( left ) and the singular matrix ( right ) . ]
( solid ) and @xmath200 ( pluses ) for the moler matrix ( left ) and the singular matrix ( right ) . ]
( solid ) and @xmath200 ( pluses ) for the jordan blocks matrix ( left ) and the modified jordan blocks matrix ( right ) . ]
( solid ) and @xmath200 ( pluses ) for the jordan blocks matrix ( left ) and the modified jordan blocks matrix ( right ) . ]
( solid ) for the frank matrix . ]
the main purpose of ptk s method of nondiscrete induction is to derive error bounds for iterative processes that can be tight throughout the iteration .
as demonstrated for the newton iteration for the matrix polar decomposition and the matrix square root , the approach can indeed give very tight , even sharp bounds in practical applications .
however , we have also seen that the approach has some limitations , in particular for ( highly ) nonnormal matrices .
the results for the matrix square root iteration applied to the frank matrix illustrate an observation we made : for a general nonnormal matrix it appears to be very difficult to determine ( easily and a priori ) an initial matrix @xmath209 that satisfies the assumptions of theorem [ thm : main ] .
the derivation of strategies for computing such @xmath209 is beyond the scope of this paper .
we also did not analyze more closely the observations about the tightness of the bound ( [ eqn : bound ] ) for the two jordan blocks matrices .
such investigations are subject of future work .
note that the difficulty of nonormality does not appear in case of the newton iteration for the matrix polar decomposition , since it effectively acts on diagonal matrices .
convergence of an iteration means that a solution of a certain equation exists . in case of the newton iteration for the matrix square root
this equation is @xmath206 for the given ( square ) matrix @xmath99 .
finding an initial matrix @xmath209 that satisfies the assumptions of theorem [ thm : main ] guarantees that the matrix @xmath99 has a square root ( `` iterative existence proof '' ) .
the fact that existence and convergence are studied and proven simultaneously may be one reason why the ( sufficient ) conditions in theorem [ thm : main ] are more restrictive than other types of conditions for convergence of the iteration under the assumption that a square root exists . from a practical point of view
it would be of interest to obtain variants of theorem [ thm : main ] with weaker conditions on @xmath209 . as described in section [
sec : method ] , an essential constructive idea in the method of nondiscrete induction is the comparison of consecutive iterates .
the method therefore is suited best for iterations where a new iterate is generated using information only from the previous step , and newton - type methods are ideal examples in this respect .
methods that generate new iterates based on information from several or all previous steps possibly can be analyzed using the concept of _ multidimensional rates of convergence _ described in ( * ? ? ?
* chapter 3 ) .
it also would be interesting to appropriately generalize the theory of potra and ptk to scaled or , more generally , parameter - dependent iterations . as briefly mentioned in sections [ sec : numerpp ] and [ sec : numer ] , scalings are used in practice for numerical stability properties or for accelerating newton - type matrix iterations . including parameters that are not known a priori within the method of nondiscrete induction is an interesting subject of further work . in the framework established by theorem [ thm : potpta ]
it may be possible to include parameters depending only on the latest iterate @xmath14 by appropriately defining the sets @xmath21 .
more sophisticated parameter - dependence requires extensions of theorem [ thm : potpta ] .
such extensions may also lead to new convergence results for nonlinear iterative schemes like krylov subspace methods ( see , e.g. , @xcite ) , where new iterates are generated using information from all previous steps and parameters that are not known a priori .
, _ newton methods for nonlinear problems .
affine invariance and adaptive algorithms _ , vol .
35 of springer series in computational mathematics , springer , heidelberg , 2011 .
first softcover printing of the 2006 corrected printing . | vlastimil ptk s method of nondiscrete induction is based on the idea that in the analysis of iterative processes one should aim at rates of convergence as functions rather than just numbers , because functions may give convergence estimates that are tight throughout the iteration rather than just asymptotically . in this paper
we motivate and prove a theorem on nondiscrete induction originally due to potra and ptk , and we apply it to the newton iterations for computing the matrix polar decomposition and the matrix square root .
our goal is to illustrate the application of the method of nondiscrete induction in the finite dimensional numerical linear algebra context .
we show the sharpness of the resulting convergence estimate analytically for the polar decomposition iteration and for special cases of the square root iteration , as well as on some numerical examples for the square root iteration .
we also discuss some of the method s limitations and possible extensions .
nondiscrete induction , matrix iterations , matrix polar decomposition , matrix square root , matrix functions , newton s method , convergence analysis 65f30 , 65h05 , 65j05 |
one of the main goals of quantum theory formalism is a description of time evolution of quantum systems .
usually the evolution is regarded as a two step procedure which includes reversible , unitary evolution and irreversible `` measurement '' handled by the so called projection or reduction postulate .
there are some models which come across attempts to unify both , qualitatively different steps as for example @xcite .
however some of the major questions are not satisfactory solved so far .
the main directions of the investigations are related to a few major modifications and interpretations of quantum theory like consistent histories approaches @xcite , modal interpretations @xcite , quantum jumps and collapse ideas @xcite and different types of models based on stochastic equations @xcite .
the measurement process is treated differently depending on interpretation of quantum mechanics .
there is no place here to discuss all the possible interpretations .
a good , but because of very dynamic development of the field , not quite up - to - date review can be found in @xcite , see also @xcite and references therein .
the fundamental part of description of measurement seems to be the projection postulate which , from the orthodox point of view , determines the state of the system just after measurement .
however , this procedure leads to inconsistencies with the schrdinger equation .
in other words , the usual description of a process of quantum time evolution requires a coexistence of two different evolution mechanisms .
this leads to additional problems like the time duration and dynamics of the `` wave function collapse '' @xcite , a general problem of physical reality before and after measurement @xcite and non local like behavior in case of spatially separated quantum subsystems @xcite .
this short introduction is , of course , far from being complete , but we can conclude that the problems mentioned above are still opened and their explanation requires probably fundamental changes of the evolution postulates of quantum theory .
the goal of this paper is to propose an alternative model of quantum evolution . in the following , we work in a nearly standard framework of quantum mechanics replacing only the unitary time evolution and projection postulate by a projection evolution ( _ pev _ )
it means that _ we reject the schrdinger equation _ as the fundamental equation of motion .
it can be recovered as a special case of the projection evolution .
the proposed evolution procedure unites both the unitary evolution and the process of measurement . within this idea
the evolution is only a sequence of some projections made randomly by nature , and with a system dependent probability distribution .
the sequence of projections is dependent on structure of the physical system under consideration .
the procedure can be extended to a povm scheme @xcite , as well .
the main difference among our proposal and other major models is that the _ pev _ formalism allows to treat time on equal footing with the spatial variables .
time is not a parameter but a dynamical variable_. this feature makes the model hardly comparable with the ones in which the physical time is the evolution parameter .
nevertheless , some aspects of the _ pev _ idea are similar to well known models . in this sense , this idea lies somewhere between the decoherent histories approach @xcite and some modal interpretations ( see pp . :
108 , 179 , 199 , 253 in ) @xcite .
in fact , in the decoherent histories approach the paths of evolution are determined partially by a family of decompositions of unity .
on the other hand , the preferred basis can define such decomposition in the modal theories .
these decompositions of unity describe the physical , basic properties of quantum systems and play a role the evolution operator within the _ pev _ approach .
one can also find some formal similarities of the projection evolution to the analysis of the quantum zeno and anti - zeno effects e.g. , @xcite .
however , the basic mechanism we propose does not lead to a sequence of subsequent measurements only , but to random choices of the next states of the system , where the probability distribution is a function dependent on previous states of the system .
the idea of the _ pev _ can be treated as a special case of more general concept : 1 .
the states of a quantum system are described by either density operators or the appropriate functionals , and the term `` quantum evolution '' is understood as a sequence of state changes ordered by the the nature .
+ this idea requires introducing of an ordering parameter which orders the subsequent changes of the state of a physical system .
this parameter , denoted by @xmath0 will be called the evolution parameter .
+ the above statements requires additional explanations .
usually changes are related to the time parameter . in general , in our approach , there is no direct relation between both , the time and state changes .
we suggest that the natural change of the state of a physical is the primitive notion .
the space and time are the additional structures of our universe .
they are dynamical variables in the theory .
+ the standard non relativistic quantum mechanics can be considered as approximation in which both the evolution parameter and the physical time are in one to
one correspondence . in this case , the time is not a dynamical * variable * , but a parameter which can be identified with the evolution parameter of the _ pev _ .
+ both , the evolution parameter and the physical time can be approximately related one to another for the physical systems described by the states which are strongly localized in the time variable .
this can explain the successes of the traditional approach but it leads also to some well known paradoxes .
we will consider this type of quantum systems in the the section containing a few examples .
+ many problems like time energy uncertainty principle , time of arrival , life
time problems , the young type experiments for time slits @xcite and similar phenomena suggest that the physical time should be the dynamical variable not a parameter .
it is much more natural assumption .
+ in addition , the relativistic quantum mechanics , in fact , requires that both notions the evolution parameter and the time variable should be different .
note , that mixing of the space and time coordinates by the lorentz transformations results in the undetermined status of transformed coordinates .
they are neither parameters nor variables . the same one can say even about galilean transformations .
+ we claim that the physical time should be treated on the same footing as the other spatial coordinates .
it should be also the dynamical variable which spans the additional dimension in the physical space .
the physical states should determine localization of the system in space and time according to quantum rules .
+ we suppose that the evolution parameter @xmath0 is a common parameter for the whole universe . in this sense
it parameterizes a kind of the `` internal clock '' of the universe . because @xmath0 is the parameter it can not be affected by physical interactions ( the physical time can ! ) + for simplicity we assume that @xmath0 is a real number parameter .
2 . there exists a world lottery mechanism ( chooser ) which chooses randomly the next state of quantum system in a way dependent on physical structure of this quantum system .
+ the realization of the chooser is based on the so called selected collapse projection postulate ( see @xcite and references therein ) .
our chooser is not driven by the random noise but it is a kind of the _ quantum state dependent world lottery_. + the evolution parameter @xmath0 is an internal parameter of the `` world lottery '' mechanism . for every @xmath0 the nature draws lots to get next state of the universe .
+ it means , that the projection evolution ( _ pev _ ) can be considered as a stochastic process ( in respect to @xmath0 , instead of time ) determining , at each step of evolution , a physical state of the system represented by a density operator @xmath1 .
the probability distribution for the chooser is determined , in some way , by the previous states of the quantum system .
the evolutions of a larger quantum system and its subsystems have to be consistent .
+ because we assume that the evolution parameter @xmath0 is common for our universe , the universe itself changes states as a whole .
it means that the effective evolution for a given subsystem should be induced by the global evolution .
+ the description of the global evolution ( for the whole universe ) is in practice not available , and in fact , we are able to write down only evolution operators for larger or smaller subsystems .
these effective operators have to satisfy some scaling conditions . in the next section a proposal for scaling condition
is discussed .
however , this is still open problem . because we expect that such fundamental mechanism as evolution of quantum system should be a rather simple physical law , consistent with known results of quantum mechanics , we propose a rather natural realization of this idea in the next paragraph .
the approach is _ observer free_. we do not need neither to split the physical world into _ observer and the physical objects _ which have to be observed nor _ to introduce a mind _ , which allows to choose the state after observation , as is required in some approaches ( collapse of states is the most natural process in the _ pev _ ) .
the _ pev _ do not require also _ splitting of the universe into classical and quantum worlds_. we do not need also any strange assumptions about _ our `` knowledge in which way '' _
e.g. , the particle passes through the slits .
these results are a consequence of a single principle postulated within the _ pev _ approach .
let us consider a quantum system described by states represented by the quantum density operators @xmath1 . in the following by @xmath0
we denote the evolution parameter which orders the evolution of the universe .
in addition , for each @xmath0 we define a family of projectors which gives an orthogonal resolution of unity i.e. , roughly speaking , for each @xmath0 they fulfill the following conditions : @xmath2 where @xmath3 denotes the unit operator .
+ the operators @xmath4 should represent the essential constraints of the physical system , responsible for its time evolution .
the hamiltonian plays the same role in traditional approach . in this sense
, they play a role of the evolution operators because they force changes of the physical system which they describe .
the indices @xmath5 represent here the sets of quantum numbers labelling the projection operators ( [ resunit ] ) , uniquely . instead of
usual unitary evolution we postulate that the new state @xmath6 of the physical system , at the evolution parameter @xmath0 , is created from the previous state @xmath7 by _ randomly chosen , projection _
we will show a few examples how to construct such families of projectors , in next section .
the projectors @xmath4 are chosen randomly according to the following probability distribution : @xmath8 , \label{(!e!2 ) } \ ] ] where @xmath5 runs over the sets of quantum numbers describing the system at the evolution parameter @xmath0 and @xmath9 is the set of quantum numbers which describes the previous state .
it means , that the nature applies the projection postulate to get a new state from the previous one : @xmath10}. \label{quant.3 } \ ] ] to simplify our notation let us imagine that the projections @xmath4 are constant functions of the evolution parameter on the intervals @xmath11 , where @xmath12 . using this notation the idea of _
pev _ leads to the following recurrence equation for states of the system under consideration : @xmath13}. \label{(!e!3 ) } \ ] ] in other words , the equation ( [ quant.3 ] ) describes the following mechanism : having the state of a physical system given by @xmath14 at the evolution parameter @xmath15 , the next state at @xmath16 is chosen randomly , with the probability distribution ( [ ( !
e!2 ) ] ) , from all possible states ( [ ( !
e!3 ) ] ) , where @xmath17 runs over the whole range required by ( [ resunit ] ) . the applied projection postulate is in the form of the so called `` selective collapse '' @xcite . an important feature of the formula ( [ ( !
e!2 ) ] ) is a conservation of `` purity '' of states .
if the previous state is pure , the next one is also a pure state .
this allows to omit the problem of `` ensemble of worlds '' as mentioned in @xcite .
it is important to note that in this approach we do not need any `` environment '' ; the same procedure should be applied for microscopic and macroscopic systems with appropriate projection evolution operators , related by the scaling condition mentioned in the introduction and which can be postulated as follows : having a density operator @xmath18 for larger system @xmath19 , it is commonly assumed that the density operator @xmath20 of the subsystem @xmath21 can be obtained by partial trace over the residual subsystem @xmath22 i.e. , @xmath23 . to have consistent description both @xmath19 and @xmath24 one needs to fulfill the following obvious condition ( scaling property ) : @xmath25 } = \frac{{\mathsf{e}\kern-1pt\llap{$\vert$}(a,\tau;\mu)}\rho_a(\tau - d\tau ) { \mathsf{e}\kern-1pt\llap{$\vert$}(a,\tau;\mu ) } } { { \textrm{tr}}_a[{\mathsf{e}\kern-1pt\llap{$\vert$}(a,\tau;\mu)}\rho_a(\tau - d\tau){\mathsf{e}\kern-1pt\llap{$\vert$}(a,\tau;\mu ) } ] } , \label{(!e!3a ) } \ ] ] where @xmath26 are required effective _ pev _
operators for the subsystem @xmath24 with the appropriate sets of quantum numbers @xmath27 corresponding , in some way , to the quantum numbers @xmath5 of the larger system @xmath19 .
it means that drawing of lots required by the _ pev _
approach should be correlated within the connected parts of the universe . in principle , having the state @xmath18 and @xmath28 for the universe the condition ( [ ( !
e!3a ) ] ) should allow obtaining the appropriate states and the _ pev _ operators for required subsystems .
till now , we have no proof for uniqueness of solutions of the equations ( [ ( !
e!3a ) ] ) in respect to @xmath26 . the general equation ( [ ( ! e!3 ) ] ) suggests that the process of evolution can follow various paths @xmath29 considered as the series of projections chosen randomly at the evolution parameters @xmath30
. then one can ask about the conditional probability of choosing of the state @xmath14 under condition , that the previous states chosen by the nature are described by @xmath31 . at this point
, one needs to observe that the _ pev _ generates a set of `` quantum histories '' consisted of products of projection operators similar to those in @xcite .
however , the histories are ordered in respect to the evolution parameter , not by the physical time . in this case
there is no algebra of histories because the only available `` histories '' are those defined by the _
operator which does not allow for logical operations on the evolution paths .
for example , in notation of @xcite , the negation of the * compound * `` history '' like @xmath32 has no meaning in _ pev _ , because _ pev _ represents a step by step stochastic process and @xmath33 can not be ordered by the evolution parameter . the same one can say about the other logical operations on the paths . according to the _ pev _
idea , the main physical interest is not in the products of projections only , but also in series of states which they produce .
more precisely , the system is fully described at the evolution parameter @xmath0 if we have the state and the last projection operator responsible for generating this state from the previous one a kind of history of the system is a sequence of pairs of quantum states and the appropriate projections .
the probability of finding a given path of evolution at the evolution parameter @xmath0 can be calculated using equations ( [ ( !
e!3 ) ] ) and ( [ ( !
e!2 ) ] ) and can be written as : @xmath34 .
\label{(!e!4 ) } \ ] ] in this way we have defined the evolution of quantum systems as a kind of a stochastic game played by nature .
this game is not parameterized by the physical time but the evolution parameter .
within this idea we treat this game as a fundamental process a law of nature .
in this section we show three simple examples of description in terms of the projection evolution .
the most important seems to be a reproducing of the schrdinger like evolution .
it is the purpose of the next subsection in which we show how to derive the schrdinger evolution using a rather general method of generating operator .
the other examples concern a toy model of mach - zehnder interferometer and the evolution of a particle passing through the device .
let us introduce a set of mutually commuting hermitian operators @xmath35 which describes the essential properties of the system within the space time and other degrees of freedom .
it is important to notice that , due to the spectral theorem , there exists a common decomposition of unity for all the operators @xmath36 , @xmath37 .
this resolution of unity should give the _ pev _
the operators @xmath38 will be called the evolution generating operators .
let us consider the 3d harmonic oscillator , but in the non
relativistic space time . for this purpose
we introduce the single particle state space ( without spin ) @xmath39 in the four dimensional real space @xmath40 .
the scalar product in @xmath41 is defined as : @xmath42 and it contains the integration over the time .
it means that the time is dynamical variable .
now we define the evolution generating operators as : @xmath43 where @xmath44 denotes the usual harmonic oscillator hamiltonian , @xmath45 is the square of the angular momentum operator and @xmath46 denotes the third component of the angular momentum operator .
all the operators @xmath47 , @xmath48 are hermitian and they are the subject of the spectral theorem . the first operator @xmath49 has the continuous spectrum , the second one and the third have the discrete spectrum .
it allows to write the common spectral measure as a product of spectral measures of these three operators : @xmath50 where @xmath51 denotes the spectral measure for @xmath49 , @xmath52 , the decomposition of unity for @xmath45 operator and @xmath53 the corresponding decomposition of unity for @xmath46 .
the suggested evolution operator should be of the following form : @xmath54 and , in fact , it is independent of evolution parameter @xmath0 .
as usually , there are known difficulties for the continuous spectrum case .
these problems can be solved in many ways e.g. , by making use of the smeared out operators method @xcite .
the projection evolution operator ( [ projevho ] ) leads to the evolution
parameter independent evolution . in this case
, the first random choice decides about the state of the system which do not evolve further with increasing @xmath0 . for every @xmath0 the state chosen by the nature is of the following form : @xmath55 where @xmath56 are numerical coefficients .
the vectors @xmath57 are the common eigenvectors of the three operators : the harmonic oscillator hamiltonian @xmath44 , @xmath45 and @xmath46 with the eigenvalues @xmath58 , @xmath59 and @xmath60 , respectively .
the `` only '' difference between the standard approach and _ pev _ is that the functions ( [ gener : eq.4 ] ) are functions of time , where time is a dynamical variable .
it means that such states are not localized in time .
as it was mentioned above , the vectors ( [ gener : eq.4 ] ) can be normalized to the dirac delta distribution .
however , in reality one can expect the states in the form of wave packets : @xmath61 for example , the measurement of position in time ( it is a special kind of interaction within the space time ) leads , by definition , to the localization in time . for @xmath62
the states ( [ wavepackt ] ) are `` perfectly '' localized in time @xmath63 : @xmath64 the states ( [ lwavepackt ] ) correspond exactly to the traditional schrdinger type solutions for the harmonic oscillator .
however , one needs to realize that within the _ pev _ these solutions represent the states which are localized in the fixed time @xmath63 .
it means , that one needs to have some additional interactions to `` shift '' these states in time .
this is what we are doing in the traditional unitary evolution approach . within the standard quantum mechanics one assumes implicitly , that there exists a `` time machine '' which evolves the state perfectly localized in time to later moments
this mechanics describes the state vectors projected onto the subsequent moments of time loosing possible non local time correlations . on the other hand
, the state ( [ wavepackt ] ) describes a full time behavior of the system as a single vector which allows to calculate typical statistical characteristics like averages , variations and other statistical moments .
for example , the average time position of the harmonic oscillator described by the wave packets of the form @xmath65 can be obtained in the usual way by the integration : @xmath66 obviously , this example can be generalized to any arbitrary system which one can describe in the traditional way . in this paragraph
we consider a very simplified mach - zehnder interferometer .
the aim of the considerations is to show the temporal behavior of the interferometer in two models .
the first model does not correspond to the schrdinger type motion because the beam splitters localize the particle .
the second model is based on more traditional thinking about the beam splitters as some unitary devices . in both cases
we get the similar results which differs mainly in temporal part of the description .
the main disadvantage of both models is that they do not show , in general possible , spreading in time of states of the system .
because of this , within these toy models , the time of arrival of a particle to the detector is a sharp number , without random distribution
. however , this type of models can be analyzed on the elementary level and they are able to show some interesting features of the formalism . the mach - zehnder interferometer shown in [ fig1 ] allows to show the quantum interference .
the particles are produced in the source _ z_. the first beam splitter @xmath67 splits an incoming beam into two separate channels .
next , there is a phase shifter
_ ps _ and the second beam splitter @xmath68 .
@xmath69 and @xmath70 denote the detectors which detect the particle in the first or the second output channel .
however , for our purpose we consider a simplified version of the mach - zehnder interferometer .
it is shown in the figure [ fig2 ] it is able to simulate , to some extend , the mach zehnder interferometer shown in the figure [ fig1 ] for example , it happens for the @xmath71spin particles or even polarized photons . in the following we are using the toy model ( like in sect .
2.5 of @xcite ) in which the space is discrete .
however , in our approach the time is also a dynamical variable and we have to use not one dimensional space but two dimensional space
time for which both the time @xmath72 and the coordinate @xmath73 takes on only a finite numbers of integer values : @xmath74 like in @xcite we assume the `` periodic boundary conditions '' for both the time and the space coordinates , so that last sites for @xmath72 and @xmath73 are adjacent to the first sites . in this way ,
our space
time * x*=@xmath75 is the cartesian product of two discrete `` circles '' .
first one representing the time t consists of @xmath76 points and the second one x contains @xmath77 space coordinate points , respectively . in addition , we assume an extra two valued variable @xmath78 which determines two possible channels in which each space time point can be located ( see ch .
12 of @xcite ) .
the space @xmath79 of quantum states consists of complex functions : @xmath80 the scalar product in the state space @xmath79 is : @xmath81 again , it is very important to keep in mind , that time is not a parameter but a dynamical variable .
it does not enumerate the subsequent events and does nt describe the evolution of the system in this sense .
more , it is even possible that two subsequent , ordered by the evolution parameter @xmath0 , events occur in the inverse order in time . in this way we can have so called backward causation required for quantum world by some authors @xcite . to construct the projection evolution operator we have to analyze the physical structure ( in required approximation ) of devices shown in the [ fig2 ] .
let us assume that the source _ z _ produces the particles localized in both the time @xmath72 and the position @xmath82 i.e. , in the states @xmath83 which are defined as the `` eigenfunctions '' of time , position and channel operators . other words , the states @xmath84 represent the functions localizing the particle in point @xmath85 of the space time and in the channel @xmath86 : @xmath87 note that @xmath88 and @xmath89 are here the quantum numbers but @xmath90 and @xmath91 denote the corresponding variables .
starting from the source , the particle is created in one of the possible states @xmath92 i.e. , at a randomly chosen time @xmath72 , at fixed place @xmath82 and in the @xmath93 channel .
there are also possible other , less interesting processes which are represented here by a set of projection operators denoted by @xmath94 ( other processes ) .
together with the creation of particle they complete a resolution of unity . at this moment
we do not need to analyze them more carefully .
the assumptions allow to write the following projection evolution operator for the source : @xmath95 note that the operator is independent of @xmath0 . as the next step
we have to construct the toy model of free evolution .
this type of evolution is required to describe behavior of the particle outside of devices i.e. , in the free space time . in our toy model ( i )
we assume the free evolution of projective type but which resembles the schrdinger s type of evolution .
for this purpose we define the unitary transformation @xmath96 ( see the toy models in @xcite ) : @xmath97 where the ket @xmath98 represents the function @xmath99 , see ( [ locfun ] ) .
the operator @xmath100 is the unitary operator within the space spanned by the kets @xmath98 i.e. , spanned by the functions @xmath99 .
the last property allows to define the set of vectors which allow to define the schrdinger s type of evolution .
these vectors will be denoted by : @xmath101 where @xmath102 is the local ( it is a function of time ! ) unitary operator in ( [ stransform ] ) .
we assume that the set of functions @xmath103 , where @xmath104 is an integer , give an orthonormal basis in the space of all functions dependent only on time : @xmath105 the functions @xmath103 determine the main part of the amplitude describing the fact that the particle can be localized in time ( see remarks below the eq .
( [ prpsi ] ) ) .
it implies that the functions @xmath106 form an orthonormal basis in the space of states @xmath79 .
because these vectors are independent of the evolution parameter , the projection evolution operator for free evolution can be considered as a single random choice ( like in the first example ) : @xmath107 observation : the action of the operator ( [ pevfree ] ) on a vector state @xmath108 gives a schrdinger
type function : @xmath109 the nature chooses by chance ( with the distribution determined by the last state ) the actual state labelled by @xmath5 , which characterizes mainly the potential localization of the particle in the physical time . in this model , time and space positions are treated on the same footing .
it means that there are possible situation when particle can not be found in some regions of time .
however , this situation can not be interpreted in the usual way that such particle does not exist .
existence or non existence of the particle is determined here by the total state considered in the full space time .
next device in our system is the beam splitter which can be thought as a device with two input channels and two output channels . in this model
we assume that the beam splitters are able to localize particles in the space time .
it means that they are not the unitary type devices . to construct the appropriate operators ,
let us define a family of vectors which allows to `` send '' a particle into both channels simultaneously : @xmath110 \nonumber \\
{ |{bs_2;t\ x_{bs}\ \varepsilon}\rangle } = & - \sin\varepsilon\ , \left[\frac{1}{\sqrt{2 } } \left(-{|{t , x_{bs}+1,b}\rangle } + { |{t , x_{bs}+1,a}\rangle}\right ) \right ] + \cos\varepsilon\ , { |{t\ x_{bs}\ b}\rangle } \nonumber \\
{ |{bs_3;t\ x_{bs}\ \varepsilon}\rangle } = & - \sin\varepsilon\ , { |{t\ x_{bs}\ a}\rangle } + \cos\varepsilon\,\left[\frac{1}{\sqrt{2 } } \left({|{t , x_{bs}+1,b}\rangle } + { |{t , x_{bs}+1,a}\rangle}\right ) \right ] \nonumber \\ { |{bs_4;t\ x_{bs}\ \varepsilon}\rangle } = & \cos\varepsilon\ , \left[\frac{1}{\sqrt{2 } } \left(- { |{t , x_{bs}+1,b}\rangle } + { |{t , x_{bs}+1,a}\rangle}\right ) \right ] + \sin\varepsilon\ , { |{t\ x_{bs}\ b}\rangle } \label{bsvect } \ ] ] as the beam splitter has two input ( entrance ) channels , its projection evolution operator should take into account these two possibilities and can be written as the resolution of unity as follows : @xmath111 where @xmath112 denotes the first or the second beam splitter .
the beam splitters are placed in different positions @xmath113 and @xmath114 .
the parameter @xmath115 is changing continuously from @xmath116 to @xmath117 , while the particle is inside the device .
the parameter @xmath118 is a monotonic , increasing function of the evolution parameter .
more detailed structure of the function @xmath118 is not needed . the continuous series of projections , while @xmath115 is changing from @xmath116 to @xmath117 , lead to the following effective projection evolution operator : @xmath119 where @xmath120 denotes the first or the second beam splitter . as the label for this effective step of evolution one can choose any evolution parameter @xmath121 chosen from the interval of the evolution parameter required to perform the evolution of the particle within the phase shifter . in case of the first beam splitter each of the input channels @xmath93 and @xmath122
is transformed into a superposition of both output channels @xmath93 and @xmath122 .
next the particle passes the phase shifter described below and there is the second beam splitter where two input channels @xmath93 and @xmath122 are again transformed into the proper superposition of two output channels @xmath93 and @xmath122 .
the phase shifter in the one dimensional model as presented in fig.[fig2 ] acts on two channels @xmath93 and @xmath122 simultaneously . to construct its evolution operator
one can use the same idea as in the construction of ( [ pevbs ] ) and ( [ pevbseffect ] ) . for this purpose
we define a family of the following vectors : @xmath123 where @xmath124 labels both channels and @xmath125 denotes the position of the phase shifter .
the evolution operator for _ ps _ is presented here by the continuous family of the following projections : @xmath126 where the parameter @xmath115 is changing continuously from @xmath116 to @xmath117 , while the particle is inside the device .
the parameter @xmath118 is a monotonic , increasing function of the evolution parameter .
the proper effective projection evolution operator for the phase shifter can be created in the same way as in case of the beam splitter : @xmath127 as the label for this effective step of evolution one can choose any evolution parameter @xmath121 chosen from the interval of the evolution parameters required to perform the evolution of the particle within the phase shifter .
the detectors are the last devices we have to describe .
we assume that the first detector is placed in @xmath128 and it is able to detect particle in the channel @xmath93 and the second detector placed in @xmath129 detects the particle in the channel @xmath122 .
we assume also that both detectors localize particle in the space and time .
these assumptions determine uniquely the form of the projection evolution operator for detectors : @xmath130 of course , the process of evolution is described by one evolution operator collected from all of the above .
they describe the subsequent events enumerated by the evolution parameter @xmath0 : @xmath131 the first line describes the emission of particle in the source _ z_. the second line gives the free motion between the source and the first beam splitter .
it can be labelled by any evolution parameter @xmath132 because the projection of the projection is again the same projection .
the third line describes the beam splitter and it can be replaced by the effective operator ( [ pevbseffect ] ) assuming the evolution parameter equal e.g. , to @xmath133 .
in fact , one can choose any evolution parameter from the interval @xmath134 .
next , we have free motion between the beam splitter and the phase shifter . for the phase shifter , described by the next line
, one can also introduces the effective operator ( [ pevpseff ] ) and the evolution parameter @xmath121 .
next two lines describe the second beam splitter which can be interpreted in terms of the effective operators at the evolution parameter @xmath135 and two detectors which detect , or not , the particle at the evolution parameter @xmath136 . in the following we are not interested in full analysis of all possibilities given by the projection evolution operator ( [ pevtotal ] ) .
we want to find the states and the appropriate probabilities of the particle detected at the evolution parameter @xmath136 .
according to general rules of _ pev _ the interesting states can be calculated as : @xmath137 } \label{staterho } \ ] ] the most interesting path of the evolution is given by the following sequence : + @xmath138 .
+ obviously , the whole path is chosen randomly in step by step procedure , as described in _ pev _ .
because the final state is a pure state , it can be immediately written as : @xmath139 the distribution of times @xmath140 of detection of the particle in the channel @xmath141 can be calculated as the conditional probability : @xmath142 .
\label{probdet } \ ] ] the conditional probability ( [ probdet ] ) gives the probability of detection of the particle in the state labelled by @xmath143 under the conditions that the particle is created in @xmath144 an then evolves through @xmath145 and @xmath146 . now assuming the initial state is the pure state denoted by : @xmath147 the particle will be created from the source with the probability @xmath148 and after the creation
, the state of the particle is @xmath149 let us calculate a scalar product which will be useful in further considerations : @xmath150 after the particle is created from the source , it undergoes a free evolution . making use of ( [ prpsi ] ) and ( [ usescpr ] )
one can write the state of the particle as : @xmath151 the probability that the particle is found in such a state , under the condition that it was is in the state ( [ sourcestate ] ) , is equal @xmath152 .
the coefficient @xmath153 is the normalization factor which is not needed for further considerations . next we can derive a state of the particle in the next step of the evolution , i.e. in the beam splitter bs1 .
once again we calculate explicitly the scalar product @xmath154 this scalar product is not equal zero , for the time @xmath155 only , i.e. it must satisfy the condition + @xmath156 .
it allows to write the state of the particle ( for this step of the evolution ) as : @xmath157 in the similar way one can find the state in the phase shifter ps after a free evolution f2 .
it is chosen by the nature , with the conditional probability @xmath158 in the following form : @xmath159 similarly , the state of the particle in the second beam splitter bs2 , after a free evolution f3 , with the conditional probability @xmath160 is represented by @xmath161 finally , according to ( [ finalstate ] ) , the state in the detector d after a free evolution f4 is : @xmath162 the corresponding conditional probabilities of finding the particle in the output channels @xmath93 and @xmath122 are : @xmath163 and @xmath164 respectively .
observation : if there were no phase shifter on the evolution path , the second detector would never detect the particle , which is in agreement with the experimental observation . making use of the expression ( [ probdet ] )
one can write the total , conditional probability of detecting the particle at @xmath136 , in either @xmath93 or @xmath122 channel : @xmath165 .
\label{probdfull } \ ] ] the probability ( [ probdfull ] ) can be nonzero only for @xmath166 .
it implies that the final state ( [ stated ] ) is chosen by the nature , with the probability ( [ probdfull ] ) , as one of two vectors : @xmath167 because the physical time is here a dynamical variable one can construct different observables characterizing time behavior of our system .
one of the most interesting is the observable which measure a possibility of finding the particle at given time in the channel @xmath86 .
the corresponding decomposition of unity can be written as : @xmath168 where @xmath169 and @xmath89 .
the conditional probability of finding the particle at given time in the channel @xmath86 , while the particle is in a state @xmath1 , can be calculated according to the usual formula : @xmath170 in our , very simple model the conditional probability ( [ prob_rho_sigma ] ) calculated for the state ( [ statedsigma ] ) is non zero only if @xmath171 and @xmath172 . in this case
the probability is equal to one for @xmath166 .
it means that @xmath140 can be interpreted as time of detection . in more realistic model
the final state can have quite complicated distribution in time .
full probability of detection of the particle which passed the evolution path @xmath173 ( shown above the equation ( [ finalstate ] ) ) is the product of all the conditional probabilities along the evolution path and the conditional probability given by ( [ prob_rho_sigma ] ) . in our case
this procedure gives , in fact , the probability ( [ probdfull ] ) . in this toy model
the states of the particle are localized in time by all devices .
the only exception is free evolution which gives states spread over the time .
all the moments which appear in this evolution process like @xmath174 , which represents the time of particle emission from the source , @xmath175 and @xmath140 which denote the times of arrival to the appropriate devices , are random variables . in addition ,
their probability distributions are not independent e.g. , the probability of choosing by the `` world lottery '' the time of arrival to the first beam splitter @xmath176 is dependent on time of emission from the source @xmath174 , and so on .
these are special features of the model considered above . as an example
let us calculate the average time at which the particle can be localized in @xmath113 ( the first beam splitter ) under the condition that the particle was emitted from the source at the fixed time @xmath174 . in general , for any evolution parameter @xmath177 , where @xmath178
, the state can be obtained by two - step evolution according to the evolution operator ( [ pevtotal ] ) .
it can be written as : @xmath179 to calculate the average time of finding the particle in the place where is the located the first beam splitter , we have to define the appropriate operators .
first , we define the decomposition of unity which is able to determine the position of particle in the space time but independently of channels : @xmath180 the observable which measure the time at fixed coordinate point @xmath73 can be thus defined as : @xmath181 the average time of finding the particle at @xmath113 can be calculated according to the usual rules as : @xmath182 this time can be also called the time of arrival because the particle is emitted at the fixed time @xmath174 and the difference @xmath183 should give the average time of fly between the source and the first beam splitter .
we obtain the expected result , though with the additional factor @xmath184 .
the time of arrival is proportional to distance between the source and the beam splitter and it is what we expected , because in this very simple model there is no spreading of `` wave packets '' over the space .
the additional coefficient is equal to the probability of finding the particle at a given point of time axis . in the case of schrdinger equation , as an equation of motion ( implemented by the projection operators like those for free motion , see ( [ phibase],[pevfree ] ) ) , the functions @xmath185 .
taking into account that times of emissions @xmath174 of particles , in principle , are also random variables , the probability that a particle being in the initial state @xmath186 , can be localized at a given time in @xmath113 is the product of @xmath187 and @xmath188 .
this allows to derive the average time @xmath189 of finding the particle at @xmath113 averaged , in addition , over the emission times @xmath174 as : @xmath190 as the second model we consider the same one dimensional mach zehnder interferometer as in the fig .
we assume the same space of states as in the model i. the main difference between both models is that in the second one we treat both beam splitters and phase shifter as a single device .
it is described by the evolution operator similar to the description of free motion in the first model .
the source of the particles we describe exactly by the same decomposition of unity as in the model i , see ( [ pevz ] ) . on the other hand , both beam splitters and the phase shifter we describe by the operator : @xmath191 where @xmath192 is given by the equation ( [ phibase ] ) , but with another operator @xmath193 .
the operator @xmath100 describes all the three devices at once , see @xcite : @xmath194 it means that a single drawing of lots by the nature decides about the evolution through all three devices . in this case , there are no intermediate times like @xmath195 and @xmath196 which are effects of subsequent steps of the evolution . the last step of the evolution is again determined by the detectors and it is described by the operator ( [ pevd ] ) . in this case
the projection evolution operator is shorter and can be written as @xmath197 in this model there are only 3 steps of the evolution .
the system starts from the emission of the particle , passes all the three devices ( the first beam splitter , the phase shifter and the second beam splitter ) and ends at the detectors : @xmath198}. \label{staterho2 } \ ] ] the probability distribution of finding the particle at the time @xmath140 , in a given output channel , assuming the same initial state ( [ initstate ] ) , when it follows a given evolution path @xmath199 is @xmath200 = \nonumber \\ & & \left|{\langle t_z , x_z , a \vert \psi_0 \rangle}\right|^2 { \left| \chi_{\nu_{dev}}(t_z ) \right|}^2 { \left| \chi_{\nu_{dev}}(t_d ) \right|}^2 \left|{\langle{x_{d},\sigma_d}|}s^{\delta t}{|{x_z , a}\rangle}\right|^2 \label{probdet2 } \ ] ] where @xmath201 denotes the difference between the time of emission of the particle and the time of its detection .
let us explicitly write the action of the unitary operator @xmath202 onto the state @xmath203 : @xmath204 using above expression , the total conditional probability ( [ probdet2 ] ) can be easily rewritten as : @xmath205 .
\label{probdfull2 } \ ] ] the result is very similar to those from the model i , see ( [ probdfull ] ) , especially for the case when the functions @xmath206 . as in the previous case ,
the interval between the emission of the particle by the source and its detection is determined by distance between the source and the detector .
the main difference between both models is in timing of evolution steps . in the model ii , besides of @xmath174 and @xmath140 there are no intermediate times ( which are random variables ) and additional factors in probabilities related to them .
lack of spreading in time is determined here by the `` toy '' form of the operator @xmath96 ( [ sdev ] ) . in this notes
we have compared only two models of the mach - zehnder interferometer within the projection evolution approach . in practice , they give nearly the same results of the evolution .
however , it seems that up to date experiments show that , the beam splitters and phase shifters are closer to the second model than to the first one . in both models the physical time is a dynamical variable which is changing in rather primitive ( nearly no smearing in time ) way . because of this , they can not show all features of dynamics of time which is in principle a random variable in the evolution process .
in this paper , we propose a fundamental mechanism of quantum evolution based on the idea of `` natural measurements '' performed by the nature at each step of changes ( evolution ) of the physical systems . as a special case ,
our postulate allows to reproduce the schrdinger equation and go beyond it . in this way
one can reproduce not only the schrdinger equation but also other quantum equations of motions e.g. , the relativistic equations . obviously , the _ pev _ postulates require many tests , yet .
however , it seems that for each physical system one can find an appropriate set of projection operators which allows to apply the _ pev _ postulate to get physical states .
a possible method to construct the _ pev _ operators is the method of generating operators .
it is also important to notice that the _ pev _ postulate gives the unique opportunity to treat space and time on equal footing .
this feature can open some new directions of development of quantum theory . n. g. van kampen , physica a**153 * * , 97 ( 1988 ) .
l. e. ballentine , found.phys .
* 20 * , 1329 ( 1990 ) .
r. gambini , quant - ph/9803025 .
r. l. stratonovich , v. b. belavkin , int . j. theor
. phys . * 35 * , 2215 ( 1996 ) .
quant - ph/0503165 v2 a. gd , m. dbicki , quant - ph/0308018 .
a. p. balachandran and s. m. roy , quant - ph/0102019 .
; w. r. wharton : backward causation in quantum mechanics , preprint . | we propose a model of time evolution of quantum objects which unites the unitary evolution and the measurement procedures .
the model allows to treat the time on equal footing with other dynamical variables . |
the @xmath0 parameter serves to parameterize the weak hadronic matrix element responsible for @xmath5 mixing .
since this mixing gives us the only cp violation observed to date , @xmath0 is a crucial link between the measured quantity @xmath6 and the parameters of the standard model .
lattice calculations are well suited for the study of @xmath0 parameter , and it has by now received much attention . after an early round of calculations@xcite ,
the statistics have now been raised to a level which allows one to examine some of the fine points of the calculation , such as checks on the reliability of one - loop lattice perturbation theory @xcite , the chiral behavior and nondegenerate quark masses @xcite , the dependence of @xmath0 on the lattice spacing @xcite and the number of dynamical flavors@xcite . in this note
we offer more information on these latter two points .
.ensembles for @xmath2 study [ cols="<,<,<,<",options="header " , ] for creating kaons ( at rest ) we use a wall of u(1 ) noise on timeslice @xmath7 , i.e. complex random numbers @xmath8 at each space point such that @xmath9 .
this is statistically equivalent to computing a collection of delta - function sources . in particular
, our wall creates only pseudoscalars .
7.2 cm we use a lattice duplicated in the time direction , with periodic boundary conditions in space and time ( see fig .
[ fig : period ] ) . computing propagators on the doubled lattice , we obtain forward- and backward - going propagators which we use for computing @xmath0 . that is , if @xmath10 is the @xmath11 propagator on the doubled lattice , then our operator correlation functions are schematically of the form @xmath12 , where @xmath13 or 48 .
we employ three kinds of operators : landau gauge , gauge invariant , and tadpole improved .
landau gauge operators are defined by fixing the gauge and omitting explicit links in non - local operators . for gauge - invariant operators we supply the links , averaging over all shortest paths .
tadpole - improved operators are gauge - invariant operators , but with all links rescaled by @xmath14 , where @xmath15 , @xmath16 is the average plaquette , and @xmath17 is the number of links needed to connect fermion fields .
we opted for tadpole - improved operators on all configurations , using the others on a subset of configurations for checks .
the matching between continuum and lattice operators is of the form @xmath18 where @xmath19 is the one - loop anomalous dimension matrix , and @xmath20 are finite coefficients , which can be sizable .
we take these from the calculations of for the continuum scheme , we choose ndr , quoting results either at scale @xmath21 or at @xmath22 .
we use the @xmath23 coupling constant @xmath24 , defined as @xmath25 7.2 cm to check how well the perturbation theory works , we computed all three operators on a subset of the @xmath26 ensemble , finding that after one - loop corrections are put in , the matrix elements agree within our statistical error . for the bulk of the calculation we used tadpole - improved operators exclusively
. 7.2 cm
figs . [ fig : bkq ] and [ fig : bkd ] show the results for @xmath0 on three ensembles of configurations .
values at 9 quark mass points are fitted to the form expected from chiral perturbation theory , @xmath27 .
the and curves are similar in shape , while the quenched curve crosses between the other two .
while this is perfectly allowed , we should also inject a small note of caution our ensembles have the same @xmath28-masses , but these masses are presumably affected to some degree by the finite volume .
if this effect is sizable and depends significantly on @xmath2 , our curves could shift a little .
7.2 cm taking the results at face value , we note that the @xmath26 and @xmath29 results lie nearly on top of each other at the kaon mass , consistent with our earlier results @xcite . also , most of the @xmath26 data lie below @xmath29 , consistent with the observation by other groups that quenching seems to increase @xmath0 slightly ( see , e.g. ref .
however , the @xmath30 data turn this picture upside down .
7.2 cm fig .
[ fig : bkvsnf ] shows our final values for @xmath0 , obtained at the physical kaon mass and by extrapolation to the chiral limit .
we see that the interpolated @xmath31 result is a few percent higher than quenched .
performing the same analysis on the quenched ensembles , we obtain the result shown in figure [ fig : bkvsa ] , where we plot @xmath32 versus the scale as determined from @xmath33 .
the data are well fit by the quadratic form @xmath34 , where the scale of the power correction parameters turns out to be typical of qcd : @xmath35 , @xmath36 .
alternatively , we note that we can avoid making reference to the possibly problematic @xmath33 by using the scaling form @xmath37 where we take @xmath38 here to be the @xmath23 coupling
. this amounts to shuffling around the @xmath39 corrections , and in practice tends to straighten the data out .
that is to say , much of the curvature in figure [ fig : bkvsa ] might be ascribed to scaling violations in @xmath33 itself . to quote a final value
we make the conservative choice of a linear fit to the four points with @xmath40 , and obtain @xmath41 7.2 cm
from the dynamical comparison , we find that @xmath42 is ( @xmath43)% larger than @xmath44 . combining with the @xmath4 extrapolation we we quote our current central value @xmath0 in the real world : @xmath45 remaining uncertainties include possible finite - size effects in the dynamical ensemble , higher order perturbative corrections in the matching , and higher order chiral ( @xmath46 ) effects .
a study of hadronic weak matrix elements relevant for @xmath47 using the same techniques and ensembles is currently underway .
* acknowledgements . *
we thank the columbia collaboration for access to the dynamical configurations .
cray t3d time was supplied by the ohio supercomputer center and the los alamos advanced computing laboratory .
99 g. kilcup , s. sharpe , r. gupta and a. patel , _ phys .
* 64 * ( 1990 ) 25 .
s. sharpe , r. gupta and g , kilcup , _ nucl .
b ( proc . suppl . ) _ 26 ( 1992 ) 197
. s. sharpe , _ nucl .
b ( proc . suppl . ) 34 _ ( 1994 ) 403
. n. ishizuka _
et al . _ , _ phys .
_ 71 ( 1993 ) 24
. m. klomfass and w. lee , lattice 1995 , p. 469 .
jlqcd collaboration , lattice 1995 , p. 465 .
jlqcd collaboration , these proceedings .
g. kilcup , _ phys .
_ 71 ( 1993 ) 1677 .
d. chen , these proceedings .
n ishizuka , y. shizawa , _ phys .
_ d49 ( 1994 ) 3519 .
s. sharpe and a. patel , _ nucl .
_ b417 ( 1994 ) 307 .
a. soni , lattice 95 , p. 43 | we present results of a study of @xmath0 using tadpole improved gauge - invariant staggered operators . using three ensembles of @xmath1 configurations with varying numbers of dynamical flavors , we observe a small dependence on @xmath2 . using 7 quenched ensembles at different values of @xmath3
, we extrapolate to @xmath4 . |
the optical ( e.g. ghez et al 2000 ; eckart et al 1997 ; gebhardt et al 2000 ; ferrarese & merritt 2000 ) and radio ( e.g. miyoshi et al 1995 ) evidence for compact central masses in the nuclei of many nearby galaxies is now very clear .
the compact objects are consistent with being supermassive black holes .
the data do not however probe the strong gravity regime of these holes , since the observed stars and gas orbit at more than 40,000 gravitational radii from the centre .
much of the x - ray emission from accreting black holes should emerge from within a few 10s of gravitational radii .
aspects of this emission can provide us with a probe of the strong gravity of black holes .
in particular the rapid variability , or in some cases quasi - periodic oscillations , and soft thermal emission from an accretion disc demonstrates the small size of the x - ray emission region . here
i concentrate on the broad iron emission lines seen in the spectra of some seyfert 1 galaxies .
the line profile and its variability can reveal the geometry of the accretion flow , and strong gravitational effects , within a few gravitational radii of the black hole .
the dominant expected line is that due to iron , as shown in fig . 1 ( george & fabian 1991 ; matt , perola & piro 1991 ) .
the spectrum shown is the result of a monte - carlo simulation of the emergent spectrum when a flat surface has been irradiated with a power - law x - ray spectrum .
the gas is assumed to be neutral , although the important issue here is that the metals retain their k and l - shell electrons .
the major emission line is iron k-@xmath0 , due to its relatively high abudance ( a cosmic mix is assumed ) and high fluorescent yield ( which increases as @xmath1 ) .
doppler broadening of this emission line , produced by the irradiation of the surface of an accretion disc by hard x - rays from its corona , together with thr special relativistic effects of beaming and the transverse doppler effect and the general relativistic effect of gravitational redshift , lead to the observed profile being very broad and skew ( fabian et al 1989 : figs . 2 , 3 ) .
the blue horn of the line is most sensitive to the inclination angle of the disc ( i.e. the importance of doppler broadening ) and the red wing is most sensitive to the inner radius ( i.e. the gravitational redshift ) .
a clear broad iron line was first seen with data from a 1994 asca observation of the seyfert 1 galaxy mcg6 - 30 - 15 ( tanaka et al 1995 : fig .
this was the result of a 4.5 day exposure on the source and is shown with the dominant power - law continuum removed .
the data are well fit ( fabian et al 1995 ) by emission from an accretion disc stretching from about 6 gravitational radii ( i.e. @xmath2 ) to about 40 @xmath3 and an inclination of about 30 deg .
the surface emissivity falls as radius @xmath4 .
a similar clear line has also been seen in ngc3516 ( nandra et al 1999 : fig .
the shape of these lines is entirely consistent with the emitting region being a flat disc in keplerian rotation close to a black hole .
k. nandra ( priv . comm . ) has compiled the line spectra of several more seyfert 1 galaxies ( fig .
5 ) which generally are weaker in flux , or have shorter exposures than for mcg6 - 30 - 15 .
a broad red wing is seen in most
. further work on the strength and width of iron lines in seyferts is reported by lubiski & zdziarski ( 2000 ) .
occasionally the iron line in mcg6 - 30 - 15 changes its profile , as in a deep minimum ( iwasawa et al 1996 ) or after a flare ( iwasawa et al 1999 ) , and appears to stretch more to the red .
this is possible evidence that the black hole is spinning rapidly ( iwasawa et al 1996 ; dabrowski et al 1997 ) , although matter streaming from the innermost stable orbit around a non - spinning black hole might mimic this result ( reynolds & fabian 1997 ; young et al 1998 ) .
xmm - newton , which has a much larger effective area than asca , has now observed mcg6 - 30 - 15 and several other seyfert 1 galaxies . the iron line profile , kindly supplied by the pi of the observation , joern wilms , is shown in fig .
the data in this ratio plot resemble the shape of the iron line from asca ( fig . 2 ) .
in the case of mrk205 ( fig .
9 ) the line appears to peak around 6.7 kev , which may indicate an ionized disc ( i.e. the metals have only k - shell electrons ) .
the sharp peak at 6.4 kev may be a fluorescent line from more distant cold gas within the galaxy
. models of ionized discs have been made for constant density ( e.g. ross & fabian 1993 ) and more recently for atmospheres in hydrostatic equilibrium ( nayakshin , kazanas & kallman 2000 ; ballantyne , ross & fabian 2001 : fig .
an example of a constant density model fitted to the asca mcg6 - 30 - 15 data is shown in fig
the models predict that the iron line shifts to 6.7 kev in the rest frame and oxygen and other features should be present below 1 kev .
whether these low energy features are detectable depends on their strength relative to the steep continuum .
broad iron lines are generally not seen in objects with luminosity exceeding a few time @xmath5 in the 210 kev band ( iwasawa & taniguchi 1993 ; nandra et al 1997 ) .
an example from xmm - newton of a narrow iron line from a powerful distant quasar , pks0537 - 286 , is shown in fig .
9 , from reeves et al 2001 ) .
narrow iron lines are expected in many agn from reflection by outer gas at parsecs and beyond ( ghisellini et al 1994 ) and are seen in some seyferts ( e.g. yaqoob et al 2000 ) .
the intensity of the iron line in mcg6 - 30 - 15 does vary ( iwasawa et al 1996 ; 99 ) , but not in any obvious way in response to the continuum .
a long observation made simulataneously with asca and rxte in 1997 ( lee et al 2000 ) confirms this .
vaughan & edelson ( 2000 ) have re - analysed the rxte data with the results shown in fig .
there is real variability in the iron line flux but it is not correlated with the continuum flux . the continuum power - law slope is correlated in the sense that the spectrum steepens as the flux increases .
further work by matsumoto et al ( 2000 ) on a 10 day asca observation in 1999 extends the peculiar behaviour .
the rms variability decreases towards high energy , particularly around the iron line ( fig .
various ideas have been mooted in order to explain the variability .
possibilities include ; a ) transrelativistic motion of flares ( reynolds & fabian 1997 ; beloborodov 1999 ) so that an observed bright continuum is associated with flux beamed away from the disc and vice versa ( this has the problem that the spectral index trend with flux would be opposite to that observed ) ; b ) smoke from the coronal flares themselves ( merloni & fabian in prep . )
in which electron scattering in the corona smears out transmitted line emission when the regions are spatially large , and hence bright ; and c ) ionization variations ( lee et al 2000 ; reynolds 2000 ) . the lack of any clear iron line
continuum correlation so far suggest that iron line reverberation ( reynolds et al 1999 ) may be complicated .
a recent development has been the observation of seyfert galaxies at high spectral resolution , provided by the gratings on chandra and xmm - newton .
branduardi - raymont et al ( 2001 ) claim from xmm - newton rgs spectra of mcg6 - 30 - 15 the presence of relativistically - broadened oxygen , nitrogen and carbon lines . as already noted , emission lines from these elements
are expected from an ionized disc .
the surprise is that they might be so strong .
moreover , the spectrum of the source below 2 kev has been previously well modelled by a variable warm absorber ( fabian et al 1994 ; otani et al 1996 ) . in their original paper ,
branduardi - raymont et al ( 2001 ) rely on some unspecified mechanism for the observed spectral turndown below 2 kev and argue from the weak , broad ( @xmath6 ) absorption lines seen in their spectra that the column density of any warm absorber is too low to provide an edge large enough to account for the spectral jump at @xmath7 ( fig .
12 ) , moreover the jump occurs in the wrong place , requiring a redshift of @xmath8 .
chandra hetg spectra provide a different view of the same object ( lee et al 2001 ; fig .
the 3 times higher spectral resolution shows that the linewidths are @xmath9 and , from curve of growth analysis of the ovii line series , argues for the presence of a significant warm absorber .
part of the redshift of the ovii photoelectric edge is then accounted for by the convergence of the ovii absorption line series .
the poster paper on the xmm - newton rgs spectrum by sako et al ( 2001 ; these proceedings ) now includes a warm absorber , together with relativistically - broadened hydrogenic , c , n and o lines .
absorption by neutral iron is suspected to be a further important ingredient by the chandra team .
the neutral iron presumably being in the form of dust , which must be present to explain the high optical / uv reddening seen in the object ( reynolds et al 1997 ) . since the conference , lee et al ( 2001 ) have accounted for a major feature around the ovii edge in terms of neutral iron absorption .
of the three edges of iron - l , known as l3 , l2 and l1 , the absorption near threshold of fe l3 is a deep trough in laboratory data ( kortright & kim 2000 ) and in the spectrum of cyg x-1 ( schulz et al 2001 ) .
this trough exactly matches the broad absorption feature seen in the chandra spectrum of mcg6 - 30 - 15 at 0.71 kev ( fig . 13 ) and
accounts for the apparent redshift of the ovii edge ( fig .
fe l2 and ovii absorption together explain the remaining features between 0.7 and 0.75 kev .
much more detailed modelling is required , covering a wider spectral band , but it is now clear that mcg6 - 30 - 15 does have a significant dusty warm absorber .
it is plausible that there is in addition some emission from ionized c , n and o within the disc , but debatable as to whether it is strong enough to create detectable features .
the relativistic line broadening parameters quoted by branduardi - raymont et al ( 2001 ) do not yet agree with those found for the time - averaged asca spectra .
the asca results ( fig .
4 ) appear to be in accord with the xmm - newton epic data ( fig .
8) taken simultaneously with the rgs spectra .
the chandra spectrum does not show two spectral jumps , one of which can be explained by photoelectric absorption and the other by the blue wing of an emission line .
i conclude that the large spectral jump seen around 0.7 kev is dominated by absorption features from a dusty warm absorber .
whether relativistically - broadened emission features are also present remains to be seen .
the broad iron lines seen in mcg6 - 30 - 15 and ngc3516 are well explained by x - ray irradiation of an accretion disc extending very close to a black hole .
the skewed shape of the lines is mainly due to the strong gravitational redshift in this region .
the variability of the iron line and the lack of correlation with continuum variations is not understood .
the emission region is plausibly complex with ionization , bulk motion and scattering effects likely .
the case for a broad oxygen line in mcg6 - 30 - 15 is unclear .
i thank david ballantyne , claude canizares , kazushi iwasawa , julia lee , raquel morales , patrick ogle , randy ross , norbert schulz and simon vaughan for help and chiho matsumoto , paul nandra , james reeves , masao sako and joern wilms for sending me plots in advance of publication .
ballantyne d.r . ,
, fabian a.c . , 2001 , mnras in press , ( astro - ph/0102268 ) beloborodov a. , 1999 , apj , 510 , 123 branduardi - raymont g. , m. sako , s.m .
kahn , brinkman a.c .
, kaastra j.s .
, page m.j . , 2001 , a&a , 365 , l140 dabrowski , y. et al . , 1997 , mnras , 288 , l11 eckart a. , genzel r. , 1997 , mnras , 284 , 576 fabian a.c . , et al 1994 , pasp , 46 , l59 fabian a.c . , nandra k. , reynolds c.s . , brandt w.n . ,
otani c. , tanaka y. , inoue h. , iwasawa k. , 1995 , mnras , 277 , l11 fabian a.c . , rees m.j . , stella l. , white n.e . , 1989 ,
mnras , 238,729 fabian a.c . , reynolds c.s . , iwasawa k. , young a. , 2000 , pasp , 112,1145 ferrarese l. , merritt d. , 2000 , apj , 539 , l9 gebhardt k. , 2000 , apj , 539 , l13 george i.m .
, fabian a.c . , 1991 , mnras , 249 , 352 ghez a.m. , morris m. , becklin e.e . , tanner a. , kremenek t. , 2000 , nature 407 , 349 ghisellini g. , haardt f. , matt g. , 1994 , mnras , 267 , 743 iwasawa k. , taniguchi y. , 1993 , apj , 413 , l15 iwasawa , k. et al . , 1996 ,
mnras , 282 , 1038 iwasawa k. , fabian a.c .
, young a. , matsumoto c. , inoue c. , 1999 , mnras , 306 , l19 kortright j.b . , kim s.k . , 2000 , phys rev .
b , 12216 laor , a. , 1991 , apj , 376 , 90 lee j. , fabian a. c. , reynolds c. s. , brandt w. n. , iwasawa k. , 2000 , mnras , 318 , 857 lee j. , ogle p.m. , canizares c.s . ,
marshall h.l . ,
schulz n.s . , morales r.m . , fabian a. c. , iwasawa k. , 2001 , apj , submitted ( astro - ph/0101065 ) matt , g. , perola , g.c . , piro l. , 1991 , aaa , 247 , 25 miyoshi m. , et al 1995 , nature , 373 , 127 nandra k. , george i.m . , mushotzky r.f . ,
turner t.j .
, yaqoob y. , 1997 , apj , 488 , 91 nandra k. , george i. m. , mushotzky r. f. , turner t. j. , yaqoob t. , 1999 , apj , 523 , l17 nayaksin s. , kazanas d. , kallman t. r. , 2000 , apj , 537 , 833 otani c. et al . , 1996 , pasj , 48 , 211 reeves j.n . , et al 2001a , a&a , 365 , l116 reeves j.n . , et al 2001a , a&a , 365 , l134 reynolds c. s. , 2000 , apj , 533 , 811 reynolds , c.s . , ward , m.j .
, fabian a.c . ,
celotti a. , 1997 , mnras , 291 , 403 reynolds , c.s . , begelman , m.c . , 1997 , apj , 488 , 109 reynolds , c.s .
young a.j . ,
begelman m.c . , fabian a.c . , 1998 ,
apj , 514 , 164 ross , r.r . , fabian , a.c . , 1993 , mnras , 261 , 74 schulz n.s . , et al 2001 , in preparation tanaka , y. et al . , 1995 ,
nature , 375 , 659 vaughan s. , edelson r. 2000 , apj , 548 , 694 yaqoob t. , george i. m. , nandra k. , turner t. j. , serlemitsos p. j. , mushotzky r.f . , 2000 , apj , 546 , 759 lubiski p. , zdziarski a.a . , 2001 , mnras , in press ( astro - ph/0009017 ) | the observations and interpretation of broad iron lines in the x - ray spectra of seyfert 1 galaxies are reviewed .
the line profiles observed from mcg6 - 30 - 15 and ngc3516 show extended red wings to the line explained by large gravitational redshifts .
the results are consistent with the emission expected from an x - ray irradiated flat accretion disc orbiting very close to a black hole .
results from xmm - newton and chandra are presented and the possibility of broad oxygen lines discussed . |
the mainz mikrotron mami at the institut fr kernphysik in mainz is an accelerator to study the hadron structure with the electromagnetic probe .
the machine has been upgraded to 1.5gev electron beam energy by a harmonic double - sided microtron @xcite .
this fourth stage , called mami - c , was completed in 2007 . with mami - c
the threshold beam energy for associated strangeness production off protons was reached for the first time at mami .
the electromagnetic production of kaons off the nucleon provides an important tool for understanding the dynamics of hyperon - nucleon systems .
theoretically , the process is described often by effective lagrangian models , commonly referred to as `` isobar '' approach . in the isobaric models
the reaction amplitudes are constructed from lowest - order ( so - called tree - level ) born terms with the addition of extended born terms for intermediate particles , @xmath1 , @xmath2 , or @xmath3 resonances , exchanged in the @xmath4- , @xmath5- , and @xmath6-channels .
a complete description of the reaction process would require all possible channels that could couple to the initial and final state .
most of the model calculations for kaon electrophoto - production have been performed in the framework of tree - level isobar models @xcite , however , few coupled - channels calculations exist @xcite .
the drawback of the isobaric models is the large and unknown number of exchanged hadrons that can contribute in the intermediate state of the reaction .
depending on which set of resonances is included , very different conclusions about the strengths of the contributing diagrams for resonant baryon formation and kaon exchange may be reached .
the kaon - maid and saclay - lyon models characterise our present understanding of kaon photoproduction reactions at photon energies below 1.5gev .
the model sla is a simplified version of the full saclay - lyon model in which a nucleon resonance with spin-5@xmath72 appears in addition .
the electro - production of strangeness introduces two additional contributions , that are vanishing for the kinematic point at @xmath8 0 : the longitudinal coupling of the photons in the initial state , and the electromagnetic and hadronic form factors of the exchanged particles . concerning the differential cross - section the model variations are strongest at small kaon centre - of - mass angle . to conclude
, it is fair to say that new experimental data on strangeness production will challenge and improve our understanding of the strong interaction in the low energy regime of qcd .
.experimental setting of the kaon electro - production beam - time in 2009 .
the beam energy was 1.508mev , beam intensity 14@xmath9a , beam raster size @xmath10 1 mm , and the @xmath11h@xmath12 target length 48 mm . [ cols="^,^,^,^,^,^,^,^,^ " , ] [ tab : kinematics ] first experiments on the electro - production of kaons off a liquid hydrogen target
were performed since 2008 .
positive kaons were detected in the spectrometer in coincidence with the scattered electron into spectrometer b. the central spectrometer angle of the kaon arm was 31.50@xmath13 with a large angular acceptance in the dispersive plane of @xmath14 2143@xmath13 .
the spectrometer setting for electron detection was fixed at the minimum in - plane angle of @xmath15 15@xmath13 , thereby maximising the virtual photon flux .
the invariant momentum transfer , @xmath16 , was very low , the virtual photon energy , @xmath17 , was near the maximum of the kaon photoproduction cross - section at 11.2gev , which excites a total energy in the virtual - photon - nucleon center - of - mass system , @xmath18 , of 1.61.7gev .
the kinematic conditions are summarised in table [ tab : kinematics ] .
the cryogenic target in the spectrometer hall that was used for the kaon electro - production experiments consists of a 49.5 mm long target cell and is made of a 10@xmath9 m havar walls .
the geometry of the cell is optimized for enhancing the luminosity while keeping low the energy - losses .
kaon identification is based on specific energy - loss and time - of - flight , electron identification on a signal in a gas cherenkov counter .
measured time - of - flight and specific energy - loss spectra are shown in fig .
[ fig : tofdedx ] as a function of particle momentum .
the time - of - flight was corrected for the path - length dispersion assuming the particle was a pion .
the data sample is dominated by background particles ( p and @xmath19 ) because of a kaon - to - background ratio of 1:200 and the event sample was enriched in kaon events by a cut in missing mass to show the position of the band of kaons in relation to the two bands from the dominant particle species .
the flight time difference between protons and kaons is 1015ns , between pions and kaons 510ns .
the energy - loss separation between kaons and pions is small , whereas the separation between kaons and protons is of the order of 25mev@xmath20 cm .
after electron and kaon identification the measured momenta , in magnitude and direction with respect to the incoming beam , allow for a full reconstruction of the missing energy and missing momentum of the recoiling system .
the missing mass @xmath21 is related to the four - momentum @xmath22 of the virtual photon and the four - momentum @xmath23 of the detected kaon according to @xmath24 , where @xmath25 is the target four - momentum .
the missing energy and missing momentum can then be calculated using @xmath26 and the missing mass in terms of the kinematic variables @xmath27 , @xmath28 , and @xmath29 as well as of the reconstructed kaon energy , momentum , and scattering angle , can be written as @xmath30 a preliminary missing mass spectrum is shown in fig . [
fig : missingmass](left ) .
the overlaid histograms show the missing mass distributions in two averaged @xmath31 coincidence - time side - bands with the appropriate weights and additional kaon background .
the large momentum acceptance of the spectrometer covers the free electro - produced hyperons @xmath32 and @xmath33 .
the well - known masses of the two hyperons can serve as an absolute mass calibration .
[ fig : missingmass](right ) shows the background - subtracted missing mass distribution compared to the monte carlo spectrum .
the simulation includes radiative corrections , energy - loss in the target , and a model for the resolutions . reaction .
left : missing mass distribution with overlaid histograms showing random coincidences in two averaged @xmath31 coincidence - time side - bands with the appropriate weights and additional kaon background .
the @xmath32 and @xmath33 hyperon peaks are evident .
right : background - subtracted missing mass distribution compared to the monte carlo spectrum .
the simulation includes radiative corrections , energy - loss in the target , and a model for the resolutions.,title="fig:",scaledwidth=49.0% ] reaction .
left : missing mass distribution with overlaid histograms showing random coincidences in two averaged @xmath31 coincidence - time side - bands with the appropriate weights and additional kaon background .
the @xmath32 and @xmath33 hyperon peaks are evident .
right : background - subtracted missing mass distribution compared to the monte carlo spectrum .
the simulation includes radiative corrections , energy - loss in the target , and a model for the resolutions.,title="fig:",scaledwidth=49.0% ]
the kaon events that correspond in the missing mass to the two hyperon channels are used to extract the kaon yield .
the preliminary yields of identified p@xmath34 events as a function of the centre - of - mass kaon scattering angle are shown in fig .
[ fig : kyield ] . events as a function of the cosine of the kaon centre - of - mass scattering angle .
the identified kaon counts in both hyperon channels were corrected for background counts to obtain the yields . the 2009 data covered angles in the range of @xmath35 4090@xmath13 and @xmath36 -2020@xmath13 . in 2011 scattering angles down to zero degree are available for forward - angle measurements of the process.,scaledwidth=60.0% ] the further analysis required a detailed monte carlo simulation of the experiment in order to extract cross - section information from the data . for an unpolarised electron beam and an unpolarised target ,
the five - fold differential cross - section for the p@xmath37 process can be written , see _ e.g. _ @xcite , in a very intuitive form : @xmath38 where the virtual photo - production cross - section is conventionally expressed as @xmath39 the kaon polar angle @xmath40 is given in spherical coordinates in the hadronic centre - of - mass system . in this system
@xmath41 is the energy of the virtual photon .
the terms indexed by @xmath42 are the transverse , longitudinal and interference cross - sections .
the transformation between the differentials @xmath43 is incorporated into the virtual photon flux @xmath44 via the jacobian : @xmath45 the experimental yield can then be related to the cross - section by @xmath46 where @xmath47 is the experimental luminosity that included global efficiencies such as dead - times and beam - current dependent corrections such as the tracking efficiency , @xmath48 is the acceptance function of the coincidence spectrometer set - up , and @xmath49 is the correction due to radiative or energy losses . with a run - time of 244h
using 2@xmath9a beam current ( at 14% dead - time ) and 38h using 4@xmath9a beam current ( at 45% dead - time ) the accumulated and corrected luminosity for the 2009 data - taking campaign was @xmath50 2300fbarn@xmath51 .
the geometrical acceptance of the spectrometer set - up , the path - length from target to detectors , kaon decay in flight , and kaon scattering were determined by a monte carlo simulation using the simulation package geant4 . within the a1 collaboration a different simulation package ( simul++ ) for the experiments was developed in the past .
this code allows to simulate , according to a chosen kinematics , the phase - space accepted by the spectrometers . in alternative
, the events can be generated sampling a distribution given by the cross - section of a specific process .
the output of the geant4 study was implemented in this simulation as a generalised acceptance map , leading to the phase - space accepted by the combination of the kaon spectrometer with the electron spectrometer together with the radiative corrections . by virtue of the monte carlo technique used to perform this integral the quantities @xmath48 and @xmath49 are not available separately or on an event - by - event basis .
the general acceptance can be written in terms of the cross - section at a given point @xmath52 , called the scaling point , at @xmath53 , @xmath54 , @xmath55 , and @xmath56 , when studying the dependence on the remaining variable @xmath57 : @xmath58 then , the behaviour of the cross - section across the acceptance needs to be scaled according to a given theoretical description .
the kaon - maid @xcite isobar model was implemented in this analysis , however , one can as well use the full ( sl ) and the simplified version ( sla ) of the so - called saclay - lyon model @xcite .
the scaling point for the data set from 2009 is at @xmath59 0.036(gev@xmath60 , @xmath61 1.750gev , @xmath62 0.4 and @xmath63 0 .
finally , the cross - section is extracted dividing the yield by the luminosity and the evaluated integral .
the angular distribution of the cross - section gives information on the angular dependence of the production mechanisms for each hyperon .
preliminary differential cross - sections of kaon electro - production in the @xmath32 and @xmath33 channels are shown in fig .
[ fig : crosssections ] .
the data points are compared to the kaon - maid model , to a variant of that model , and to the saclay - lyon model .
it is expected that the final results will constrain the existing phenomenological models and generate theoretical interest .
the difference between photo- and electroproduction data at small @xmath27 is very important as various models predict different transitions between the photoproduction point and the electroproduction cross - section .
if , _ e.g. _ , the kaon - maid model was right , then the electroproduction data would be at @xmath64 0.5 a factor 2.5 larger than the photoproduction data .
the data then could help to fix the unknown longitudinal coupling constants in this kinematical region .
using the recently installed magnetic spectrometer and a high - resolution spectrometer in coincidence it was possible measure elementary kaon electro - production at mami in a kinematic regime not covered by other experiments .
the most interesting kinematic region is at at very forward laboratory angles .
precise data from very forward angle kaon production are urgently required for calculating hypernuclear production cross - sections . in this kinematics the elementary amplitude serves as the basic input , which determines the accuracy of predictions for hypernuclei @xcite .
the detection of kaons in this angular range and the tagging of nearly zero - degree electrons will be achieved at mami in the very near future by steering the primary beam through the spectrometer after passing through a magnetic chicane comprising two compensating sector magnets .
the compact design of and its capability to detect negatively and positively charged particles simultaneously under forward scattering angles complement the existing spectrometers . fig . [ fig : kaos ] shows the layout of the coming experiments with the spectrometer at zero degree . in order to use as a zero - degree double - arm spectrometer a new electron - arm focal - plane detector system was built .
it consists of two vertical planes of 18,432 fibres .
detectors and electronics for the 4,608 read - out and level-1 trigger channels have been installed and tested . for the year 2011 elementary kaon electroproduction measurements with the spectrometer at zero degree scattering angle using the commissioned beam chicane and a first hypernuclear decay - pion spectroscopy experiment using the spectrometer as kaon tagger are planned .
a feasibility study on the decay - pion spectroscopy in a single - arm measurement was performed in 2010 . | during the last two years several experimental approaches to strange systems have been realized at the spectrometer facility of the mainz microtron mami .
an instrument of central importance for the strangeness electro - production program is the magnetic spectrometer that was installed during 20038 and is now operated by the a1 collaboration in @xmath0 reactions on the proton or light nuclei . since 2008 kaon production at low four - momentum transfers off a liquid hydrogen target
was studied .
the measurements were sensitive to details of the phenomenological models describing the reaction .
two very prominent isobar models , kaon - maid and saclay - lyon a , differ in the number of contributing nucleon resonances and their longitudinal couplings at the kinematics measured at mami . in order to use as a zero - degree double - arm spectrometer
a magnetic chicane comprising two compensating sector magnets was constructed and a new electron - arm focal - plane detector system was built . |
large time projection chambers ( tpc ) have been used in high energy particle experiments as central trackers and more recently for identification of heavy nuclei and radioactive beams @xcite , @xcite . at lower
energy , gas tpc were operated both for particle tracking and calorimetry .
a tpc has been used in an experiment searching for @xmath4-e conversion @xcite . in the mev energy range ,
the tpc s involved are of smaller size .
first double beta decay spectra were obtained by elliot et al .
@xcite using a tpc of about 100 @xmath5 . in the gothard experiment @xcite , a tpc with an active volume of 200 @xmath5 provided one of the best limits on neutrinoless double beta decay in @xmath6xe .
tpc with optical readout were also considered for dark matter searches @xcite,@xcite .
the munu collaboration has built a 1 m@xmath0 gas tpc for the experimental study of @xmath7 scattering near a nuclear reactor in bugey ( france ) .
the detector is designed to identify recoil electrons and reconstruct particle tracks well below 1 mev .
the tracking capability of the tpc ( angle , threshold , and track containment ) is essential for background rejection and allows online background measurements .
the tpc vessel is made of transparent acrylic and is installed inside a 10 m@xmath0 light collection vessel , filled with liquid scintillator and viewed by phototubes . both the electrec current and the scintillation light of particles interaction in the tpc gas can be collected . in the following section , a description of the detector
is given whereas in section iii the energy and efficiency calibrations of the detector are described .
waveforms and event identification are presented in section iv . in section v the energy and angular resolutions are discussed .
incident gamma energy reconstruction is also presented .
finally , conclusions are given in section vi .
the general concept of the munu experiment and its detector have been presented previously @xcite . here
we give an update and discuss the performances of the detector .
the detector consists of a central tpc filled with cf@xmath8 gas , inside a peripheral light collection vessel filled with 10 m@xmath0 of liquid scintillator .
it is equiped with photomultipliers and can act as veto for compton and cosmic events .
the detector shown in figure [ fi : detector ] is installed inside low activity passive shieldings composed of borated polyethylene ( 8 cm thick ) and lead ( 15 cm thick ) .
the 1 m@xmath9 time projection chamber can be filled with cf@xmath8 gas at a pressure between 1 and 5 bar .
this gas was chosen because of its high density ( 1.06@xmath10 10@xmath11 electrons per cm@xmath9 at stp ) and its low z. this gives a high target density combined with relatively little multiple scattering .
most data , in particular all those reported in this paper , have been taken at a pressure of 3 bar , at which electron tracks can be reasonably well resolved down to 300 kev .
the gas is contained in a cylindrical acrylic vessel of 90 cm inner diameter , and with an inner length of 162 cm . a cathode made from high purity electrolytic copper foil
is mounted on one lid . a read - out plane consisting , from inside to outside , of a grid , an anode plane and an x - y pick - up plane ,
is mounted on the other side .
the arrangement is shown in figure [ fi : tpc_efield ] .
the grid wires have a diameter of 100 @xmath4 m .
the distance between grid wires is 4.95 mm .
the anode plane has 20 @xmath4 m anode wires , separated by 100 @xmath4 m potential wires .
the distance between anode and potential wires is also 4.95 mm .
the distance from the anode to the x - y plane is 3 mm whereas the grid and the anode plane are separated by a 8.5 mm gap .
the spacing between x strips on one side of the supporting pet foil , or y strips on the other side , is 3.5 mm , giving in total 256 x and 256 y strips .
the drift field is defined by the cathode , the read - out plane , primarily the grid , and the field shaping rings mounted on the outside of the acrylic vessel .
the x - y plane is grounded through the read - out preamplifiers . the drift time and drift velocity
are optimized in order to get the best spatial resolution . at 3 bar ,
we are operating the tpc with a drift field of 206 v / cm .
this gives a drift velocity of 2.14 cm.@xmath4s@xmath13 , and a total drift time of 74.8 @xmath4s .
this was determined from the total drift time of muon tracks spanning the entire chamber , from cathode to grid .
the voltages applied to the last field shaping ring , the grid , the anode and the potential wires must lead to sufficient amplification around the anode wires .
moreover they must provide a symmetric field configuration around the anode wires to minimize mechanical strain and sagging , and a homogeneous drift field all the way down to the grid over the entire area .
the garfield code was used to find a configuration meeting these requirements .
it is shown in figure [ fi : tpc_efield ] .
gas purity is essential for good charge collection , but also to minimize contributions to the background from radon emanations .
the cf@xmath8 gas is therefore constantly circulated at a flow rate of 500 l / h through a commercial high temperature getter made by saes @xcite followed by a cold trap .
the cold trap consists of a coil tubing followed by a a cell filled with 50 g of active charcoal .
it is immersed in an ethanol bath kept at a temperature of 190 @xmath14k .
the signals from the x and y strips and the anode are amplified in current voltage preamplifiers and sampled at 12.5 mhz ( 80 ns sampling time ) in 12 bit fadc s with logarithmic response .
electronics noise on the x and y strips is serious , because of the length , over 2 m , of the flat signal cables from the tpc to the preamplifiers located outside the polyethylene , lead and steel shielding .
a simple procedure allows to significantly reduce it .
first a fourier transform of the signals is performed .
frequency peaks identified as noise sources are eliminated .
the loss in spatial resolution is acceptable .
to give an idea of the quality of the tracks , an example is shown in figure [ fi:2460_11 ] .
the tpc vessel is immersed in a cylindrical steel tank painted with tio2 for diffuse reflection and filled with a mineral oil based liquid scintillator ( ne235 , provided by nuclear enterprise , scotland ) .
the scintillator has an attenuation length of 8 m at 430 nm and is viewed by 48 hemispherical photomultipliers of 20 cm diameter ( emi 9351 low activity series , provided by hamamatzu photonics ) , subdivided in two groups of 24 each on the anode and cathode sides .
the mean quantum efficiency of the photocathodes is around 25 % .
the pmt s are sampled with the same fadc s as the tpc anode and x - y strips .
radioactive sources can be introduce into the light collection vessel at an axial distance of 26 cm from the center of the detector , on the anode side through an acrylic pipe .
the radial position can be as close as 1 cm from the tpc lateral wall .
the main role of the external detector is to identify and to reject the compton electrons induced by @xmath15 rays coming from outside as well as from inside the tpc .
since the tpc walls are made of transparent acrylic , the scintillation light of the cf@xmath1 can also be collected . within the tpc
, the primary scintillation light of cf@xmath1 can be emitted anywhere while avalanche scintillation light is produced only in the vicinity of the anode plane during the amplification process .
the scintillation light spectrum of cf@xmath1 has a narrow maximum at 160 nm and a broader one around 300 nm @xcite-@xcite .
the light is detected by the phototubes through the acrylic walls and the liquid scintillator . due to the transmittance of the acrylic and of the liquid scintillator ,
only a fraction of about 10 % of the spectrum with wavelengths larger than 400 nm makes significant contribution to the light signal in our detector .
as already mentionned , anode , x - y strips and phototube signals are read out with 8 bits flash adc s with logarithmic response .
the memory size is 1024 words .
the sampling interval is 80 ns and is given by a common clock .
this sampling time is well ajusted to the electric pulses from the anode which have a relatively slow rise time of few hundred ns .
signal pulses from the phototubes associated with compton electrons are much faster ( @xmath16 30 ns ) and are shaped to meet the sampling frequency .
the digitized data of the flash adc s provide the pulse shape of the signals delivered by the phototubes , the anode , and the x - y strips .
this way , projection on the x - y planes can be reconstructed .
time bins can be converted to relative position coordinate along the tpc z - axis using the drift velocity in the tpc ( 2.14 cm.@xmath4s@xmath13 in normal running conditions ) .
the scintillation light produced in the liquid scintillator or the primary scintillation light of the cf@xmath1 which is strong enough to be seen for @xmath2 particles and not for nearly minimum ionizing electrons and muons , are collected within one or two sampling bins of 80 ns .
charges arrive to the anode with a time delay proportional to their distances from the anode plane .
absolute z position can thus be defined for events with prompt scintillation light . for events such as confined single electrons , which have no primary scintillation light in cf@xmath1 and no interaction with the liquid scintillator
, only relative z positions can be obtained .
in the following , we first discuss the light collection and energy calibration for the liquid scintillator detector .
in particular , we discuss the @xmath15 ray detection efficiency and we compare the measurements to simulations .
we then describe the energy calibration of the tpc . in the light collection vessel ,
@xmath15 rays undergo multiple compton scattering and are in most cases totally absorbed within the scintillator liquid . radioactive sources ( @xmath3mn , @xmath17cs , @xmath18na )
are used for light collection and energy calibrations . along the cylindrical vessel axis ( z ) , a convenient position parameter is the assymetry ( i.e. difference - to - sum ratio ) of light signals collected by phototubes of the cathode and anode sides . the light collection has been measured as function of this parameter using the total absorption peak of a @xmath3mn source .
experimental data agree with simulations based on known emission spectrum and attenuation length of the liquid scintillator , quantum efficiency of the phototubes , and optical properties of other components . at the central part of the vessel ,
the response is quite homogeneous over a large region ( @xmath19 40 cm from the center along the detector axis ) where about 7.5@xmath20 of photons emitted in the liquid scintillator are collected . at 80 cm from the center , close to both ends of the tpc , the fraction of light collected is 9.5@xmath20 .
it increases to 15 @xmath20 at 120 cm from the center of the detector .
position dependence of the light collection can be corrected for an optimum energy resolution or exploited for event localization .
the quantity of photoelectrons collected was also evaluated from total absorption spectra of @xmath3mn measured for each individual phototube and compared to single photoelectron spectra . for an energy deposition of 1 mev at the central part of the detector , 144 @xmath19
8 photoelectrons were collected in the 48 phototubes while 150 were predicted .
total absorption spectra obtained with different radioactive sources and corrected for position dependence are compared to simulations in figure [ fi : mnac ] .
the energy resolution is 13@xmath20 at 1 mev . in the following text ,
only corrected values of energy will be used .
cf@xmath1 is composed of low z atoms , only a small amount of very low energy @xmath15 rays will lose the entire energy in the gas . most of them undergo a single compton scattering in the gas and are absorbed in the scintillator . the munu detector has been designed for an optimum detection of @xmath15 rays associated with a compton electron inside the tpc . to minimize @xmath15 absorption ,
the lateral walls of tpc are made of very thin ( 5 mm ) acrylic . at an energy , threshold set at 100 kev which corresponds to a total collection of 15 photoelectrons , the contributions of dark noise from individual phototubes
is not negligeable . for this reason
a multiplicity criteria asking for at least 5 phototubes hit is required for a trigger .
the counting rate for the 10 m@xmath9 of liquid scintillator within the shielding is 710 s@xmath13 including 270 s@xmath13 coming from cosmic muons .
the energy spectrum measured at @xmath21 100 kev ( low ) threshold shown in figure [ fi : ac1 ] corresponds to a counting rate of 0.05 s@xmath13 kg@xmath13 of liquid scintillator .
signals from the scintillator can be applied to collect or to veto tpc events associated with a scattered @xmath15 ray .
( similarly , a signal at an energy threshold of @xmath21 8 mev electron equivalent can be applied to veto cosmic muons ) .
the detection efficiency of scattered @xmath15 has been measured by using a @xmath3mn source of 15 kbq inserted at 1 cm from the tpc wall .
electron energy spectra were collected with and without the anti - compton veto operating at 100 kev hardware threshold .
the anti - compton efficiency is evaluated from the ratio of the two electron spectra , corrected from background contributions measured in the absence of the radioactive source .
only 3@xmath20 of the @xmath15 associated with a compton electron of more than 300 kev escape detection .
the scattered photon is lost when it does not interact with the liquid scintillator or when it is absorbed by inert material inside the vessel(acrylic walls , cables ) .
figure [ fi : eff ] shows the comparison of the experimental results to simulations at two thresholds : 100 and 125 kev @xcite . for a @xmath15 ray coming from outside the light collection vessel , the inefficiency is much smaller , since it has to travel twice through about 50 cm of liquid scintillator .
taking also into account the solid angle and the absorption due to the liquid scintillator , the simulation shows that the probability to miss a @xmath15 ray leaving more than 300 kev in the tpc is less than 10@xmath22 .
particle tracks in the tpc produce charges which drift to the anode plane . there
the charge is amplified and avalanche light is produced .
light emission during the multiplication phase in cf@xmath1 has been previously observed@xcite . in our detector , about 2.10@xmath23 photoelectrons
are collected per mev of electron .
electric current signals are more sensitive to electromagnetic noises and have larger pedestal fluctuations .
current and light waveforms from the same electron event will be shown in the following section .
avalanche light signals have better bin to bin pedestal stability and are used instead of the electric current signals for energy evaluation and for determination of track ends .
an event per event comparison shows the linear correlation between the amplitudes the two types of signals as given in figure [ fi : tpcac ] .
compton electron spectra were obtained using a @xmath3mn source both from the electric current and the light and are compared to simulations in figure [ fi : mntpc ] .
mean energy loss of muon events per unit length in the tpc is used to monitor day - to - day gain variations ( figure [ fi : gainmoni ] ) .
the gain changes on the anode plane are mapped with alpha and muon events and are presented in figure [ fi : gainxy ] .
the experimental area in bugey has an overhead shielding in heavy concrete equivalent to 20 m of water . compared to ground level
, the cosmic muon flux is reduced by a factor of five .
the muon counting rate is expected to be 65 s@xmath13 in the tpc and 270 s@xmath13 in the light collection vessel .
a cosmic muon attaining the tpc has to cross more than 100 cm of liquid scintillator . with a mean energy lost of about 1.8 mev / cm
, it leaves a very large prompt signal in the phototubes .
this signal can be used as trigger to select muon events or to veto muon correlated events .
the energy loss of cosmic muons in the cf@xmath1 covers an energy range from hundred kev to a few mev .
the muon mean energy loss per unit length in the tpc can be used to continuously monitor the gain variations in a complementary way to the measurements with radioactive sources .
muon events crossing both cathode and anode have x - y projection confined within the tpc radius and are used to evaluate the drift velocity . the time difference between the beginning and the end of the track correponding to a maximum drift path of 1.62 m. the x - z
, y - z projections of a muon event are shown together with the x - y projection in figure [ fi : muon ] .
alpha particles from natural radioactivity have a range shorter than 1 cm in cf@xmath1 at 3 bar : 4.1 mm at 5 mev and 8.6 mm at 8 mev .
the spatial extensions of @xmath2 events are limited to a few x - y strips and a few time bins .
while no primary scintillation light was seen for compton electrons at a threshold as low as 30 kev electron equivalent , the light from @xmath2 particles was detected .
a prompt signal equivalent an 153 kev electron ( 22 photoelectrons ) is produced in the liquid scintillator by a 5.3 mev @xmath2 particles . in taking into account the light collection and photocathode quantum efficiency ,
we find 1100 photons per @xmath2 particle or 207 @xmath19 30 photons / mev .
this is compatible with values of 220 photons per mev in the range of 360 - 600 nm measured at lower pressures @xcite .
+ the time difference between primary light and avalanche light signals of alpha particles emitted from the cathode provides an other measurement of the drift velocity .
the total charge ( or avalanche light ) collected for @xmath2 particles is much smaller than for electrons of the same energy .
a suppression factor as high as 28 has been measured for @xmath2 particle emitted close to the anode plane .
this factor however decreases when the drift time ( i.e. distance from the anode ) increases ( figure [ fi : alphag ] ) . for alpha particles emitted from the cathode ,
the suppression factor is only 5.6 .
the charge suppression is very likely due to an screening effect in the avalanche region while the change of the suppression factor may be a consequence of electron diffusions along transverse(t ) as well as longitudinal(l ) directions . in the munu tpc , for a drift length of 1.62 m from the anode to the cathode ,
the dispersion @xmath24 or @xmath25 is 2 mm , about half of the range of @xmath2 particles at 5 mev .
the smearing of the charge density along the drift path could than attenuate the charge space screening effect in the avalanche region .
radon was introduced accidentally into the tpc by one of the oxysorb @xcite purifiers used to clean the cf@xmath1 from oxygen and water . a maximun rate of 37 s@xmath26 was measured .
alpha particles detected in the tpc came from the decay of @xmath27rn ( e@xmath28=5.48 mev ) and its radioactive daughters . in the radon @xmath27rn decay chain ,
the beta from @xmath29bi is followed by the @xmath2 from @xmath29po(half life 160 @xmath4s ) .
such delayed coincidence events can be observed easily within the 80 @xmath4 s range of our flash adc s . a correlated @xmath30 event attributed to the @xmath29bi- @xmath29po decay cascade
is shown in figure [ fi : bipo ] . when the oxysorb was removed from the gas circuit ,
both the rates of @xmath2 events and of correlated @xmath31-@xmath2 events from @xmath29bi-@xmath29po decay cascade decreases with a period of 3.2 days compatible with that of @xmath27rn decay period ( figure [ fi : decay ] ) .
after the cf@xmath1 gas change , residual @xmath2 events were still observed at a steady level of @xmath21 1 mbq / m@xmath9 of @xmath27rn in the gas and an activity of 17 @xmath4bq/@xmath32 from the cathode .
for these last events , a well defined drift time and a well defined energy peak at 940 kev electron equivalent were observed .
these events can be attributed to @xmath2 particles from @xmath33po decay , a radioactive daughter of @xmath33pb ( half life 22 years ) .
radioactive daughters of @xmath27rn were very likely drifted to the surface of the copper cathode but apparently not deeply implanted into the metal .
the cathode copper foil has now been replaced by a new one , the surface of which was chemicaly etched with radiopure chemicals before installment .
the rate of @xmath2 from the cathode was reduced by a factor of 10 .
compton electron events generated within the tpc have in most cases an associated scattered @xmath15 ray detected in the light collection vessel .
the sequence for a typical event , shown in figure [ fi : gammae ] is the following : + - first , scintillation light due to the absorption of the scattered @xmath15 ray in the liquid scintillator is detected , + - then , when ionisation charges of the compton electron have drifted to the anode plane , avalanche light is measured with a time delay proportional to the distance from the anode , + - almost simultaneously electric current signals are measured on the anode and x - y strips .
+ the display of the event is completed by the reconstruction of the projections in the x - z / y - z and x - y planes .
the localization of the gamma along the axis of the anti - compton can be evaluated from the assymetry of the light collection(see
section iii.1 ) .
the distribution of time differences of compton electrons and associated scattered @xmath15 s is flat along the tpc which indicates that the background is homogeneous along the tpc .
a sharp drop corresponds to the end of the chamber .
the drift velocity extracted this way is compatible with the muon result .
for such events , charges are created next to the anode wires .
thanks to a larger electric field and thus faster drift time between the grid and the anode the charges created in this region are collected much faster .
the result is a charge blob at the end of the track on the anode side .
if the electron stops inside the gas volume , the normal blob due to increased ionization before stopping will be seen in addition at the true end , resulting in a `` double blob '' event .
normal blobs have a higher integrated charge than anode blobs while anode blobs have charges accumulated in the first or the second time bins at the start of the track .
the avalanche light pulses are used to detect the sharp rise of pulse height and identify tracks originating from the anode plane . a typical `` double blob ''
event is shown in figure [ fi : twoblob ] . + a distribution of pulse heights for an unselected set of electron events is shown in figure [ fi : scrise ] together with a distribution of compton electron events which do not cross the anode .
this last criteria is fullfilled by requiring an associated scattered @xmath15 ray signal well separated from the avalanche signal . for a
well ajusted cut ( 1500 mv or @xmath21 20 kev ) , only 1.5@xmath20 of normal compton electron will be wrongly identified as originating from the anode .
the direction distribution of `` double blob '' events , projected in the yz plane(figure [ fi : da2blob ] ) shows that they are indeed dominantly originated from the anode side .
electrons emitted from lateral walls can be identified from their xy projections . in practice , a fiducial radius is defined to reject events having tracks touching the outermost x or y strips .
only electrons coming from the cathode wall are not distinguishable from contained electrons of the same direction .
+ the munu set - up is optimized for the detection of low energy single electron events in particular those from neutrino electron scattering .
the intense electron anti - neutrino flux from the 2800mw power reactor at 18.6 m amounts to 1.2 10@xmath34 @xmath35 /@xmath36.s .
the direction of neutrinos to the detector is well defined.(@xmath19 5@xmath37 ) a neutrino candidate is a contained single recoil electron in the forward solid angle . in data taking , we take dvantage of the properties of the light signals to apply hardware and online cuts to reject undesired types of events : muon associated events , `` micro electric flashes '' , alpha particles and compton electrons .
electron tracks down to 300 kev have been measured .
the gross counting rate in the tpc ( 70 s@xmath13 ) is dominated by cosmic muons , compton electron events amounting to 0.14 s@xmath13 .
after successive cuts : muon rejection , fiducial volume cuts ( events touching the lateral walls or the anode plane ) , then anti - compton rejection ( 100 kev threshold ) , the counting rate is 780 day@xmath13 .
the counting rates are shown in table [ tb : rates ] .
the recoil energy spectrum of contained electron events is shown in figure [ fi : estpc ] .
+ for neutrino electron scattering events induced by far away reactor , a solid angle cut can further be applied . for a forward angle confined within ( @xmath38/4 ) ,
the residual counting rate is 97 day@xmath13 per 11 kg of cf@xmath1 at an energy threshold of 300 kev or 8.8 ( kg.mev.day)@xmath13 .
+ the recoil energy spectrum of contained electron events does not show any precise structure .
very likely , the remaining background does not have a dominant origine but may have different components : @xmath15 contaminants of the walls and the electrodes of the tpc or single electron from the gas itself ( cosmic activation or presence of radioactive gas ) .
this is under further investigations .
the energy resolution of a wire chamber in proportional regime usually depends on the primary ionization statistics and mainly on the single electron avalanche response spectrum , which dominates in the present case of pure cf@xmath8 through a polya parameter @xmath39 already measured in ref @xcite .
+ moreover this gas reveals a strong attachment for values of the reduced field between 40 and 140 v.@xmath32 .
while no attachment occurs during the drift phase due to the low value of drift field , nearly @xmath40 of the drifted electrons are attached in the early stage of the avalanche , affecting the resolution @xcite .
+ for a primary deposition of 300 kev in the tpc only 2@xmath20 of the primary ionization electrons will survive and reach the multiplication phase of the avalanche , the same number as for a 6 kev deposition and no attachment ; the fluctuations should then be of the same order of magnitude .
+ however the energy of a track is not measured through one single avalanche , but through successive ones ( up to 100 independent avalanches for a 300 kev track , depending on track length , energy loss per unit length and angle in respect to the drift direction ) .
the overall fluctuations are then reduced when compared to a single measurement .
a complete calculation including the previous parameters and simulated tracks with energy loss tables predicts a resolution of @xmath41 at 640 kev in agreement with measurements .
+ due to the different contributions , the resolution is not expected to follow a simple @xmath42 law ; a fit to the simulations at different energies rather shows an empirical @xmath43 law , in better agreement with the low energy part of the compton spectra .
+ the angular resolution has been measured through @xmath44 compton scattering from various sources .
the diffused @xmath15 is measured in the scintillator , the recoiling electron track in the tpc .
the time difference between the @xmath15 and the drifted electron gives the absolute localization along the drift direction , allowing ( together with the xy track projection ) a full 3d determination of the vertex of the interaction and then the exact incident @xmath15 ray direction .
+ the electron scattering angle is obtained by fitting the direction of the first centimeters of the track since the angular resolution is mainly limited by multiple scattering in the first few centimeters of sampled track . in the present case the electron track recognition , the vertex determination and the direction were obtained via a fully automatic image processing which will be described elsewhere .
+ figure [ fi : resolang ] shows the reconstructed 835 kev @xmath15 ray from a @xmath3mn source measured at 3 bars with a recoiling electron energy threshold of 300 kev .
the @xmath45 of the reconstructed energy peak at 835 kev is 220 kev , although the shape is far from a gaussian shape .
the overall angular resolution averaged over the compton recoil spectrum above 300 kev threshold is @xmath46 = 39.7@xmath47 2.2 . to our knowledge
this is the first time that a @xmath21 1 mev photopeak is reconstructed by measuring the compton scattering in a gaseous detector .
this result is encouraging in view of using a gas tpc to detect solar neutrinos .
it demonstrates that the @xmath48be solar neutrino peak at 862 kev could be reconstructed even at 3 bars by measuring the neutrino - electron scattering with a detector based on ths same principe as ours . by lowering the pressure to 1 bar , we shoud
be able to measure the pp solar neutrino which has the end - point at 420 kev .
the munu detector is a 1 m@xmath0 tpc enclosed within a 10 m@xmath0 light collection vessel .
it has been built to study of @xmath49 scattering at the bugey reactor and to set a lower experimental limit on the neutrino magnetic moment .
it is particulary dedicated to detect low energy electrons and to reconstruct their energy and direction .
one of the originalities of munu is the presence of the external light collection vessel .
it sees the light produced by ionizing particles in its scintillator .
this allows to eliminate cosmics related events as well as compton events occuring in the tpc .
in addition the primary scintillation light produced by alphas is well resolved , which allows to identify these particles .
for all these events the measurement of the primary light provides an absolute determination of the track position along the tpc axis .
the primary light from minimum ionizing particles is too weak to be seen , but the light emitted at the anode during the amplification process gives a strong signal .
it complements that from the anode signal , having even lower noise .
it is very useful in characterising the topology of tracks , which will allow for a powerful event selection when analysing the data in terms of neutrino scattering .
installed within low activity shieldings , the detector is now collecting neutrino induced electron events at the bugey reactor at a recoil electron threshold of 300 kev .
+ the result of the reconstruction of the incident energy obtained with the 835 kev @xmath3mn source is encouraging and it shows that the spectroscopy of low energy neutrino is attainable .
the background has to be however further suppressed .
the ultimate goal is to use an upgraded munu detector underground , where the background conditions are optimal , and to look for solar neutrinos .
+ aknowledgement + this work has been supported by the institut national de physique nuclaire et de physique des particules ( in2p3/cnrs ) , infn , le fond national suisse .
+ we are grateful to the staff of the bugey nuclear power plant ( edf ) for their hospitality and help . + the support of the technical staff of the participating laboratories , in particular that of c. barnoux , b. guerre - chaley , l. eraud , m. marton , r. blanc , d. schenker , j.m .
vuilleumier , d. filippi , d. maniero , m. negrello is greatly aknowledged + g.bagieu , r. brissot , j.m .
laborie , g. gervasio took part in the early phase of the experiment , we thank for their contributions .
j. baechler , nucl .
inst . and meth .
409(1 - 3)(1998)9 - 13 v. hlinka et al .
inst . and meth .
419(2 - 3)(1998)503 d. a.
bryman et al .
55 ( 1985 ) 465 s. r. elliott et al.j .
17 supplement ( december 1991 ) s145-s153 j.c .
vuilleumier et al .
phys . rev .
d 48 - 3(1993)1009 k.n .
buckland et al .
73 - 8 ( 1994)1067 u. titt et al . nucl .
inst . and meth .
a , 416 - 1 ( 1998)85 c.amsler et al .
meth . a 396(1997)115 oxisorb filter by messer griesheim gmbh,47805 krefeld , germany saes pure gas , inc .
san luis obispo , california(usa ) monotorr purifiers j.m .
laborie , thesis univ .
de grenoble 1998 c. cerna , thesis univ .
de grenoble 2000 h.a .
van sprang et al . , chemical physics 35 ( 1978)51 a. pansky et al .
inst . and meth . a 354(1995)262 c. broggini , nucl .
meth . a 361(1995)543 l. g. christophorou et al . , j. phys .
chem . ref .
data , vol 25(5 ) ( 1996)1341 - 1375 j. lamblin , thesis univ .
de grenoble ( in process ) written by r. veenhof at cern j.vavra et al nim a324 ( 1993 ) 113 w.s.anderson et al .
nim a323 ( 1992 ) 273 | we report on the performance of a 1 m@xmath0 tpc filled with cf@xmath1 at 3 bar , immersed in liquid scintillator and viewed by photomultipliers .
particle detection , event identification and localization achieved by measuring both the electric current and the scintillation light are presented .
particular features of @xmath2 particle detection are also discussed . finally , the @xmath3mn photopeak , reconstructed from the compton scattering and recoil angle is shown
. + 2 |
a principal goal in sensory neuroscience is to understand how properties of our environment are reflected in neural response patterns .
this can be achieved by constructing encoding models that explicitly link stimuli to responses via an intermediate feature space . in literature ,
several feature spaces have been proposed such as linear @xcite and non - linear @xcite feature spaces .
more recently , research has focused on the development of feature spaces that are themselves learnt from a large amount of training data which embodies the statistical invariances in our environment @xcite . to date , the best predictions of neural responses to naturalistic stimuli have been achieved using deep neural networks @xcite .
an alternative strategy for extracting relevant features from a complex naturalistic stimulus is to directly annotate stimuli in terms of their semantics .
this approach has been used by huth et al .
@xcite , who hand - labelled individual movie frames according to the entities that make up the wordnet ontology @xcite .
it was demonstrated that these entities can be used to predict neural responses across cortex .
this is close in spirit to the work by mitchell et al .
@xcite , who showed that neural responses to presented nouns could be predicted using a feature space that couples nouns to their associated verbs . the goal of the present work is to take the sketched approach to the next level by using semantic vector space models of language that represent individual words in terms of continuous vector representations to predict neural responses to naturalistic stimuli @xcite .
these word embeddings have proven highly successful in the sense that the estimated continuous vector representation can be shown to accurately capture semantic information .
if we are able to use these word embeddings to predict neural responses to naturalistic stimuli then this may provide new insights into the neural representation of semantic information .
moreover , it may offer an elegant way to reconstruct semantic information from patterns of brain activity . here , we tested this approach by annotating individual images with semantic labels ( i.e. words ) and using the associated word embeddings to predict neural responses as people experienced those stimuli .
we show that neural responses in downstream areas are accurately predicted by the word embeddings , demonstrating for the first time that the semantic space implied by word embeddings provide insights in the neural representation of complex naturalistic stimuli .
in order to examine the suitability of word embeddings for predicting neural responses , we made use of the data set that was originally published in @xcite . for each of two male subjects ( s1 and s2 ) , five sessions of data were collected as subjects were presented with natural images . training and test data were collected in the same scan sessions .
the total number of images used for training and testing were 1750 and 120 , respectively .
each training image was repeated two times , and each test image was repeated 13 times . in this study , we analyzed the data from one subject ( s1 ) .
stimuli consisted of grayscale natural images drawn randomly from different photographic collections .
subjects fixated on a central white square .
stimuli were flashed at 200 ms intervals for 1 s followed by 3 s of gray background .
data were acquired using a 4 t inova mr scanner and a quadrature transmit / receive surface coil .
eighteen coronal slices were acquired covering occipital cortex ( slice thickness 2.25 mm , slice gap 0.25 mm , field of view 128@xmath0128 mm@xmath1 ) .
fmri data were acquired using a gradient - echo epi pulse sequence ( matrix size 64@xmath064 , tr 1 s , te 28 ms , flip angle 20@xmath2 , spatial resolution 2@[email protected] mm@xmath3 ) .
fmri scans were coregistered and used to estimate voxel - specific response timecourses .
after deconvolution of these timecourses from the time series data , an estimate of response amplitude was obtained for each presented unique image in each voxel .
voxels were assigned to visual areas using retinotopic mapping data acquired in separate sessions .
additionally , anatomical and functional volumes were coregistered manually .
_ _ freesurfer _ _ and _ _ mrtools _ _ were used for cortical surface reconstruction and visualization , respectively . in order to associate word embeddings with naturalistic image stimuli ,
the following procedure was used .
first , each of the total of 1870 images was manually annotated by an annotator who labeled each image with the first word that came to mind when inspecting the contents of those images .
this led to a total of 512 unique labels for the stimuli in the training set and 81 unique labels for the stimuli in the test set ( 73 overlapping labels between the training and test set ) .
next , each word was converted to a continuous vector representation . to obtain continuous word embeddings we used different models based on the skip - gram word - to - vec ( w2v ) model described in @xcite which learns word representations by predicting the surrounding word of a given word and the global vectors for word representations ( glove ) model described in @xcite which learns word representations by factorizing a word co - occurance matrix . for the w2v model
, we used the pretrained 300-dimensional word representations trained on 100 billion work google news dataset provided online . for the glove model
, we used eight pretrained word representations trained on 6 billion word wikipedia 2014 + gigaword 5 ( 50- , 100- , 200- and 300-dimensional ) and 27 billion word twitter ( 25- , 50- , 100- , 200-dimensional ) datasets provided online .
for a stimulus @xmath4 , we use @xmath5 to denote the word embedding associated with the label @xmath6 assigned to that stimulus . as a control model , to compare low - level feature spaces with high - level feature spaces , we used the gabor wavelet pyramid ( gwp ) model described in @xcite .
the continuous word embeddings were used to predict neural responses to individual stimuli by using them as input to a linear response model .
a separate response model was trained for each voxel using regularized linear regression .
the used estimation procedure is described in detail in @xcite . for each voxel @xmath7 , after estimating the regression coefficients @xmath8 , we obtain @xmath9 as the predicted response of that voxel to input stimulus @xmath4 .
voxel response models were estimated using the entire training set and evaluated on the test set .
a five - fold cross - validation on the training set was used for model selection ( regularization parameters and significant voxels ) .
we discarded the voxels whose cross - validated prediction accuracies on the training set were not significantly higher than chance level ( @xmath10 , student s t - test , bonferroni correction ) . to quantify
how well the used feature space predicts voxel responses , we defined the encoding performance of a model as the pearson correlation coefficient ( @xmath11 ) between the observed and predicted responses on the test set .
we also projected the encoding performance back onto the cortical surface to get an estimate of which regions are sensitive to the semantic information that is embodied by the continuous word vector representations .
we compared the encoding performance of multiple models across the union of the significant voxels for the models being compared since different voxels can be significant for different models .
we set the encoding performance for the nonsignificant voxels that are in the union to zero .
we first compared the mean encoding performance of the 9 different semantic vector space models ( figure 1 ) . recall that the models differed in architecture ( w2v or glove ) , dimensionality ( 25 , 50 , 100 , 200 or 300 dimensions ) , training set ( google news , wikipedia 2014 + gigaword 5 , twitter ) and training set size ( 100b , 6b or 27b words ) . the 300-dimensional w2v model trained on the 100 billion word google news dataset
had the highest mean encoding performance ( 0.2064 ) , whereas the 25-dimensional glove model trained on the 27 billion word twitter dataset had the lowest mean encoding performance ( 0.1655 ) .
the difference in architecture , dimensionality and training set size did not have as much an impact on the mean encoding performance as did the dataset selected for training the model .
for example , the encoding performance of the 300-dimensional w2v model trained on the 100 billion word google news dataset was significantly higher than that of the 50- and 100-dimensional glove models trained on the 27 billion word twitter dataset but not the 6 billion word wikipedia 2014 + gigaword 5 dataset .
, one - way analysis of variance , tukey s range test ) .
for example , the mean encoding performance of model a is significantly higher than that of models f , g and h but not models b , c , d , e and i. ] we then analyzed how the results of the best performing model ( w2v model ) change across the voxels in v1 , v2 , v3 , v3a , v3b , v4 , lo and anterior to v1-lo .
the fraction of the significant voxels for the w2v model and the mean encoding performance of the w2v model significantly increased from upstream to downstream areas .
the lowest and highest fraction of significant voxels were in v3 ( 3.4% ) and lo ( 19.1% ) , respectively .
the lowest and highest mean encoding performance were in v2 ( 0.1679 ) and lo ( 0.2672 ) , respectively . however , 54.9% of the significant voxels were anterior to v1-lo .
the mean prediction accuracy for these voxels was 0.2657 . to validate our findings , we compared the mean encoding performance of the w2v model with that of the gwp model ( figure 3 ) .
recall that the models differ in selectivity to level of stimulus features ( low- or high - level ) .
the difference in mean encoding performance of the models decreased from low- to mid - level areas ( v1 , v2 , v3 , v4 ) and increased from mid- to high - level areas ( v3a , v3b , lo and anterior to v1-lo ) .
the w2v model outperformed the gwp model in high- but not low- and mid - level areas .
in each of the areas , the mean encoding performance of one of the models was significantly higher than that of the other model ( @xmath10 , @xmath12 or @xmath13 , two - sample student s t - test ) .
, * * : @xmath12 and * * * : @xmath13 , two - sample student s t - test .
( b ) projection of the encoding performance of the w2v vs gwp models to the cortical surface .
the dashed curve separates the voxels in v1-lo ( posterior ) from the rest of the voxels ( anterior ) . ]
semantic vector space models that learn a projection of words to a vector space that preserves the meaning of words , similarities between words and analogies have been very successful in computational linguistics .
however , there has been no study on the relationship between such word embeddings and human brain activity . here
, we investigated this relationship by testing the extend to which learned embeddings of visual stimulus labels in a vector space are predictive of human brain activity .
we showed that stimulus - driven voxel responses in downstream visual areas are accurately captured by such word embeddings .
that is , we demonstrated for the first time the relationship between the semantic space implied by word embeddings and the neural representation of complex naturalistic stimuli .
our findings are in line with the previous studies which show that individual concepts are represented in the inferior temporal cortex but expand on them by showing that a continuous , distributed , low - dimensional vector space can be underlying these representations .
future work should provide insights into how concepts are represented in the brain in the absence of complex visual stimulation ( e.g. encoding visual or auditory words ) or in combination with other concepts ( e.g. encoding visual or auditory sentences ) .
we consider this research to be the next step in the quest to elucidate how brains encode their environment .
furthermore , it could one day provide the basis for exciting new applications in brain - computer interface technology .
the data used in this paper is archived at crcns.org under digital object identifier http://dx.doi.org/10.6080/k0qn64ng .
this research was supported by vidi grant number 639.072.513 of the netherlands organization for scientific research ( nwo ) . | encoding models have as their objective to predict neural responses to naturalistic stimuli with the aim of elucidating how sensory information is represented in the brain .
this prediction is achieved by representing the stimulus in terms of a suitable feature space and using this feature space to linearly predict observed neural responses . here , we investigate to what extent semantic vector space models can be used to predict neural responses to complex visual stimuli .
we show that these models provide good predictions of neural responses in downstream visual areas , improving significantly over a low - level control model based on gabor wavelet pyramids .
the outlined approach provides a new way to model and map high - level semantic representations across cortex . |
the ability to measure accurate chemical abundances in h ii regions in our own and other galaxies is vital for our understanding of their chemical evolution .
emission lines emitted by the nebular gas photoionised by massive stars provide us with powerful metallicity indicators for near and intermediate redshift galaxies .
the major complication is posed by the critical task of determining the physical parameters of the nebula , in particular the electron temperature , @xmath0 .
failure to achieve realistic estimates may lead to gross errors in the final abundance determination .
temperature fluctuations within the nebula may be an important cause of error .
collisionally excited lines ( cels ) , which are routinely used in abundance studies , are naturally weighted towards hotter regions and may therefore lead to underestimating of the real abundances .
recombination lines ( rls ) are less affected by errors in temperature determinations , and should in theory yield more accurate results .
one of the outstanding problems in nebular astrophysics is the discrepancy between abundances and electron temperature estimates obtained from cels or rls in planetary nebulae ( pne ) ( e.g. rubin et al . 2002 ; wesson , liu & barlow 2005 ; ercolano et al .
2005 ; liu et al .
2006 ; peimbert & peimbert 2006 and references therein ) , galactic and magellanic h ii regions ( e.g. peimbert , peimbert & ruiz 2000 ; peimbert 2003 ; tsamis et al .
2003 ; esteban et al . 2004
; garca - rojas et al .
2005 , 2006 ) and extragalactic h ii regions ( e.g. peimbert & peimbert 2003 ; peimbert , peimbert & ruiz 2005 ) . the cause of this discrepancy is still uncertain ; temperature fluctuations and chemical inhomogeneities in the gas have both been advocated to explain the discrepancy .
photoionisation models including chemical inhomogeneities have been successful in matching the observed cel and rl spectra of some pne ( e.g. ercolano et al .
2003b ) and h ii regions ( tsamis & pquignot 2005 ) .
however the discrepancy between the abundances derived from recombination lines ( rl ) and those derived from cels has been also been attributed to temperature fluctuations _ not _ caused by abundance inhomogeneities ( e.g. garca - rojas et al .
the @xmath1 parameter ( peimbert , 1967 ; peimbert & costero 1969 ; peimbert 1971 ) , introduced in order to quantify such fluctuations , can be derived empirically by the comparison of temperatures obtained from the depth of the balmer jump to those obtained from cels .
although the situation for h ii regions is notoriously less worrying than for pne , the problem still remains that empirically determined @xmath1 values are larger than those indicated by chemically homogeneous photoionisation models , indicating that the causes of the proposed temperature fluctuations are still not understood .
aside from temperature fluctuations the task of obtaining reliable abundance estimates from cel spectra of h ii regions is further complicated by the fact that often the complete set of emission lines needed for a direct measurement of the electron temperatures is not available for all sources ; the 4363 line of @xmath2o iii@xmath3 $ ] and the 5755 line of @xmath2n ii@xmath3 $ ] are weak and therefore not detected if the spectra do not have good signal - to - noise or if the metallicities are high , implying low electron temperatures .
a large effort has been made to provide metallicity and temperature indicators based on various combinations of strong lines ( e.g. pagel et al .
1979 ; alloin et al . 1979 ; mc gaugh 1991 ; storchi - bergmann et al . 1994 ; vlchez & esteban 1996 ; van zee et al .
1998 ; pilyugin 2001a ; pettini & pagel 2004 ; prez - montero & daz 2005 ) .
hgele et al .
( 2006 ) have recently proposed a new methodology for the production and calibration of empirical relations between the different line temperatures based on observational data alone .
however the data sets available are still too limited to provide usable indicators .
currently , calibrations of metallicity indicators and ionic temperature relations are generally not based on observational data alone , but make use of grids of photoionisation models , run for a range of metallicities and ionisation parameters ( e.g. mcgaugh 1991 , charlot & longhetti 2001 , kewley & dopita 2002 ) . given the vast parameter space generally under investigation ,
most studies so far have been carried out with spherically symmetric or plane parallel geometries , with major limitations on the allowed density distribution and with the spatial distribution of the ionising sources not being investigated .
nearby h ii regions , however , show complex structures in the distribution of gas and stars , which are often intermixed . in this paper
we carry out a theoretical investigation of the importance of the effects due to the spatial distribution of the ionising sources via the construction of a number of 3d photoionisation models , using the mocassin code ( ercolano et al . 2003a , 2005 ) for simple gas density distributions and three spatial configurations for the ionisation sources .
our modelling strategy is described in section 2 .
temperature fluctuations are estimated for our models by computing theoretical @xmath1 and @xmath4 values ( peimbert , 1967 ) for each of our models and the results are given and discussed in section 3 . in this section we also compare integrated emission line spectra from such configurations to search for systematic errors which may be caused by the simplifying assumption of a single , central location for all ionising sources .
we also test the robustness of a number of commonly used ionic temperature relations .
a summary of our main results is presented in section 4 .
cccc z / z@xmath5 & @xmath6 & + & @xmath2e40 erg / sec@xmath3 $ ] & + & & sphere & shell + 2.0 & 3.30 & 2.65 & 3.45 + 1.0 & 3.00 & 2.85 & 3.55 + 0.4 & 2.64 & 3.20 & 3.75 + 0.2 & 2.30 & 3.45 & 3.90 + 0.05 & 1.98 & 3.75 & 4.20 + [ tab : mods1 ] lcccclcccc model & star distrib . & geometry & z / z@xmath5 & symbol & model & star distrib . & geometry & z / z@xmath5 & symbol + csp2.0 & central & sphere & 2.0 & small magenta circle & csh0.4 & central & shell & 0.4 & small red square + hsp2.0 & half distrib & sphere & 2.0 & medium magenta circle & hsh0.4 & half distrib & shell & 0.4 & medium red square + fsp2.0 & fully distrib & sphere & 2.0 & large magenta circle & fsh0.4 & fully distrib & shell & 0.4 & large red square + csh2.0 & central & shell & 2.0 & small magenta square & csp0.2 & central & sphere & 0.2 & small green circle + hsh2.0 & half distrib & shell & 2.0 & medium magenta square & hsp0.2 & half distrib & sphere & 0.2 & medium green circle + fsh2.0 & fully distrib & shell & 2.0 & large magenta square & fsp0.2 & fully distrib & sphere & 0.2 & large green circle + csp1.0 & central & sphere & 1.0 & small black circle & csh0.2 & central & shell & 0.2 & small green square + hsp1.0 & half distrib & sphere & 1.0 & medium black circle & hsh0.2 & half distrib & shell & 0.2 & medium green square + fsp1.0 & fully distrib & sphere & 1.0 & large black circle & fsh0.2 & fully distrib & shell & 0.2 & large green square + csh1.0 & central & shell & 1.0 & small black square & csp0.05 & central & sphere & 0.05 & small green circle + hsh1.0 & half distrib & shell & 1.0 & medium black square & hsp0.05 & half distrib & sphere & 0.05 & medium green circle + fsh1.0 & fully distrib & shell & 1.0 & large black square & fsp0.05 & fully distrib & sphere & 0.05 & large green circle + csp0.4 & central & sphere & 0.4 & small red circle & csh0.05 & central & shell & 0.05 & small green square + hsp0.4 & half distrib & sphere & 0.4 & medium red circle & hsh0.05 & half distrib & shell & 0.05 & medium green square + fsp0.4 & fully distrib & sphere & 0.4 & large red circle & fsh0.05 & fully distrib & shell & 0.05 & large green square + [ tab : mods ] . ] . ] the aim of the current investigation is to uncover possible _ systematic _ differences on the temperature structure and emission line spectra of nebulae ionised by a centrally concentrated set of stars or clusters versus those ionised by the same set of sources which are randomly distributed within the half or full volume , whose respective strmgren spheres do or do not overlap . in particular
we want to test which of the stellar configurations considered produces the largest temperature fluctuations and of what magnitude . at this stage we do not attempt to assess the absolute strengths and weaknesses of one metallicity indicator or ionic temperature relation over others , a task that may only be carried out by a systematic investigation of the large parameter space . in an attempt to isolate the effects of the stellar distribution from those due to the gas density distribution , we consider two extremely simple geometries - a homogeneous spherical volume and a homogeneous spherical shell , both of constant hydrogen number density , @xmath7 = 100 @xmath8 .
the shell models have inner radii of 2.8@xmath910@xmath10 cm and strmgren radii ( corresponding to the case when all stars are located at the centre ) as listed in table [ tab : mods1 ] .
the total number of ionising photons is constant for all models and is @xmath11 = 3.80@xmath910@xmath12 sec@xmath13 . for each density distribution
we are interested in a comparison between the centrally concentrated and the half and fully distributed source cases . in the remainder of this paper
we will refer to models ionised by a central concentration of stars as models c , those ionised by the same set of stars distributed in the half and full volumes will be referred to as models h and f , respectively . case c is equivalent to models ionised by a single central source with a total bolometric luminosity equal to the sum of the bolometric luminosities of the individual sources ( as given in table [ tab : mods1 ] ) and a spectral shape given by the superposition of the spectra of all sources .
these models could also be performed using a 1d code , with a spherically symmetric gas density distribution . similarly
a 1d code could also be employed for case f models where the strmgren spheres do not generally overlap ; the left panel of figure [ fig : dist ] shows a 3d representation of the strmgren sphere distribution for case f , plotted as iso - surfaces where the ionisation fraction of hydrogen is 0.95 .
the adjacent panel shows a projected map of the ionic abundance of h@xmath14 . in the intermediate case , case h
, the strmgren spheres of most h ii regions partially overlap ; the determination of the radiation field in the overlap region is not possible without the application of a 3d code .
as we are interested in a comparison of the three cases ( c , h and f ) , self - consistency is crucial and for this reason we have run all models with the same 3d code , mocassin .
our models include five metallicities ( @xmath15 = 0.05 , 0.2 , 0.4 , 1.0 , 2.0 ) .
the `` solar '' abundance set uses the values from grevesse & sauval ( 1998 ) with the exception of c , n , o abundances which are taken from allende prieto et al 2002 , holweger 2001 and allende prieto et al 2001 , respectively .
these were scaled to lower and higher metallicities considering the empirical abundance trends observed in h ii regions by izotov et al .
( 2006 ) . in order to maximise the effects of spatially distributed sources whose
strmgren spheres may overlap ( completely or partially ) , or be totally independent , we consider the limiting case of an ionising set composed of two stellar populations , a 37 m@xmath5 and a 56 m@xmath5 population , with half of the total ionising photon output [ s@xmath13 ] being emitted by each population .
these two stellar masses were chosen as they have very different @xmath17/@xmath11 ratios , where @xmath11 is the total output [ s@xmath13 ] of h - ionising photons ( energy @xmath18 1 ryd ) and @xmath17 is the total output [ s@xmath13 ] of he - ionising photons ( energy @xmath18 1.8 ryd ) , and are likely to produce the largest effects on the temperature structure and sharpness of the ionisation front .
the ionising spectra for single - mass stars were computed with the starburst99 spectral synthesis code leitherer et al .
( 1999 ) with the up - to - date non - lte stellar atmospheres implemented by smith , norris & crowther ( 2002 ) , using single isochrones for the appropriate stellar masses .
the models were calculated at metallicities consistent with the nebular gas and were obtained for an instantaneous burst , at an age of 1 myr . at solar metallicities for the 37 m@xmath5 stars , the stellar atmosphere models emit 32.3% of luminosity in the h - ionising continuum and 7.7% in the he - ionising continuum , while for the 56 m@xmath5 stars these percentages are 47.9% and 13.7% , respectively .
the exact percentages vary with stellar metallicity , nevertheless the values above are given as a guide to appreciate the different spectral hardness of the two populations
. some other defining parameters of our models , together with their nomenclature and associated symbols are given in table [ tab : mods ] .
the 3d photoionisation code mocassin ( ercolano et al . 2003a , 2005 ) uses a monte carlo approach to the radiation transport problem and it is therefore completely independent of geometry and density distribution .
both the stellar and diffuse components of the radiation field are treated self - consistently , without the need of approximations .
multiple ionisation sources can be located at arbitrary positions in the simulation grid with the only limit being imposed by computing resources .
the atomic data used is frequently updated and include sets of energy levels , collision strengths and transition probabilities from version 5 of the chianti database ( landi et al . , 2005 ) and the improved h i , he i and he ii free - bound continuous emission data recently published by ercolano & storey ( 2006 )
. a public version of the fortran 90 code , which is fully parallelised using the message passing interface ( mpi ) libraries , can be obtained from b. ercolano .
version 2.02.38 was used for the models presented in this work .
we simulate one quadrant of each model , using the technique described and tested by ercolano et al .
( 2003b ) , whereby the positive x - y , y - z and x - z planes intersecting the z- , x- and y- axes respectively at zero , act as mirrors , reflecting the incoming photons back into the simulated cube .
the full volume is finally described by 10@xmath19 cells for the shell models and by 125000 cells for the spherical models ionised by 240 sources .
the number of energy packets used varies during the course of our simulations , but typically 1 - 10 million packets are sufficient for our grids to reach convergence within 10 - 20 iterations .
we experimented with higher resolution grids and a larger number of energy packets and found our results to be virtually unaffected .
as stated above , the aim of this paper is not to provide new calibrations to abundance diagnostic ratios nor to assess their absolute accuracy .
it is therefore not in our intention to create models that fit any particular observations . while trying to maximise the effects of a complex ionising field distribution , it is however still necessary to ensure that the ionisation and temperature structures and , hence , emission line ratios we obtain from our models are in the range of those observed in nature . in figure
[ fig : val1 ] we plot our results in a number of line ratio diagrams , including those following veilleux & osterbrock ( 1987 ) and osterbrock , tran & veilleux ( 1992 ) , that show the variation in @xmath2n ii@xmath3 $ ] and @xmath2s ii@xmath3 $ ] excitation parametrised in terms of @xmath2o iii@xmath3/h\beta$ ] ( top panels ) and those following @xcite showing the variation in terms of r23o iii@xmath3$]5007,4959 + @xmath2o ii@xmath3$]3726,3729)/h@xmath20 .
similarly , in figure [ fig : val2 ] we plot the h ii region ionisation sequence parametrised in terms of oxygen and sulphur line ratios. = ( @xmath2o ii@xmath3$]3726,3729/@xmath2o iii@xmath3$]5007,4959)/(@xmath2s ii@xmath3$]6716,6731/@xmath2s iii@xmath3$]9069,9532 ) ] the grey dots represent giant h ii regions in spiral galaxies , taken from garnett & kennicutt ( 1994 ) , garnett et al .
( 1997 ) , van zee et al .
( 1998 ) , bresolin et al .
( 1999 , 2004 , 2005 ) , and kennicutt et al .
( 2003 ) , spanning metallicities from 12 + log(o / h ) @xmath21 8 to @xmath21 8.8 .
figure [ fig : val1 ] and the top right panel of figure [ fig : val2 ] also include a few points from metal - poor emission line galaxies from izotov et al .
( 2006 ) , which extend to 12 + log(o / h ) @xmath21 7.2 ; it was not possible to use this data for the other three panels of figure [ fig : val2 ] , due to the lack of simultaneous detection of all the oxygen and sulphur lines needed .
in general , our models fall within or near the locus of observed h ii regions , with the exception of our lowest metallicity ( @xmath22 ) and highest metallicity ( @xmath23 ) models , which is not surprising given that the observational sets available did not include many data points at such extreme values of @xmath24 ; furthermore we did not attempt any systematic variation of the ionisation parameter with @xmath24 , while observations of giant h ii regions argue that the ionisation parameter decreases as @xmath24 increases ( see e.g. mc gaugh 1994 ) .
however this does not affect the achievement of our aims .
the temperature structure and emission line spectra obtained by our models are analysed here in detail .
temperature fluctuations , which may introduce errors in the empirical calculations of abundances , are examined .
the robustness of commonly used abundance diagnostics and ionic temperature relations are also tested against possible errors introduced by geometrical effects . [ cols="<,^,^,^,^,^,^",options="header " , ] [ tab : abun ] we have shown that the temperature structure of models with centrally concentrated ionising sources , case c , may vary compared to those of similar models where the sources are distributed within the half volume ( case h ) , which have partially overlapping strmgren spheres , and to those with sources distributed within the full volume ( case f ) , which have fully independent strmgren spheres . case c to case f variations in the temperature structure of the models may have implications for a number of ionic temperature relations .
these scaling laws are often employed in abundance studies when observational data for a given ionic zone is missing .
some of the most popular relations we found in the literature include t@xmath25(s iii ) vs t@xmath25(ar iii ) , t@xmath25(o iii ) vs t@xmath25(ar iii ) , t@xmath25(o ii ) vs t@xmath25(n ii ) .
these ratios are expected to be around unity and we found them to be very little affected by the shifts in temperatures due to the spatial distribution of stars .
this is not surprising , given that the ionic species involved are both high ( o iii , s iii , ar iii and n iii ) or low ( o ii and n ii ) , and therefore while the absolute temperature values in each case may shift to higher or lower values , the resulting ratios remain virtually unaffected .
scaling laws have also been derived for t@xmath25(o iii ) vs t@xmath25(s iii ) and t@xmath25(o iii ) vs t@xmath25(o ii ) , by garnett ( 1992 ) , based on his own grids of photoionisation models and those by stasiska ( 1982 ) .
more recently , izotov et al . ( 2006 ) produced somewhat different relations between t@xmath25(o ii ) and t@xmath25(o iii ) , and between t@xmath25(s iii ) and t@xmath25(o ii ) ( their eqs .
14 and 15 , valid only for metal poor cases to 12 + log(o / h ) @xmath21 8.2 ) , based on a set of up - to - date ( but still spherically symmetric ) photoionisation models that reproduce the observed trends of metal - poor galaxies .
they found that the observed values of t@xmath25(o ii ) , t@xmath25(o iii ) and t@xmath25(s iii ) in a large sample of h ii galaxies reproduce the theoretical relations , but with a large scatter ( not only attributable to observational errors in the case of t@xmath25(o ii ) , t@xmath25(o iii ) , see their figures 4a and b ) .
figure [ fig : temprel ] shows the behaviour of our z / z@xmath5 = 1.0 and 2.0 ( left panel , black and magenta symbols respectively ) and z / z@xmath5 = 0.05 , 0.2 and 0.4 ( right panel , blue , green and red symbols , respectively ) metallicity models for cases c , h and f with regards to the t@xmath25(o ii ) versus t@xmath25(o iii ) scaling laws of garnett ( 1992 , black solid line in both panels ) and izotov et al .
( 2006 , coloured curves in the right panels ) .
the symbols are as described in table [ tab : mods ] .
our lowest metallicity bin z / z@xmath5 = 0.05 ( corresponding to 12 + log(o / h)= 7.39 ) falls in between izotov s 12 + log(o / h)= 7.2 and 7.6 metallicity bins .
their t(o iii ) versus t(o ii ) relations given in their eqns 14 are therefore plotted in blue in the right panel of our figure [ fig : temprel ] , with the thinner and thicker curves indicating the 7.2 and 7.6 metallicity bin , respectively .
our z / z@xmath5 = 0.4 metallicity bin ( red symbols ) corresponds to 12 + log(o / h)= 8.29 , slightly above izotov s 12 + log(o / h ) = 8.2 bin which , nevertheless is represented by the red line in figure [ fig : temprel ] .
a mixed relation such as t@xmath25(o iii ) vs t@xmath25(o ii ) is much more affected by the fact that low and high ionic temperatures shift in opposite directions for intermediate ( red points ; z / z@xmath5 = 0.4 ) and high metallicities ( magenta and black points ; z / z@xmath5 = 2 and 1 ) .
this will contribute to the scatter noticed observationally for this relation .
for example , kennicutt , bresolin & garnett ( 2003 ) presented a study of those h ii regions in m 101 for which direct measurements of the nebular auroral lines could be obtained .
they found the t@xmath25(o iii ) vs t@xmath25(s iii ) relation to be matched closely by their observations , whereas the t@xmath25(o iii ) and t@xmath25(o ii ) temperatures turned out to be rather uncorrelated .
it is also worth noting at this point tht another factor that could contribute to the scatter in the t@xmath25(o iii ) vs t@xmath25(o ii ) relation is the dependency of t@xmath25(o ii ) on the electron density ( prez - montero & daz 2003 ) .
the scatter found by our models is for a given electron density , and therefore it may be even larger for a different sample covering a wider range of densities .
the ionising flux distribution is certainly a factor that contributes to weakening the correlation , however , as pointed out by stasiska ( 2005 ) , at high metallicities , the temperature derived from the @xmath2o ii@xmath3 $ ] line ratio may by strongly in error due to the contribution of recombination from o@xmath26 .
we have assessed that the geometrical distribution of stars within an h ii region plays a role in the temperature structure of the gas , the magnitude of these effects will clearly be dependent on the total mass of the stars and the concentration level .
it is important now to verify the robustness of commonly used metallicity indicator against these temperature shifts .
in particular we are interested in identifying systematic trends with stellar distribution rather than absolute errors .
table [ tab : abun ] lists the _
empirical _ oxygen abundances obtained using 6 different indicators .
we used the following calibrations for the 6 indicators : ( 1 ) o@xmath27 pilyugin ( 2000 , 2001b ) , ( 2 ) o@xmath28n@xmath29 stasiska ( 2006 ) , ( 3 ) n@xmath29 pettini & pagel ( 2004 ) , ( 4 ) s@xmath27 prez - montero & daz ( 2005 ) , ( 5 ) s@xmath28o@xmath28 stasiska ( 2006 ) , ( 6 ) ar@xmath28o@xmath28 stasiska ( 2006 ) .
it is worth reiterating at this point that metallicity indicators should , and generally are , only used for statistical studies as the error on a single region may be very large ; the models presented in this work do not attempt to cover the full parameter space occupied by h ii regions and therefore do not aim at identifying the most accurate indicator _ in absolute terms_. here we are mainly interested in studying a possible systematic error in the derived abundances introduced by the 3d stellar distribution . for this reason for each trio of models ( c , h and f for a given density distribution and metallicity )
we compute @xmath30 , the largest difference between the oxygen abundance in log units derived from cases c , h and f. the results are summarised in table [ tab : abun ] , where the mean values @xmath31 are also listed . for all metallicity indicators apart from o@xmath27 , there is a clear trend for higher metallicity being derived from models ionised by fully distributed sources , case f , than from case c and h models .
the reverse is true for o@xmath27 , where metallicities derived from case f models are smaller than those derived from case c and h models .
the @xmath30 values at a given metallicity vary from one indicator to the next , however they are rarely below @xmath210.1 dex ( only for the highest metallicity case ) , with more representative values around 0.3 dex for the spherical case , but often larger than 0.4 ( and larger than 1.1 dex in one case ) .
these deviations are slightly smaller for the shell density distributions ; this is obvious as even for f cases most of the ionising radiation in these models will be emitted from the central cavity , reducing the differences between c and f cases .
we note that in some cases the values given by the case h models are lower than those given by the respective case c models ( rather than being equal or in the middle between c and f ) .
we have analysed these deviations statistically and can confirm that the small differences are simply due to the variance of our monte carlo models and do not bear any physical significance .
the reason for the systematic effect we see in the _ empirical _ abundance determinations can be understood by re - examining figure [ fig : val2 ] , where the ionisation sequence of our model is parametrised in terms of oxygen and sulphur emission lines .
the c ( small symbols ) , h ( medium symbols ) and f ( large symbols ) form a sequence , with the case f models generally showing a lower ionisation parameter than case c and h models .
the differences in the abundances derived from case c to f with the various indicators clearly reflect the different dependence of each indicator on the ionisation parameter .
this can be simply shown for an idealised system .
the ionisation parameter of a pure - hydrogen spherical volume of gas with number density @xmath32 ionised by a source emitting @xmath33 hydrogen - ionising photons per second is defined as @xmath34 = @xmath33/(4@xmath35 ) , where @xmath36 is the strmgren radius .
in such a system , from the ionisation balance equation ( e.g. osterbrock , 1989 , eqn 2.19 ) @xmath36 is directly proportional to @xmath37 and to @xmath38 , neglecting the temperature dependence of the hydrogen recombination coefficient . for simplicity
we compare the ionisation parameter of the system above , @xmath39 , ionised by only one source , to that of a similar system ( f ) ionised by @xmath40 identical sources each emitting @xmath33/@xmath40 hydrogen - ionising photons per second . with these assumptions
it is easy to show that the ionisation parameter , @xmath41 , for system f , measured at the strmgren radius of each individual source is simply related to @xmath39 by @xmath42 and therefore always smaller than @xmath39 for @xmath43 the magnitude of the errors in the metallicity derived by the strong line methods are significant .
it is , however , true that the effects reported here represent a worst case scenario , and our extreme assumptions on stellar populations create a large dispersion in the resulting ionisation parameters , which in some cases exceeds the observed range , as seen in figures [ fig : val1 ] and [ fig : val2 ] . aside from the magnitude of the @xmath30 values , however , a worrying aspect is the fact that the discrepancies between cases c , h and f are systematic .
this can have an impact on galactic metallicity gradients determined via strong line methods , if these are calibrated via photoionisation models .
in fact , if compact clusters ( close to case c ) and loose associations ( close to case f ) are randomly distributed throughout a given galaxy , then the systematic errors would only cause a larger scatter in the observed metallicities , but , given sufficient number statistics , they would not affect the measured metallicity gradient .
however if the ratios of compact clusters to loose associations is somewhat dependent on the galactocentric distance , then the systematic errors due to stellar geometrical distributions may indeed introduce a bias on the measured galactic metallicity gradient , if the abundances are obtained from strong line methods calibrated on ab - initio models which do not reproduce the observed excitation of h ii regions .
for example , recent work by rosolowsky et al .
( 2007 ) , presenting high resolution molecular gas maps of m33 , showed a truncation in the mass distribution of giant molecular clouds ( gmcs ) at a galactocentric distance of 4 kpc . a recent study on the demographics of young star - forming clusters in m33 by bastian et al .
( 2007 ) also shows the same cut - off at 4 kpc for the clusters detected .
we could interpret this as tentative evidence of different star formation environments from the centre to the edge of m33 , however , we prefer to postpone this discussion until more compelling observational evidence becomes available .
following our theoretical investigation on the effects of the spatial configuration of ionisation sources on the temperature structure of h ii regions , we summarise our conclusions as follows : 1 . for intermediate to high metallicities ( 0.4@xmath44 2 ) , for a given gas density distribution , abundance and ionising spectral shape and intensity , a model with a central concentration of stars ( case c ) will result in higher ionic temperatures for high ionisation species ( o@xmath45 , s@xmath45 etc .
) , compared to the same model with stars fully distributed within the volume ( case f ) .
the opposite is true for low ionisation species ( e.g. o@xmath46 , n@xmath46 ) .
this results in a shallower gradients in the electron temperature distribution across the ionic species zones .
2 . low metallicity models ( z / z@xmath47 0.2 ) do not show the temperature inversion from low to high ionic species zones , rather a shift in the temperature is experienced by all ionic species zones , resulting in case f models being cooler than case c and h models .
3 . at intermediate to high metallicity ( @xmath48 ) ,
models with stars distributed within the full volume are more isothermal ( show lower @xmath1 values ) than the same models with a central concentration of stars .
these temperature `` fluctuations '' obtained for case c models are a simple consequence of a large temperature gradient .
multiple ionising sources of different temperatures at central or non - central locations do not produce significant temperature fluctuations in the ionised gas of models with @xmath48 .
4 . at low metallicities ( @xmath49 ) , models with stars distributed within the full volume ( case f )
show _ larger _ @xmath1 values than the same models with a central concentration of stars . here
we are seeing the effects of true temperature `` fluctuations '' for case f models .
the magnitude of @xmath50 remains however too small to have any significant effect on derived abundances .
multiple ionising sources of different temperatures at central or non - central locations are not the cause of significant temperature fluctuations in the ionised gas of our models with @xmath51 .
the relation @xmath1(o@xmath14 ) = @xmath1(o@xmath45 ) , often used in empirical studies , is not verified by our models .
extreme care should be taken to account for the uncertainties introduced by the use of this relation in studies seeking to apply corrections to cel - derived abundances making use of an empirical estimation of temperature fluctuations .
for our @xmath15 = 2 models , @xmath1(o@xmath45 ) is always a factor of 2 or more higher than @xmath1(o@xmath14 ) , while for lower metallicities @xmath1(o@xmath45 ) becomes lower than @xmath1(o@xmath14 ) sometimes by large factors ( up to approximately 10 ) .
7 . for intermediate to high metallicity models , electron temperatures in the o@xmath45 and o@xmath46 ionisation zones
are shifted in opposite directions , contributing to the scatter observed in the @xmath52 versus @xmath53 relation .
we confirm that h ii region abundances derived on the basis of the t@xmath25(o ii ) alone should , therefore , be considered highly uncertain . 8 .
metallicity indicators calibrated by grids of spherically symmetric photoionisation models may suffer a systematic bias , due to their dependence on the ionisation parameter of the system .
for the same input parameters case f models will always result in smaller ionisation parameters than case c and h models .
the errors estimated in this work ( typically 0.3 dex , but larger in some cases ) are likely to represent the worst - case scenario , but nevertheless their magnitude and their systematic nature does not allow them to be ignored . *
acknowledgments : * be would like to thank the organisers , fabio bresolin and lisa kewley , and all the participants to the metals07 workshop on metallicity calibrations for gaseous nebulae ( held at the institute for astronomy , of the university of hawaii on jan 22 - 26 2007 ) , for productive discussion and advice for this work
. we would also like to thank rob kennicutt for his many suggestions and christophe morisset and luc jamet for their comments and for their careful study of our results .
be was partially supported by _
grants go6 - 7008x and go6 - 7009x .
nb was supported by a pparc postdoctoral fellowship .
nb gratefully acknowledges the hospitality of the harvard - smithsonian center for astrophysics , where a significant part of this work took place .
the authors wish to thank the referee jorge garca - rojas for helpful comments and constructive discussion .
osterbrock , d. e. 1989 , research supported by the university of california , john simon guggenheim memorial foundation , university of minnesota , et al .
mill valley , ca , university science books , 1989 , 422 p. | spatially resolved studies of star forming regions show that the assumption of spherical geometry is not realistic in most cases , with a major complication posed by the gas being ionised by multiple non - centrally located stars or star clusters .
geometrical effects including the spatial configuration of ionising sources affect the temperature and ionisation structure of these regions .
we try to isolate the effects of multiple non - centrally located stars , via the construction of 3d photoionisation models using the 3d monte carlo photoionisation code mocassin with very simple gas density distributions , but various spatial configurations for the ionisation sources .
our first aim is to study the resulting temperature structure of the gas and investigate the behaviour of temperature fluctuations within the ionised region .
we show that geometry affects the temperature structures in our models differently according to metallicity . for the geometries and
stellar populations considered in our study , at intermediate and high metallicities , models with ionising sources distributed in the full volume , whose strmgren spheres rarely overlap , show _ smaller _ temperature fluctuation than their central ionisation counterparts , with fully overlapping concentric strmgren spheres .
the reverse is true at low metallicities .
finally the true temperature _ fluctuations _ due to the stellar distribution ( as opposed to the large - scale temperature gradients due to other gas properties ) are small in all cases and not a significant cause of error in metallicity studies .
emission line spectra from h ii regions are often used to study the metallicity of star - forming regions , as well as providing a constraint for temperatures and luminosities of the ionising sources .
empirical metallicity diagnostics must often be calibrated with the aid of photoionisation models .
however , most studies so far have been carried out by assuming spherical or plane - parallel geometries , with major limitations on allowed gas and dust density distributions and with the spatial distribution of multiple , non - centrally located ionising sources not being accounted for .
we compare integrated emission line spectra from our models and quantify any systematic errors caused by the simplifying assumption of a single , central location for all ionising sources .
we find that the dependence of the metallicity indicators on the ionisation parameter causes a clear bias , due to the fact that models with a fully distributed configuration of stars always display lower ionisation parameters than their fully concentrated counterparts .
the errors found imply that the geometrical distribution of ionisation sources may partly account for the large scatter in metallicities derived using model - calibrated empirical methods .
( ism ) : h ii regions ; galaxies : abundances ; radiative transfer |
in the last several years the use of the overlap lattice fermion has become more popular because the conceptual and technical clarity that results from its exact chiral symmetry is seen to overcome its superficially higher computational cost as compared to conventional lattice formulations of the fermion .
furthermore , the disparity in numerical intensity can be mitigated by using coarse lattice spacing , given that good scaling and localization properties are established .
hernndez , jansen , and lscher @xcite have shown numerically that neuberger s overlap operator is local on pure - glue backgrounds ( using the wilson gauge action on fine lattices ) .
the decay rate ( inverse of the range ) @xmath1 of the overlap kernel can be interpreted as the mass of unphysical degrees of freedom .
recently , golterman , shamir , and svetitsky @xcite have argued that , in practice , these can be dangerously light .
they provide an indirect estimate for pure - gauge ensembles ( for wilson , iwasaki , and dbw2 gauge actions ) at different lattice spacings . for cutoff
@xmath2 they estimate @xmath3 , which is acceptably large . for @xmath4 , however , they raise an alarm with their estimate of @xmath5 for the wilson case , @xmath6 for iwasaki , and @xmath7 for dbw2 . however , their estimate is indirect , and follows a chain of approximations at the end of which they assume that @xmath8 , where @xmath9 is the `` mobility edge '' of a dirac spectrum .
here we come to a very different conclusion using the most direct approach which takes advantage of the fact that it is not expensive to _ calculate _ , rather than _ estimate _ , @xmath10 .
( by comparison , it would be much more expensive to calculate the overlap propagator @xmath11 which must be computed by nested iterations . ) in contrast to the results of golterman , shamir , and svetitsky @xcite , who speculate that overlap simulations with a cutoff of @xmath0 ( such as @xcite ) might have a range as long as 4 lattice units , here we show directly that such is not the case ; with appropriate choice of negative wilson s mass , the range is about 1 lattice unit ( in euclidean distance or 2 units of `` taxi - driver '' distance ) .
all is well .
a preliminary version of these results was presented in @xcite .
durr , hoelbling , and wenger @xcite came to similar conclusions using the overlap operator with a uv filtered ( `` thick link '' ) wilson kernel .
we use the renormalization - group - improved iwasaki @xcite gauge action , on three different lattices ; for each , the lattice size , lattice spacing , and number of configurations used are tabulated in table [ table : lattices ] .
.[table : lattices ] lattice size , lattice spacing ( as set by the sommer parameter @xmath12
@xcite ) , lattice spacing ( as set by @xmath13 @xcite ) , and number of configurations ( for scaling / locality ) .
[ cols="<,<,<,>",options="header " , ] so even at lattices as coarse as @xmath14 ( @xmath15 scale ) , the range is about 1 lattice unit ( measured using euclidean distance , or 2 units using taxi - driver distance ) .
no unphysical degrees of freedom are induced at distances longer than the lattice cutoff .
we include here a useful check of our calculations and expectations . with the same program used for the interacting - field case , we have calculated @xmath16 for source @xmath17 and
sink @xmath18 for the free - field case , that is , with all @xmath19 matrices on the links set equal to the identity matrix .
we compared the output with an alternative calculation in terms of fourier series and found complete agreement .
as an aside , we then repeated the calculation of the overlap operator range for the free - field case .
free field case for the @xmath20 lattice ( cut=1 ) .
@xmath21 is displayed ( red points ) for all coordinates ; the average @xmath22 over all equivalent points with a given taxi - driver distance @xmath23 is shown in blue . ] for the @xmath20 lattice , we display in fig . [
fig : free ] a plot for the free - field case similar to that of the interacting case ( fig .
[ fig : taxi_16 ] ) , and using the same value to cut the data . an important difference , however , is that we display the value of @xmath21 for _ all _ points , that is , the maximum is not taken as in eq .
[ taxi_def ] .
one sees in fig .
[ fig : free ] that , in contrast to the interacting case , the maximum values for each taxi - driver distance do not form a smooth envelope
. presumably , this is a consequence of the lack of rotational invariance for the free case .
this lack of smoothness makes it difficult to obtain a quantitative estimate of the range .
a much smoother curve is obtained by taking the average ( mean ) @xmath22 , rather than the maximum , of all equivalent points which have the same taxi - driver distance .
but alas , this is different from what has been presented for the interacting case . to make a direct comparison of the interacting to the free case , we plot in fig . [
fig : ratio_free ] the ratio of @xmath24 for the interacting to @xmath25 for the free - field case . here , for each configuration , the average is taken over all coordinates which have a given taxi - driver distance @xmath23 .
then an average and standard error are calculated over all configurations .
ratio of interacting to free field case ( cut=1 ) . ] for large taxi - driver distance @xmath23 , we expect an exponential decay @xmath26 , and thus expect the ratio to decay exponentially as @xmath27 where @xmath28 the graph is fit quite nicely with a long straight line giving a value @xmath29 .
notice that a positive value indicates that @xmath30 .
initially , this was not expected from preliminary mobility - edge arguments @xcite .
but it is now acknowledged that the connection between mobility edge and overlap range is severed already in the free - field case @xcite .
furthermore , a crude heuristic argument in support of the observation can be made .
to wit , regard @xmath16 as a sum of contributions from various paths of links connecting source @xmath17 to sink @xmath18 . in the free case ,
the product of these links is the identity . in the interacting case
, it is a path - dependent matrix with norm less than one
. then one would expect @xmath31 for each @xmath23 .
if each decays exponentially at large @xmath23 , then one can not have @xmath32 without contradiction .
at lattice 2004 , davies _ et al . _ @xcite collected world data to demonstrate that different quenched quark formulations
could have a consistent continuum limit .
`` aoki '' plot for various quenched data , as obtained from @xcite .
our data ( solid circles ) are labeled `` overlap '' . ] the conclusion , as illustrated in fig .
[ fig : aoki ] , is that they could .
but we emphasize that the constrained fit demanded that there exist a global continuum limit ( by design ) .
furthermore , the global continuum limit differed substantially from the continuum limit obtained solely from the wilson formulation where the discretization errors are the largest .
the lesson learned is that with large discretization errors , it is quite possible to be misled when extrapolating to the continuum limit , even with high statistics and many lattice spacings .
it is important to seek a formulation with very small discretization errors to be able to trust the continuum extrapolation .
here we add our data `` overlap '' ( solid circles in fig .
[ fig : aoki ] ) to the global quenched spectrum data and find that of all the formulations , its discretization errors are smallest , allowing for viable computation at surprisingly coarse lattice spacing . as another example of the efficacy of the overlap formulation
, we made a non - perturbative computation @xcite of the renormalization constants of composite operators on the @xmath33 lattice using the regularization independent scheme .
we found that the relations @xmath34 and @xmath35 agree well ( within 1% ) above the renormalization point @xmath36 .
the @xmath37 and @xmath38 corrections of the renormalization are small ; the mass dependence is less than about 3% up to @xmath39 .
it is viable to simulate quenched overlap fermions at surprisingly coarse lattice spacing .
locality is well under control ; the range ( characteristic exponential decay length ) is about one lattice unit ( of euclidean distance , or about two lattice units of taxi - driver distance ) for lattice spacing as coarse as @xmath40 ( @xmath15 scale ) , such as that used in @xcite , and trends to zero ( in physical units ) in the continuum limit .
scaling is excellent ; the aoki plot is almost flat up to @xmath40 .
the overlap fermion outperforms the other formulations , that is , discretization errors are the smallest for overlap .
non - perturbative renormalization of operators show little mass dependence @xcite ; e.g. less than about 3% up to @xmath39 for the renormalization factors .
this work is supported in part by the u.s .
department of energy under grant de - fg05 - 84er40154 .
the work of nm is supported by u.s .
doe contract no .
de - ac05 - 06or23177 , under which jefferson science associates , llc operates jefferson lab . | the overlap fermion offers the considerable advantage of exact chiral symmetry on the lattice , but is numerically intensive .
this can be made affordable while still providing large lattice volumes , by using coarse lattice spacing , given that good scaling and localization properties are established . here ,
using overlap fermions on quenched iwasaki gauge configurations , we demonstrate directly that , with appropriate choice of negative wilson s mass , the overlap dirac operator s range is comfortably small in lattice units for each of the lattice spacings 0.20 fm , 0.17 fm , and 0.13 fm ( and scales to zero in physical units in the continuum limit ) . in particular ,
our direct results contradict recent speculation that an inverse lattice spacing of @xmath0 is too low to have satisfactory localization .
furthermore , hadronic masses ( available on the two coarser lattices ) scale very well .
overlap fermion , locality 11.15.ha , 12.38.gc |
in theories where matter confines to a 4-d brane and gravity is the only massless field able to propagate along the extra dimensional volume , one generally expects short distance corrections to the usual 4-d newtonian potential .
the shape and distance at which these corrections become relevant generally depend on the geometry and size of the extra dimensional volume , thus allowing for distinctive signals dependent of the particular content of the theory .
for instance , in the single - brane randall - sundrum scenario @xcite , where a 4-d brane of constant tension @xmath4 is immersed in an infinitely large ads@xmath5 volume , a zero mode graviton @xmath6 localizes about the brane .
this zero mode is exponentially suppressed away from the brane with a warp factor @xmath7 , where @xmath8 is the distance from the brane along the fifth extra - dimensional direction .
the newtonian potential describing the gravitational interaction between two bodies of masses @xmath9 and @xmath10 in the brane , and separated by a distance @xmath11 , is then found to be @xcite @xmath12 where @xmath13 is newton s constant .
the correction @xmath14 springs out directly from the way in which gravitons propagate in an ads@xmath5 spacetime .
if the tension @xmath15 is small enough as compared to the planck mass @xmath16 , then it would be possible to distinguish this type of scenario from other extra - dimensional models in short distance tests of gravity .
present tests @xcite give the robust constraint @xmath17 m .
it is therefore sensible to ask how other braneworld scenarios may differ from the randall - sundrum case at short distances , especially within the context of more realistic models . in what follows
we address this question for a fairly general class of supersymmetric braneworld scenarios with a bulk scalar field , where the geometry of the extra - dimensional space differs from the usual ads profile .
we show that the newtonian potential for this type of models picks up a non - trivial correction at scales comparable to the tension of the brane @xcite , that differs dramatically from the one shown in eq .
( [ v - rs ] ) .
let us consider a 5-d spacetime @xmath18 , where @xmath19 is a fixed 4-d lorentzian manifold and @xmath20 is the orbifold constructed from a circle with points identified through a @xmath21-symmetry .
@xmath22 is bounded by two 3-branes , @xmath23 and @xmath24 , located at the fixed points of @xmath20 .
there is a bulk scalar field @xmath0 with a bulk potential @xmath25 and boundary values @xmath26 and @xmath27 at the branes .
additionally , the branes have tensions @xmath28 and @xmath29 which are given functions of @xmath26 and @xmath27 ( see fig.[fig ] ) . with a bulk potential @xmath25 .
additionally , the bulk space is bounded by branes @xmath23 and @xmath24 located at the orbifold fixed points .
the branes are characterized by tensions @xmath28 and @xmath29 , and may contain matter fields @xmath30 and @xmath31 respectively . ]
the total action of the system is @xmath32 - \frac{3 m_{5}^{3}}{2 } \ , \int_{\sigma_{1 } } \ !
\lambda_{1 } ( \phi^{1 } ) - \frac{3 m_{5}^{3}}{2 } \ , \int_{\sigma_{2 } } \
! \lambda_{2 } ( \phi^{2 } ) .
\label{c2 : s - g}\end{aligned}\ ] ] here @xmath33 is the short notation for @xmath34 , where @xmath35 is the determinant of the 5-d metric @xmath36 of signature @xmath37 ( a similar convention follows for @xmath38 ) .
@xmath39 is the 5-d fundamental mass scale and @xmath40 is the 5-d ricci scalar .
additionally , @xmath41 and @xmath42 are the brane tensions .
our interest is focused on a class of models embedded in supergravity , where the bulk potential @xmath25 and the brane tensions @xmath41 and @xmath42 satisfy a special relation so as to preserve half of the local supersymmetry near the branes @xcite .
this is @xmath43 where @xmath44 is the superpotential of the system . under these conditions the system presents an important property : there is a bps vacuum state consisting of a static bulk background in which the branes can be allocated anywhere , without obstruction . indeed , suppose a metric @xmath45 , where @xmath8 parameterizes the extra - dimension and @xmath6 is the induced metric on the 4-d foliations of @xmath22 parallel to the branes .
if the bulk fields depend only on @xmath8 and @xmath46 ( with @xmath47 the minkowski metric ) then one finds that the entire system is solved by functions @xmath48 and @xmath49 satisfying @xmath50 where @xmath51 .
remarkably , boundary conditions at the fixed points are also given by these two equations .
thus , the presence of the branes forces the system to acquire a domain - wall - like vacuum background , instead of a flat 5-d minkowski background .
let us mention here that in order to have the right phenomenology in this type of scenarios , it is important to have the moduli @xmath52 and @xmath53 stabilized ( in the 4-d effective low energy theory , they are found to be massless and dangerously coupled to matter @xcite ) .
this can be done in a simple way by breaking supersymmetry on the branes in an appropriate manner @xcite .
it is possible to compute the modifications to newton s law arising from the way in which bulk gravitons propagate in a background given by eq .
( [ bps - equ ] ) . in general , we can write the newtonian potential describing the gravitational interaction between two masses @xmath9 and @xmath10 on the visible brane as @xmath54 $ ] , where @xmath55 is a function of @xmath11 whose shape is dictated by the form of @xmath1 . for concreteness ,
let us consider the case of dilatonic braneworlds @xmath56 where @xmath57 is some fundamental mass scale .
then , the background geometry of the system is @xmath58 , \qquad \mathrm{and } \qquad \omega(z ) = \left [ 1 - \alpha^{2 } w_{0 } z \right]^{1/4\alpha^{2}}.\end{aligned}\ ] ] notice the presence of a singularity @xmath59 at @xmath60 . without loss of generality , one may take the position of @xmath23 at @xmath61 ( since @xmath62 , this is a positive tension brane ) .
then , @xmath24 can be anywhere between @xmath61 and @xmath60 .
interestingly , the relevant case of 5-d heterotic m - theory @xcite corresponds to @xmath63 .
if , for simplicity , we further assume that the visible brane is @xmath23 while the second brane @xmath24 is very close to the bulk singularity , then it is possible to find three different solutions for @xmath55 , depending on the value of @xmath3 ( see @xcite for a detailed analysis of this ) @xmath64 + y_{\nu - 1}^{2}[m / k ] } & \qquad \textrm{if $ \alpha^{2 } < 1/4 $ } \\
\frac{32}{9 \pi w_{0 } } \int_{\sqrt{b}}^{\infty } dm \frac{\sqrt{m^{2 } - b}}{m } e^{-mr } & \qquad \textrm{if $ \alpha^{2 } = 1/4 $ } \\
\frac{4}{3 } \frac{4 \alpha^{2 } -1}{1 + 2 \alpha^{2 } } \sum_{n } e^{- k \ , u^{\mu+1}_{n } r } & \qquad \textrm{if $ \alpha^{2 } > 1/4 $ } \end{array } \right.\end{aligned}\ ] ] in the previous expression we have defined @xmath65 , @xmath66 , @xmath67 and @xmath68 , where @xmath69 with @xmath52 the value of @xmath0 at the positive tension brane .
additionally , @xmath70 is the @xmath71-th zero of the bessel function @xmath72 $ ] , that is @xmath73 = 0 $ ] . observe that eq .
( [ v - rs ] ) is recovered for @xmath74 .
we have shown the corrections to newton s law for susy braneworld models arising from the way in which gravitons propagate in a bulk with a geometry that differs from the more commonly studied ads@xmath5 .
a sensible question regarding this type of models is whether there are any chances of observing short distance modifications of general relativity in the near future . to explore this , notice that the relevant energy scale at which the corrections become significant is @xmath69 , instead of the more fundamental mass scale @xmath75
. typically one would expect @xmath76 which has to be above tev scales to agree with particle physics constraints .
nevertheless , the factor @xmath77 leaves open the possibility of bringing @xmath78 up to micron scales . in the case of 5-d heterotic m - theory ( @xmath63 ) one
has @xmath79 , where @xmath80 is the volume of the calabi - yau 3-fold in units of @xmath39 . in order to have an accessible scale @xmath81 m
, it would be required @xmath82 , where we assumed @xmath76 . on the other hand , newton s constant
is given by @xmath83 , which implies @xmath84 .
thus , to achieve @xmath82 one requires the following values for @xmath39 and @xmath80 @xmath85 which are in no conflict with present phenomenological constraints coming from high energy physics .
in particular , non - zero kaluza - klein modes coming from the compactified volume @xmath80 would have masses of order @xmath86gev . on the other hand ,
if @xmath39 is of the order of the grand unification scale @xmath87gev , then corrections to the newtonian potential would be present at the non - accessible scale @xmath88 m .
d. j. kapner , t. s. cook , e. g. adelberger , j. h. gundlach , b. r. heckel , c. d. hoyle and h. e. swanson , phys .
lett . * 98 * , 021101 ( 2007 ) .
e. g. adelberger , b. r. heckel , s. hoedl , c. d. hoyle , d. j. kapner and a. upadhye , phys .
* 98 * , 131104 ( 2007 ) .
g. a. palma , jhep * 0009 * , 091 ( 2007 ) .
e. bergshoeff , r. kallosh and a. van proeyen , jhep * 0010 * , 033 ( 2000 ) .
g. a. palma and a. c. davis , phys .
d * 70 * , 064021 ( 2004 ) ; phys .
d * 70 * , 106003 ( 2004 ) .
g. a. palma , phys .
d * 73 * , 044010 ( 2006 ) [ erratum - ibid .
d * 74 * , 089902 ( 2006 ) ] .
p. horava and e. witten , nucl .
b * 475 * , 94 ( 1996 ) .
a. lukas , b. a. ovrut , k. s. stelle and d. waldram , nucl .
b * 552 * , 246 ( 1999 ) . | in braneworld models coming from string theory one generally encounters massless scalar degrees of freedom moduli parameterizing the volume of small compact extra - dimensions . here
we discuss the effects of such moduli on newton s law for a fairly general 5-d supersymmetric braneworld scenario with a bulk scalar field @xmath0 .
we show that the newtonian potential describing the gravitational interaction between two bodies localized on the visible brane picks up a non - trivial contribution at short distances that depends on the shape of the superpotential @xmath1 of the theory .
in particular , we compute this contribution for dilatonic braneworld scenarios @xmath2 ( where @xmath3 is a constant ) and discuss the particular case of 5-d heterotic m - theory .
address = deutsches elektronen - synchrotron desy , theory group , notkestrasse 85 , d-22603 hamburg , germany |
anyons @xcite are two - dimensional particles whose wave function @xmath0 obeys the interchange conditions @xmath1 where @xmath2 denotes continuous anticlockwise interchange of particles @xmath3 and @xmath4 , such that no other particles are encircled , and @xmath5 is the statistics parameter , which may be any real number . for a given hamiltonian ,
not explicitly dependent on @xmath5 and with a discrete spectrum for any @xmath5 , the energy eigenvalues @xmath6 and ( usually ) the corresponding eigenfunctions @xmath7 will be continuous functions of @xmath5 .
a continuous function @xmath6 , for a fixed index @xmath8 and @xmath9 , is what we will refer to as an ( anyon ) trajectory . for an earlier discussion of the @xmath5 dependence of the states ,
see @xcite . since the exponent in ( [ int ] ) is the only place where @xmath5 appears in this formulation , the spectrum and the set of energy eigenstates are obviously periodic in @xmath5 with period 2 ; thus , all the information is contained in the sets @xmath10 and @xmath11 for @xmath12 .
however , individual trajectories are not generally periodic in @xmath5 ; one way to see this is to observe that when @xmath5 increases continuously , the kinetic angular momentum @xmath13 changes according to the formula @xmath14 with @xmath15 an integer rotation of the whole system multiplies the wave function by a phase factor , which equals , on the one hand , @xmath16 $ ] , and on the other hand , @xmath17 $ ] , since each of the @xmath18 pairs of anyons is interchanged twice . ] , so that changing @xmath5 by 2 will not bring us back to the same state .
it is assumed that the hamiltonian is rotationally invariant , so that @xmath13 is a good quantum number .
thus , @xmath19 , where @xmath20 , will be the energy of some state with statistics parameter @xmath5 , but the index of that state will be different from @xmath8 : @xmath21 where @xmath22 depends on @xmath8 and @xmath23 : @xmath24 . at any given @xmath5 , the trajectories @xmath8 and @xmath22 look different , having different angular momenta @xmath25 and @xmath26 , but when viewed on the interval @xmath27 , they are seen to be copies of one and the same trajectory , shifted by @xmath28 along the @xmath5 axis . to represent the trajectories in a way which is not redundant , it is convenient to take @xmath13 as a parameter instead of @xmath5 .
@xmath13 is directly related to @xmath5 by equation ( [ angm ] ) , it is an observable ( gauge invariant ) quantity , and the dependence @xmath29 is the same for all the trajectories that are shifted copies of one another .
we arrive at the viewpoint that the states of an @xmath30-anyon problem may be obtained from the set of trajectories , i.e. , functions @xmath29 for @xmath31 .
this provides a more economical and unified way of decribing the @xmath30-anyon spectra . for each trajectory
there is an integer value @xmath15 such that the statistics is bosonic at angular momenta @xmath32 .
we may always choose @xmath33 .
the value of @xmath15 groups the set of trajectories into @xmath34 classes .
only trajectories from the same class can cross , i.e. have the same @xmath35 and @xmath13 at the same statistics .
a part of a trajectory between any two neighboring bosonic points corresponds to what is usually referred to as one anyon state on the interval @xmath12 .
it is useful to have a geometric picture of this : if one plots a trajectory for @xmath36 and then wraps the plot around a cylinder of circumference @xmath34 , then points which correspond to @xmath5 differing by an even number ( that is , to the same statistics ) will fall on the same vertical line . on the surface of the cylinder one will see the set of all @xmath30-anyon states corresponding to the trajectory .
thus , an infinite number of pieces of trajectories ( or states in the usual terminology ) on the interval @xmath12 are shown to make up one single trajectory on the interval @xmath37 .
in other words , certain anyon ( and in particular , boson or fermion ) states that are usually considered entirely different , are in fact parts of one and the same continuous pattern .
this is reminiscent of the method of organizing the spectra of particle physics into regge trajectories .
in fact , by applying the concept of regge trajectories to a two - dimensional , two - particle system , one obtains the same grouping of bosonic / fermionic states .
it follows that if some trajectory @xmath29 is analytic , then it is in principle sufficient to find only one of its pieces .
the rest is uniquely determined by analytic continuation .
this means that it is sufficient to find one anyonic state in some range of @xmath5 and then an infinite number of others are obtained `` automatically '' by analytic continuation and periodicity .
however , trajectories are not always analytic . with our assumption that the hamiltonian does not explicitly depend on @xmath5 ( and
involves only non - singular interactions ) both energies and wave functions will depend analytically on @xmath5 , except possibly at bosonic points .
the source of non - analyticity is that the relative angular momentum of one pair of particles becomes zero , which can only happen at bosonic points .
when there is non - analytic behavior at some point with degeneracy , one must determine how each trajectory continues through this point by investigating the corresponding wave functions .
they should change continuously across these points .
we are going now to apply this reasoning to the problems of two and three anyons in a rotation symmetric harmonic potential .
we scale the variables so that @xmath38 , the mass @xmath39 and the angular frequency @xmath40 .
two more preliminary remarks are in order .
first , the potential is parity invariant ; hence , if @xmath0 is an eigenfunction of the hamiltonian and satisfies ( [ int ] ) , then its complex conjugate @xmath41 is an eigenfunction of the hamiltonian with the same energy and satisfies ( [ int ] ) with @xmath42 instead of @xmath5 . in particular
, this implies that all information about the states is in fact contained in the interval @xmath43 $ ] .
second , there is the tower structure of the spectrum @xcite .
it has been observed that , for any number of anyons , all the states come in towers , the angular momentum being the same for all members of a tower and the energies being @xmath44 , @xmath45 , @xmath46 , where @xmath44 is the energy of the lowest , `` bottom '' state .
this sequence of levels is due to radial excitations .
the radial coordinate @xmath47 is defined by @xmath48 , with @xmath49 the center - of - mass coordinates .
thus , it is sufficient to find only the bottom states ( consequently , bottom trajectories ) ; from now on , we will always mean these , unless otherwise specified .
now we will demonstrate that in the two - anyon problem there is only _
one _ ( bottom ) trajectory , namely the continuation of the ground state .
recall that the one - particle spectrum in the harmonic potential consists of states with @xmath50 and @xmath51 ( so the degeneracy equals the energy ) .
define complex coordinates by @xmath52 .
since the center - of - mass motion is trivial , we will always concentrate on the relative motion only . for two anyons , the complete set of ( not normalized ) solutions
may be written as @xmath53 with @xmath54 , @xmath55 an integer , @xmath4 a non - negative integer , and @xmath56 standing for @xmath57 if @xmath58 and for @xmath59
if @xmath60 ; the energy and the angular momentum are @xmath61 respectively .
towers consist of states with the same @xmath55 and different @xmath4 ; in accordance with the aforesaid , it is sufficient to consider the bottom states only , for which @xmath62 and @xmath63 _ irrespective _ of @xmath55 . indeed ,
all the states in question belong to one and the same trajectory , because the relation ( [ period ] ) does hold , in the form @xmath64 ( and the same for the wave function ) .
for @xmath65 this is the ground state , @xmath66 thus , _ all _ bottom states are continuations of _ one _ ground state .
equation ( [ el ] ) exemplify our previous remark that a trajectory may become non - analytic when the relative angular momentum of one pair of particles ( here @xmath13 itself ) becomes zero . by adding the pair potential @xmath67 to the hamiltonian , ( [ el ] )
is changed to @xmath68 , which demonstrates that singular interactions may lead to different behaviours .
we go now to the problem of three anyons , which is extremely interesting due to its nontriviality , on the one hand , and the possibility of a more or less exact analysis , on the other hand .
recall some results available .
a state starting with @xmath69 at @xmath70 ( @xmath4 integer ) may reach one of the following points at @xmath71 : @xmath72 , @xmath73 , @xmath74 , @xmath75 @xcite .
in other words , in any interval from a bosonic to a fermionic point or from a fermionic to a bosonic point , each state is characterized by an ( average ) slope @xmath76 where @xmath77 is the change in energy , and @xmath78 is the change in angular momentum .
for the @xmath79 states the dependence @xmath29 is linear , so that @xmath80 is the slope at any point in the interval , and their wave functions may be written down exactly @xcite .
for the @xmath81 states the dependence is nonlinear , so a slope of @xmath82 is indeed only an average slope .
now , there are exact expressions for the multiplicities of states with all slopes @xcite .
if @xmath83 denotes the number of bottom states in the relative motion spectrum that go from @xmath69 to @xmath84 as @xmath5 goes from @xmath85 to @xmath86 , i.e. , from a bosonic to a fermionic point , then there are the asymptotic expressions for @xmath87 ( terms of order unity being omitted ) @xmath88 and the exact formulas @xmath89 , @xmath90 .
in each of the formulas , @xmath13 has to be within the respective interval specified and the equality @xmath91 ( mod 2 ) has to hold , otherwise @xmath92 .
recall also that for all states at any statistics , there is the inequality @xmath93 @xcite .
as one goes along a trajectory , the slope will change at certain values of @xmath13 , so any trajectory possesses a sequence of slopes and points of slope change .
we make now the following statements .
* for each trajectory there exists an @xmath94 such that @xmath95 if and only if @xmath96 . * for each trajectory there exists an @xmath97 such that @xmath98 if and only if @xmath99 .
* the change of slope from / to @xmath100 can occur at bosonic points only , while the change from @xmath101 to @xmath102 or vice versa can occur at both bosonic and fermionic points . to prove statement ( i ) , note first that the inequality @xmath103 and the fact that @xmath104 force every trajectory to have @xmath95 at least for some large positive values of @xmath13 .
now , the explicit form of the @xmath95 states @xcite is such that if one of them exists for some @xmath105 , it does exist with @xmath95 for any @xmath106 , that is for any @xmath107 ; in other words , if @xmath95 at some point , then @xmath95 everywhere to the right of that point .
this concludes the proof .
a `` mirror reversed '' reasoning proves statement ( ii ) .
statement ( iii ) is proved upon noticing that the change of slope from / to @xmath100 always implies nonanalyticity in the function @xmath29 , because it is a change between a linear and a nonlinear dependence .
such nonanalyticity can only happen at bosonic points , as discussed in sec . 1 .
this is also associated with the breakdown of regular perturbation theory dependent aharonov - bohm - type ( perturbative ) interaction .
] at such points @xcite . note that since statements ( i ) and ( ii ) mean that each trajectory has @xmath98 for some values of @xmath13 and @xmath95 , for some others , it follows that each trajectory is indeed nonanalytic at least at one point .
( this is of course equally true for the @xmath30-anyon problem , where the extreme slopes , with linear behavior , are @xmath108 , and each trajectory possesses both of them just like in the case at hand ; cf .
@xcite . )
so far we may conclude that the behavior of each trajectory has the following features : at large negative @xmath13 , slope @xmath109 , at some point @xmath110 a change to @xmath111 , for @xmath112 some sequence of slopes @xmath111 and finally a change to @xmath113 at @xmath114 and always @xmath113 at @xmath115 .
note that there is exactly one trajectory for which @xmath116 , that is which has slopes @xmath113 and @xmath109 only : the one which contains the ground state near bose statistics , @xmath117 where @xmath118 , and the overall gaussian factor is understood .
we will now prove that * each bottom trajectory may be unambiguously labeled by its @xmath114 point .
indeed , given an arbitrary bosonic point @xmath69 , it is easy to see that the number of bottom trajectories for which @xmath119 , equals @xmath120 where the inequality follows from eq .
( [ b+3 ] ) .
the equality is reached , meaning that a trajectory with @xmath119 exists , for each @xmath69 such that @xmath121 and @xmath122 ( mod 2 ) . like the equation itself ,
this statement is true up to terms of the order of unity .
1 shows the exact picture of the distribution of the @xmath114 points on the plane ( the bullets and triangles will be explained later ) .
we proceed to show how all the linear parts of the trajectories ( and thus all the three - anyon linear states ) are constructed as excitations of the `` ground trajectory '' , the one which contains the ground state . by virtue of symmetry , it is enough to consider @xmath95 states only .
it is convenient to use the coordinates proposed in @xcite , which are the discrete fourier transform of the complex particle coordinates .
the relative motion is described by @xmath123 where @xmath124 .
introduce the creation and annihilation operators @xmath125 and , with @xmath126 , the same relations for @xmath127 , @xmath128 , @xmath129 , @xmath130 .
then the relative hamiltonian and angular momentum operator become @xmath131 the commutation relations are @xmath132= [ b_u , b_u^{\dagger}]=[a_v , a_v^{\dagger}]=[b_v , b_v^{\dagger}]=1 $ ] , and all other commutators vanish . consequently
, if @xmath0 is a common eigenstate of @xmath133 and @xmath13 with quantum numbers @xmath69 , then @xmath134 is also a common eigenstate with @xmath135 .
however , to yield true anyonic eigenstates , a combination of the creation operators must be fully symmetric and produce wave functions that are not singular at the points where the positions of two particles coincide . the pairs @xmath136 and @xmath137 define two equivalent irreducible representations of the permutation group @xmath138 : for example , the pair @xmath136 transforms , under the six possible permutations , into @xmath136 , @xmath139 , @xmath140 , @xmath141 , @xmath142 , @xmath143 , respectively .
therefore fully symmetric quantities are those of the form @xmath144 with @xmath145 .
the pairs @xmath146 , @xmath147 , and @xmath137 all transform like @xmath136 . taking successive products of the above - mentioned representations of @xmath138 and decomposing them into irreducible representations ,
it is straightforward to prove that all symmetric polynomials in @xmath148 can be expressed as polynomials in the following basic symmetric polynomials : @xmath149 the meaning of the subscripts is that @xmath150=p\,c^{\dagger}_{pq } , \qquad [ l , c^{\dagger}_{pq}]=q\,c^{\dagger}_{pq},\ ] ] so @xmath151 changes the energy by @xmath152 and the angular momentum by @xmath153 .
it remains to check for possible singularities at @xmath154 .
all the @xmath113 states have the form , up to the gaussian factor @xmath155 , @xmath156^{\nu } = p(u,\bar{u},v,\bar{v})(u^3-v^3)^{\nu}\ ] ] with @xmath157 a polynomial and @xmath158
. therefore the only source of singularity can be differentiation of the last factor .
this immediately implies that @xmath159 , which involves @xmath160 only , is always regular , and so is @xmath128 ; hence , @xmath161 and @xmath162 are regular .
now , @xmath163 shows that @xmath164 by itself may be singular , due to the last term .
nevertheless , @xmath165 is regular .
further , @xmath166 but @xmath167 , @xmath168 and @xmath169 are singular . a generic ( not normalized ) linear state of slope @xmath95 may then be written as @xmath170 with @xmath171 ( for the @xmath109 states , of course
, @xmath172 would take the place of @xmath151 . )
this is a classification of _ all _ linear three - anyon states ( cf .
@xcite ) . however , some of these states are tower excitations and some are different parts of the same trajectories .
first , the operator @xmath173 is nothing but the tower raising operator .
it never produces singularities when acting on any state , linear or nonlinear @xcite , thus providing the tower structure of the whole spectrum .
hence put @xmath174 .
second , note that @xmath175 and consequently @xmath176 \psi_{00mn}(u , v;\nu ) \\ & \!\!\ ! = \!\!\ !
& 8(u^3-v^3)^2 \psi_{00mn}(u , v;\nu ) \\ & \!\!\ ! = \!\!\ !
& 8\psi_{00mn}(u , v;\nu+2);\end{aligned}\ ] ] in other words , the wave functions ( [ 39 ] ) , being linearly independent for @xmath177 , are no longer so for @xmath178 , therefore some of them do not lead to new trajectories and should be excluded from the trajectory counting . as the last formula shows , it is sufficient to restrict to @xmath179 . thus , a bottom trajectory may be labeled by three numbers @xmath180 and @xmath181 .
this is completely equivalent to the labeling by @xmath114 , because it follows from ( [ 40])([41 ] ) that @xmath182 since the point of slope change to @xmath113 here is @xmath183 , and it is straightforward to see that for any point @xmath114 there is no more than one set @xmath184 such that these two equations are satisfied .
the set of bottom trajectories being _ two_-parametric ( plus a `` double degeneracy '' due to @xmath4 ) is due to the fact that a bottom state is identified by _ three _ quantum numbers ( six degrees of freedom , minus two for the center of mass and one for the tower excitations ) .
two quantum numbers ( say , @xmath185 and @xmath94 ) identify a trajectory , and the third one ( @xmath13 ) chooses a point on that trajectory . to compare , in the two - anyon problem , only _
one _ quantum number , @xmath13 , is enough to identify a bottom state , consequently there is only one trajectory . in general , for the @xmath30-anyon problem , the family of trajectories will be @xmath186-parametric .
let us return once more to fig . 1 ; there are two copies of one and the same two - dimensional pattern , made of bullets ( @xmath62 ) and triangles ( @xmath187 ) , and increasing @xmath3 or @xmath55 by 1 means moving within the pattern ( 3 units to the right and 3 units up , or 1 unit to the right and 3 units up , respectively ) .
thus the counting of trajectories is complete , but to find the wave functions , certain modifications of eq .
( [ 39 ] ) are still necessary .
first , we need only such functions for which @xmath188 , where @xmath189 is the tower lowering operator , but since @xmath190 does not commute with the @xmath191 s , even for @xmath174 some of the functions ( [ 39 ] ) will contain an admixture of non - bottom states with the same @xmath35 and @xmath13 . to correct for this ,
it turns out to be sufficient to replace @xmath192 in ( [ 39 ] ) by another operator , writing @xmath193 where @xmath194^l \label{43}\ ] ] and @xmath195 is sen s supersymmetry operator @xcite .
clearly , @xmath196 is symmetric , and it is straightforward to check that @xmath197 always [ @xmath190 commutes with @xmath198 and gives zero when acting on @xmath199 .
also , @xmath200 changes @xmath35 and @xmath13 by the same amount as @xmath192 , so that the formulas ( [ 45 ] ) and ( [ 46 ] ) remain valid .
second , a wave function that is continuous at the point @xmath114and thus corresponds to a trajectory is , in general , not one of the form ( [ 42 ] ) but a linear combination of several functions of that form .
indeed , for any fixed @xmath201 and @xmath202 we have @xmath203 degenerate @xmath95 states @xmath204 , where @xmath205 $ ] .
these are linearly independent , but not orthogonal ; and which of their linear combinations to choose as a basis in the relevant @xmath203-dimensional subspace , makes no difference anywhere except at the point @xmath183 . here , the states with @xmath206 continue as linear to @xmath207 , and the basis in the relevant @xmath208-dimensional subspace may still be chosen at will , but one state `` decouples '' , i.e. , becomes nonlinear , and that is the one orthogonal to this subspace .
in general , it will be not @xmath209 but a sum @xmath210 , where the numbers @xmath211 , @xmath206 , are determined by the @xmath208 orthogonality conditions . consider the simplest example @xmath212 , @xmath213 , @xmath62 , which corresponds to the bosonic point @xmath214 , see fig .
the two relevant states of the form ( [ 42 ] ) are @xmath215 and @xmath216 . at @xmath217
, it makes no difference which two linear combinations of them to choose ; but at @xmath183 , @xmath218 continues to be linear , while the state which actually `` decouples '' and continues to @xmath207 as nonlinear , is the one orthogonal to @xmath218 at @xmath183 .
an elementary calculation shows that it is @xmath219 .
such reasoning has to be repeated at each bosonic point with @xmath220 , and general expressions for the coefficients @xmath221 apparently do not exist .
we have shown that each trajectory is uniquely identified by its @xmath114 point ; the point being given , one should in principle be able to reconstruct the behavior of the whole trajectory , and in particular to find all the bosonic and fermionic points it passes through .
in fact , it would be sufficient to find @xmath110 and the behavior between that point and @xmath114 [ where @xmath29 is nonlinear ] . to see whether this can be achieved , we plot the low - lying trajectories , using the exact multiplicities given in @xcite , on fig .
( there are six different values of @xmath15 and consequently six plots , one for each class of trajectories . )
the nonlinear pieces are shown schematically only .
the continuity of the trajectories is confirmed by the fact that at bosonic and fermionic points , where they cross , the total number of trajectories coming from the left and from the right is always the same .
the main question is how to identify which piece on the left is a continuation of which piece on the right .
if at a certain crossing point perturbation theory works and the first - order corrections to energy are different for all the states involved , then the rule is that the order of trajectories , by increasing energy , on the right is opposite to that on the left ( because the first - order corrections have opposite signs ) .
this argument does not work ( a ) if the first - order corrections to some of the states are equal and ( b ) when perturbation theory breaks down , at certain bosonic points , as mentioned above . a way out then is to find the wave functions numerically , using the method of @xcite and to identify them by comparing their limits as the crossing point is approached . as it appears , the above rule
the trajectory which is the @xmath4-th from above on the left is the @xmath4-th from below on the right
holds for all crossings analyzed . in other words ,
all crossings appear to be _ true _ crossings .
when this rule is used , it becomes possible to identify the trajectories completely without actually finding the wave functions .
it turns out that for all the trajectories there is at most one point of slope change , apart from @xmath114 and @xmath110 , which we denote by @xmath222 , the change being from @xmath101 to @xmath102 as @xmath13 increases through @xmath223 ; clearly , @xmath224 .
a generic behavior of a trajectory is , therefore , @xmath225 ( numbers on top of the arrows meaning slopes ) , but in fact it may be @xmath226 or @xmath227 ( for the ground state , @xmath228 ) .
table 1 shows the points of slope change for all the trajectories with @xmath229 .
@xmath230 & @xmath231 & @xmath232 & @xmath233 & & @xmath234 & @xmath235 & @xmath230 & @xmath231 & @xmath232 & @xmath233 + 2 & 0 & 2&0 & 2&0 & & 9 & @xmath2365 & 7 & 1 & 9 & 7 + 6 & @xmath2364 & 4&2 & 4&2 & & 10&@xmath2368 & 8 & @xmath2362 & 10&4 + 7 & @xmath2365 & 5&1 & 5&1 & & 10&@xmath2366 & 8 & 0 & 10&6 + 5 & @xmath2363 & 4&0 & 5&3 & & 10&@xmath2364 & 8 & 2 & 10&8 + 4 & @xmath2362 & 4&@xmath2362 & 6&4 & & 11&@xmath2363 & 10&0 & 11&3 + 11&@xmath2369 & 7&3 & 7&3 & & 13&@xmath2367 & 10&2 & 11&5 + 5 & @xmath2361 & 5&@xmath2361 & 7&5 & & 13&@xmath23611 & 9 & 1 & 11&7 + 8 & @xmath2364 & 7&@xmath2361 & 8&2 & & 7 & @xmath2363 & 7 & @xmath2363 & 11&9 + 8 & @xmath2362 & 7&1 & 8&4 & & 14&@xmath23612 & 10&0 & 12&6 + 8 & @xmath2366 & 6&0 & 8&6 & & 14&@xmath23610 & 10&2 & 12&8 + 9 & @xmath2367 & 7&@xmath2361 & 9&5 & & 14&@xmath2368 & 10&4 & 12&10 + table 1 .
the behavior of the low - lying trajectories .
the regularity is not obvious here , but it becomes visible when one looks at sufficiently high - lying trajectories .
introduce the quantities @xmath237 that is , the `` numbers of revolutions '' , or the numbers of bosonic points that a trajectory passes through having slopes @xmath102 and @xmath101 , respectively ; these may be integer or half - integer , as @xmath223 may correspond to either a bosonic or a fermionic point .
the following obvious relations hold @xmath238 the results , deduced from the numerical analysis , are the following : @xmath239 the formulas , as well as the inequality signs , are valid asymptotically for @xmath240 , in the same sense as ( [ b+3])([b+1 ] ) .
one can verify the compatibility of ( [ n+])([n- ] ) with ( [ b+3])([b+1 ] ) ; also , a simple consistency check comes from the symmetry requirement a trajectory @xmath241 must have its partner @xmath242 : this is satisfied as well .
note that @xmath243 depends continuously on @xmath94 , but there is a discontinuity in @xmath244 at the point @xmath245 .
also , the exact results exhibit certain periodic structure with period 8 by @xmath185 and 4 by @xmath94 . explaining these features remains an open item . a semiclassical interpretation rather rough , although of some features of the trajectories may be adduced .
when two particles are close together and the third one is far from them , that is , say , @xmath246 [ @xmath247 , one gets two independent anyonic oscillators , one ( @xmath248 ) with the statistics parameter @xmath5 , the other ( @xmath249 ) with the parameter @xmath250 @xcite .
the energy in this approximation is @xmath251 , where in general @xmath252 ; as @xmath5 increases through @xmath253 , the slope changes from @xmath109 to @xmath101 ( if @xmath254 ) or from @xmath102 to @xmath113 ( if @xmath255 ) ; the other two points of change , however , can not be described correctly , nor can the formulas ( [ n+])([n- ] ) be explained .
perhaps a more accurate semiclassical approximation would be able to explain them .
we have developed the concept of trajectories for anyons , noting that many different many - anyon states are in fact continuations of each other .
we discuss in particular the harmonic oscillator external potential . apart from the center - of - mass and tower excitations , in the two - anyon problem
there is only one trajectory .
for the three - anyon problem , we have worked out the classification and the main features of the trajectories : the slope of a trajectory always changes as @xmath256 ( where the @xmath102 and/or @xmath101 pieces may be missing ) , and the point @xmath114 of the last change may be used to label the trajectory .
the wave function corresponding to the linear part of the trajectory can be written down exactly by applying excitation operators to the ground trajectory .
concerning the nonlinear part , we conjecture the formulas , based on a numerical analysis , expressing the lengths of intervals of slope @xmath101 and @xmath102 in terms of @xmath114 .
qualitatively , the picture will be the same for the @xmath30-anyon problem : again , each trajectory will have the points @xmath114 and @xmath110 such that its @xmath29 dependence is linear with slopes @xmath108 for @xmath96 ( @xmath99 ) and nonlinear in between . here
the family of trajectories will be @xmath186-parametric .
it is quite plausible that in this case as well , the sequence of slopes will be regular , from @xmath257 up to @xmath258 in steps of two .
however , to gain a more precise understanding of the behavior of the trajectories remains an open problem .
s. m. would like to express his sincere thanks to the institutt for fysikk at the nth , trondheim , and the theory division of the institut de physique nuclaire , orsay , where parts of this work were done , for the kind hospitality and support .
fig . 1 . the distribution in the @xmath69 plane of the points @xmath114 at which the trajectories turn linear with slope @xmath95 .
the bullets define the set @xmath259 , @xmath260 , and the triangles define the set @xmath261 , @xmath262 , with @xmath263 and @xmath264 in both cases .
fig . 2 , a
the six classes of trajectories in the @xmath69 plane , with @xmath265 ; solid and dashed vertical lines mean bose and fermi statistics , respectively , dashes show the multiplicities of the linear states . | we develop the concept of trajectories in anyon spectra , i.e. , the continuous dependence of energy levels on the kinetic angular momentum .
it provides a more economical and unified description , since each trajectory contains an infinite number of points corresponding to the same statistics . for a system of non - interacting anyons in a harmonic potential ,
each trajectory consists of two infinite straight line segments , in general connected by a nonlinear piece .
we give the systematics of the three - anyon trajectories .
the trajectories in general cross each other at the bosonic / fermionic points .
we use the ( semi - empirical ) rule that all such crossings are true crossings , i.e. the order of the trajectories with respect to energy is opposite to the left and to the right of a crossing .
hep - th/9507034 + july 1995 * anyon trajectories * * and the systematics * * of the three - anyon spectrum * stefan mashkevich institute for theoretical physics 252143 kiev , ukraine jan myrheim , kre olaussen , ronald rietman institutt for fysikk , nth , universitetet i trondheim n7034 trondheim , norway |
consider a favorable gambling game , such as betting at 3:1 odds on the outcome of a fair coin toss . if we are allowed to play this as many times as we like ( decided adaptively ) , we can eventually increase our winnings to any desired target amount , with certainty .
for instance , proportional betting strategies such as the kelly criterion ( see kelly @xcite or breiman @xcite ) , have long been known to accomplish this . breiman @xcite `` hopefully conjectured '' the following .
suppose our goal is to achieve a set target bankroll , say $ 1 m , starting with a fraction @xmath0 of that amount .
let @xmath1 be the expected number of rounds we have to play before we attain our goal .
then there exists a threshold @xmath2 , and an optimal strategy of the following form : * when the current bankroll is less than @xmath3 , bet to optimize @xmath4bankroll@xmath5 .
this is achieved by betting a particular fraction of the current bankroll , which is only dependent on the proposition being offered .
this is sometimes known as `` kelly betting . ''
* when the current bankroll is at least @xmath3 , bet to reach the target bankroll in the current round
. we will call such strategies `` threshold strategies . ''
breiman describes this conjecture as `` expressing a moderate faith in the simplicity of things . ''
indeed , his proposed strategy seems quite plausible .
however , our main result is a proof that this strategy is not optimal .
[ thm : main ] there exists a favorable gambling game and initial bankroll @xmath0 , for which the optimal @xmath1 is at most @xmath6 times that of any threshold strategy .
we have not attempted here to optimize the constant @xmath6 , but it would be nice to know what its best possible value is .
we will use the following version of doob s optional stopping theorem ( see ( * ? ? ? * theorem 10.10 ) ) .
[ thm : opt - stop ] let @xmath7 be a supermartingale .
let @xmath8 be a stopping time with @xmath9 , and suppose there is a constant @xmath10 such that , for all @xmath11 , @xmath12
. then @xmath13 is integrable and @xmath14 .
for our gambling game , consider a biased coin which comes up heads with probability @xmath15 .
suppose we are allowed to bet on heads , and be paid off at @xmath16 odds .
that is , for each unit bet , the net change in our bankroll is @xmath17 with probability @xmath18 , and @xmath19 with probability @xmath15 . in this case , it is easily checked that the kelly criterion says to bet @xmath20 of the current bankroll at each timestep , so that the new bankroll will be : half the current bankroll with probability @xmath18 , and twice the current bankroll with probability @xmath15 . for an initial bankroll , @xmath21 , let @xmath22 be the expected number of plays of this game until a bankroll of at least @xmath23 is achieved , under the strategy minimizing this quantity [ lem:3k ] for our example game , when @xmath24 , for @xmath25 a positive integer
, the ( unique ) optimal strategy is to bet @xmath20 of the current bankroll every round until a bankroll of @xmath23 is reached .
therefore , @xmath26 .
first we observe that the given strategy results in an expectation of @xmath27 rounds until the bankroll reaches @xmath23 .
this is because the base-2 logarithm of the bankroll is a biased random walk on the negative integers , that moves one unit towards @xmath28 with probability @xmath15 at each timestep , and one unit away with probability @xmath18 .
the analysis of the hitting time to @xmath28 for this random walk is standard . to see that this is the best possible result , consider any possible strategy , and let @xmath29 denote the bankroll after @xmath30 steps
. let @xmath31 denote the hitting time to the target bankroll : @xmath32 .
define @xmath33 .
observe that , regardless of the strategy chosen , @xmath34 is a supermartingale , since kelly betting maximizes the conditional expectation of @xmath35 given @xmath36 , and , for kelly betting , @xmath36 would be a martingale .
further , note that @xmath37 , and @xmath38 .
now , we may assume @xmath39 , since otherwise this strategy is clearly worse than kelly betting .
suppose for now that @xmath40 is bounded almost surely . then applying the doob s optional stopping theorem [ thm : opt - stop ]
, we have @xmath41 , which proves that , again , the kelly strategy is superior .
finally , we will show that any strategy with @xmath40 unbounded can be strictly improved upon by another strategy with @xmath42 , and therefore our previous assumption that the steps taken by @xmath43 are bounded was made without loss of generality . first note that @xmath44 absolutely , since even if we bet the entire bankroll , we can not more than triple our stake .
suppose , for some value of @xmath29 , this strategy bets more than @xmath45 of the bankroll .
let s look at the expected number of rounds needed until @xmath46 .
with probability @xmath15 , @xmath47 , but with probability @xmath18 , the bankroll initially drops by a factor of @xmath48 , and hence , no matter what , it will take more than @xmath49 rounds to return to its initial value of @xmath29 .
thus , in expectation , it takes more than @xmath50 rounds to exceed @xmath29 for the first time . and ,
moreover , the value of @xmath51 is in the interval @xmath52 .
now , note that kelly betting quadruples the stake in an expected @xmath50 rounds , which clearly dominates the above strategy .
so , it would be strictly superior to `` bet to double '' until the stake reaches @xmath53 , and then proceed optimally from that point onward .
thus , any strategy that ever bets more than @xmath45 of its bankroll can be improved upon by one that does not , which gives us bounds of @xmath54 on the revised strategy , which completes the proof .
lemma [ lem:3k ] is interesting because it shows that kelly betting is in fact optimal for infinitely many starting bankrolls .
furthermore , when coupled with our next two ( easy ) lemmas , this actually implies that there exist threshold strategies for which @xmath1 is within a constant factor of optimal .
these results are not specific to the example game chosen for this paper ; precise statements and proofs are left as an exercise to the reader . on the flip side ,
as we will see , lemma [ lem:3k ] is also the key to our proof that every threshold strategy is actually suboptimal .
[ lem : decr ] [ lem:3/2 lb ] in general , @xmath22 is a decreasing function of @xmath21 .
furthermore , for our example game , @xmath55 for all @xmath56 . clearly @xmath22 is non - increasing , since extra money can always be ignored with no penalty .
we omit a detailed proof that @xmath8 is strictly decreasing , noting only if there is extra money being ignored , then after a sufficiently long sequence of consecutive losses under , say , a proportional betting scheme , this extra money will become the vast majority of the bankroll , at which point we can appeal to lemma [ lem:3k ] to see that the expected hitting time to @xmath23 is strictly better than without the extra money . to see that @xmath55 , note that we can not achieve a bankroll of @xmath23 without winning at least one coin toss .
but the expected number of coin tosses until the first heads is @xmath57 .
so this is clearly a lower bound on the hitting time to @xmath23 .
the inequality is strict , since if we keep losing , eventually our bets must become too small to guarantee reaching @xmath23 on the first heads .
[ lem : bet - to-1 ] for our example game , any optimal strategy always bets to @xmath23 when the bankroll is @xmath58 .
additionally , @xmath59 and @xmath60 .
suppose for a bankroll @xmath61 , and our strategy bets to some value @xmath62 .
then , since the first coin flip either results in a bankroll of @xmath63 , or a bankroll @xmath64 , we have @xmath65 this implies @xmath66 .
but , by lemma [ lem:3k ] , we know that @xmath67 , so we have a contradiction to the fact that @xmath8 is a decreasing function ( lemma [ lem : decr ] ) .
thus the correct strategy must be to bet to @xmath23 . since , from @xmath68 or @xmath69 , the strategy of betting to @xmath23 either results in winning directly , or reaching a bankroll of @xmath20 , and since lemma [ lem:3k ] tells us that @xmath67 , an easy calculation yields the values of @xmath70 and @xmath71 , for which we shall have a use later .
[ lem:1/3 - 1/2 ] for our example game , if breiman s conjecture were true , then the critical threshold @xmath3 would necessarily be in @xmath72 $ ] .
since winning bets are paid off at @xmath16 odds , it is impossible to `` bet to @xmath23 '' with a bankroll of less than @xmath18 . moreover ,
unless the bankroll is strictly greater than @xmath18 , we can not bet to @xmath23 without risking the entire bankroll , in which case @xmath73 .
so @xmath74 .
on the other hand , if @xmath75 , then betting the kelly criterion ( risking half the bankroll ) is too much , as winning results in a bankroll exceeding @xmath23 , which has no added utility .
so @xmath76 .
( lemma [ lem : bet - to-1 ] also implies @xmath76 . )
[ lem:7/18 ] for any @xmath77 $ ] , playing breiman s strategy results in @xmath78 .
we consider two cases . *
case a : * @xmath79 . in this case , we first bet to @xmath23 , winning the game with probability @xmath15 , and otherwise losing @xmath80 , for a new bankroll of @xmath81 . assuming we lost , we will now risk half the bankroll at each step , until our bankroll has doubled up to a value of @xmath15 .
( note that , by lemma [ lem:1/3 - 1/2 ] , @xmath74 . ) note that , by lemma [ lem:3k ] , doubling up thrice takes @xmath82 rounds in expectation .
since lemma [ lem : bet - to-1 ] tells us @xmath59 , we can now compute @xmath83 * case b : * @xmath84 . in this case
, we first bet to double up to @xmath85 , which , by lemma [ lem:3k ] takes 3 rounds in expectation .
since lemma [ lem : bet - to-1 ] tells us @xmath86 , we therefore have @xmath87 just as in case a , completing the proof .
combining lemmas [ lem:1/3 - 1/2 ] and [ lem:7/18 ] shows us that , if breiman s conjecture were true , then @xmath78 for our example game .
our next result contradicts this . for our example game , @xmath88 .
on the first bet , suppose we bet @xmath89 .
if we win , our bankroll goes up to @xmath15 .
if we lose , our bankroll goes down to @xmath90 .
since lemmas [ lem:3k ] and [ lem : bet - to-1 ] tell us that with optimal play , @xmath91 and @xmath59 , we therefore have @xmath92 combining the above lemmas shows us that breiman s conjecture is false , and with a tiny bit more work , that any strategy meeting breiman s form has an expected cost at least @xmath93 of optimal starting from the initial bankroll of @xmath94 , thus proving theorem [ thm : main ] it would be nice to know the `` price '' for playing a threshold strategy .
we have seen that this is at least @xmath93 for our example game .
although we have made no serious attempt to improve this value , it seems likely that it is not far off the mark .
from an initial bankroll of @xmath21 , in our example game . , width=642 ] figure [ fig : bcopt ] shows a graph of upper and lower bounds on the optimal @xmath22 .
it was obtained by a computer program , which , beginning with the upper and lower bounds that follow immediately from lemma [ lem:3k ] and [ lem : decr ] , recursively derives better and better upper and lower bounds in terms of the previous ones .
all of the bounds the program works with are step functions , which allow for exact calculations , up to the precision of the machine arithmetic . for efficiency reasons ,
the program frequently approximates these step functions with more conservative ones , so as to avoid an exponential growth in the number of distinct function values it has to keep track of .
note that the upper bound is very close to all the values we know exactly from our lemmas , suggesting that the lower bound remains too conservative .
also note that the `` bump '' visible in the graph at @xmath95 is not an artifact . for comparison , @xmath96 .
we have no good explanation for this feature of the graph , but it seems to suggest that optimal betting has some interesting structure which remains to be understood .
i would like to thank yuval peres and evangelos georgiadis for introducing me to this problem and for helpful comments . | consider a gambling game in which we are allowed to repeatedly bet a portion of our bankroll at favorable odds .
we investigate the question of how to minimize the expected number of rounds needed to increase our bankroll to a given target amount . specifically , we disprove a 50-year old conjecture of l. breiman @xcite , that there exists a _ threshold strategy _ that optimizes the expected number of rounds ; that is , a strategy that always bets to try to win in one round whenever the bankroll is at least a certain threshold , and that makes kelly bets ( a simple proportional betting scheme ) whenever the bankroll is below the threshold . optimal betting , kelly betting , algorithms , counterexample , optional stopping , computer assistance |
it is generally acknowledged that the time - dependent hartree - fock ( tdhf ) theory provides a useful foundation for a fully microscopic many - body theory of low - energy heavy - ion reactions @xcite .
the success of the tdhf method is predicated on the expectation that the pauli principle plays an important role in simultaneously building up a time - dependent mean - field and suppressing the propagation of the strong @xmath0 interaction terms . in recent years
there has been a significant progress in the study of heavy - ion collisions using the tdhf theory @xcite .
this progress is partially due to the available computational power which allows calculations without resorting to simplifying but somewhat unphysical assumptions used in the past , as well as to the considerable improvements made in the parametrization of effective interactions @xcite .
one of the issues faced in the reduction of the finite many - body problem to be described in terms of single - particle degrees of freedom and a mean - field is the center - of - mass ( c.m . )
this is due to the fact that the resulting wavefunction , say a slater determinant , does not factorize into a product of one single center - of - mass wave - function and a wavefunction of the internal coordinates @xcite .
thus the calculated energies include a contribution from the center - of - mass motion in a complicated way .
another way of describing the problem is that the coordinates of the nucleons in the intrinsic wavefunction , @xmath1 , should be defined with respect to a fixed origin , which means the coordinates are not independent and satisfy @xmath2 .
the second problem faced in tdhf has to do with exit channel properties for collisions where more than one final fragment exists , e.g. deep - inelastic collisions . in this case , although we may have two well separated fragments , the final slater determinant can not be transformed to a block diagonal form thus making the identification of asymptotic channels unclear .
this is known as the _ cross - channel coupling _ problem . in the sections below
we shall discuss the ramifications of these issues to the interpretation of the results obtained from tdhf calculations .
in particular we shall discuss each problem in a more detailed fashion and show the magnitude of the effect and try to establish the conditions under which these issues may have a minimal effect on the calculated observables .
for many - body wavefunctions obtained using variational calculations there is no well defined prescription for eliminating the center - of - mass coordinate .
most of the suggested methods can be described by the general expression given by lipkin @xcite @xmath3 where the function @xmath4 is the general wavefunction depending on the coordinates of @xmath5 nucleons , the function @xmath6 is the weight function for integrating over the center - of - mass coordinate @xmath7 , and the result is the wavefunction @xmath8 depending on the @xmath9 internal coordinates .
an equivalent expression in momentum space is given in ref .
one common derivative of the above is to project out the @xmath10 part of the many - body wavefunction .
the proper way of choosing the weight function , @xmath6 , is such that the expectation value of the internal hamiltonian ( i.e. the hamiltonian after the removal of the center - of - mass dependence ) for the wavefunction ( [ eq : wave ] ) is a minimum . a proof for the above statement is given in section 4 of ref .
@xcite . in practice
this turns out to be a very difficult task .
alternatively , one can specify the functional form for @xmath6 but include variational parameters , similar to the lagrange parameters for constraints , in the definition .
the variational minimization then would also include these parameters thus giving the best @xmath6 for the chosen functional form @xcite . while these methods may be useful for the calculation of static properties , a generalization to dynamical calculations poses serious technical problems and therefore are not used in practice . as discussed below
, tdhf initial wavefunctions actually need the center - of - mass wavefunction to allow for boosts .
if one is primarily interested in energies an alternate procedure is the subtraction of the center - of - mass energy from the total kinetic energy of the system .
if this is done during the minimization procedure one may effectively achieve the above task .
formally , this can be written as : @xmath11\sum_{i=1}^{a}\mathbf{p}_i^2-\frac{1}{2ma}\sum_{i \neq j}\mathbf{p}_i\cdot\mathbf{p}_j\;.\end{aligned}\ ] ] the last part of this expression is numerically difficult as it includes double integrals and is usually neglected .
consequently , most skyrme forces use the simple correction to the kinetic energy given in the first term .
there are various skyrme fits that include the quadratic term as well , such as the sly6 parametrization @xcite .
the general statements made about this correction is that it is not very sensitive to deformation but it can be substantial in magnitude .
a comprehensive study of the influence of the center - of - mass correction in mean - field studies is given in ref .
this problem has been recently raised in ref .
@xcite in the context of calculating fission barriers , where a corrective procedure was suggested . in tdhf ,
the static hartree - fock slater determinant is boosted by multiplying each slater determinant with an exponential phase factor @xmath12 where @xmath13 is the hf wavefunction for nucleus @xmath14 , @xmath15 is the corresponding center of mass coordinate @xmath16 and @xmath17 is the boost momentum chosen to be tangent to the coulomb trajectory at an initial separation distance .
the target and projectile are placed to be in the center - of - mass frame for the entire system and satisfy @xmath18 thus , the total momentum for the system has expectation value zero , and the relative momentum is correctly given by @xmath19 . here , the assumption is that the physics of the collision is determined by the _ relative momenta_. this is one of the reasons why we expect the contributions coming from the wave packets contained in the initial slater determinants to a large extent cancel out . however , as we discuss below , there are other reasons as to why projecting out the center - of - mass part of the wavefunctions is not appropriate in tdhf calculations . the ratio @xmath20 plotted as a function of mass number @xmath5 , for two collision times . ] in tdhf the initial boosted states are interpreted as wave packets because the wave packet is already contained in the static slater determinants @xcite .
the result of multiplication by the phase factor given above is to shift the center of momentum of the wave packet by the vector * k*. however , there are fundamental differences between a boosted tdhf solution and a schrdinger wave packet .
since we are using a single slater determinant we are not free to change the wave packet nature of the solution .
the two are coupled via the minimization process and this goes beyond just the contribution / correction to static energies .
this coupling results in a _ non - spreading _ wave packet for the tdhf evolution @xcite . it can be shown that the tdhf wave packet is identical to the schrdinger wave packet when the terms causing the spreading in the schrdinger wave packet are neglected @xcite .
the condition that the wave packet behave completely classically over the time scale of the collision requires @xcite ( i.e. the spreading of the wave packet never catches up with its kinematical motion ) @xmath21^{-1}\ ; , \label{eq : wp}\ ] ] where @xmath22 is the root - mean - square width of the gaussian wave packet representing the nucleus , and @xmath23 is the total mass of the nucleus .
if we take the initial wave packet to be characterized by the size of the nucleus we can write @xmath24 corresponding spreads in c.m . momentum and energy are given by @xcite @xmath25 with the corresponding spread in beam - energy @xmath26 where @xmath27 is the beam energy per particle .
for light ions this spread could be as much as 10% but reduces for heavier systems . in fig .
[ fig : wp ] we plot the dependence of the ratio in eq .
( [ eq : wp ] ) on the mass number for two different reaction times . the longer time scale of @xmath28 fm / c refers to the whole reaction , while the shorter time scale of @xmath29 fm / c describes the initial evolution until the nuclei begin to overlap .
we see that for heavy systems the ratio is small and the tdhf wave packet may be considered as the classical limit of a schrdinger wave packet . on the other hand , for long reaction times and light systems the correspondence does not seem to be very good . however , this argument may be flawed since one can take the position that we are only interested in the evolution of the wave packet until the two nuclei begin to have substantial overlap @xcite . in this case
the collision times are much smaller thus resulting in a smaller ratio for most mass numbers , as shown in fig .
[ fig : wp ] . in this context
one can interpret the tdhf results from a classical perspective , which is the correct limit for a non - spreading wave packet .
consequently , center - of - mass corrections are not appropriate in tdhf calculations and have always been neglected .
the effect of using skyrme forces fitted with the simple c.m .
correction term without this correction in tdhf will be discussed further in the next section .
schematic illustration of the initial and final many - body states .
the initial state is block diagonal whereas the final state is a full slater determinant . ]
implications of the above uncertainties depend on the details of the collision process under study .
reactions that are predominantly determined by the entrance channel , such as the initial formation of the compound system in fusion , or the calculation of fusion barriers using the density - constrained tdhf ( dc - tdhf ) method @xcite are expected to be most reliable . on the other hand
longer time processes such as deep - inelastic scattering will be prone to higher uncertainties . for example , in calculating wilczynski plots ( double - differential cross - section with respect to scattering angle and energy ) for deep - inelastic collisions we may expect a spread in energy , as much as @xmath30 mev for light systems and a spread in impact parameter of order @xmath31 fm .
the so - called _ cross - channel coupling _ describes the problem that in deep - inelastic tdhf collisions the final fragments remain _
entangled_. this complicates the identification of exit channel fragments as distinct nuclear systems @xcite . in tdhf ,
the entrance channel asymptotic conditions are set up correctly as discussed in the previous section . in practice
, the initial nuclei are calculated using the static hartree - fock theory without the center - of - mass correction .
the resulting slater determinants for each nucleus comprise the larger slater determinant describing the colliding system during the tdhf evolution , as depicted in fig .
[ fig : tdhf_collision ] .
nuclei are assumed to move on a pure coulomb trajectory until they reach the initial separation between the nuclear centers used in the tdhf evolution . using the coulomb trajectory we compute the relative kinetic energy at this separation and the associated translational momenta for each nucleus in the center - of - mass frame .
the nuclei are then boosted by multiplying each hf determinant by an exponential phase as in eq .
( [ eq : boost ] ) so that the relative momentum is the correct one calculated for the coulomb collision .
the galilean invariance of the tdhf equations with the full skyrme force results in the evolution of the system without spreading and the conservation of the total energy for the system . in other words
, a translating slater determinant does not dissipate its kinetic energy .
in tdhf , the many - body state remains a slater determinant at all times .
the final state is a filled determinant , even in the case of two well separated fragments ( does not go back to block - diagonal form ) .
this phenomenon is commonly known as the _ cross - channel coupling _ and indicates that it is not possible to identify the well separated fragments as distinct nuclei since each single particle state will have components distributed everywhere in the numerical box . in this sense
it is only possible to extract _ inclusive _ ( averaged over all states ) information from these calculations . in order to test the validity of the initial tdhf setup and the use of the skyrme interaction without the c.m .
correction term used in the parametrization we utilize the density constrained tdhf method @xcite . in this method we constrain the instantaneous tdhf density and minimize the energy .
this in effect corresponds to the extraction of the internal excitation energy from the evolving system thus tracing the dynamical trajectory on the multi - dimensional static energy surface of the composite nuclear system . in ref .
@xcite we have shown that the ion - ion potential barrier could simply be calculated by subtracting the binding energies of the two nuclei ( calculated without the c.m .
correction term ) as @xmath32 .
this in principle tests both the accuracy of the entrance channel evolution and the energy calculation since for relatively large distances the result must agree with the point coulomb potential ( for spherical systems ) . in fig .
[ fig : vr]a we show the results calculated for the head - on ( zero impact parameter ) collision of the @xmath33o+@xmath33o system at @xmath34 mev . also shown is the double folding result for the same system , using densities obtained from electron - scattering and the m3y effective nn interaction . in fig .
[ fig : vr]b we show the barriers calculated for @xmath35ca+@xmath36zr systems . comparison with the point coulomb expression as shown in fig .
[ fig : vr]b is excellent with differences on the order of @xmath37 kev , until nuclear effects come into play .
thus , we can conclude that not including the c.m .
correction energy in tdhf calculations does not seem to alter the results when energy differences are considered .
furthermore , we have also performed the same calculations by changing the collision energy with no appreciable change in results indicating that the entrance channel tdhf dynamics is not prone to uncertainties present in the long - time evolution case .
tdhf calculations in general do not use the center - of - mass correction terms present in most parametrizations of the skyrme interaction . in this paper
, we discuss this issue along with the problem of cross - channel coupling .
we also provide estimates and numerical tests to understand the impact of these assumptions .
this work , together with the recent investigations provided in ref .
@xcite , where the authors discuss the conservation of angular momentum in tdhf theory , find that the entrance channel dynamics of a heavy - ion collision is expected to be well described by the tdhf time evolution .
also , when energy differences are taken ( e.g. in the calculation of heavy - ion potentials ) the omission of the center - of - mass correction terms do not seem to alter the results .
uncertainties are expected for long - time evolution resulting in well separated final fragments .
this work has been supported by the u.s .
department of energy under grant no .
de - fg02 - 96er40963 with vanderbilt university
99 negele j w 1982 _ rev .
phys . _ * 54 * 913 umar a s and oberacker v e 2006 _ phys .
c _ * 73 * 054607 maruhn j a , reinhard p - g , stevenson p d , stone j r and strayer m r 2006 _ phys .
c _ * 74 * 027601 simenel c , chomaz p and france g de 2001 _ phys .
lett . _ * 86 * 2971 chabanat e , bonche p , haensel p , meyer j and schaeffer r 1998 _ nucl . phys . _ * a635 * 231 ; _ nucl .
_ * a643 * 441 deshalit amos and feshbach herman 1974 _ theoretical nuclear physics _ , ( wiley , new york ) lipkin h j 1958 _ phys
. rev . _ * 110 * 1395 peierls r e and thouless d j 1962 _ nucl .
phys . _ * 38 * 154 bouten m , bouten m c and cornelissens t 1985 _ j. phys .
g _ * 11 * 231 bender m , rutz r , reinhard p - g and maruhn j a 2000 _ eur .
. j. _ * a7 * 467 skalski j 2007 _ phys
c _ * 76 * 044603 flocard h , koonin s e and weiss m s 1978 _ phys . rev .
c _ * 17 * 1682 irvine j m 1980 _ j. phys
* 6 * l175 umar a s , strayer m r , ernst d j and sandhya devi k r 1984 _ phys .
c _ * 30 * 1934 goldberger m l and watson k m 1975 _ collision theory _ , ( robert e. krieger publishing company , inc .
, ny ) p.64 .
umar a s and oberacker v e 2006 _ phys .
c _ * 74 * 021601(r ) koonin s e , davies k t r , maruhn - rezwani v , feldmeier h , krieger s j and negele j w 1977 _ phys .
c _ * 15 * 1359 cusson r y , reinhard p - g , strayer m r , maruhn j a and greiner w 1985 _ z. phys .
a _ * 320 * 475 umar a s , strayer m r , cusson r y , reinhard p - g and bromley d a 1985 _ phys
c _ * 32 * 172 guo lu , maruhn j a , reinhard p - g and hashimoto y 2008 _ phys . rev .
c _ * 77 * 041301(r ) | we provide a discussion of issues related to the center - of - mass motion and cross - channel coupling in applications of the time - dependent hartree - fock ( tdhf ) theory to heavy - ion collisions .
we find that the entrance channel dynamics of a heavy - ion collision as described by tdhf does not seem to be significantly influenced by these effects , whereas the long - time evolution may be less reliable .
= 10000 = 10000 = 10000 |
the remarkable discovery of weak ferromagnetism in lightly doped divalent hexaborides has been hailed as a great surprise in the physics of magnetism @xcite . in spite of the absence of
partially filled d- or f- orbitals , which are usually required for magnetism , these materials undergo a ferromagnetic transition at a relatively high curie temperature ( 600k ) and with a small magnetic moment in a narrow range of doping .
_ @xcite have put forward the proposal of a doped excitonic insulator to explain the unexpected ferromagnetism in these materials , a theory that has a long and rich history @xcite , but for which hexaborides seem to be the first experimental realization .
band structure calculations @xcite of @xmath0 where @xmath1 indicate a small band overlap or small band gap at the three x points in the brillouin zone , a condition favorable to the formation of an exciton condensate .
the coulomb attraction between particles and holes can overcome the small band gap and lead to the spontaneous condensation of electron - hole pairs . as a result of hunds rule , a triplet exciton condensate is energetically favored over a singlet , giving rise to a local spin polarization within the unit cell but no net magnetization .
the connection of exciton condensation to ferromagnetism was first pointed out by volkov _
_ @xcite . upon doping ,
a spin singlet exciton is formed in presence of the triplet order , breaking translational as well as time - reversal symmetry , and hence resulting in ferromagnetism . in more physical terms
, the extra carriers added into the system are distributed asymmetrically between the two spin projections to preserve the most favorable pairing condition for one spin orientation .
although appealing , the excitonic model has been largely studied only for highly idealized cases . in this paper , we relax some of these approximations to extract generic predictions applicable to excitonic condensates . in particular , we are interested in explaining recent angle resolved photo emission spectroscopy ( arpes ) @xcite measurements and de haas - van alphen ( dhva ) @xcite measurements on the fermi surface in @xmath2 and @xmath3
. we study in some detail the zero temperature phase diagram as a function of the doping level in the presence of imperfectly nested electron and hole fermi surfaces .
we show a possible new phase , the excitonic metal ( em ) , can reconcile the existence of a fermi surface and an excitonic condensate . furthermore , as an upshot of the em
, the ungapped fermi surface leads to a compensated magnetic moment , a mechanism for partial polarization markedly different from the one proposed by zhitomirsky _
et al_. the paper is structured as follows .
first , in the next section we introduce a two - band model hamiltonian and discuss its physical relevance to the excitonic state . in section iii , we analyze the model for the case of perfectly nested fermi surfaces . using the techniques of mean - field theory ,
we recast the problem into a form closely related to bcs theory .
we show that excitonic states with collinear spin polarization are favored in the ferromagnetic phase . in section iv
, the effect of incommensurability of the fermi surface on the emergence of ferromagnetism is studied .
we report numerical results on the magnetic moment and density of states in the excitonic phases .
the effect of the excitonic order on the optical conductivity is considered in section iv .
as argued by balents _ et al _ @xcite , the binding of electron and hole is most effective for states with similar group velocity at equal momenta .
hence , in the hexaborides , the electron - hole pair will be more strongly bound if both components are drawn from the same x point of the brillouin zone .
thus , we neglect the three - pocket band structure appropriate to the hexaborides to study a two band model of conduction and valence electrons .
to analyze the problem of excitonic ferromagnetism , we introduce the green s function defined by the equality @xmath4 here @xmath5 and @xmath6 are operators annihilating and creating an electron with spin @xmath7 ( @xmath8 ) at point @xmath9 at imaginary time @xmath10 . in the two band model
, we can express the field @xmath11 in terms of @xmath12 and @xmath13 , the creation operator in the conduction and valence band , respectively : @xmath14 where @xmath15 and @xmath16 are the bloch functions with quasimomentum @xmath17 of electrons in the conduction and valence bands , respectively .
the hamiltonian of the system can be written as @xmath18 , where @xmath19 is the non - interacting part of the hamiltonian . @xmath20 and @xmath21 are the kinetic energy of conduction and valence electrons , respectively . the interacting part of the hamiltonian is given by the coulomb interaction @xmath22 where @xmath23 . the exciton bohr radius @xmath24 sets the length scale of the problem , the physics being controlled by momentum exchange processes of the order of @xmath25 . in the weak binding limit ,
i.e when the energy of an exciton @xmath26 is much smaller than the bandwidth @xmath27 , @xmath24 is much larger than the interlattice spacing @xmath28 and the interaction is dominated by the long range part of the coulomb interaction . in this case , we can treat the interaction in the so - called dominant term approximation @xcite , @xmath29 , where @xmath30 here @xmath31 is the long range part of the coulomb interaction .
notice that unlike the full crystal hamiltonian @xmath32 , @xmath33 conserves _ separately _ the charge and the spin of the conduction and valence electrons .
the order parameter @xmath34 that corresponds to the pairing of a valence and conduction electrons is a matrix in spin space . because the approximate hamiltonian conserves separately the particle number of each species , the energy of the system is invariant under a constant shift of the overall phase of the order parameter .
similarly , the @xmath35 symmetry leads to the degeneracy of the excitonic triplet and singlet states .
the full crystal hamiltonian will in general break these symmetries . however
, we can justify the use of the approximate hamiltonian by arguing the neglected terms will affect the formation of the excitonic condensate only at a very low energy scale . indeed ,
these terms involve matrix elements of the coulomb potential at wavevector @xmath36 and hence can be treated as small perturbations compared to the long range coulomb interaction part of the interaction at wavevector @xmath25 .
it is interesting nonetheless to study how they lift the degeneracy .
there are three classes of terms that we have ignored .
the first one contains electron - hole exchange terms such as @xmath37 .
these terms break the @xmath35 symmetry explicitly .
they are ultimately responsible for the splitting of the ground state of the hamiltonian in favor of the triplet state @xcite .
another class of terms left aside are matrix elements of the kind @xmath38 . in case of the hexaborides ,
these terms are prohibited at small @xmath39 due to lattice symmetry .
the third class contains interband pair scattering terms that create and destroy simultaneously two valence and conduction electrons in the brillouin zone
. this term will select the _ phase _ of the order parameter .
however , the effect of these virtual processes are limited by the low density of excited pairs .
first , we study the hamiltonian @xmath33 in a mean - field ( mf ) treatment . for large exciton radius @xmath24 ,
the excitons overlap strongly and we expect mf theory to give correct results .
the excitonic phase is characterized by the pairing of valence and conduction electrons . to emphasize the similarity of our treatment to the nambu - gorkov formalism for superconductors , we introduce the green s function in the bloch basis @xmath40 the excitonic gap @xmath41
is related to the off - diagonal green s function @xmath42 ignoring for the moment the possibility of ferromagnetic ordering , the mf green s function is given by @xmath43 where @xmath44 and @xmath45 and @xmath46 are pauli matrices in the band space .
[ mf1 ] must be solved in conjunction with the gap equation ( eqn .
[ gap ] ) to obtain a self - consistent solution . at this point , it is convenient to introduce the decomposition of @xmath41 in terms of the singlet @xmath47 and triplet @xmath48 order parameter .
@xmath49 where @xmath50 and @xmath51 are respectively the @xmath52 unit and pauli matrices in the spin space .
because the hamiltonian @xmath33 is invariant under separate rotations of the valence and conduction electrons , the triplet states are degenerate with the singlet states .
consider the case where the order parameter is collinear with the spin , i.e @xmath53 . with this replacement ,
the mf hamiltonian decouples the spin up and spin down electrons .
it is straightforward to obtain the one - particle excitation spectrum of spin @xmath54 electrons @xmath55 where @xmath56 . in the theory of excitonic insulator
, it is often assumed that the electronic dispersion relation are nested and isotropic , i.e. @xmath57 . as made apparent in equation [ mf1 ] , the particle - hole transformation @xmath58 maps the problem onto two copies of the bcs model in a zeeman field ( a copy for each spin polarization ) .
the zeeman field takes the role of the chemical potential .
this problem has been worked out , within a mf treatment , by larkin _
_ @xcite and fulde _ et al .
assuming strong screening of the coulomb interaction , i.e. @xmath59 , we see from equation @xmath60 , that the gap equation becomes momentum independent . at stoichiometry , we obtain a bcs - like state with a gap @xmath61 where @xmath62 is a cutoff energy around the fermi surface and @xmath63 is the density of states per spin at the fermi level .
the lower band @xmath64 is fully occupied whereas the upper band @xmath65 is empty : the system is an excitonic insulator with a one - particle excitation gap of @xmath66 .
turning on the chemical potential , the hole and electron fermi surfaces shift with respect to one another .
this loss of nesting lowers the condensation energy and ultimately the excitonic state disappears . for a uniform order parameter ,
the transition occurs at @xmath67 via a first order transition .
the condensation energy per spin direction relative to the normal state for @xmath68 , is @xmath69 at @xmath70 , the ground state is degenerate and _
each _ spin polarization can be in a state where @xmath71 or @xmath72 .
although the ground state has generally a 2-fold degeneracy at a first order transition point , the degeneracy for _ each _ spin species is non - generic . up to this point
, the mf treatment has neglected the effect of the intraband coulomb interaction . while it is appropriate to ignore the effect of the direct part of the coulomb interaction for uniform states , its exchange part yields a ferromagnetic interaction .
the spin of the electrons in the conduction and valence band , respectively @xmath73 and @xmath74 , interact via the heisenberg hamiltonian @xmath75 where @xmath76 . in our mean - field theory , @xmath77 .
however , the exchange constant tends to be poorly approximated by the direct coupling constant because the true exchange is screened by the dielectric function of the crystal .
hence we take @xmath78 to be an adjustable parameter interaction ( although still of order @xmath79 ) .
phenomenologically , @xmath78 is related to the antisymmetric landau - fermi liquid parameter @xmath80 via the relation @xmath81 @xcite .
notice that there are additional terms in the full crystal hamiltonian that couple to the excitonic order parameter but are negligibly small compare to @xmath79 and hence can be safely ignored in first approximation . to take into account the exchange energy ,
we introduce a generalized mean - field theory , replacing @xmath82 and similarly for the valence electrons .
we find @xmath83 where the self - consistent equations are @xmath84 , \nonumber \\
\vec{h}_b= \frac{j}{2 } \sum_{k } tr\left [ \vec{\sigma } { \cal g}_{bb}({\bf k},\tau=0 ^ - ) \right].\end{aligned}\ ] ] we have studied the energetics for collinear spin polarization and perfectly nested fermi surfaces . while the spin branches are no longer decoupled , we find the condensation energy of the system @xmath85 is given by a simple maxwell construction @xmath86 where @xmath87 .
minimizing this functional , we find that an excitonic ferromagnet ground state is favored in a range of chemical potential @xmath88 ( see fig .
[ maxwell ] ) .
this excitonic polarized state is characterized by having one spin species , say spin up , in the normal state whereas the spin down electrons are paired ( @xmath89 ) .
one can understand the origin of the magnetism by studying its effect on the fermi surfaces .
the magnetic field favors the pairing of the spin down electrons by increasing the coincidence of the hole and electron fermi surfaces for the spin down .
this gain in condensate energy is _ linear _ in @xmath90 near the first order transition .
while this redistribution of the charge carriers comes at the expense of a loss in kinetic energy for the spin up electrons , this increase in energy is quadratic in the field therefore favoring ferromagnetic ordering .
we have investigated the energetics of order parameters non - collinear with the spin .
an instance of such of state is @xmath91 .
this particular state has been proposed by zhitomirsky _
@xcite to account for ferromagnetism in the hexaborides .
unlike in the collinear states , the ferromagnetic moment , @xmath92,\ ] ] is perpendicular to the exciton polarization @xmath93 .
for non - collinear spin polarization , the simple eqn.[maxwell ] is no longer valid . instead ,
we find that the gain in condensate energy in the ferromagnetic non - collinear state is _ quadratic _ in @xmath90 , so it is never the lowest energy state .
an inquiring reader might object that such a state could lower its energy by taking advantage of terms that we ignore in the dominant - term approximation . while the energy difference of the non - collinear and collinear state is of the order of @xmath94 , the correction brought by the neglected terms are of the order @xmath95 where @xmath96 . in strong coupling where @xmath97
, one needs to reconsider the problem and it was found that the excitonic order in the ferromagnetic regime may be of non - collinear type @xcite .
physically , the key signature of the non - collinear state @xmath98 is the spontaneous generation of spin currents , markedly absent in collinear spin polarized states .
however , it is known to be notoriously difficult to observe spin currents experimentally . a possible way to distinguish the two phases
is by studying their collectives modes although this warrant further investigations .
although the previous results are suggestive , a complete understanding of the ferromagnetism and its applicability to the hexaborides is still lacking . the large number of compounds exhibiting weak ferromagnetism discovered to this day argues for the robustness of the phenomenon .
however , most of theoretical results seemingly hinge on the perfect nesting of the hole - electron fermi surface to lead to an excitonic insulator
. moreover , the existence of a fermi surface as seen in arpes and dhva data rules out this simple scenario .
a crucial step to circumvent these difficulties is to realize the hole and electron bands are not related by any symmetry principle and in real materials , electron and hole fermi surfaces are not generally perfectly nested . to study the effect of incommensurability on the excitonic states ,
we introduce the anisotropic dispersion relations for the valence and conduction bands @xmath99 here @xmath100 is the fermi velocity , @xmath101 is the fermi wavevector , @xmath102 is a dimensionless parameter that controls the degree of antinesting and @xmath103 is the azimuthal angle . for @xmath104 , we recover the isotropic and nested dispersion relations whereas for finite @xmath102 the fermi surface is described by two non - coinciding ellipsoids ( see fig .
the parameter @xmath105 is determined by imposing charge neutrality at stoichiometry at @xmath106 .
i.e. @xmath107= \sum_k 1,\ ] ] where @xmath108 is the heaviside function .
this toy model characterization of the energy dispersion describes simply the effect of anisotropy as well as allows us to make analytic progress .
furthermore , the large uncertainties in the available data for the effective masses and gap values did not allow for an accurate theoretical description although lda calculations @xcite on @xmath109 would suggest @xmath102 to be positive . based upon our previous analysis
, we consider only collinear states and minimize the energy functional @xmath110= \sum_{\alpha,\sigma=\pm } \left [ \int \frac{d^3 k}{(2 \pi)^3 } ( \xi^\alpha_{k,\sigma}- \sigma h ) \theta(-\xi^\alpha_{k,\sigma}+\sigma h ) \right ] \nonumber \\
+ \frac{2 h^2}{j}+ \frac{\delta_+^2+\delta_-^2 } { v_0 } , \end{aligned}\ ] ] where @xmath111 is the energy eigenvalue determined in eqn .
[ eigen ] .
the integral over wavevector can be expressed in terms of elliptical functions . from the three parameters @xmath112 minimization search ,
we have determined the phase diagram at zero temperature as a function of @xmath113 , the chemical potential and @xmath102 , the antinesting parameter ( see fig.[pd ] ) .
not surprisingly , for small values of @xmath102 , we find that the system behaves analogously to the nested case . at stoichiometry ,
the system is in the excitonic insulating ( ei ) phase . increasing the chemical potential ,
the ei state becomes unstable to a ferromagnetic phase before disappearing into the normal state via a first order transition .
however , larger values of @xmath102 allows for the possibility of gapless excitations in the one particle excitation spectrum .
hence , while the excitonic metal ( em ) exhibits a non - zero order parameter i.e. ( @xmath114 or @xmath115 ) , the presence of an incompletely gapped fermi surface modifies radically its electronic properties . to illustrate this point ,
we have calculated the electronic density of states ( dos ) , @xmath116,\ ] ] at stoichiometry in the insulating and metallic phases . as shown in fig .
[ dos ] , in spite of the excitonic ordering , the dos in the metallic phase does not show the typical shoulder edge present in the insulating excitonic phase .
the em / ei transition is continuous and the excitonic gap varies smoothly across the transition line defined by @xmath117 .
however , the metallic instability is signaled by the density of states at the fermi level which show a singular behavior across the transition line , increasing as @xmath118 .
the introduction of charge carriers in the em phase raises the fermi level , leading to changes in the topology of the fermi surface .
recent results on dhva experiments on divalent hexaborides have shown the presence of fermi sheets in the undoped compounds @xcite .
the fermi surface is thought to consist of two pieces , an electron ring and a hole lens centered around the x - point in the brillouin zone . in the la - doped @xmath109 ,
dhva data show the hole pocket dropping out , yielding a single electron sheet .
these results are consistent with our model . as seen in fig .
[ fs ] , in the undoped case , the fermi surface is gapped at the intersection of the hole and electron fermi surface , leaving two fermi sheets , a ring and a lens . for positive values of @xmath102 ,
the lens and ring are hole - like and electron - like , respectively .
as electrons are added to the system , the hole pocket shrinks and eventually becomes gapped out due to the excitonic gap , leaving a single electron ring , consistent with experimental data .
however , further experiments on cleaner samples are needed to clearly determine the topology of the fermi surface .
an upshot of the finite density of states at the fermi level is the possibility of distributing the extra carriers asymmetrically between spin species . as shown in fig.[gap ] , the em phase gives way to a _ partially _ polarized ferromagnet where the extra carriers carry only a _ fraction _ of a bohr magneton .
when @xmath102 becomes of the order of @xmath119 , the incommensurability of the hole and electron fermi surfaces becomes too large and hinders the gap formation .
however , this state may be itself unstable to the formation of non - homogeneous states with spin and/or charge density waves@xcite .
the result of our calculation of the magnetic moment per dopant for the doped excitonic metal is shown in fig.[mm ] .
in contrast to the case of a doped excitonic insulator , the ferromagnetic moment per carrier is smaller than a bohr magneton due to the electrons spilling over both branches of the spin species . unlike in the ei / fpfm transition ,
the electronic charge density can in principle evolve smoothly across the em / ppfm boundary , thus one might expect a second order ( continuous ) transition to arise .
however , we find in mft a first order transition .
moreover , we suspect this may be a general result due to singular quasiparticle - mediated interactions .
an argument in favor of a first order transition is found by considering the system at low but non - zero temperature . in the neighborhood of the transition , were it second order , it should be possible to develop a landau expansion of the free energy in terms of the magnetic order parameter @xmath90 with @xmath120 already non - zero .
this expansion must be analytic at @xmath121 @xmath122 the parameters a and b are related to two- and four - point spin correlation functions respectively . as @xmath123 , we find @xmath124 and @xmath125 hence precluding the validity of such an expansion and therefore of a continuous transition
. it would be desirable to develop a proper field theoretic argument posed directly at @xmath106 .
a consequence of the direct first - order transition is the discontinuous jump of the electronic density across the boundary line .
since experiments are performed at fixed charge density , phase separation occurs for a range of doping .
however , the long range coulomb interaction will likely frustrate macroscopic phase separation and an inhomogeneous state is expected .
a powerful technique that can be used to probe the excitonic state is the infrared optical conductivity . just as for superconductors ,
the excitonic insulator signature on the optical conductivity is the suppression of absorption for frequencies twice below the excitonic gap @xmath126 .
however , one of the puzzling aspects of optical conductivity measurements on the undoped hexaborides is the absence of hard optical gap and strong doping dependence @xcite .
one expects the presence of gapless quasiparticles in the excitonic metal state to be responsible for this result . to investigate this issue quantitatively we calculate the optical conductivity tensor @xmath127 which is related to the current - current correlation function @xmath128 @xmath129 where @xmath130 \rangle $ ] and @xmath131 is the current operator .
the greek indices refer to the spatial components .
we calculate the correlation function in the matsubara formalism and obtain the desired retarded function by analytic continuation . to allow for electronic scattering due to impurities and/or lattice imperfections
, we introduce a finite lifetime to the quasiparticle by performing the substitution @xmath132 in the green s function @xmath133 where for concreteness , we consider a pure triplet order parameter .
ignoring vertex corrections to the conductivity , the correlation function is given by the bubble diagram , @xmath134 , \end{aligned}\ ] ] where @xmath135 is the velocity of the quasiparticles . in order to evaluate the correlation function , we use the standard spectral representation for the green s function @xmath136.\ ] ] 2 in the physical regime where the scattering rate @xmath137 is much smaller than @xmath120 , the imaginary part of the correlation function in real frequencies is given by @xmath138=\sum_{\sigma } \delta_{\mu \nu } \frac{(e v_f)^2}{3 } \int \frac{d^3k}{(2\pi)^3 } \frac{2 \gamma}{\omega^2 + \gamma^2 } \frac{1}{4 \chi^2 + \gamma^2 } \big [ \left ( \theta(\xi^{+}_{\sigma})- \theta(\xi^{+}_{\sigma}+\omega ) -\theta(\xi^{-}_{\sigma})+
-\omega ) \right ) \nonumber \\ \frac{(2 \chi^2)(\chi^2 + ( \chi+\omega)^2 ) + 4 \chi ( \chi+\omega)(\chi^2 - 2\delta^2)}{(\omega+2\chi)^2+\gamma^2 } - ( \omega \rightarrow -\omega ) \big],\end{aligned}\ ] ] 2 where @xmath139 .
we evaluate the correlation in different limits . for low frequencies ,
i.e. @xmath140 , we find a drude - like conductivity , @xmath141 where @xmath142 hence , the scattering of gapless excitations generates absorption in the low energy sector of the optical conductivity . for high frequencies ,
the dominant contribution to the integral comes from exciton pair breaking , i.e. when @xmath143 .
thus for @xmath144 @xmath145.\ ] ] 2 in the clean limit , the threshold for exciton pair breaking is given by @xmath146 .
in this case the absorption peak shows a square root singularity @xmath147 near the threshold due to the large amount of phase space available in @xmath148 space .
however , in presence of scattering , the wavevector @xmath148 is no longer a good quantum number and the peak is smeared out on a width of order @xmath137 ( see fig [ abs ] ) .
we conjecture that the absence of hard optical gaps in experimental data is due to metallicity and scattering that shifts the absorption to lower frequencies .
notice that because the order parameter for the spin up and spin down is generally different in the ferromagnetic state , in clean materials , the absorption peak would be split into two peaks upon entering the ferromagnetic phase .
its experimental observation in the hexaboride materials would provide considerable support for the theory of excitonic ferromagnet .
excitonic ordering near a semiconductor - metal transition offers a natural explanation of weak ferromagnetism in the doped hexaborides .
we have generalized previous studies on the excitonic ferromagnet by considering the effect of imperfect nesting on the excitonic state . adding this significant ingredient
, we showed that a novel phase , the excitonic metal , is stable and can account for the fermi surface seen in the hexaborides .
a ferromagnetic instability still occurs in the excitonic metal near the first order transition .
furthermore , the small ferromagnetic moment observed in lightly doped hexaborides is naturally explained as a result of compensated moments on the imperfectly gapped fermi surface .
although this model qualitatively explains experiments , further studies will be needed to determine the effect of the gapless excitations on the electronic transport and thermodynamic properties of the system .
l.b . and m.y.v .
were supported by the nsf dmr9985255 and the sloan and packard foundations .
young , d. hall , m. e. torelli , z. fisk , j. l. sarrao , j. d. thompson , h. r. ott , s. b. oseroff , r. g. goodrich and r. zysler , nature * 397 * 412 , ( 1999 ) .
m. e. zhitomirsky , t. m. rice and v. i. anisimov , nature * 402 * , 251 ( 1999 ) .
t. jarlborg , phys .
lett . * 85 * , 186 , ( 2000 ) .
l. balents and c. m. varma , phys .
lett . * 84 * , 1246 ( 2000 ) . c. o. rodriguez , r. weht and w. e. pickett , phys .
lett . * 84 * , 3902 , ( 2000 ) .
v. barzykin and l. p. gorkov , phys .
. lett . * 84 * , 2207 ( 2000 ) .
t. ichinomiya , phys .
b * 63 * , 045113 ( 2001 ) . b. i. halperin and t. m. rice , solid state physics * 21 * , 115 ( 1968 )
. s. massida , a. continenza , t.m .
de pascale , and r. monnier , z. phys b * 102 * , 83 ( 1997 ) .
b. a. volkov , yu . v. kopaev and a. i. rusinov , sov .
jetp . * 41 * , 952 ( 1975 ) .
j. d. denlinger , j. a. clack , j. w. allen , g. -h .
gweon , d. m. poirier , c.g .
olson , j.l .
sarrao , z. fisk , cond - mat/0009022 , ( 2000 ) .
d. hall , d.p .
young , z. fisk , t.p .
murphy , e.c .
palm , and a. teklu , r.g .
goodrich , cond - mat 0104184 , 2000 .
a. i. larkin and yu .
n. ovchinnikov , sov .
. jetp . * 20 * , 762 ( 1965 ) .
p. fulde and r. a. ferrel , phys . rev . *
135 * , a550 ( 1964 ) .
a. a. abrikosov , l. p. gorkov , and i. e. dzyaloshinski , in _ methods of quantum field theory in statistical physics _ , prentice - hall , new jersey , ( 1963 ) | we investigate a model of excitonic ordering ( i.e electron - hole pair condensation ) appropriate for the divalent hexaborides .
we show that the inclusion of imperfectly nested electron hole fermi surfaces can lead to the formation of an undoped excitonic _ metal _ phase .
in addition , we find that weak ferromagnetism with compensated moments arises as a result of gapless excitations .
we study the effect of the low lying excitations on the density of states , fermi surface topology and optical conductivity and compare to available experimental data .
+ + pacs : 71.35.cc , 71.35.-y , 71.20-b 2 |
it has been widely accepted that gamma - ray burst ( grb ) emission is produced in a shocked optically thin plasma via the synchrotron process ( synchrotron shock model ; ssm ) .
the resultant emission spectra is a convolution of synchrotron emission from electrons with distributed energies , and can be described as a broken power - law function with a broad @xmath1 peak . while the overall spectral shape , particularly at energies around the peak of the spectral energy density and above , depends critically on the energy spectrum of the source electrons , the shape of the low - energy spectrum
is bound to converge to a power - law function , @xmath2 , with index @xmath3 , regardless of the source electron spectrum .
this is because each of the individual source electron produces emission with the same low - energy asymptotic spectral form , namely a power - law function with an index of @xmath3 .
most of the observed spectra of grbs are known to be well described by the `` band function '' ( band et al . 1993 ) , which is essentially an empirical functional form consisting of two smoothly connected power - law continua , where the spectral index is more negative at higher energies ( preece et al .
the band function has three parameters to characterize the spectral shape : the low - energy photon spectral index , @xmath4 , the high - energy photon spectral index , @xmath5 , and the break energy , @xmath6 . here
, we use the convention that indices @xmath4 and @xmath5 are positive in the power - law function , @xmath7 , following preece et al .
if a photon spectrum produced by the optically thin synchrotron shock is fitted to the band function , the low - energy photon spectral index , @xmath4 , can not be larger than @xmath8 , because at any part of the ssm spectrum the photon slope is more negative than @xmath8 ( e.g. , tavani 1996 ) .
preece et al .
( 1998 ) examined the time - resolved energy spectra of over 100 bright grbs observed by the batse experiment on cgro , and found 23 bursts in which the spectral index limit of the ssm was violated . since the spectral measurement of grbs at energies below 30 kev was difficult for batse , preece et al .
( 1998 ) used the `` effective spectral index '' at 25 kev , i.e. the slope of the power - law tangent to the band function at the chosen energy ( 25 kev ) . furthermore , frontera et al .
( 2000 ) found several grbs showing a similar low - energy depletion in the range 220 kev , especially during the initial part of the grb by the bepposax . in this paper
we present the spectral observation of grb 020813 with the hete-2 satellite .
the fregate and wxm detectors on hete-2 cover a wide energy range of 2400 kev , and are well suited to study the spectral shape in the x - ray regions .
grb 020813 has one of the highest spectral peak energies , @xmath9 , among the grbs localized by hete-2 ( sakamoto et al .
we found that part of its time - resolved spectra exhibits a flat low - energy slope that violates the `` death line '' of the ssm . with its long duration , high fluences and characteristic spectra ,
grb 020813 presents an ideal case to study the low - energy part of a hard grb . because of its unusual hardness ( see figure 4 ) , it is also interesting to study the spectral shape at high energies .
that aspect will be presented in a separate paper using the data from another spacecraft , which covers energy ranges higher than those available with hete-2 . in section 2 ,
the properties of grb 020813 observed by hete-2 are presented , including a detailed analysis of the time - resolved spectra . in section 3 , we discuss the possible process that causes the violation of the ssm limit on the low - energy power - law index . in particular , synchrotron self - absorption and compton - scattering processes are critically examined .
the trigger for this event came from the fregate instrument , at 02:44:19.17 utc ( 9859.17 s ut ) on 2000 august 13 .
the trigger occurred in the 30@xmath0400 kev band , on the 1.3 s time scale .
the wxm flight software localized the burst , and its position was ra = @xmath10 , dec = @xmath11 ( j2000.0 ) , with a 90@xmath12 error radius of 14@xmath13 which includes statistical and systematic errors .
the flight location was reported in a grb coordinate network ( gcn ) position notice at 10:50:48 ut , 4 minutes 14 s after the burst .
ground analysis of the wxm data for the burst produced a refined location , which was reported in a gcn position notice 117 min after the burst .
the wxm location can be expressed as a 90@xmath12 confidence circle that is 4@xmath13 in radius and is centered at ra = @xmath14 , dec = @xmath15 .
ground analysis of the sxc data for the burst also produced a refined location , which was reported in a gcn position notice 184 min after the burst .
the sxc location can be expressed as a 90@xmath12 confidence circle that is 60@xmath16 in radius and is centered at ra = @xmath17 , dec = @xmath18 .
grb 020813 was also observed by interplanetary network ( ulysses , konus - wind , and mars odyssey ) and the resulting ipn localization was fully consistent with the wxm error circle ( hurley et al .
2002a , 2002b ) . an optical afterglow ( ot )
was observed 112 min after the burst at ra = @xmath19 , dec = @xmath20 ( fox et al .
2002 ) inside the wxm error circle , and follow - up observations were carried out by dozens of telescopes distributed around the world .
the light curve of the optical afterglow has shown a break at 14 hr after the burst ( e.g. , urata et al .
2003 ) . in the framework of jetted fireballs
, this break corresponds to a jet half - opening angle of [email protected] @xmath22 [email protected] ( covino et al .
moreover , optical spectroscopic observations at the keck observatory have determined its redshift as @xmath23 ( barth et al .
figure 1 shows the wxm and sxc localizations , which are consistent with the location of the ot .
confidence limits.,width=302 ] figure 2 shows the complex light curve of grb 020813 in the four wxm energy bands ( 2@xmath05 , 5@xmath010 , 10@xmath017 , and 17@xmath025 kev ) and the three fregate energy bands ( 6@xmath040 , 40@xmath080 , and 32@xmath0400 kev ) .
there are more than three peaks in the wxm light curve and at least four spiky peaks in the fregate light curve .
the photon counts in the peaks are higher at later phases .
an inspection of the burst light curve in the wxm and fregate energy bands shows that the fourth peak is much harder than the others .
table 1 gives @xmath24 and @xmath25 durations of grb 020813 in the 2@xmath05 , 5@xmath010 , 10@xmath025 , and 2@xmath025 kev wxm bands and the 6@xmath040 , 40@xmath080 , 32@xmath0400 , and 6@xmath0400 kev fregate energy bands .
moreover , the timescale of the temporal structure seems to be shorter at higher energies , which is a feature commonly observed in many grbs ( e.g. , fishman et al .
1992 , link et al .
1993 , fenimore et al .
we can see a number of `` shots '' in the 32@xmath0400 kev light curve .
-2 : wxm data ( 1 ) 2@xmath05 kev , ( 2 ) 5@xmath010 kev , ( 3 ) 10@xmath017 kev , and ( 4 ) 17@xmath025 kev bands ; and fregate data ( 5 ) 6@xmath040 kev , ( 6 ) 40@xmath080 kev , and ( 7 ) 32@xmath0400 kev bands.,width=302 ] lccc instrument & energy & @xmath25 & @xmath24 + & ( kev ) & ( s ) & ( s ) + hete-2 wxm & 2@xmath025 & @xmath26 & @xmath27 + & 2@xmath05 & @xmath28 & @xmath29 + & 5@xmath010 & @xmath30 & @xmath31 + & 10@xmath025 & @xmath32 & @xmath33 + hete-2 fregate & 6@xmath0400 & @xmath34 & @xmath35 + & 6@xmath040 & @xmath36 & @xmath37 + & 40@xmath080 & @xmath38 & @xmath39 + & 32@xmath0400 & @xmath40 & @xmath41 + two types of data sets ( `` @xmath42 data '' and `` @xmath43 data '' ) are provided by the fregate detector on board hete-2 ( atteia et al . 2003 ) .
the burst data are only available when fregate triggered on the grb , whereas continuous , monitoring data are always recorded whenever hete-2 is operating . in a spectral analysis of fregate data , we usually use burst data , since it provides time- and energy - tagged photons with a much finer time resolution ( 6.4 @xmath44s ) than the continuous data , which is accumulated every 5 s. however , due to the long duration of grb 020813 , the memory was full and burst data were only available at 0@xmath4570 s. therefore , we constructed the energy spectrum for the remaining part of the grb by using continuous data over the full energy range of fregate ( 7@xmath0400 kev ) , which was also discrete data and only available during three separate intervals : @xmath46 s , 40@xmath45120 s , and 220@xmath45300 s ( figure 3 , left ) .
note that the differences between the @xmath42 data and the @xmath43 data are its time resolution and energy resolution .
the quality of the derived spectra is exactly the same for both types of data in the following analysis .
the background level at an arbitrary time was estimated by interpolating the @xmath47-burst @xmath43 data ( @xmath46 s ; bg1 ) and the @xmath48-burst @xmath43 data ( 220@xmath45300 s ; bg2 ) with a linear function of @xmath49 , where @xmath50 is the time after the burst in seconds ( figure 3 , @xmath51 ) .
we then calculated a weighted mean of the bg1 and the bg2 spectra at each time @xmath50 , and subtracted this from the fregate spectral data .
light curve of the @xmath43 data in the fregate energy band .
@xmath52 background of the @xmath43 data with a linear function.,title="fig:",width=226 ] light curve of the @xmath43 data in the fregate energy band .
@xmath52 background of the @xmath43 data with a linear function.,title="fig:",width=226 ] to analyze the wxm data , we used the `` tag data '' generated at the burst trigger time in the full 2@xmath025 kev energy range .
the background was integrated over 110 s , which was extracted from 118 s to 8 s before a burst . in the wxm spectral analysis
, we considered only those events that registered on wires in the x- and the y - detectors illuminated by the burst .
furthermore , since the variation in the gain is not uniform at the ends of the wires in the wxm detectors ( shirasaki et al .
2003 ) , we used only the photon counts that registered in the central @xmath53 mm region of the wires to construct reliable spectra of the burst .
as one can see in figure 2 , the light curve of grb 020813 shows a very complex structure , such that four distinct peaks , at least , are visible in the light curve . considering the fast time variability , we analyzed the spectrum of grb 020813 every 5 s , which is also the limit imposed by the time resolution of the fregate continuous data . before
going into the detailed spectral analysis , we first review the spectral evolution during a burst , by comparing the photon counting rates in the low energy ( wxm ) band with those in the high energy ( fregate ) bands .
the hardness ratio of the fluxes , @xmath54(30@xmath0400 kev ) to @xmath55(2@xmath030 kev ) , is shown in figure 4 ( third panel ) .
one can see that grb 020813 possibly consists of two distinct bursts that may have different physical origins .
the border between the first and second bursts is approximately given by @xmath56 s. note that this corresponds to the epoch when the first gradual burst decayed to the initial ( background ) level , and succeeding flares started to appear in the light curve ( figure 2 ) .
we therefore denote the former part as `` p1 '' and the latter part of the flare as `` p2 '' , just for convenience .
the time intervals for p1 and p2 correspond to 17@xmath052 s and 67@xmath0109 s from the burst trigger time , respectively .
note that the peak of p1 shows larger values of @xmath54(30@xmath0400 kev)/ @xmath55(2@xmath030 kev ) than that of p2 , meaning that energy spectra in the peak of p1 are harder than those of p2 . in a more detailed analysis
, we considered three different models to reproduce the observed spectra : ( 1 ) a power - law function , ( 2 ) a cutoff power - law function and ( 3 ) a band function ( band et al .
we first fitted the data with a single power - law function , but this model did not represent the spectra well , except in low photon statistics regions .
this is because the spectrum of grb 020813 is not a simple power - law , but bends sharply at high energies , as is often reported in other grbs ( e.g. , band et al .
1993 ) . on the other hand , both the cutoff power - law and the band function provide improved fits for all regions . due to the limited photon statistics , it is hard to say which model is a better representation of the grb 020813 spectra for each 5 s segment .
we thus fit the time - averaged spectrum with a cutoff power - law and the band function , and compared their results .
when the photon statistics are sufficiently high , the band function gives a better fit with a reduced @xmath57 of 1.18 for 140 dof , compared to 1.33 for 141 dof for the cutoff power - law model . in the following analysis
, we therefore use the band function to discuss the evolution of the spectra for all time intervals .
table 2 presents the results of a time - resolved spectral analysis of grb 020813 .
we used the xspec version 11.2.0 software package to do the spectral fits .
the photoelectric absorption in the direction of grb 020813 is 7.0 @xmath58 @xmath59 @xmath60
( vanderspek et al . 2002 ) , which is negligible even in the wxm energy range above 2 kev .
therefore , we do not consider @xmath61 as a parameter in the spectral analysis .
figure 4 summarizes the time evolution of spectral parameters together with the wxm and the fregate light curves measured at 2@xmath025 kev and 7@xmath0400 kev , respectively ( the first and the second panels ) .
the time variations of the low and high - energy photon indices , @xmath4 and @xmath5 , are plotted in the fourth and fifth panels .
the break energy in the spectrum @xmath62 is given in the sixth panel , and the fluxes measured in the 2@xmath0400 , 2@xmath025 and 30@xmath0400 kev energy bands are plotted in the remaining panels .
as shown in figure 4 , grb 020813 spectra have a clear break around 100 kev , meaning that most of the power is emitted in the hard x - ray and the soft gamma - ray energy bands . for the time - averaged spectrum of p1
, we found that @xmath63 ergs @xmath60 and @xmath64 ergs @xmath60 .
thus , the logarithmic fluence ratio , defined as log@xmath65 , is @xmath00.96 .
similarly , we obtain log@xmath66 for p2 , with @xmath67 ergs @xmath60 and @xmath68 ergs @xmath60 .
these results confirm that grb 020813 belongs to the class of long classical grbs according to the hete classification method ( sakamoto et al . 2005 ) .
however , we note that log@xmath69 is quite large compared to other grb populations ( sakamoto et al . 2005 ) and would be at the extreme end of the classical grb population .
this is also indicated by the unusually @xmath70 of low - energy ( below @xmath62 ) photons observed in p1 .
in fact , the low - energy power - law indices , @xmath4 , range from @xmath71 to 0.0 , some of which are much larger than expected from the `` death line '' ( @xmath3 ) .
we will thus consider the origin of this low - energy depletion in the next section .
rrrrr & & & & + & & & & + 7.0 ( -0.7 14.8 ) & 0.6@xmath72 & -2.3@xmath73 & 44@xmath74 & 1.0 ( 18 ) + 17.4 ( 14.9 19.9 ) & -0.0@xmath75 & -1.5@xmath76 & 40@xmath77 & 1.2 ( 27 ) + 22.4 ( 19.9 24.9 ) & -0.2@xmath78 & -1.8@xmath79 & 118@xmath80 & 1.1 ( 35 ) + 27.7 ( 25.1 30.2 ) & -0.3@xmath81 & -2.0@xmath82 & 106@xmath83 & 0.8 ( 47 ) + 32.8 ( 30.2 35.4 ) & -0.5@xmath81 & -1.9@xmath84 & 115@xmath85 & 1.3 ( 57 ) + 38.1 ( 35.5 40.7 ) & -0.5@xmath81 & -9.4@xmath86 & 95@xmath87 & 1.2 ( 58 ) + 43.6 ( 41.0 46.2 ) & -0.7@xmath81 & -9.4@xmath88 & 65@xmath89 & 0.9 ( 45 ) + 48.7 ( 46.2 51.3 ) & -0.9@xmath78 & -3.1@xmath90 & 45@xmath91 & 0.8 ( 29 ) + 53.9 ( 51.3 56.5 ) & -1.2@xmath92 & -9.4@xmath93 & 59@xmath94 & 0.8 ( 30 ) + 59.4 ( 56.8 62.0 ) & -1.1@xmath95 & -9.4@xmath93 & 42@xmath96 & 1.5 ( 30 ) + 64.7 ( 62.0 67.3 ) & & & & + 69.9 ( 67.3 72.5 ) & -1.4@xmath81 & -9.3@xmath97 & 87@xmath98 & 0.8 ( 30 ) + 75.2 ( 72.5 77.7 ) & -1.2@xmath81 & -9.4@xmath99 & 138@xmath100 & 0.8 ( 74 ) + 80.4 ( 77.7 83.0 ) & -0.9@xmath81 & -1.8@xmath81 & 57@xmath101 & 1.0 ( 74 ) + 85.7 ( 83.0 88.2 ) & -0.9@xmath81 & -1.7@xmath102 & 138@xmath103 & 1.2 ( 73 ) + 90.9 ( 88.2 93.5 ) & -0.8@xmath81 & -1.9@xmath102 & 109@xmath104 & 1.2 ( 73 ) + 96.1 ( 93.5 98.7 ) & -0.8@xmath81 & -1.8@xmath81 & 80@xmath105 & 0.8 ( 73 ) + 101.4 ( 98.7 104.0 ) & -1.2@xmath81 & -9.4@xmath88 & 111@xmath106 & 0.9 ( 68 ) + 106.6 ( 104.0 109.2 ) & -0.8@xmath81 & -1.6@xmath81 & 100@xmath107 & 1.3 ( 73 ) +
as we have discussed in the previous section , grb 020813 shows a depletion of low - energy photons , especially in the p1 region .
such flat spectra , exceeding the `` death line '' ( @xmath3 ; preece et al .
2000 ) , have already been reported in a number of batse ( burst and transient source experiment ) grbs , though they were not conclusive due to the limited sensitivity of batse at low energies .
frontera et al .
2000 , using the bepposax wfc and grbm , reported that a few grbs show significant low - energy depletion in 220 kev , though their energy coverages of wfc and grbm are not completely continuous , and the energy resolution of grbm was somewhat limited .
therefore , the hete-2 observation of grb 020813 presents an independent measurement of a similar characteristic spectra with higher reliability over the wide and continuous energy range ( 2400 kev ) of hete-2 . in the following ,
we examine various scenarios to understand the origin of the x - ray emission from grb 020813 .
first , we consider the spectral hardening due to synchrotron self - absorption ( ssa ) .
we can define a critical frequency , @xmath108 , where the plasma becomes optical thick , resulting in a hard x - ray spectrum with @xmath4 = @xmath109 well below @xmath108 ( e.g. , sari et al . 1998 ) .
meanwhile , the high - energy spectral index of grb 020813 , @xmath110 , is that expected from a `` fast cooling '' of electrons with its power - law index , @xmath111 (= 2@xmath5 + 2 ) .
therefore , a sharp spectral break , as observed in grb 020813 , is expected only when the minimum synchrotron frequency , @xmath112 , is @xmath113 with respect to @xmath108 , such that @xmath114 . in this particular case ,
the observed ssa frequency is approximately given as , where @xmath115 is the doppler beaming factor , @xmath116 is the minimum electron lorentz factor and @xmath117 is the bulk lorentz factor . assuming a viewing angle between the grb and observer @xmath118
, we can derive @xmath119 .
furthermore , we assume @xmath116/@xmath117 @xmath120see , equation ( 1)@xmath121 $ ] is of order unity as often assumed in modeling the grb emission . ]
@xmath122 where @xmath123 is the redshift , @xmath124 ( @xmath125 ) the radius of emission volume , @xmath126 ( @xmath127 ) the electron number density , and @xmath128 ( gauss ) the magnetic field strength ( dermer et al . 2000 ) .
then , the peak frequency of the synchrotron radiation in the observer s frame is given as @xmath129 ( rybicki , lightman 1979 ) .
combining equations ( 1 ) and ( 2 ) to eliminate b , and normalizing to typical values for grbs , the electron number density is given by @xmath130 in figure 5 , we plot the allowed parameter space for the region size @xmath124 and the electron number density @xmath126 of grb 020813 . here , we have assumed @xmath108 = @xmath112 = @xmath131 hz . equation ( 3 ) is represented by line ( i ) .
then , from the electron energy density , @xmath132 , and the magnetic field energy density , @xmath133 , with @xmath134 , we obtain @xmath135 equations for the cases of @xmath136 0.1 , 1 , and 10 are represented by lines ( ii ) in figure 5 . finally , from the relation @xmath137
, we obtain @xmath138 where @xmath139 is the fraction of the shock energy given to the electrons and @xmath140 is the total explosion energy of grb 020813 , @xmath141 ergs ( 2@xmath0400 kev ) , calculated from the observed fluence and duration of p1 .
equations for the cases of @xmath142 0.001 , 0.1 , and 0.5 are represented by lines ( iii ) in figure 5 .
as can be seen in figure 5 , the most likely parameters for the ssa model would be a region defined by the crossing of three different lines .
these are @xmath143 @xmath144 , @xmath145 cm , @xmath146 , and @xmath147 gauss , assuming @xmath148 and @xmath149 .
using these parameters , the peak flux is @xmath150 ergs @xmath60 s@xmath151 , calculated from @xmath152 $ ] , where @xmath153 is the synchrotron emission coefficient and @xmath154 is the self - absorption coefficient ( e.g. , rybicki , lightman 1979 ) . it is more than 15 orders of magnitude different from the observed value , @xmath155 ergs @xmath60 s@xmath151 . in other words , in order to observe @xmath108 in the hard x - ray band , we have to choose unusual parameters , such as @xmath156 @xmath157 @xmath158 , in contrast to the standard grb fireball model ( e.g. , piran 1999 ) . ) allowed by the ssa model with @xmath159 of @xmath131 hz and @xmath140 of @xmath141 ergs . line ( i )
was derived from the ssa frequency . lines ( ii )
were derived from relation of the energy density of the electrons and the magnetic field . here ,
we used @xmath136 0.1 , 1 , and 10 ( from bottom to top ) . lines ( iii ) were derived from a calculation of the total energy of the electrons . here , we used @xmath160 0.001 , 0.1 , and 0.5 ( from left to right ) .
the crossing point of the three expressions indicated the allowed parameters for the ssa model.,width=302 ] as another possible scenario , we consider the special case that the hard x - ray spectrum is the low - energy end of the inverse compton emission , which could be a `` mirror '' of the synchrotron self - absorption spectrum in the radio band ( e.g. liang et al . 1997 ) .
for this case , the synchrotron emission spectrum is boosted by a factor of @xmath161 via inverse compton scattering of synchrotron photons .
therefore , we assume @xmath162 hz and @xmath163 hz . in figure 6
, we show the allowed values of @xmath126 as a function of @xmath124 for the synchrotron self - compton ( ssc ) model . as in figure 5 , we found parameters of the allowed region : @xmath164 @xmath144 , @xmath165 cm , @xmath166 , and @xmath167 gauss , assuming @xmath148 and @xmath149 .
note that a @xmath168 cm radius is close to the typical value of the `` afterglow '' emission , rather than the prompt grb emission ( e.g. , piran 1999 ) .
we therefore conclude that the ssc model for the hard x - ray spectrum is also unlikely in the standard fireball scenario . ) allowed by the ssc model with @xmath169 of @xmath131 hz , @xmath170 of @xmath131 hz , and @xmath171 of @xmath172 ergs . lines ( i ) , ( ii ) , and ( iii ) represent the same relations as in figure 5 .
the crossing point of the three expressions indicates the allowed parameters for the ssc model.,width=302 ] thus , it is difficult to explain the p1 spectrum of grb 020813 by either the ssa or ssc model .
therefore , we consider other radiation processes , such as `` jitter '' radiation and compton drag effects .
medvedev ( 2000 ) pointed out that the radiation emitted by ultra - relativistic electrons in highly nonuniform , small - scale magnetic fields is different from synchrotron radiation if the electron s transverse deflections in these fields are much smaller than the beaming angle .
he showed that the spectral power distribution of the radiation produced by the power - law distributed electrons is well described by a sharply broken power - law , @xmath173 for @xmath174 and @xmath175 for @xmath176 , where @xmath177 is the electron power - law index and @xmath178 is the jitter break frequency ( medvedev 2000 ) .
therefore , the low - energy depletion ( @xmath4 is larger than the death - line ) with a spectral break may be naturally explained by jitter radiation . in this model ,
the spectral break due to jitter will be observed at @xmath179 where @xmath180 is the relative lorentz factor of two colliding shells , @xmath181 , and @xmath182 is the initial effective thermal lorentz factor , @xmath183 ( medvedev 2000 ) . if @xmath116 is close to 1000 , @xmath112 can be observed in the x - ray band . in the realistic case ,
the emergent spectrum is determined by the statistical properties of the magnetic field in the emission region .
if the magnetic field is highly inhomogeneous , jitter radiation may overcome the synchrotron radiation , whereas the reverse applies for the uniform magnetic field cases .
therefore , it is possible that the x - ray spectrum of grb 020813 ( p1 region ) may be a mixture of jitter radiation with the usual synchrotron radiation component , but detailed modeling of the spectrum is beyond the scope of this paper . as an alternative idea , lazzati et al .
( 1999 ) and ghisellini et al .
( 2000 ) proposed that the gamma - ray photons that characterize the prompt emission of grbs are produced through the compton drag process , caused by the interaction of a relativistic fireball with a very dense soft photon bath .
this is an interesting assumption , because some of the grbs are indeed associated with supernovae , where the expanding star can provide enough soft photons to make the radiative drag effective .
this model accounts for the basic properties of grbs , i.e. , the overall energetics and the peak frequency of the spectrum , with a radiative efficiency of more than 50@xmath12 .
also note that the grb should have a very hard spectrum @xmath184 , as expected from the rayleigh - jeans law below the peak frequency of the emission .
in fact , our spectral shape is in good agreement with the spectra produced by compton drag for @xmath185 ( ghisellini et al .
the major weakness of this model is that the predicted spectrum is basically thermal , and a typical power - law spectrum is not always produced .
thus , we find that both the jitter radiation and the compton drag may reproduce the low - energy depletion observed in grb 020813 . however , we note that a complete theoretical interpretation of the data is beyond the scope of this paper
. there are other theoretical models that can explain the spectra with low - energy depletion .
the average p2 spectrum is well described by the band function with @xmath186 and @xmath187 .
in contrast to p1 , @xmath4 falls in the range allowed by the standard synchrotron shock model ( ssm ) .
as shown by sakamoto et al .
( 2005 ) , integration of the batse and the hete-2 data revealed that the `` typical '' grb spectrum is well represented by a broken power - law ( band ) function with @xmath188 and @xmath189 .
therefore , the observed break energy corresponds to the peak energy ( @xmath190 ) in @xmath191 space , which is where most of the grb power is emitted . for the case of p2 in grb 020813 , however , the spectrum is still @xmath192 up to @xmath193 400 kev . in the standard picture of ssm
, such a break could be explained only when @xmath112 is well above the hete bandpass , which again seems to be quite different from the p1 spectrum .
we stress , however , that it does not seem to be strange even if p1 and p2 show quite different spectral features , since they constitute distinct `` peaks '' both in the light curve and in the hardness ratio ( figure 4 ) .
according to the internal shock scenario ( e.g. , piran 1999 ) , grb light curves are thought to be produced by the collisions of shells that are moving with different lorentz factors .
a fraction of the bulk kinetic energy is converted to the random energy of electrons , which emit the @xmath194 radiation as prompt grb emission . since the radiation from each two - shell collision
would be observed as a single pulse , p1 and p2 may be attributed to the collisions of different pairs of shells , which are possibly taking place in different physical conditions ( e.g. , magnetic field , density of material ) .
a detailed simulation of a collisionless shock plasma predicts that the magnetic field produced in grb shocks randomly fluctuates on a very small scale of the relativistic skin depth , which is @xmath195 cm in internal shocks ( medvedev , loeb 1999 ) .
meanwhile , the emitting ultrarelativisitic electrons have much larger larmor radii . in such situations ,
jitter radiation may overcome the usual synchrotron radiation , but it is also likely that the large scale fields are also present in the grb ejecta .
therefore , the dominance of the `` synchrotron '' and the `` jitter '' radiation would be determined by a combination of large - scale fields ( @xmath196 ) and non - uniform , small - scale magnetic fields ( @xmath197 ) .
we mention in this context that the different observed properties in p1 and p2 may reflect the different balance of large / small - scale magnetic fields , where @xmath198 for p1 and @xmath198 for p1 and @xmath199 for p2 , but further study using more data is necessary to confirm this .
in this paper we reported hete-2 wxm / fregate observations of bright and long grb 020813 .
our observational results demonstrate that the peak p1 ( 17@xmath052 s ) of the burst shows a depletion of low - energy photons below about 50 kev .
therefore , we examined various scenarios to understand the origin of such a low - energy depletion . we conclude that it is difficult to explain the depletion by either the ssa or ssc model .
one possibility is that the low - energy depletion may be understood as a mixture of `` jitter '' radiation with the usual synchrotron radiation component . on the other hand ,
the spectrum of p2 ( 67@xmath0109 s ) can be explained by the standard ssm .
we believe that the magnetic field is uniform and that synchrotron radiation may overcome the jitter radiation in this region .
+ we are grateful to r. yamazaki for giving useful advise concerning our discussions . the hete-2 mission is supported in the us by nasa contract nasw-4690 ; in japan in part by grant - in - aid ( 14079102 and 14gs0211 ) from the ministry of education , culture , sports , science , and technology ; and in france by cnes contract 793 - 01 - 8479 .
k.h . is grateful for support under mit - sc - r-293291 .
atteia , j .- l . , et al .
2003 , in gamma - ray burst and afterglow astronomy 2001 , ed .
g. r. ricker & r. k. vanderspek ( new york : aip ) , 17 band , d. , et al .
1993 , , 413 , 281 barth , a. j. , et al .
2003 , , 584 , l47 covino , s. , et al .
2003 , a&a 404 , l5 dermer , c. d. , b " ottcher , m. , & chiang , j. 2000 , , 537 , 255 fishman , g. j. , meegan , c. a. , wilson , r. b. , horack , j. m. , brock , m. n. , paciesas , w. s. , pendleton , g. n. , & kouveliotou , c. 1992 , in gamma - ray bursts : huntsville , 1991 , ed .
w. s. paciesas & g. j. fishman ( new york : aip ) , 13 fenimore , e. e. , in t zand , j. j. m. , norris , j. p. , bonnell , j. t. , & nemiroff , r. j. 1995 , , 448 , l101 fox , d. w. , blake , c. , & price , p. 2002 , grb coord .
circ . , 1470 frontera , f. , et al .
2000 , , 127 , 59 ghisellini , g. , lazzati , d. , celotti , a. , & rees , m. j. 2000 , , 316 , 45 hurley , k. , et al . 2002a , grb coord .
1482 hurley , k. , et al .
2002b , grb coord .
, 1483 lazzati , d. , ghisellini , g. , celotti , a. , & rees , m. j. 2000 , , 529 , l17 liang , e. , kusunose , m. , smith , i. a. , & crider , a. 1997 , , 479 , l35 link , b. , epstein , r. i. , & priedhorsky , w. c. 1993 , , 408 , l81 medvedev , m. v. 2000 , , 540 , 704 medvedev , m. v. , & loeb , a. 1999 , , 526 , 697 piran , t. 1999 , , 314 , 575 preece , r. d. , briggs , m. s , mallozzi , r. s. , & pendleton , g. n. 1998 , , 506 , l23 preece , r. d. , briggs , m. s , mallozzi , r. s. , pendleton , g. n. , paciesas , w. s. , & band , d. l. 2000 , , 126 , 19 rybicki , g. b. & lightman , a. p. 1979
, radiative processes in astrophysics ( new york : wiley - interscience ) sakamoto , t. , et al . 2005 , , 629 , 311 sari , r. , piran , t. , & narayan , r. 1998 , , 497 , l17 shirasaki , y. , et al . 2003 , , 55 , 1033 tavani , m. 1996 , , 466 , 768 urata , y. , et al
. 2003 , , 595 , l21 vanderspek , r. , marshall , h. l. , ford , p. g. , & ricker , g. r. 2002 , grb coord .
circ . , 1504 | a bright , long gamma - ray burst ( grb ) was detected and localized by the instruments on board the high energy transient explorer 2 satellite ( hete-2 ) at 02:44:19.17 utc ( 9859.17 s ut ) on 2002 august 13 .
the location was reported to the grb coordinates network ( gcn ) about 4 min after the burst . in the prompt emission ,
the burst had a duration of approximately 125 s , and more than four peaks .
we analyzed the time - resolved 2@xmath0400 kev energy spectra of the prompt emission of grb 020813 using the wide field x - ray monitor ( wxm ) and the french gamma telescope ( fregate ) in detail .
we found that the early part of the burst ( 17@xmath052 s after the burst trigger ) shows a depletion of low - energy photons below about 50 kev .
it is difficult to explain the depletion with by either synchrotron self - absorption or comptonization .
one possibility is that the low - energy depletion may be understood as a mixture of `` jitter '' radiation the usual synchrotron radiation component . |
the history of single spin asymmetries ( ssa ) can date back to the 1970s when significant ssa were observed in @xmath0 @xcite . in the early 1990s ,
large asymmetries in @xmath1 were found at fnal @xcite .
however , there were no satisfactory theory to describe the phenomena , and pqcd theory had nothing to do then . in recent years
, ssa phenomenon were also observed in semi - inclusive deep inelastic scattering ( sidis ) processes , which had attracted many interests , particularly in the case where the transverse polarized targets are used .
for example , hermes collaboration @xcite has reported the observation of the azimuthal asymmetries in single - pion productions on both longitudinally and transversely polarized hydrogen targets .
more recently , compass collaboration @xcite has published their results off a transversely polarized deuterium targets as a complementary measurement to hermes experiment . on the theoretical side , by taking into the account of the transverse parton momenta inside the nucleon @xcite , these asymmetries are now assumed to have correlations with the concept of transversity @xcite which we are not familiar with so far . the idea ( though not the term ) of transversity was put forward by ralston and soper via the drell - yan process @xcite , where they introduced the concept of parton transverse polarization .
a clarification of transversity on the role of chiral - odd parton distributions and the general twist was provided by jaffe and ji in ref .
detailed twist clarification for various parton distributions can be found in refs .
now we know that at the leading twist ( twist 2 ) , three fundamental quark distributions provide a complete description of quark momentum and spin in the nucleon . in the last forty years , two of them , the unpolarized and longitudinal polarized parton distributions(@xmath2 and @xmath3 )
have been precisely measured , yet the third type , the `` transversity distribution ( @xmath4 ) '' , is still little known both theoretically and experimentally .
the difficulty in experiments lies in its chiral - odd property . in the helicity basis
, @xmath4 represents a quark helicity flip , which can not occur in any hard process for massless quarks within qed or qcd @xcite .
this chiral - odd property makes it inaccessible in inclusive deep inelastic scattering ( dis ) .
however , several researches have shown that the transversity distribution can manifest itself in semi - inclusive deep inelastic scattering ( sidis ) reactions @xcite , where the transversity distribution function couple with a chiral - odd fragmentation function the collins function @xcite .
this was verified by observing the ssa in hermes and compass experiments recently .
results from both hermes and compass have offered us a first glimpse at transversity distributions on u and d quarks in proton . if studying ssa more explicitly , we would find that these asymmetries can be explained @xcite in terms of both the sivers @xcite and collins @xcite effects .
the sivers effect involves the so called sivers function and the ordinary fragmentation function , while the collins effect involves the transversity distribution function and the collins fragmentation function .
the two competing contributions have different kinematic dependence .
the collins effect depends on _ y _ and strongly correlated with _ z _ , but the sivers effect is independent of _ y _ and not strongly correlated with _
z_. also , they have different azimuthal angle dependence : the sivers effect is proportional to @xmath5 , while the collins effect is proportional to @xmath6 , where the definition of @xmath7 and @xmath8 can be found in ref.@xcite .
thus we can distinguish the two effects in experiment without much difficulty . in our paper
, we are interested in the transversity information provided by ssa , so we only concentrate on the collins effect in this paper . in the collins effect
, the transversity distribution function describes the correlation between the spin of quarks and spin of nucleons , and the so called collins function describes the fragmentation of transversely polarized quarks into unpolarized hadrons .
both functions are important in our numerical calculations and will be discussed in the next section . with this understanding ,
a lot of studies @xcite have attempted to explain the hermes measurement or to give more predictions .
particularly , the article @xcite predicted the collins azimuthal asymmetries ( fig .
[ fig1 ] ) on a transversely polarized hydrogen target in hermes kinematics , yet the prediction is not consistent with the new released hermes results @xcite .
however , as the authors pointed out in the paper that the `` unfavored '' process might lead to a sizable effect , which is in conflict with our general understanding and might result in the inconsistency .
recently , the authors of ref .
@xcite proposed that the polarized `` unfavored '' collins function is approximately equal to the `` favored '' one , but with an opposite sign .
they fitted the hermes data @xcite and obtained new sets of parametrization .
we will shown in this paper that the updated results of ref .
@xcite with these new parametrization of the collins function , are also consistent with the hermes data ( fig .
[ fig2 ] ) .
both calculations imply that the `` unfavored '' collins function may play an important role in the fragmentation processes . at jefferson lab ( jlab ) , experiments e06010 and
e06011 will perform the measurement of single spin asymmetries with a transversely polarized @xmath9he target which is an effectively transversely polarized neutron target , so it may provide a direct measurement of the neutron transversity distributions . in our paper , we will present the predictions of the azimuthal asymmetries in the jlab kinematics region .
the influence due to the `` unfavored '' collins function is included in the prediction .
we conclude that the comparison of the data with the prediction will be able to provide constraints on both the collins functions and transversity distributions , but not a pure measurement of the transversity distributions .
at the leading twist , the differential cross section for a sidis reaction includes the contributions from the unpolarized part , the collins effect and the sivers effect .
it can be written as @xmath10 where the beam is not polarized and the target is transversely polarized . under the factorization assumption , the cross sections in eq .
( [ duu ] ) can be expressed as product of parton distribution and fragmentation functions @xcite .
here we only give the unpolarized and the collins terms as follows , @xmath11 here @xmath12 is the target transverse polarization ; @xmath13 is the charge of the quark with flavor @xmath14 .
@xmath15 is the angle between the hadron lane and the lepton plane .
@xmath16 is the angle between the @xmath17 and the lepton plane . @xmath18 and @xmath19 represent the momentum and transversity distribution functions of the nucleon target respectively . by comparing the cross sections with the opposite polarized target , one obtains the single spin asymmetry @xmath20 after integrating the intrinsic transverse momentum @xmath21 and the kinematical variable @xmath22 and @xmath23 , we can get the formula for calculating the @xmath24 dependence of the collins asymmetry ( denoted as @xmath25 ) @xmath26 @xmath27 is the fragmentation function for an unpolarized quark with flavor @xmath28 into a hadron , defined as @xmath29 @xmath30 is the so called collins function , defined as @xmath31 the kinematical variable @xmath22 can be calculated from @xmath32 with @xmath33=51.8 gev@xmath34 . to get the transversity distribution function , we use two models in this paper , the quark - diquark model @xcite and a pqcd based counting rule analysis @xcite . in the quark diquark model case ,
the transversity distributions are given by @xmath35w_s(x)-\frac{1}{6}d_v(x)w_v(x);\nonumber\\ \delta d_v(x)&=&-\frac{1}{3}d_v(x)w_v(x).\end{aligned}\ ] ] @xmath36 and @xmath37 are the melosh - wigner rotation factors for spectator scalar and vector diquarks , which come from the relativistic effect of quark transversal motions @xcite . for the pqcd based analysis , we adopt the parametrization @xmath38 where the superscripts `` th '' means the theoretical calculation in the pqcd analysis @xcite , and `` para''means the input from parametrization .
the cteq @xcite parametrization is served as the input for both models to get the unpolarized parton distribution functions .
detailed constructions of the quark distributions can be found in refs .
@xcite .
both of the two models have successfully predicted the longitudinal polarized parton distribution functions @xcite .
but two models give different distribution functions at @xmath39 : the pqcd based counting rule analysis @xcite predicts @xmath40 , while the su(6 ) quark - spectator - diquark model @xcite predicts @xmath41 . in a recent literature @xcite ,
the author extracted the transversity distribution for @xmath42 and @xmath43 quarks from the now available data .
they found that @xmath44 and @xmath45 turned out to be opposite in sign , with @xmath46 smaller than @xmath47 .
this seems to be coincidence qualitatively with the su(6 ) quark - diquark model .
but there it predicts @xmath48 when @xmath49 , which is coincidence quantitatively with neither models we used in our paper .
the correctness of different parametrization is still unclear , and need to be checked by more experiments . as to the ordinary fragmentation functions for @xmath50 , we follow the parametrization of @xmath51 given by kretzer , leader and christova @xcite : @xmath52 and for @xmath53 , we have @xcite @xmath54 where @xmath51 denotes the `` favored '' fragmentation function and @xmath55 denotes the `` unfavored '' one . the collins fragmentation function @xmath56 , which describes the transition of a transversely polarized quark into a pion , is theoretically little known and has not yet been measured . in ref .
@xcite , collins suggested a parametrization : @xmath57 where @xmath58 gev . in hermes analysis , @xmath59 gev is taken as a rough estimate @xcite .
the transverse momentum dependence of @xmath60 is assumed to have a gaussian type shape : @xmath61 one can obtain @xmath62 with @xmath63 , and @xmath64 being the mean square momentum that the detected hadron acquires in the quark fragmentation process with @xmath65 @xmath66 according to hermes @xcite .
recently , vogelsang and yuan suggested another model @xcite which assumes a stronger constraint for the pion collins functions : @xmath67 if we consider the isospin and charge symmetry relations between the different fragmentation functions , we can have the following relations , @xmath68 . \label{hu}\end{aligned}\ ] ] substituting ( [ hu ] ) into ( [ h1 ] ) , one obtains @xmath69 which means that the `` unfavored '' collins function is approximately equal to the `` favored '' one , but with an opposite sign .
this is in conflict with almost all the theoretical analysis before , and it still needs to be further checked by experiment . in ref .
@xcite , the parametrization of collins functions is advanced in the forms , @xmath70 the authors of ref .
@xcite fitted the parametrization to the hermes data @xcite and obtained @xmath71 the difference between these two sets of parametrization is that in set i , both favored and unfavored collins functions are parameterized in terms of favored unpolarized quark fragmentation function , while in set ii the unfavored collins function is parameterized in terms of the unfavored unpolarized quark fragmentation function .
also in the literature @xcite , the authors provide their own parametrization to the collins functions , with more parameters compared to the parametrization in ref . @xcite . as the author pointed
that their parametrization is agree with the extractions obtained in ref .
@xcite , we will use the parametrization with less parameters provided in ref .
@xcite in our paper .
this parametrization was obtained from the pion data in ref .
@xcite . in our paper , we assume that this parametrization is also right for the kaon case . in other words , we assume that with this parametrization , the information depending on the final hadron states is only contained in the ordinary fragmentation functions .
figs . [ fig1 ] and [ fig2 ] show our calculated results on hermes measurement as function of @xmath24 , using different sets of parametrization .
first we find that the quark - diquark model and the pqcd based counting rule analysis give almost the same result in the intermediate region of _
x_. fig .
[ fig1 ] presents the results when we use the parametrization as eq .
( [ col ] ) shows .
however , the calculated results are overestimated compared to the experimental data . fig .
[ fig2 ] presents the results when we use two sets of parametrization of collins functions given by ref .
@xcite , which shows that our calculations fit the hermes data well .
as @xmath42 quark dominates in the proton , so @xmath42 to @xmath72 is favored while @xmath42 to @xmath73 is unfavored .
the calculation indicates that both the `` favored '' and `` unfavored '' processes have the sizable effect .
this is the direct consequence of eq .
[ set ] where the favored and unfavored collins functions have the approximately equal size but with the opposite signs .
this is quite different from the ordinary fragmentation functions where the favored process plays much more important roles than the unfavored process .
if we ignore the opposite sign , the numerator of formula ( [ a1 ] ) is nearly of the equal size for both the @xmath72 and @xmath73 productions , but the denominator is larger in the @xmath72 production case than that in the @xmath73 production case .
so we expect a larger asymmetry in the @xmath73 production and this can also be seen from fig.[fig2 ] . besides these
, we also notice that both the two sets of parametrization shown in eq .
( [ set ] ) give similar predictions , so more precise experiments are needed to give the constraints .
the jlab experiments e06010 and e06011 will measure the target single spin asymmetries in the semi - inclusive deep - inelastic reaction off a transversely polarized @xmath9he target for both @xmath50 and @xmath53 . duo to the special structure of @xmath9he , which can be considered as a nearly free polarized neutron , jlab will provide a measure of the neutron structure . applying the same technique ,
we make predictions of azimuthal asymmetries @xmath74 for @xmath50 and @xmath53 productions at jlab kinematics . at jlab ,
the kinematics region is : @xmath75 the azimuthal asymmetry of @xmath9he target can be expressed by the neutron and proton asymmetries : @xmath76 where @xmath77 and @xmath78 are the effective polarizations of the proton and the neutron within the @xmath9he nucleus , and @xmath79 is the polarization of the @xmath9he target which is assumed to be @xmath80 in the experiment .
a three - body fadeev calculation @xcite shows that in inelastic scattering reactions , @xmath81 and @xmath82 .
@xmath83(@xmath84 ) in the above formula represents the ratio of @xmath85 events on neutron ( proton ) over the total @xmath9he@xmath85 events . at jlab ,
one expects : @xmath86 and @xmath87 . for the case of kaon production
, we adopt the same value as a rough estimate .
[ fig3 ] shows the prediction of the asymmetry ( @xmath74 ) on proton target .
compared to fig .
[ fig2 ] , we find that the predictions are almost the same .
[ fig4 ] shows the neutron case , where @xmath43 quark dominates , so the neutron data will be more sensitive to the d quark in comparison with the hydrogen target .
here @xmath43 to @xmath73 is favored and @xmath43 to @xmath72 is unfavored .
again we get the result that the two processes both have the sizable effect , although the asymmetry in @xmath72 production seems not larger than that in the @xmath73 production as we argued in the above section due to the different electric charges between @xmath42 and @xmath43 quarks .
[ fig5 ] is for the @xmath9he target , and we expect a similar prediction as the neutron case shown in fig
. [ fig4 ] , for @xmath9he can be considered as a nearly free polarized neutron .
but the magnitude of the asymmetry is much smaller than the proton result as the hermes experiments showed , because some small coefficients should be multiplied according to eq .
( [ he ] ) .
this is the difficulty measuring the structure of neutron for there is no such a free neutron as a proton .
we expect precise experiments that will be done at jlab can give us some information on neutron . from fig .
[ fig3 ] to fig .
[ fig5 ] , we find that the quark - diquark model and the pqcd based analysis still give similar predictions .
also the two sets of parametrization predict similarly .
so the pion case is insensitive to the models , thus can give constraints on both the collins functions and transversity distributions that are little known yet , rather than give direct information on transversity distributions .
next we present the predictions on @xmath53 productions as fig .
[ fig6 ] shows .
@xmath88 comprises a @xmath42 and an @xmath89 quark , thus the main contribution for the @xmath88 production comes from the valence @xmath42 quark with the favored process and the valence @xmath43 quark with the unfavored process .
this is quite similar to the case of @xmath72 production .
but for @xmath90 , the result is quite different from the others .
@xmath91 comprises a @xmath92 quark and an @xmath33 quark , so the favored process do not contribute for the case , because the transversity is mainly the behavior of valence quarks .
consequently , the asymmetry of @xmath91 production only comes from the unfavored process .
so the kaon production can give us information about the unfavored collins function , and we are looking forward to the measurement . the predictions on kaon productions are shown in fig .
this time , we can see clearly that the quark - diquark model and the pqcd based analysis still give similar results on @xmath88 , but the two different approaches of collins functions seem to have differences .
thus we expect the experiments to give constraints on the collins functions through @xmath88 production . for the @xmath91 case ,
not only the two different approaches of collins functions , but also the two models of distributions predict differently .
so it is concluded that we can get information on collins functions through @xmath88 production , and then distinguish the two models of distributions through @xmath91 production .
however we should notice that the asymmetries for kaon productions are very small , and the differences between the models are also small , approximately the same order as the asymmetries .
since the measurements at jlab can not reach such precision , we have to admit that distinguishing models through experiments is still difficult , and high precision experiments are needed to clarify the details .
the results of jlab on a polarized @xmath9he target can give complementary results to hermes hydrogen measurement .
compared with compass deuterium measurement which is an indirectly measurement of neutron , jlab data will have a unique advantage because of the higher bjorken _ x _ , since the transversity property is mainly a valence behavior .
in addition , we expect that jlab results will give more detailed information on fragmentation functions .
using quark - diqurk model and pqcd based analysis for distributions , and new sets of parametrization of collins functions , we reanalyzed the hermes experiment and found that with the new parametrization of the collins functions advanced by @xcite , the predictions are consistent with the new released hermes data @xcite , which implies that the `` unfavored '' collins functions may play an important role . furthermore , we calculated the azimuthal spin asymmetries of pion and kaon productions in semi - inclusive deep - inelastic scattering of an unpolarized charged lepton beam on a transversely polarized @xmath9he target in the jlab kinematics region . due to the lack of independent measurement of collins functions , it is difficult to obtain transversity distribution functions directly from the measurement on pion productions .
the comparison of the data with the prediction can give constraints on both the collins functions and transversity distributions , but not a direct measurement of the transversity distributions .
using different models for distribution functions and collins functions , we found that the @xmath88 production is sensitive to different sets of collins functions but insensitive to different models of distributions and the @xmath91 production is sensitive to both different approaches of distributions and collins functions .
so we suggest distinguishing the quark - diquark model and pqcd based analysis through this process after the collins functions being constrained by the @xmath88 production .
but due to the small magnitude of the asymmetry , the measurement through this suggestion can only be achieved by experiments with much higher precision compared to the coming jlab experiment .
the @xmath50 productions are sensitive to neither the different approaches of distributions and collins functions , so they might not give much exciting results , but these processes can be useful to give constraints on the results .
additionally , we point here that the asymmetry of @xmath91 production is contributed by pure unfavored processes , thus is an ideal process to study the unfavored collins functions . for the semi - inclusive @xmath72 and @xmath73 productions in deep inelastic scattering of unpolarized charged lepton on a transversity polarized @xmath93 target in the hermes kinematics region , with the the parametrization for the collins function of ( [ col ] ) .
the solid and dashed curves correspond to the calculated results for quark - diquark model and the pqcd based analysis , respectively.,title="fig:",width=302 ] for the semi - inclusive @xmath72 and @xmath73 productions in deep inelastic scattering of unpolarized charged lepton on a transversity polarized @xmath93 target in the hermes kinematics region , with the the parametrization for the collins function of ( [ col ] ) .
the solid and dashed curves correspond to the calculated results for quark - diquark model and the pqcd based analysis , respectively.,title="fig:",width=302 ] for the semi - inclusive @xmath72 and @xmath73 productions in deep inelastic scattering of unpolarized charged lepton on a transversely polarized @xmath93 target in the hermes kinematics region .
we use the parametrization of equations ( [ seti ] ) .
the upper row corresponds to the set i parametrization while the lower row corresponds to the set ii parametrization .
the solid and dashed curves correspond to the calculated results for quark - diquark model and the pqcd based analysis , respectively.,title="fig:",width=302 ] for the semi - inclusive @xmath72 and @xmath73 productions in deep inelastic scattering of unpolarized charged lepton on a transversely polarized @xmath93 target in the hermes kinematics region .
we use the parametrization of equations ( [ seti ] ) .
the upper row corresponds to the set i parametrization while the lower row corresponds to the set ii parametrization .
the solid and dashed curves correspond to the calculated results for quark - diquark model and the pqcd based analysis , respectively.,title="fig:",width=302 ] for the semi - inclusive @xmath72 and @xmath73 productions in deep inelastic scattering of unpolarized charged lepton on a transversely polarized @xmath93 target in the hermes kinematics region .
we use the parametrization of equations ( [ seti ] ) .
the upper row corresponds to the set i parametrization while the lower row corresponds to the set ii parametrization .
the solid and dashed curves correspond to the calculated results for quark - diquark model and the pqcd based analysis , respectively.,title="fig:",width=302 ] for the semi - inclusive @xmath72 and @xmath73 productions in deep inelastic scattering of unpolarized charged lepton on a transversely polarized @xmath93 target in the hermes kinematics region .
we use the parametrization of equations ( [ seti ] ) .
the upper row corresponds to the set i parametrization while the lower row corresponds to the set ii parametrization .
the solid and dashed curves correspond to the calculated results for quark - diquark model and the pqcd based analysis , respectively.,title="fig:",width=302 ] for the semi - inclusive @xmath72 and @xmath73 productions in deep inelastic scattering of unpolarized charged lepton on a transversely polarized @xmath93 target with the jlab kinematics cut .
we use the parametrization of equations ( [ seti ] ) .
the upper row corresponds to the set i parametrization while the lower row corresponds to the set ii parametrization .
the solid and dashed curves correspond to the calculated results for quark - diquark model and the pqcd based analysis , respectively.,title="fig:",width=302 ] for the semi - inclusive @xmath72 and @xmath73 productions in deep inelastic scattering of unpolarized charged lepton on a transversely polarized @xmath93 target with the jlab kinematics cut .
we use the parametrization of equations ( [ seti ] ) .
the upper row corresponds to the set i parametrization while the lower row corresponds to the set ii parametrization .
the solid and dashed curves correspond to the calculated results for quark - diquark model and the pqcd based analysis , respectively.,title="fig:",width=302 ] + for the semi - inclusive @xmath72 and @xmath73 productions in deep inelastic scattering of unpolarized charged lepton on a transversely polarized @xmath93 target with the jlab kinematics cut .
we use the parametrization of equations ( [ seti ] ) .
the upper row corresponds to the set i parametrization while the lower row corresponds to the set ii parametrization .
the solid and dashed curves correspond to the calculated results for quark - diquark model and the pqcd based analysis , respectively.,title="fig:",width=302 ] for the semi - inclusive @xmath72 and @xmath73 productions in deep inelastic scattering of unpolarized charged lepton on a transversely polarized @xmath93 target with the jlab kinematics cut .
we use the parametrization of equations ( [ seti ] ) .
the upper row corresponds to the set i parametrization while the lower row corresponds to the set ii parametrization .
the solid and dashed curves correspond to the calculated results for quark - diquark model and the pqcd based analysis , respectively.,title="fig:",width=302 ] , but for the @xmath94 target.,title="fig:",width=302 ] , but for the @xmath94 target.,title="fig:",width=302 ] + , but for the @xmath94 target.,title="fig:",width=302 ] , but for the @xmath94 target.,title="fig:",width=302 ] , but for the @xmath9he target.,title="fig:",width=302 ] , but for the @xmath9he target.,title="fig:",width=302 ] + , but for the @xmath9he target.,title="fig:",width=302 ] , but for the @xmath9he target.,title="fig:",width=302 ] , but for the @xmath88 and @xmath91 productions.,title="fig:",width=302 ] , but for the @xmath88 and @xmath91 productions.,title="fig:",width=302 ] + , but for the @xmath88 and @xmath91 productions.,title="fig:",width=302 ] , but for the @xmath88 and @xmath91 productions.,title="fig:",width=302 ]
we are grateful to xiaodong jiang , haiyan gao , and xin qian for useful discussions .
this work is partially supported by national natural science foundation of china ( nos . 10421503 , 10575003 , 10528510 ) , by the key grant project of chinese ministry of education ( no .
305001 ) , by the research fund for the doctoral program of higher education ( china ) . 99
g. bunce _ et al . _ ,
phys . rev . lett .
36 , 1113 ( 1976 ) .
et al . _ , [ compass collaboration ] , phys .
lett . * 94 * , 202002 ( 2005 ) ; e.s .
et al_. [ compass collaboration ] , hep - ex/0610068 ; anna martin ( on behalf of the compass collaboration ) , hep - ex/0702002 .
f.e . close , phys .
43 b * , 422 ( 1973 ) ; nucl .
* b 80 * , 269 ( 1974 ) ; r. carlitz , phys . lett .
* b 58 * , 345 ( 1975 ) ; j. kaur , nucl . phys . *
b 128 * , 219 ( 1977 ) ; a. schfer , phys . lett . * b 208 * , 175 ( 1988 ) ; f.e . close and a.w .
thomas , _ ibid _ * b 212 * , 227 ( 1988 ) ; n. isgur , phys . rev . *
d 59 * , 034013 ( 1999 ) . | we use a new set of collins functions to update a previous prediction on the azimuthal asymmetries of pion productions in semi - inclusive deep inelastic scattering ( sidis ) process on a transversely polarized nucleon target .
we find that the calculated results can give a good explanation to the hermes experiment with the new parametrization , and this can enrich our knowledge of the fragmentation process .
furthermore , with two different approaches of distribution and fragmentation functions , we present a prediction on the azimuthal asymmetries of pion and kaon productions at the kinematics region of the experiments e06010 and e06011 planned at jefferson lab ( jlab ) .
it is shown that the results are insensitive to the models for the pion case . however , the results for kaon production are sensitive to different approaches of distribution and fragmentation functions .
this is helpful to clarify some points in the study of the azimuthal spin asymmetries and fragmentation functions in hadronization processes . |
recent advances in optical microscopy @xcite have inspired numerous analysis methods aiming to quantify the motion of individual molecules in live cells @xcite .
the resolution afforded by current optical microscopes allows researchers to more reliably measure two - dimensional ( 2d ) @xcite and three - dimensional ( 3d ) @xcite position vs. time data in single particle tracking ( spt ) experiments .
this permits researchers to probe _ in vivo _ forces without introducing external perturbations into the system .
techniques capable of reliably quantifying the _ in vivo _ forces experienced by single - molecules ( without ensemble averaging ) offer the potential to gain new molecular - level understanding of various complex biological processes including cell division @xcite , virus assembly @xcite , endocytosis @xcite and drug delivery @xcite . in this article
, we demonstrate how the hierarchical dirichlet process switching linear dynamical system ( hdp - slds ) framework developed by fox and co - workers @xcite can be used to deduce the direction and magnitude of different forces that contribute to molecular motion in living cells @xcite .
the utility of combining the hdp - slds with spt was motivated by experiments aiming to quantify the time varying forces driving chromosome dynamics .
the approach presented shows promise in both ( i ) accelerating the scientific discovery process ( i.e. , statistically significant changes in dynamics can be reliably detected ) and ( ii ) automating preprocessing tasks required when analyzing and segmenting large spt data sets .
the technique introduced is applicable to various scenarios where spt trajectories are sampled frequently in time and particles can be accurately tracked over multiple frames , e.g. @xcite .
extracting accurate and reliable force estimates from noisy position vs. time data in the aforementioned setting requires one to account for numerous complications inherent to experimental spt data in living cells .
for example , nonlinear and/or time changing systematic forces need to be differentiated from thermal fluctuations ( i.e. , random forces ) , both of which contribute to motion at the length and time scales measurable in living systems @xcite .
furthermore , additional measurement noise ( consisting of localization error amongst other factors @xcite ) induced by the optical measurement apparatus must be systematically accounted for since this noise source varies substantially between and within single trajectories ; inaccurate effective measurement noise estimates can appreciably influence estimates of kinetic parameters as well as statistical decisions about the underlying physical system @xcite .
finally , extracting forces from position vs. time data requires one to explicitly or implicitly make numerous assumptions about the underlying effective dynamics .
we believe these assumptions should be systematically tested directly against experimental data before one trusts kinetic quantities inferred from experimental data @xcite .
however , in live cell spt studies , reference `` ground - truth '' is rarely available .
hence , techniques checking statistical assumptions directly against data are attractive ( e.g. , through goodness - of - fit hypothesis testing @xcite ) .
the hdp - slds approach combines hidden markov modeling ( hmm ) , kalman filtering , @xcite , and more recent ideas from dirichlet process modeling @xcite .
we demonstrate how the hdp - slds method can be used to reliably identify the time at which a state change occurs as well as the number of states implied by a specific time series . in the hdp - slds approach @xcite ,
the number of underlying states are inferred from the data via nonparametric bayesian techniques @xcite .
inferring the number of states jointly with the parameters determining the dynamics ( i.e. , in a single fully bayesian computation ) is useful because the number of underlying effective states is rarely known _ a priori _ in live cell spt applications due to inherent heterogeneity between and within trajectories @xcite .
since the hdp - slds framework directly infers the number of states from observed data , the user does not need to provide an accurate upper bound on the number of states or worry about _ a posteriori _ model selection issues @xcite .
this allows the hdp - slds approach to readily identify a wide range of distinct dynamical regimes which may occur within a single trajectory .
once the experimental trajectory is segmented into distinct kinetic states , one can use classic maximum likelihood estimation ( mle ) techniques to infer kinetic parameters @xcite . carrying out
mle estimation after segmentation mitigates the bias induced by prior assumptions .
it is emphasized throughout that the procedure described in this work can systematically account for measurement noise and complex spatial - temporal variations in thermal fluctuations & force / velocity fields . in the particular application that we studied ,
the heuristics suggested for generating hdp - slds prior parameter input required modification before accurate results could be obtained in spt applications .
specifically , prior specification heuristics employed by the original hdp - slds bayesian analysis ( using guidelines outlined in ref .
@xcite ) required substantial modification ( prior specification substantially influences both state inference and molecular motion parameter estimation ) .
in particular , the hyperparameters " governing the measurement and thermal noise ( parameterizing components of the `` base - measure '' @xcite ) required more careful calibration / tuning before high accuracy results could be obtained . in s1 text , we discuss how ideas in refs .
@xcite can be used for this type of hyperparameter tuning .
we note that the experimental applications are meant to serve as a proof - of - concept .
the results presented are not intended to provide an exhaustive study of the forces involved in chromosome movement in metaphase .
the problem is amenable to the hdp - slds approach since the gfp tagged experiments experienced abrupt state changes in the experimental trajectories before other downstream spt computations could be reliably carried out ( e.g. , force and diffusion coefficient estimation ) .
not all `` abrupt '' were visually obvious to human observers , hence the approach is promising in both data - mining and preprocessing contexts .
it should also be noted that two recent articles have discussed the utility of nonparametric bayesian ideas in single - molecule data analysis @xcite ; ref .
@xcite is similar in spirit to this manuscript except the focus is exclusively on simulation data mimicking features of spt data and ref .
@xcite focuses on using nonparametric bayesian modeling to detect the number of states in hmm models .
note that in ref .
@xcite each observation is assumed to be drawn independently from a distribution indexed by a latent state and the approach employed does not assume explicit temporal dependence between successive spt time series observations whereas this work accounts for such temporal dependence ( allowing quantitative approximations of spatially or temporally dependent effective forces and/or velocities ) .
this article is organized as follows : the mathematical models and experimental methods used are presented in sec .
[ sec : meth ] ; results obtained when analyzing chromatid dynamics in yeast are presented in sec . [
sec : expres ] .
conclusions are presented in sec .
[ sec : discussion ] . we have also included supporting material ( s1 and s5 texts ) where additional technical details are provided and control simulation studies
are analyzed ( results shown in s1-s7 figs . ) .
a companion publication @xcite compares the hdp - slds approach to an open - source spt hidden markov model assuming a finite number of states through simulation studies .
we begin by reviewing the main technical ideas underlying the hdp - slds introduced in ref .
note that the slds models assumed allow researchers to decouple noise induced by thermal fluctuations , measurement noise , and spatially anisotropic velocity / forces .
forces are computed by leveraging the overdamped langevin model @xcite ( this relationship is made explicit in sec .
[ sec : oddesc ] and s3 text ) . in the early spt works , the spatial and temporal resolution afforded by the measurement device led researchers to focus mainly on mean - square - displacement ( msd ) type analyses to analyze single - molecule data @xcite .
msd approaches have many undesirable features , namely they tend to introduce unnecessary temporal averaging ( i.e. , they ignore the natural time ordering of the trajectory measurements ) and they have a difficult time accounting for spatially varying forces ( a common occurrence in live cells @xcite ) .
advances in spatial and temporal resolution have inspired many researchers to develop new techniques aiming at more reliably extracting single - molecule level information out of measurements @xcite .
the previously cited works are most similar in spirit to the work presented , but all of the works encounter technical difficulties when the number of `` states '' is not known in advance .
additional technical complications arise when position estimates are obscured by non - negligible measurement noise " @xcite or force / velocity fields exhibiting temporal dependence @xcite .
the method of fox et al .
@xcite overcomes the difficulties mentioned above by assuming that a discrete time series model of the form : @xmath0 can be used to describe the dynamics of each unique state .
the position of the molecule or particle at time @xmath1 is denoted by the vector @xmath2 and the measured value of the position at this same time is denoted by @xmath3 ( subscripts are used to index time ) ; the position is not directly measurable due to `` localization noise '' and other artifacts induced by the experimental apparatus .
the term `` effective measurement noise '' is meant to include the net measurement noise induced by finite photon counts , background fluorescence , motion blur , etc . @xcite .
effective measurement noise is modeled as a mean zero normal random variable with covariance @xmath4 ; the expression @xmath5 conveys that the random vector , @xmath6 , is distributed according to the normal distribution @xmath7 .
techniques for checking the validity of this modeling assumption via hypothesis testing technique are discussed in ref .
@xcite .
the term @xmath8 represents a ( constant ) velocity vector " experienced by the particle multiplied by the observation time @xmath9 ; the matrix @xmath10 accounts for systematic spatial variations in forces or velocity ; random thermal fluctuations are modeled by @xmath11 .
the `` drift terms '' @xcite , i.e. @xmath8 and @xmath12 , can be used to quantify active ( linear ) force and velocity fields ; spatial anisotropy in the force and velocity fields are accounted for by @xmath12 .
the physical assumptions we employ for inferring forces from position vs. time data are discussed in sec .
[ sec : oddesc ] .
the net parameter vector characterizing the ( discrete ) dynamics of a single state is given by @xmath13 .
note that the hdp - slds presented in ref .
@xcite assumed that all observations are uniformly spaced by @xmath9 time units .
although the molecular position , @xmath14 , is not directly observable , the discrete state - space model above allows one to readily plug directly into the established kalman filtering equations @xcite to infer the dynamics of @xmath14 using the observations @xmath15 .
the kalman filtering framework allows one to systematically treat measurement noise as well as spatially dependent particle forces and/or velocities ( this spatial variation can induce classic `` confinement '' or `` corralling '' effects or be indicative of molecular binding in spt data @xcite ) .
the importance of properly accounting for measurement noise is demonstrated in s2 text and is also presented elsewhere @xcite
. an illustrative trajectory of the ( observable ) measurements @xmath15 and ( unobservable ) @xmath14 is shown in fig .
[ fig : illeg ] ; this figure will also be used to also illustrate the `` latent state '' modeling discussed in the next paragraph .
note that the presence of @xmath10 slightly complicates physical interpretation of the parameters @xcite , but the discrete model above can be readily mapped to a continuous time stochastic differential equation ( sde ) studied in ref .
@xcite ( where diffusion coefficients , effective friction , and instantaneous force terms associated with an overdamped langevin equation can be readily extracted ) .
the equations mapping discrete time parameters used in the hdp - slds framework to continuous time sde parameters are presented in s3 text for the reader s convenience .
the ability to map between discrete and continuous time models is important for both specifying physically inspired priors and for interpreting parameter estimates since classical quantities like diffusion coefficients and forces are typically defined via continuous time sde models @xcite . in the discrete time model above
, @xmath16 contains the parameters ( vectors and matrices ) required to specify the stochastic dynamics defining the evolution equations of the system in a single `` state '' . however , simple linear evolution equations are not expected to be valid for the entire duration of the trajectory in live cell measurements @xcite . in this article
, we assume that @xmath16 can change abruptly over time or space and hence the system s state " can change over time .
we will use the notation @xmath17 to denote the state at time @xmath18 and will use @xmath19 to denote the parameter vector characterizing the dynamics of @xmath14 when it is in state @xmath20 .
the standard goal of an hmm inference procedure is to infer the state sequence @xmath21 from an observation sequence @xmath22 .
the typical hmm framework assumes temporal transitions between the states are governed by markovian transition probabilities . if there are @xmath23 fixed states , the transition vector associated with state @xmath18 in the traditional hmm framework is prescribed by the vector @xmath24 .
note that each component provides the probability of the state transition from state @xmath18 to state @xmath25 and the sum of components of this @xmath23 dimensional vector is one ( hence defines a proper discrete probability distribution ) ; the collection @xmath26 defines the classical hmm `` transition matrix '' . in bayesian inference of standard hmms
, the dirichlet distribution is sometimes used as a prior for @xmath27 @xcite .
additional details on the dirichlet distribution and various infinite dimensional extensions , namely the dirichlet process and the hierarchical dirichlet process , are provided in s4 text . the hdp - slds framework of fox et al .
comprehensively addresses issues not accounted for in refs . @xcite . the following modified quote from ref .
@xcite captures the essence of the hdp - slds method : _ the hdp - slds is an [ infinite discrete state space ] extension of hidden markov models ( hmms ) in which each hmm state , or mode , is associated with a linear dynamical process . " _ as in ref .
@xcite , the statistical influence of thermal fluctuations and effective measurement noise on the dynamics is accounted for using the kalman filter framework ( this allows linear spatial variations in force and/or velocity ) ; however , the hdp - slds also provides a mechanism for segmenting trajectories into chunks where dynamics are distinct .
in addition , the hdp - slds framework advocated in ref .
@xcite introduces a `` sticky parameter '' encouraging temporal state persistence ; this feature can be advantageous in many spt applications @xcite .
the ability to avoid model selection nuances @xcite and infer the number of states in a data - driven fashion within a single bayesian inference while also rigorously accounting for spatially or temporally varying stochastic dynamics are major advantages of the hdp - slds method . before discussing the major technical weaknesses of the hdp - slds method
, we discuss some strengths / weaknesses of published spt modeling approaches .
wavelet methods have the ability to identify sharp and abrupt changes in time without making too many assumptions about the underlying stochastic process @xcite , but they do not readily account for subtle spatial variations in the dynamics or noise ( e.g. , temporally varying velocity fields or diffusion coefficients ) and can have difficulty in separating diffusive noise from measurement noise .
the hdp - slds overcomes the aforementioned challenges by assuming a specific parametric model ( however , if anomalous diffusive noise is deemed important to quantify system dynamics @xcite , this can be a problem for the hdp - slds approach ) .
@xcite is one of the pioneering efforts attempting to account for spatial and temporal variations in spt signals , but the approach neglected to explicitly account for the effects of measurement noise ( the approach also focused on comparative hypothesis tests and model selection as opposed to goodness - of - fit tests directly checking the consistency of a model s distributional assumptions to experimental data ) .
the method in ref .
@xcite attempts to account for spatial variations ( allowing for correlated 2d forces ) and the influence of measurement noise , but appeals to _ ad hoc _ statistical approximations which can adversely affect state and parameter inference @xcite ; these approximations ( which can be avoided using kalman filtering ideas @xcite ) can substantially complicate downstream analysis where one would like to check the assumptions against experimental data @xcite ( also ref .
@xcite did not employ any formal hypothesis testing procedures ) .
the windowed local mle " method in refs .
@xcite can account for spatial variations in force and measurement noise @xcite and provides a procedures for goodness - of - fit testing , but does not prescribe a systematic and automated method for segmenting time series data into distinct dynamical segments .
open problems facing the hdp - slds are associated with prior specification and model validation @xcite .
the priors used are selected primarily for computational convenience .
for example , the so - called matrix normal inverse - wishart ( mniw ) and other priors proposed allow for exact markov chain monte carlo ( mcmc ) sampling @xcite , but such priors have no real connection to the physical mechanisms governing molecular motion .
the priors and their associated hyperparameters used by the hdp - slds method are discussed further in s5 text . in the results ,
we demonstrate how a variant of the method in ref .
@xcite can be used to correct for some artifacts introduced by `` bad priors '' ( i.e. , the parametric prior distributions assumed do not accurately reflect the true underlying process ) .
quantitatively prescribing a `` good prior '' is difficult because single - molecules trajectories contain a high degree of heterogeneity induced by the local micro - environment , conformational fluctuations of the biomolecules , etc . hence finding an accurate prior representative of a single trajectory is non - trivial ( finding a prior governing a population of trajectories in the spirit of ref .
@xcite is even harder in slds modeling ) .
fortunately , the nonparametric bayesian hdp - slds from ref . @xcite combined with frequentist ideas @xcite provides researchers in spt analysis a set of tools which can be used to more reliably and systematically extract information from complex _ in vivo _ measurements as we demonstrate in the results and supporting text . in what follows ,
we present a technical discussion explicitly pointing out how we infer forces from a sequence of position measurements .
particle position is modeled using the overdamped limit of the langevin equation ( i.e. , a first order sde ) .
the general form of the corresponding first order sde is : @xmath28 where @xmath29 is a local effective friction matrix " associated with the langevin model and @xmath30 represents the force vector @xcite ; within the slds the force is a function of both the position @xmath14 and the latent state .
the notation above was selected to facilitate comparison to refs . @xcite and @xcite .
the above model assumes particle inertia can be ignored ( i.e. , the dynamics are in the overdamped " regime or the smoluchowski limit of the full langevin equation @xcite ) and local diffusivity is constant . in text
s3 , the mapping between the continuous and discrete hdp - slds time series models is provided ( note that in s3 text , @xmath31 ) .
on the time and length scale of spt measurements , it is generally accepted that particle inertia is unimportant .
however , goodness - of - fit tests can be used to check if unobserved and unmodeled momentum dynamics affect position measurements @xcite .
if inertia is believed or empirically determined to be important , more complex ( second order sde ) models for fitting forces from position data can be entertained @xcite .
the subtler assumption implicit in our estimate of the forces from the above model is that the einstein relationship can be used to relate local effective friction to the local effective diffusion coefficient via the relationship : @xmath32 the above relationship is a classic microscopic relationship relating molecular diffusivity to local friction @xcite . on longer time scales ,
the microscopic parameters become effective parameters @xcite and the validity of the above relationship can become questionable due to phenomena like molecular crowding @xcite or other factors ( e.g. , transient binding ) inducing more complex dynamics . the temporal segmentation approach proposed aims to detect when a sudden event alters the dynamics in a statistically significant fashion , hence the odds of mixing " different dynamical regimes is reduced ( e.g. , if the obstacle density experienced in one part of the cell is vastly different than another , then the effective dynamics in each regime will be different @xcite ) .
the hdp - slds s ability to identify regimes , within a single trajectory , where locally linear sde models are consistent with the data makes the validity of eq .
[ eq : fd ] more plausible .
it should be noted that effective force " data has been backed out from noisy position measurements in multiple _ in vitro _
single - molecule manipulation experiments @xcite . in the aforementioned works , force measurements obtained via an external probe ( e.g. , afm ) could be compared to internal forces inferred from noisy position vs. time data ( more precisely molecular extension vs. time ) obtained by appealing to relationships similar to those in eqns .
[ eq : odg ] and [ eq : fd ] . despite the system
s being far out of equilibrium in refs .
@xcite , appealing to the relationship in eq .
[ eq : fd ] yielded effective `` internal forces '' consistent with those measured by external probes .
however , testing the validity of the above expression directly against _ in vivo _ spt data is non - trivial since directly measuring fores in the crowded and heterogeneous environment of live cells is difficult .
the formal statistical hypothesis testing procedure employed can only judge the consistency of the assumed drift and diffusion terms of the sde in eq .
[ eq : odg ] against observational data ; said differently , only the product @xmath33 affects the test ( the validity of eq .
[ eq : fd ] does not influence the hypothesis test in any way ) .
as we will discuss later , our force estimates obtained ( without assuming or inputting a local effective cellular viscosity ) are consistent with other _ in vivo _ force measurements @xcite . even if a researcher does not believe eq .
[ eq : fd ] should hold , our fitted models can still produce local effective velocity fields @xcite which can aid in our quantitative understanding of intracellular dynamics . before proceeding
, we explicitly mention a few more technical notes about our nominal force estimates .
our models make no assumptions about the existence of a gradient potential field @xcite or the stationarity ( temporal or spatial ) of statistics characterizing the random process @xcite .
transient structures and phenomena , which are often resolvable in live cell spt measurements @xcite , can invalidate the validity of a force obtained by taking the gradient of an energy potential field and can also complicate nonparametric estimators @xcite ( nonparametric estimators often make implicit stationarity assumptions ) .
our models sample data on millisecond timescales , but allow for the local effective forces and diffusion to change over time and/or space ( i.e. , statistics are allowed to be non - stationary ) ; we believe this feature will be critical in accurately characterizing heterogeneous high resolution optical microscopy data .
we examined the _ in vivo _ dynamics of chromatin during mitosis to determine the behavior of a region of the chromosome visualized through the binding of lac repressor fused to gfp ( laci - gfp ) to lac operator ( laco ) .
the laco is a repeated array of operator dna sequences ( 256 repeats , 10 kilobase prs . )
integrated 6.8 kb from the centromere on chromosome xv .
the spindle pole bodies ( sites of microtubulare nucleation ) are visualized through a fusion protein between a spindle pole component ( spc29 ) and rfp ( red fluorescent protein ) .
cells were grown to logarithmic phase at 24 @xmath34c in rich media .
images were acquired on a nikon eclipse ti wide - field inverted microscope with a 100x apo tirf 1.49 na objective ( nikon , melville , new york , usa ) and andor clara ccd camera ( andor , south windsor , connecticut , usa ) using continuous laser illumination .
images were streamed at the net effective camera acquisition rate of 22 frames / sec .
the ccd camera s exposure time was set nominally to be the inverse of the net frame rate .
images were acquired at room temperature with nikon nis elements imaging software ( nikon , melville , new york , usa ) . the program `` speckle tracker '' and other methods outlined in ref .
@xcite were used to estimate the centroid of gfp and rfp spots for position measurements .
in fig . [ fig : fhist ] we display a histogram of the estimated instantaneous force magnitudes as well as the force magnitude trajectory computed from a time series of measured @xmath35 chromatid positions ( the @xmath35 data and state estimates are shown in fig .
[ fig : noisejump ] ) .
the force was estimated by first determining the number of states implied by the observed trajectory and using the method described in s1 text to segment the trajectory .
for the experimental trajectories shown in this paper , the initial prior mean of the measurement noise covariance matrix , @xmath4 , was assumed to be the identity matrix multiplied by a scalar , @xmath36 , where @xmath37 .
this value was inspired by the fact that the effective measurement noise standard deviation was found to be in the @xmath38 range for these spt trajectories ( with a mode at @xmath39 ) .
we specified @xmath40 for the so - called weak limit approximation " of the hdp - slds @xcite ; this term is discussed further in s5 text and the relative insensitivity of the hdp - slds method to the @xmath23 parameter is demonstrated in fig s5 .
additional parameters required to run the hdp - slds segmentation are reported in s5 text .
after the hdp - slds segmentation was obtained , the estimated mle parameter vector of each unique state was used to obtain molecular position estimates via the kalman filter @xcite .
finally , the position estimates along with the mle parameters were plugged into the equations shown in s3 text to evaluate the instantaneous effective force ( note : this provides a collection of 2d force vectors ) .
the force vector magnitudes reported in fig .
[ fig : fhist ] observed are representative of other chromatid data sets studied using this analysis
. however , some datasets exhibited more interesting `` force regime changes '' ( as shown and discussed in the final experimental spt trajectory studied ) .
two things should be emphasized in the analysis of this trajectory : ( i ) previously published works have reported that chromatids in metaphase - like conditions experience forces in the ( relatively low ) 0.1 - 0.2 pn range @xcite .
our findings are consistent with these previous results , except the model considered here does not require an estimate of the effective viscosity ( a quantity difficult to estimate in live cells ) to infer force vectors .
local effective forces are estimated using time series analysis techniques applied to the so - called overdamped langevin equation outlined in sec .
[ sec : oddesc ] .
( ii ) the hdp - slds method aided in accurately determining a transition between two states ( physical interpretation of the identified states discussed below ) . the top left panel of fig .
[ fig : noisejump ] displays the white light image of a trajectory obtained from the yeast chromatid spt experiments .
note that we use the phrase white light image " throughout to indicate a single diffraction limited image of the yeast cell obtained with white light illumination ; the trajectory obtained using a laser frequency tuned to enhance gfp excitation was then overlaid upon this single image ( all images are recorded at different times , but the yeast cell is not expected to move substantially during the experiment , so the white light illumination image gives one an idea of the spatial environment explored by the molecule ) .
the top right panel displays the trajectory in @xmath35 space with time information color coded . the red vertical line in the bottom left panel of fig .
[ fig : noisejump ] shows the time at which a state change point was detected by the hdp - slds method .
the bottom right panel of fig .
[ fig : noisejump ] shows that state estimates of the vbspt @xcite and hdp - slds methods both identified two states , though the latter captures the state persistence more accurately .
after the temporal segmentation was formally estimated using the hdp - slds , we computed the posterior mode of @xmath16 implied by the assumed hdp - slds model for each of the states in the two identified temporal segments .
these two parameter vectors ( one vector drawn for each of the two states ) were used along with the experimental data to compute the @xmath41 test statistic @xcite and the corresponding @xmath42values .
the null hypothesis of the one - sided goodness - of - fit test using the @xmath41 statistic assumes that the time series was generated by a markovian sde ( evaluated at an `` optimal '' parameter estimate @xcite ) and the alternative hypothesis assumes that the time series was produced by any other process @xcite .
`` optimal '' parameter estimates were obtained by two means : ( i ) drawing the parameter vector obtained after @xmath43 iterations of the mcmc sampler in the hdp - slds inference ; note that this approach included prior bias and ( ii ) via mle estimation applied to hdp - slds segmented trajectories ; with this approach , bias is mitigated since priors do not influence the parameter estimates . in the first segment , using the observed data and the hdp - slds parameter vector estimate mentioned above , resulted in a @xmath42value of 0.32 ( little evidence for rejection ) , but in the second segment a @xmath42value @xmath44 0.01 was computed ( large evidence for rejection ) .
visual inspection of data s1 ( see rightmost spot ) shows the effective measurement noise has abruptly changed , however the prior assumed by the hdp - slds segmentation only allowed the components of @xmath4 to change modestly .
the default settings of the matlab code associated with ref .
@xcite assumes little dispersion about the nominally known mean @xmath4 ( all segments yielded an effective measurement noise having @xmath45 on all diagonal components via the hdp - slds method ) . using a so - called `` non - informative '' or `` uniform '' prior can potentially remedy the situation , but such a prior can introduce both technical and computational problems in nonparametric bayesian methods @xcite ; see s5 text for further discussions on this issue . alternatively , if one uses the segmentation afforded by the hdp - slds , but instead computes the mle using a variant of the technique of ref .
@xcite ( discussed in s1 text ) as opposed to using the posterior sample provided by the hdp - slds ( which contains biases induced by priors ) , the @xmath42values were found to be @xmath46 and @xmath47 for the left and right segments , respectively .
the primary difference in the estimation results was in the noise components estimated , @xmath4 and @xmath48 , the effective measurement noise and diffusion coefficient matrices , respectively ( see s3 text ) .
for example , the mles @xmath49)$ ] [ @xmath50 & @xmath51)$ ] [ @xmath52 were obtained in the left segment and @xmath53)$ ] [ @xmath50 & @xmath54)$ ] [ @xmath52 were obtained in the right segment ( i.e. , the segment occurring later in time ) using the model in ref .
hats are used to denote mle estimates and @xmath55 denotes the square diagonal matrix formed by the arguments .
note that the diffusion coefficient estimate , @xmath56 , is near machine single - precision zero in the right segment .
the gfp image stack associated with this video readily shows photobleaching occurs ( see rightmost point spread function spot in s1 data ) .
the mle predicted effectively zero diffusion for the right segment ( i.e. , the state appearing temporally after the first segment ) , quantitatively indicating that significant photobleaching had occurred and that the second state was mainly background photon noise not corresponding to an individual molecule .
the quantitative evidence for this statement is the mle parameter estimates ( effective measurement noise near diffraction limit in the `` right segment '' ) and the corresponding @xmath42values , reported in two paragraphs above , resulting from testing the fitted model against the data . in this experiment ,
the hdp - slds approach was able to automatically detect this measurement noise transition despite a poorly tuned prior ; the state change was identified without requiring subject matter expertise or manual image inspection .
[ fig : twostate ] displays another experimental trajectory with a different type of state switching which was readily identified by the hdp - slds procedure .
the posterior mode computed using gibbs sampling of the hdp - slds model contained assignments with two and three states ; in total @xmath57 posterior mcmc samples were sampled .
a majority of the mcmc draws implied two - state switching .
however , one does not necessarily need to be concerned about the number of states if one is only interested in determine change points .
the change points predicted by the hdp - slds are denoted by vertical lines ( change points were determined to occur if a state change occurred at time @xmath18 in over 25% of the posterior samples ) .
accurately identifying change points can help in quantifying the transition times between distinct physical states @xcite .
however the state switching in this particular system was likely caused by vibrations in the microscope s piezo stage .
the hdp - slds can be used to automatically identify physically relevant state changes or state changes induced by experimental artifacts . as we show in fig .
s5 , isolated outliers ( e.g. , caused by intense fluorescent background `` flashes '' in some spt experiments ) are also readily identified by the hdp - slds framework .
next we turn to an example exhibiting subtle state changes ( identified using the procedure outlined in s1 text ) in a more biologically relevant study . fig .
[ fig : fhist1 ] shows two gfp tagged chromatid trajectories during the metaphase stage of mitosis . the white light image of these trajectories is shown in fig .
[ fig : mitosis ] where the spatial location of the rfp tagged mitotic spindle pole bodies ( spb ) as well as the gfp tagged chromatid trajectories are shown .
hdp - slds state segmentation was carried out for each of the trajectories ; the state segmentation is denoted by vertical lines in the bottom panel of plots in fig .
[ fig : fhist1 ] .
the procedure used to infer forces in fig .
[ fig : fhist ] was carried out again for the pair of chromatids undergoing mitosis .
the histogram of the force magnitudes is not atypical for trajectories observed , however the evolution of the force vectors was suggestive of a dynamic regime shift . in order to probe this regime shift further , we computed the eigen - values and eigen - vectors of the mle of @xmath58 ( a matrix associated with the continuous - time overdamped langevin model ; details provided in s3 text ) .
the eigen - analysis shown in fig .
[ fig : mitosis ] provides quantitative information about variations in the effective force vector as a function of position ( i.e. , the eigen - analysis provides both magnitude and directional information about a linear vector field ) .
the upper - right panel shows that the largest magnitude eigen - value consistently decreased over time for each state identified by the hdp - slds segmentation .
the rate of decrease was similar for the pair of sister chromatids .
this phenomenon was originally identified using the crude time window segmentation of ref .
@xcite , but the hdp - slds provides a framework for more systematically dividing the trajectory into different dynamical segments . in terms of physical interpretations , several molecular components can induce `` force cross - talk '' between sister chromatids . a potential explanation of the eigen - value phenomenon described above is that microtubules of the mitotic spindle connected to the common centromere of the sister chromatids are simultaneously inducing a common force change to both chromatids .
alternatively ( or perhaps in addition ) , a common tension signal induced by chromatin loops and a network of pericentric proteins ( cohesin and condensin @xcite ) might be influencing the forces experienced by the pair of sister chromatids ; recall that the gfp dyes used are in close proximity to the centromere so molecular changes in the pericentric region can potentially influence the fluctuations observed in the tagged sister chromatids @xcite .
regardless of the underlying biological cause of the phenomenon observed , the downstream eigen - analysis substantially benefited from the temporal segmentation provided by the hdp - slds framework . using the segmentation afforded by the hdp - slds method resulted in the eigen - vectors shown in the bottom panel of fig .
[ fig : mitosis ] ; in this figure , the eigen - vectors associated with the largest magnitude eigen - value is displayed as a solid line and the eigen - vector associated with the smaller eigen - value of the estimated 2d @xmath58 matrix is shown by a dashed line ( the origin for the eigen - vectors correspond to the empirical average of the identified states ) .
the ( real - valued ) eigen - values and eigen - vectors provide spatial information about restoring force directions experienced by the tagged chromatid .
observe how the pair of chromatids both exhibit a state change occurring around 16@xmath59 .
note that the time of the change point occurring around 16@xmath59 coincides for the sister chromatids , but each trajectory was processed independently by separate hdp - slds analyses . at 16@xmath59 ,
the dominant eigen - vectors of the chromatids change from pointing towards the spindles to a direction nearly orthogonal to the plane connecting the two mitotic spindle poles .
the eigen - values for both chromatids are decreasing at roughly the same rate ( i.e. , their effective
stiffness " is increasing ) . also , recall that the pair of chromatids are subject to forces from multiple sources that can change in magnitude and direction ; for example , tugging induced by the microtubules connected to the centromere shared by the chromatids , forces associated with nucleosome remodeling ( e.g. , rna polymerase `` tugging '' on the strand as it makes rna ) , as well as other spring like forces induced by the shared centromere and percentric components @xcite .
the various explanations posited above are admittedly speculative .
what these results unambiguously show is that sister chromatids experience a force regime shift whereby the direction and magnitude of the forces ( inferred by position vs. time data ) co - occur in both direction and magnitude .
this information gives new insight into the effective forces experienced by particles in time changing , complex live cellular environment .
researchers can get a more reliable `` force map '' that contains rich dynamical information beyond that contained in a mean square displacement or correlation - based analysis .
note that this `` force map '' allows a spatially dependent vector field to change over time ( an important feature for detecting dynamical changes in single - molecule measurements ) .
a dominant force direction pointing orthogonal to the spindle axis is highly suggestive of a new dominant force ( not induced by the microtubule pulling ) being felt by both sister chromatids .
deciding the correct physical explanation of the phenomenon observed requires additional experiments , but the results obtained on the metaphase trajectories are presented in order to show the power of spt analysis coupled with hdp - slds segmentation and downstream likelihood - based time series analysis @xcite .
currently we are attempting to use this type of analysis to more systematically and quantitatively probe forces associated with the various physical phenomena described here using experiments which more carefully control the types of forces that can be experienced by the gfp tagged chromatids . before proceeding
, we should note that the gfp array used to tag the chromatids induces a gradual continuous trend in the effective measurement noise ( the finding can readily be observed in s1-s3 data ) .
a gradual time trend induced in large part by different fluorophores bleaching out has been observed and quantitatively studied in previous works ( see supp . mat . of ref .
however , the hdp - slds formulation considered assumes a constant measurement noise for the entire duration of the trajectory .
s6 - s7 figs .
explore various simulations mimicking features in the experimental data ; specifically we analyze data where the effective measurement changes linearly from 15 nm to 35 nm over the course of 400 measurements ( this induces model misspecification since the hdp - slds model assumes fixed covariance ) .
the linear change in measurement noise is consistent with the mle range of values observed in s3 data .
s6 - s7 indicate that if one uses a prior near the temporal midpoint of the time changing measurement noise that the hdp - slds segmentation can achieve accuracy similar to the ideal situation where the hdp priors match the dgp exactly . for
whatever system is being studied , it is highly recommended to employ simulations mimicking features implied by the data in order to be more confident that the segmentation output does not contain artifacts induced by features not explicitly accounted for in the underlying model .
techniques accounting for gradual time trends in certain parameters extending the hdp - slds model would be an interesting future research topic . in terms of more general applicability of this type of analysis , systematically segmenting spt trajectories into distinct dynamical regimes and then using classical statistical physics models is one alternative to using so - called anomalous diffusion models commonly used in spt analysis @xcite .
this is advantageous since fitted overdamped langevin model output can readily be interpreted in terms of force , velocity , and molecular friction .
furthermore , an overdamped langevin sde model can be tested directly against experimental data and the effects of measurement noise can be systematically accounted for by likelihood methods @xcite .
anomalous diffusion models are more difficult to physically interpret and rigorous time series analysis ( accounting for measurement noise , sampling noise , spatially dependent forces , etc . )
is challenging for various technical reasons @xcite .
we have demonstrated the utility of the hdp - slds method of fox et al .
@xcite in automatically segmenting spt data into different dynamical regimes .
the approach shows great promise in systematically processing a variety of spt data sets .
when applied to experimental data , we demonstrated how new quantitative information can be extracted about transient forces experienced during mitosis in live yeast cells ; the method explicitly accounts for the statistical effects of measurement noise on top of `` thermal '' or `` process '' noise ( both sources are non - negligible in many spt applications ) .
simulations results reported in s2 text were used to exhibit strengths and weaknesses of the hdp - slds approach within an spt context . work presented in ref .
@xcite compared the hdp - slds approach to classic ( finite state space ) hmm models . in terms of strengths ,
the hdp - slds approach can account for an unknown number of states , position dependent anisotropic active forces , and can account for the statistical effects of measurement noise by exploiting kalman filtering modeling in conjunction with hdp ideas @xcite ; these features are not accounted for in any hmm approach currently used in spt analysis and hence the hdp - slds represents a `` state - of - the - art '' technology within this domain .
the approach provides a more systematic means for the `` time window size '' selection problem discussed in ref .
the hdp - slds approach was also confirmed to be fairly robust to many modeling assumptions and hyperparameters associated with hdp modeling @xcite . in situations where a collection of transient dynamical responses are experienced within a single trajectory and the `` states '' causing the different kinetics are experimentally resolvable , the data processing procedure discussed here provides an attractive alternative to anomalous diffusion modeling @xcite since the procedure can mitigate artifacts induced by aggregating distinct kinetic states @xcite .
furthermore , the hdp - slds approach can be modified to produce output that can be readily physically interpreted in terms of classic spt models ( see sec.[sec : oddesc ] ) . a substantial weakness of the hdp - slds approach ( also discussed in ref .
@xcite ) is associated with its dependence on reliable priors for the parameters determining governing the kalman filter parameters ( i.e. , hyperparameters specifying the `` base measure '' @xcite ) . the data - driven scheme advocated in ref .
@xcite for selecting the prior mean of @xmath4 exhibits undesirable properties in many spt applications ( results presented in s2 text ) . a variant of the mle - based approach of ref
. @xcite was discussed ( s1 text ) and was demonstrated ( s2 text ) to be capable of providing pilot estimates needed to tune the base measure hyperparameters priors in spt applications .
the formal hypothesis testing methods utilized in ref .
@xcite also allow one to test the validity of a candidate segmentation and assumed dynamical model against experimental data in situations where reference ( or `` ground - truth '' ) data is unavailable . in summary
, combining recent nonparametric bayesian modeling ideas , like the hdp - slds , with frequentist ideas show great potential in quantitatively analyzing complex spt data sets . in the future , we aim to provide user - friendly software making the tools discussed herein readily accessible to the diverse set of researchers involved in spt . the development of new , statistically rigorous data processing algorithms capable of reliably extracting information about molecular motion from live cell data is becoming ever more important as _ in vivo _
microscopy techniques continue to improve in spatial and temporal resolution @xcite .
cpc was supported by internal r&d funds from ursa analytics and kb was funded by the national institutes of health r37 grant gm32238 .
arhel n , genovesio a , kim k , miko s , perret e , et al .
( 2006 ) quantitative four - dimensional tracking of cytoplasmic and nuclear hiv-1 complexes .
nat methods 3 : 817824 .
brandenburg b , zhuang x ( 2007 ) virus trafficking - learning from single - virus tracking .
nat rev microbiol 5 : 197208 .
lessard ga , goodwin pm , werner jh ( 2007 ) three - dimensional tracking of individual quantum dots .
appl phys lett 91 : 224106 .
ngerl uv , willig ki , hein b , hell sw , bonhoeffer t ( 2008 ) live - cell imaging of dendritic spines by sted microscopy .
proc natl acad sci u s a 105 : 189827 .
huang b , wang w , bates m , zhuang x ( 2008 ) three - dimensional super - resolution imaging by stochastic optical reconstruction microscopy .
science 319 : 810 - 813
. rohatgi r , milenkovic l , scott mp ( 2007 ) patched1 regulates hedgehog signaling at the primary cilium .
science 317 : 3726 .
manley s , gillette jm , patterson gh , shroff h , hess hf , et al .
( 2008 ) high - density mapping of single - molecule trajectories with photoactivated localization microscopy .
nat methods 5 : 1557 .
pavani srp , thompson ma , biteen js , lord sj , liu n , et al . (
2009 ) three - dimensional , single - molecule fluorescence imaging beyond the diffraction limit by using a double - helix point spread function .
proc natl acad sci u s a 106 : 29959 .
rohatgi r , milenkovic l , corcoran rb , scott mp ( 2009 ) hedgehog signal transduction by smoothened : pharmacologic evidence for a 2-step activation process .
proc natl acad sci u s a 106 : 3196201 .
thompson ma , casolari jm , badieirostami m , brown po , moerner we ( 2010 ) three - dimensional tracking of single mrna particles in saccharomyces cerevisiae using a double - helix point spread function .
proc natl acad sci u s a 107 : 1786471 .
sahl sj , leutenegger m , hilbert m , hell sw , eggeling c ( 2010 ) fast molecular tracking maps nanoscale dynamics of plasma membrane lipids .
proc natl acad sci u s a 107 : 682934 .
grnwald d , singer rh ( 2010 ) in vivo imaging of labelled endogenous @xmath60-actin mrna during nucleocytoplasmic transport .
nature 467 : 604609 .
ram s , kim d , ober rj , ward es ( 2012 ) 3d single molecule tracking with multifocal plane microscopy reveals rapid intercellular transferrin transport at epithelial cell barriers .
biophys j 103 : 1594603 .
gao l , shao l , chen bc , betzig e ( 2014 ) 3d live fluorescence imaging of cellular dynamics using bessel beam plane illumination microscopy .
nat protoc 9 : 1083101 .
welsher k , yang h ( 2014 ) multi - resolution 3d visualization of the early stages of cellular uptake of peptide - coated nanoparticles .
nat nanotechnol 9 : 198203 .
chen bc , legant wr , wang k , shao l , milkie de , et al .
( 2014 ) lattice light - sheet microscopy : imaging molecules to embryos at high spatiotemporal resolution .
science ( 80- ) 346 : 12579981257998 .
masson j , casanova d , turkcan s , voisinne g , popoff m , et al . ( 2009 ) inferring maps of forces inside cell membrane microdomains .
phys rev lett 102 : 48103 .
berglund aj ( 2010 ) statistics of camera - based single - particle tracking .
phys rev e 82 : 011917 .
ensign dl , pande vs ( 2010 ) bayesian detection of intensity changes in single molecule and molecular dynamics trajectories .
j phys chem b 114 : 28092 .
weber sc , thompson ma , moerner we , spakowitz aj , theriot ja ( 2012 ) analytical tools to distinguish the effects of localization error , confinement , and medium elasticity on the velocity autocorrelation function .
biophys j 102 : 244350 .
persson f , lindn m , unoson c , elf j ( 2013 ) extracting intracellular diffusive states and transition rates from single - molecule tracking data .
nat methods 10 : 265271 .
chen k , wang b , guan j , granick s ( 2013 ) diagnosing heterogeneous dynamics in single - molecule / particle trajectories with multiscale wavelets .
acs nano 7 : 863444 .
calderon cp , thompson ma , casolari jm , paffenroth rc , moerner we ( 2013 ) quantifying transient 3d dynamical phenomena of single mrna particles in live yeast cell measurements .
j phys chem b 117 : 1570113 .
stephens ad , haggerty ra , vasquez pa , vicci l , snider ce , et al .
( 2013 ) pericentric chromatin loops function as a nonlinear spring in mitotic force balance .
j cell biol 200 : 75772 .
aguet f , antonescu cn , mettlen m , schmid sl , danuser g ( 2013 ) advances in analysis of low signal - to - noise images link dynamin and ap2 to the functions of an endocytic checkpoint .
dev cell 26 : 27991 .
hajjoul h , mathon j , ranchon h , goiffon i , mozziconacci j , et al .
( 2013 ) high - throughput chromatin motion tracking in living yeast reveals the flexibility of the fiber throughout the genome .
genome res 23 : 182938 .
, dionne p , salvatico c , renner m , specht cg , et al .
( 2014 ) mapping the energy and diffusion landscapes of membrane proteins at the cell surface using high - density single - molecule imaging and bayesian inference : application to the multiscale dynamics of glycine receptors in the neuronal membrane .
biophys j 106 : 7483 .
calderon cp , weiss le , moerner we ( 2014 ) robust hypothesis tests for detecting statistical evidence of 2d and 3d interactions in single - molecule measurements .
phys rev e 89 .
n , smal i , de chaumont f , maka m , sbalzarini if , et al .
( 2014 ) objective comparison of particle tracking methods .
nat methods 11 : 281290 .
jouvenet n , simon sm , bieniasz pd ( 2011 ) visualizing hiv-1 assembly .
j mol biol 410 : 50111 .
weigel av , tamkun mm , krapf d ( 2013 ) quantifying the dynamic interactions between a clathrin - coated pit and cargo molecules .
proc natl acad sci u s a 110 : e4591600 .
fox e , sudderth eb , jordan mi , willsky as ( 2011 ) bayesian nonparametric inference of switching dynamic linear models .
ieee trans signal process 59 : 15691585 .
thompson re , larson dr , webb ww ( 2002 ) precise nanometer localization analysis for individual fluorescent probes .
biophys j 82 : 277583 .
enderlein j , toprak e , selvin pr ( 2006 ) polarization effect on position accuracy of fluorophore localization .
opt express 14 : 811120 .
gahlmann a , moerner we ( 2014 ) exploring bacterial cell biology with single - molecule tracking and super - resolution imaging .
nat rev microbiol 12 : 922 .
calderon cp ( 2013 ) correcting for bias of molecular confinement parameters induced by small - time - series sample sizes in single - molecule trajectories containing measurement noise .
phys rev e 88 : 012707 .
calderon cp , harris n , kiang c , cox d ( 2009 ) quantifying multiscale noise sources in single - molecule time series . j phys chem b 113 : 138 .
calderon cp , chen w , harris n , lin k , kiang c ( 2009 ) quantifying dna melting transitions using single - molecule force spectroscopy .
j phys : condens matter 21 : 034114 .
hamilton j ( 1994 ) time series analysis .
princeton , nj : princeton university press .
stengel r ( 1994 ) optimal control and estimation .
toronto , ontario : dover publications .
teh yw , jordan mi , beal mj , blei dm ( 2006 ) hierarchical dirichlet processes .
j am stat assoc 101 : 15661581 .
jk ( 2010 ) bayesian nonparametrics .
new york : springer - verlag .
claeskens g , hjort n ( 2008 ) model selection and model averaging .
cambridge , uk : cambridge university press .
schwantes cr , mcgibbon rt , pande vs ( 2014 ) perspective : markov models for long - timescale biomolecular dynamics .
j chem phys 141 : 090901 .
calderon cp ( 2007 ) fitting effective diffusion models to data associated with a
glassy potential " : estimation , classical inference procedures and some heuristics .
multiscale model simul 6 : 656687 .
fox e , sudderth e , jordan m , willsky a ( 2010 ) bayesian nonparametric methods for learning markov switching processes .
ieee signal process mag 27 : 4354 .
calderon c ( 2014 ) data - driven techniques for detecting dynamical state changes in noisily measured 3d single - molecule trajectories .
molecules 19 : 1838118398 .
hines ke , bankston jr , aldrich rw ( 2015 ) analyzing single - molecule time series via nonparametric bayesian inference .
biophys j 108 : 540556 .
kusumi a , sako y , yamamoto m ( 1993 ) confined lateral diffusion of membrane receptors as studied by single particle tracking ( nanovid microscopy ) .
effects of calcium - induced differentiation in cultured epithelial cells .
biophys j 65 : 202140 .
saxton mj , jacobson k ( 1997 ) single - particle tracking : applications to membrane dynamics .
annu rev biophys biomol struct 26 : 37399 .
golding i , cox ec ( 2004 ) rna dynamics in live escherichia coli cells .
proc natl acad sci u s a 101 : 113105 .
park hy , buxbaum ar , singer rh ( 2010 ) single mrna tracking in live cells .
methods enzymol ( chapter 18 ) 472 : 387406 .
montiel d , cang h , yang h ( 2006 ) quantitative characterization of changes in dynamical behavior for single - particle tracking studies .
j phys chem b 110 : 1976370 .
press s , lee j , dill ka ( 2013 ) extracting conformational memory from single - molecule kinetic data .
j phys chem b 117 : 495502 . van de meent jw , bronson
je , wiggins ch , gonzalez rl ( 2014 ) empirical bayes methods enable advanced population - level analyses of single - molecule fret experiments .
biophys j 106 : 132737 .
calderon cp ( 2007 ) on the use of local diffusion models for path ensemble averaging in potential of mean force computations .
j chem phys 126 : 084106 .
calderon cp , harris n , kiang c , cox d ( 2009 ) analyzing single - molecule manipulation experiments .
j mol recognit 22 : 356 .
jaqaman k , dorn jf , jelson gs , tytell jd , sorger pk , et al .
( 2006 ) comparative autoregressive moving average analysis of kinetochore microtubule dynamics in yeast .
biophys j 91 : 231225 .
hoze n , holcman d ( 2014 ) residence times of receptors in dendritic spines analyzed by simulations in empirical domains .
biophys j 107 : 30083017 .
calderon cp ( 2010 ) detection of subtle dynamical changes induced by unresolved conformational coordinates in single - molecule trajectories via goodness - of - fit tests .
j phys chem b 114 : 32423253 .
pokern y , stuart a , wiberg p ( 2009 ) parameter estimation for partially observed hypoelliptic diffusions .
j royal statistical society series b 71 : 4973 .
gardiner c ( 1985 ) handbook of stochastic models .
berlin : springer - verlag .
park s , schulten k ( 2004 ) calculating potentials of mean force from steered molecular dynamics simulations .
j chem phys 120 : 594661 .
holcman d , hoze n , schuss z ( 2011 ) narrow escape through a funnel and effective diffusion on a crowded membrane .
phys rev e - stat nonlinear , soft matter phys 84 : 021906 .
fisher jk , ballenger m , obrien et , haase j , superfine r , et al .
( 2009 ) dna relaxation dynamics as a probe for the intracellular environment .
proc natl acad sci u s a 106 : 92505 .
verdaasdonk js , vasquez pa , barry rm , barry t , goodwin s , et al .
( 2013 ) centromere tethering confines chromosome domains .
mol cell 52 : 113 .
fox e , sudderth e , jordan mi , willsky as ( 2011 ) a sticky hdp - hmm with application to speaker diarization .
ann appl stat 5 : 10201056 .
bouck d , bloom k ( 2007 ) pericentric chromatin is an elastic component of the mitotic spindle .
curr biol 17 : 741748 .
hong y , li h ( 2005 ) nonparametric specification testing for continuous - time models with applications to term structure of interest rates .
rev fin studies 18 : 3784 .
stephens ad , quammen cw , chang b , haase j , taylor rm , et al .
( 2013 ) the spatial segregation of pericentric cohesin and condensin in the mitotic spindle .
mol biol cell 24 : 390919 .
lubelski a , sokolov i m , klafter j ( 2008 ) nonergodicity mimics inhomogeneity in single particle tracking .
physical review letters 100 : 250602 .
michalet x , berglund a ( 2012 ) optimal diffusion coefficient estimation in single - particle tracking .
phys rev e 85 : 061916 .
ober rj , ram s , ward es ( 2004 ) localization accuracy in single - molecule microscopy .
biophys j 86 : 1185200 .
smith cs , joseph n , rieger b , lidke ka ( 2010 ) fast , single - molecule localization that achieves theoretically minimum uncertainty .
nat methods 7 : 3735 .
efron b ( 2010 ) large - scale inference : empirical bayes methods for estimation , testing , and prediction .
cambridge : cambridge university press .
at - sahalia y , fan j , xiu d ( 2010 ) high - frequency covariance estimates with noisy and asynchronous financial data .
j am stat assoc 105 : 15041517 .
calderon cp , martinez j , carroll r , sorensen d ( 2010 ) p - splines using derivative information .
multiscale model simul 8 : 15621580 .
kloeden p , platen e ( 1992 ) numerical solution of stochastic differential equations .
berlin : springer - verlag .
gardiner jd , wette mr , laub aj , amato jj , moler cb ( 1992 ) algorithm 705 ; a fortran-77 software package for solving the sylvester matrix equation axbt + cxdt = e. acm trans math softw : 232 - 238 .
neal r ( 2000 ) markov chain sampling methods for dirichlet process mixture models .
j comput graph stat 9 : 249265 .
sethuraman j ( 1994 ) a constructive definition of dirichlet priors .
stat sin 4 : 639650 . | understanding the basis for intracellular motion is critical as the field moves toward a deeper understanding of the relation between brownian forces , molecular crowding , and anisotropic ( or isotropic ) energetic forcing .
effective forces and other parameters used to summarize molecular motion change over time in live cells due to latent state changes , e.g. , changes induced by dynamic micro - environments , photobleaching , and other heterogeneity inherent in biological processes .
this study discusses limitations in currently popular analysis methods ( e.g. , mean square displacement - based analyses ) and how new techniques can be used to systematically analyze single particle tracking ( spt ) data experiencing abrupt state changes in time or space .
the approach is to track gfp tagged chromatids in metaphase in live yeast cells and quantitatively probe the effective forces resulting from dynamic interactions that reflect the sum of a number of physical phenomena .
state changes can be induced by various sources including : microtubule dynamics exerting force through the centromere , thermal polymer fluctuations , and dna - based molecular machines including polymerases and protein exchange complexes such as chaperones and chromatin remodeling complexes .
simulations aiming to show the relevance of the approach to more general spt data analyses are also studied .
refined force estimates are obtained by adopting and modifying a nonparametric bayesian modeling technique , the hierarchical dirichlet process switching linear dynamical system ( hdp - slds ) , for spt applications .
the hdp - slds method shows promise in systematically identifying dynamical regime changes induced by unobserved state changes when the number of underlying states is unknown in advance ( a common problem in spt applications ) .
we expand on the relevance of the hdp - slds approach , review the relevant background of hierarchical dirichlet processes , show how to map discrete time hdp - slds models to classic spt models , and discuss limitations of the approach .
in addition , we demonstrate new computational techniques for tuning hyperparameters and for checking the statistical consistency of model assumptions directly against individual experimental trajectories ; the techniques circumvent the need for `` ground - truth '' and/or subjective information . |
understanding the structure of the inner parsec of quasars which includes the line and continuum emitting regions has proven elusive .
reverberation mapping techniques have been applied to several low - luminosity radio - quiet agn .
though they required large amounts of telescope time , they set on a firm ground basic aspects of the broad line region ( blr ) , i.e. , the stratification of ionization ( see @xcite , @xcite , and @xcite for reviews ) .
other results are unfortunately more ambiguous .
one reason for the ambiguity resides in our poor understanding of the constraints on the inputs to the line formation processes in the blr ( @xcite ) .
statistical studies of line profiles have in principle the advantage of an approach less dependent on the poorly known physical conditions of the blr .
however they are as yet plagued by small number statistics and by the difficulty of obtaining a uniform sample , in terms of observational ( i.e. , signal - to - noise ratio , s / n ) and intrinsic properties ( such as optical luminosity ; see @xcite for a recent review ) .
radio - loud agn have been associated to an extreme in the distribution of parameters of radio - quiet agn ( @xcite ; @xcite ) . however ,
marziani et al . ( 1996 , hereafter m96 ) have shown that radio - loud agn with detected superluminal motion , and apparent velocity @xmath3 show profiles with low ew and remarkably strong red - ward asymmetry .
there is no equivalent phenomenology in radio - quiet objects .
also , widely spaced double - peaked profiles appear almost exclusively in radio - loud agn ( see @xcite for a critical review , and @xcite )
. statistical analysis of balmer line profiles suggests systemic , kinematic , and structural differences in the blr of ( at least some ) radio - quiet and radio - loud agn ( m96 ; @xcite ) .
there is therefore little evidence that the basic reverberation and profile analysis results can be extended straightforwardly to radio - loud or high luminosity agn . in the handful of radio - loud agn
known to have variable profiles , the profiles are often peculiar ( i.e. , double - peaked ) , as in the cases of 3c 390.3 ( @xcite ) , her twin 3c 382 , oq 208 ( @xcite ) , and pictor a ( @xcite ; @xcite ) .
this is different from the case for radio - quiet agn , since monitored seyfert nuclei do not show double - peaked line profiles .
the most common profile among radio - loud agn , however , is not the extremely broad , double - peaked profile observed in arp 102b ; instead , it is regular , single peaked , red - ward asymmetric with red - ward peak displacement ( m96 ; class ar , r according to @xcite ) .
the double - peaked profiles are probably not representative of the radio - loud agn as a class .
the low redshift ( heliocentric @xmath4 ; m96 ) radio quasar pg 1512 + 370 ( b2 1512 + 37 , 4c 37.43 ) has been frequently observed at optical wavelengths ( @xcite ; @xcite , among others ) . the broad profile component ( ) of pg 1512 + 370 belongs to the class ar , r for most epochs of observations , and hence it is fairly typical of radio - loud agn .
variability of broad uv lines and ly@xmath5has been reported in 1991 from iue observations ( @xcite ) .
these authors suggested the possibility of moderate line flux variations from published data ( ew ( ) appeared to vary by @xmath6 20 % ) . in this paper , we report the discovery of line profile variations in pg 1512 + 370 , and in particular variation of the blue wing and of a secondary `` blue peak '' from the analysis of spectra collected from 1988 to 1996 ( [ results ] ) . we will attempt to explain the profile and its variations as due to ( a ) accretion disk emission ( [ disk ] ) ; ( b ) radially moving gas ( [ reverber ] ) , or ( c ) a binary blr ( [ bh ] ) .
we summarize the available data in table [ tab : observations ] .
the initial reduction followed the standard iraf procedures , which include bias subtraction , flat fielding , wavelength calibration , correction for extinction and flux calibration ( also see m96 , and references therein for further details ) .
the spectra have been corrected for redshift , the individual redshifts being determined as the average of the peaks of the gaussian fits to the upper half of @xmath7 and @xmath14959,5007 . the upper panel of figure [ fig : spectra ] shows the optical spectra for pg 1512 + 370 in the @xmath7 region .
the spectra have been continuum subtracted by fitting a low order polynomial through wavelength regions uncontaminated by emission lines .
most of the weight in the continuum fit has been assigned to : ( 1 ) the region around 4200 as adopted for the quasar composite spectrum by francis et al .
( 1991 ) , and ( 2 ) to the red side of @xmath25007 .
other wavelength regions around 4020 and , with decreasing weight , around 4620 have been included in the fit when available .
the spectra have then been scaled to the same flux of @xmath24959 .
we did not choose @xmath25007 because of the intrinsically larger errors due to the presence of the telluric b band at its red side .
the scale factor determined on the basis of @xmath24959 is affected by the extended appearance of the emission ( diameter @xmath6 4 arcsecs , @xcite ) . on the other hand , the narrow profile component ( ) flux is difficult to measure , because the underlying is variable . as a result , measured fluxes
are thought to be accurate within @xmath8% at a @xmath9 confidence level .
the procedure to remove the contribution of ( which is minimal ; see , e.g. , @xcite ) , , and from the profile is described in m96 .
table [ tab : lines ] reports measurements taken on the cleaned line .
the only signicant change of _ total _ flux recorded in our spectra occurred probably between 1988 and 1990 , when it was @xmath10% of the 1988 flux .
the larger decrease is observed in the blue wing of and with a possible smaller decrease in the red wing . between 1990 and 1996
the flux remained constant within the observational uncertainties .
no shape change is noticeable in the 1990 , 1992 , and 1994 spectra . at these times
, there is no peak at @xmath11 ( the `` blue peak '' ) or blue shoulder visible and the profile can be classified as ar , r according to sulentic ( 1989 ) .
however , the line profile showed remarkable changes between 1988 and 1990 as well as between 1995 and 1996 . 1 .
the blue peak is visible in the 1988 spectrum but is apparently absent in 1990 , when the total flux is also significantly lower .
2 . the blue wing appears more prominent in the 1995 spectrum but depressed in the 1996 spectrum obtained at spm , with a blue wing fractional flux decrease of @xmath6 20% .
this variation resembles the one that occurred between 1988 and 1990 , save that the total flux did not change appreciably in the 1995 and 1996 observations .
the lower panel of figure [ fig : spectra ] shows the difference between the 1995 and 1996 spectra in the region .
the slight depression in the 4925 - 5025 region is due to the extended nature of , as different contribution from extra - nuclear probably fell within the slit in the 1995 and 1996 spectra . to compute the probability that the observed change in line profile shape is due to random noise fluctuation , we first estimated the 3@xmath12 noise level from the observed maximum noise fluctuations over the average signal in the rest frame range 5060 - 5100 ( much larger than the noise correlation length @xmath13 pixels ) .
we rebinned the spectrum ( over the wavelength range where we observe a change ) with a step @xmath14 .
we then multiplied the probability that the change in amplitude for each bin is due to noise ( which is assumed constant over the line wing ) .
we find the probability that the blue wing change between 1995 - 1996 is due to random noise fluctuations is negligibly small .
similar considerations apply to the 1990 - 1988 line profile difference ( not shown in figure [ fig : spectra ] ) .
the remaining question at this point is whether the blue peak may or may not be variable in radial velocity .
the ability to resolve the `` blue shoulder '' into a `` blue peak '' depends on the spectral resolution of the data .
the 1988 and 1996 ( asiago ) spectra have a resolution of 3 and 2 fwhm respectively .
the blue peak is well visible at @xmath15 ( in 1988 ) and @xmath16 ( in 1996 ) with respect to the radial velocity of .
the 1995 - 1996 difference has a broad gaussian shape , with maximum at @xmath17 .
radial velocity values of the blue peak measured in 1988 , 1995 , and 1996 are therefore consistent .
the possibility that the blue peak is present in the 1990 , 1992 , and 1994 spectra but not visible because of the lower resolution can not be ruled out .
after all , the blue flat top observed in the 1996 san pedro martir spectrum is resolved as a blue peak in the 1996 asiago spectrum , collected 30 days earlier , when the blue wing appeared depressed .
two interpretations seem possible . 1 .
the blue peak is a component of more variable than the rest of , at approximately constant @xmath18 . in this case
, the blue peak ( i ) may be due to a flux redistribution within unrelated to continuum changes or ( ii ) may be a structure produced by response to continuum variation .
both the 1988 and 1995 spectra ( when the blue wing was more prominent ) show the highest continuum fluxes .
this suggests that an enhancement of the blue wing is related to an increase of the continuum level .
however , as the total flux did not change significantly in all our spectra obtained between 1990 and 1996 , it is also possible , albeit less likely , that we are observing a line flux redistribution unrelated to continuum variations ( as in ngc 5548 ; @xcite ) .
model implications are discussed in [ reverber ] .
if the disappearance of the blue peak in 1990 - 1994 is genuine , the radial velocity of the peak in the blue wing may have changed to @xmath6 0@xmath19 in 1990 , 1992 , and 1994 . in [ bh ]
we discuss a possible model .
the observed profile as a whole is not consistent with the predictions of disk models at any epoch of observation .
nevertheless , if we wanted to interpret the secondary peak @xmath20 as the doppler - boosted peak of a disk contribution , we would need a disk with @xmath21 , @xmath22 following chen & halpern ( 1989 ) .
non - conventional disk models ( elliptical disks or warped disks ; @xcite ; @xcite ) may be more successful in fitting the line profile at a given epoch . however , the variation observed between 1988 and 1990 and between 1995 and 1996 rules out the possibility of `` switched - on '' disk emission , since the change appears to be restricted to the blue wing in both cases ( see figure [ fig : spectra ] ) .
the observed variation would require implausible changes of the disk parameters from one epoch to another .
the moderate change in the blue wing flux , @xmath23% , without a simultaneous response in the red wing , points toward line emission from gas whose motion is predominantly radial .
given the observed blueshift and assuming optically thin clouds , as it is generally believed to be the case for most emitting clouds , the gas must be moving outwards .
we are not saying here that the line emission is due exclusively to outflowing gas , but that a fraction related to the blue peak most probably is .
without further data , it is however not possible to elaborate on this result .
caution is needed as the temporal sampling of our spectra is quite probably not sufficient to constrain line and continuum variations ( @xcite ) .
the line luminosity is l()@xmath24 for h@xmath25 mpc@xmath26 , and q@xmath27 . under the assumptions of menzel - baker recombination theory ,
the light travel time across the blr is @xmath28 days .
spectra obtained at a frequency of @xmath29 1 month@xmath26 are desirable .
if the time scale for continuum variation is found to be comparable to @xmath30 , then `` ripples '' in the line profile could be produced by reverberation effects . at 20 cm , on a first map ( @xcite ) ,
pg 1512 + 370 appears as a double - lobed radio source , with lobes extending up to a projected linear distance of @xmath31 400 kpc from the central core , at p. a. @xmath6 105@xmath32 .
integral field spectroscopy revealed two emitting regions approximately aligned along the east - west direction with the nucleus ( @xcite ) .
our 1996 spectrum also revealed the eastern emitting region as a `` blob '' , at @xmath33 , in excellent agreement with the results of durret et al .
extended emitting regions in radio quasars show a statistical tendency to align with the radio - source axis in radio - loud , steep spectrum quasars ( @xcite ) .
extended provides some evidence in favor of outflowing gas related to the presence of radio ejecta ( see the discussion by @xcite for pg 1512 + 370 ) .
we consider it unlikely that the blue peak is due to some out - flowing gas that is episodically related to the radio jet .
the dynamical time - scale for the blr is probably shorter than 8 years ; a blob moving at @xmath34 could not have maintained the same radial velocity over that time period . within the errors in our spectra
, the profile variations can be ascribed to a stationary profile plus a superimposed component of variable radial velocity , contributing up to @xmath35% of the total line flux . in a binary blr framework ( see , e.g. , @xcite , @xcite , @xcite ) , we assume that the secondary component is observed at @xmath36 ( in any case at @xmath37 ) between 1990 and 1994 ( i.e. , not clearly observed because it underlies the ) .
this assumption is justified since the underlying the may be variable ( figure [ fig : spectra ] ) , and a slightly redshifted `` shoulder '' partly underlying may be identified in the 1996 ( spm ) spectrum .
the `` shoulder '' close to is less prominent in 1995 , when the blue wing was in turn much stronger .
this interpretation is not fully satisfactory because the 1988 and 1996 asiago spectra are of higher dispersion than the remaining spectra .
the blue peak is resolved in the 1996 asiago spectrum but not in the 1996 san pedro martir spectrum , because of the lower resolution ( the two spectra were obtained 30 days apart ) .
the same may be true for the 1990 , 1992 , and 1994 spectra .
if we accept the blue peak as a component of variable radial velocity , then the immediate implication for the orbit of a binary black hole system is that it is highly eccentric , and that the binary must have mass ratio very different from unity . from the observed - frame core - to - lobe flux ratio r = 0.091 measured on the first map ( 1.4 ghz ) we can estimate the angle between the line of sight and the radio axis .
it is known that the relationship between r and orientation is valid in a statistical sense , probably because of a large intrinsic scatter in the the lorentz factor of the jet ( @xmath38 ; @xcite ; @xcite ) , and in the ratio f between the intrinsic jet luminosity and the unbeamed luminosity ( @xcite ) .
we can write r as r = f @xmath39 , where @xmath40 is the doppler factor , and p = 2.75 for a continuous jet ( @xcite ) .
we assume @xmath415 , and we use f = @xmath42 ( @xcite ) , as appropriate for fanaroff - riley type ii radio sources .
we obtain an inclination @xmath43 . for more extreme values of @xmath44 , namely for @xmath45 and @xmath46
, we obtain i=15@xmath32 ( f= @xmath47 ) and i= 26@xmath32 ( f=@xmath48 ) respectively .
to assume @xmath43 seems therefore reasonable .
pg 1512 + 370 may have been excluded from superluminal motion searches , since its radio emission is lobe - dominated .
however , at @xmath43 and @xmath49 , we expect that pg 1512 + 370 should be superluminal with apparent transverse velocity @xmath50 ( also if @xmath51 , which implies i=26@xmath32 , we should observe @xmath52 1.7 ) . if @xmath53 , an acceptable fit to the six observed radial velocities ( @xmath541.76 ) is obtained for binary period @xmath55 yr , eccentricity @xmath56 , and semi - major axis @xmath57 cm .
this solution corresponds to a total mass for the binary of @xmath58 .
the evolution of massive black hole binaries orbital parameters ( semi - major axis , eccentricity ) for several mass ratios has been studied recently by @xcite . from figure 8 of his work
we infer that e @xmath59 for @xmath60 cm is somewhat too large for any value of the initial eccentricity of the binary , save possibly for the case of initial eccentricity @xmath61 . in this case , pg 1512 + 370 should be in a very - short lived phase of the binary evolution : high eccentricity drives large gravitational radiation losses that should lead to the merging of the two black holes in @xmath62 yr .
for the lower power fanaroff - riley type i radio source oj 287 the case for a double black hole is stronger ( e.g. @xcite ; @xcite ) , and rather similar orbital parameters are found : @xmath63 , @xmath64 yr , @xmath65 ( @xcite ) .
if the blazar nature of oj 287 and the absence of obvious line emission is due to doppler - boosted continuum emission , then pg 1512 + 370 and oj 287 may both host a binary black hole whose orbital plane is observed at different inclination .
we have presented evidence of moderate variations ( @xmath66 30% total line flux , with the largest variations occurring in the blue wing ) in the of pg 1512 + 370 from the analysis of spectra obtained from 1988 to 1996 .
a peak displaced by @xmath67 with respect to the radial velocity of ( the `` blue peak '' ) has been observed in 1988 and 1996 .
we have attempted to interpret the observations according to a number of presently fashionable blr models : accretion disk emission ( [ disk ] ) ; radially moving gas , and , in particular , emission due to gas trapped by the radio jet ( [ reverber ] ) ; a binary blr ( [ bh ] ) . only radially moving gas ( but not occasionally associated to the radio jet ) and a binary blr are probably appropriate .
this is one of the few cases where evidence from line profiles may support the presence of a binary blr ( with the cautionary note of [ bh ] ) , although the eccentricity inferred from the observed radial velocity curve may be too high for the estimated major axis . like
oj 287 , pg 1512 + 370 would then be a very peculiar object observed in a very short - lived phase of its existence .
however , we favor the idea that if variations are the response to ionizing continuum changes , the presence of the blue peak and the observed variations of the blue wing may be indicative of radial motion within the blr
. the blue peak may hence be a `` ripple '' produced by light echoes .
indeed , we expect that the blue peak will be detected by future observations at the same @xmath20 therefore ruling out the binary black hole model altogether ( as in the case of 3c 390.3 ; @xcite ; @xcite ) .
a set of high s / n ratio spectra with a good temporal coverage is necessary in order to test the proposed hypotheses .
we thank m. eracleous , t. boroson , m. moles , and p. obrien for allowing us access to their spectra in the digital form and the anonymous referee for suggestions that improved the paper .
we are also indebted to j. sulentic and m. calvani for suggesting this work and for carefully reading the manuscript .
pr acknowledges financial support from the universit di padova and the fondazione ing .
aldo gini during her visit at university of alabama .
pr also acknowledges the financial support of the department of astronomy at ohio state university .
this research has made use of the nasa / ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , caltech , under contract with the national aeronautics and space administration .
baldwin , j. a. 1997 , in emission lines in active galaxies : new methods and techniques , ed .
b. m. peterson , f .- z .
cheng , & a. s. wilson ( san francisco : astronomical society of the pacific ) , asp conference series , vol .
113 , 80 baum , s. a. , & heckman , t. m. 1989 , apj , 336 , 702 becker , r. h. , white , r. l. , & helfand , d. j. 1994 , in astronomical data analysis software and systems iii , ed .
d. r. crabtree , r. j. hanisch , & j. barnes ( san francisco : astronomical society of the pacific ) , asp conference series , vol .
61 , 165 benitez , e. , dultzin - hacyan , d. , heidt , j. , sillanp , a. , nilsson , k. , pursimo , t. , teerikorpi , p. , & takalo , l. o. 1996 , apj , 464 , l47 boroson , t.a .
, & green , r.f . 1992 , apjs , 80 , 109 brotherton , m. s. 1996 , apjs , 102 , 1 chen , k. , & halpern , j. p. 1989 , apj , 344 , 115 corbin , m. r. , & boroson , t. a. 1996 , apjs , 107 , 69 durret , f. , pecontal , e. , petitjean , p. , & bergeron , j. 1994 , a&a , 291 , 392 eracleous , m. , & halpern , j. p. 1994 , apjs , 90 , 1 eracleous , m. , livio , m. , halpern , j. p. , & storchi - bergmann , th .
1995 , apj , 438 , 610 eracleous , m. 1997 , adv space res , in press eracleous , m , halpern , j. p. , gilbert , a. m. , newman , j. a. , & filippenko , a. v. 1997 , apj , in press francis , p. j. , hewett , p. c. , foltz , c. b. , chaffee , f. h. , weymann , r. j. , & morris , s. l. 1991 , apj , 373 , 465 gaskell , c. m. 1983 , in quasars and gravitational lenses , proc .
24th lige astrophysical colloq .
( lige : inst .
dastrophysique , universit de lige ) , 473 gaskell , c. m. 1996a , apj , 464 , l107 gaskell , c. m. 1996b , in jets from stars and galactic nuclei , ed .
w. kundt ( berlin : springer ) , 165 ghisellini , g. , padovani , p. , celotti , a. , & maraschi , l. 1993 , apj , 407 , 65 halpern , j. p. , & eracleous 1994 ,
apj , 433 , l17 jackson , n. , penston , m. v. , & prez , e. 1991 , mnras , 249 , 577 lehto , h. j. , & valtonen , m. j. 1996 , apj , 460 , 207 marziani , p. , sulentic , j. w. , calvani , m. , perez , e. , moles , m. , & penston , m. v. 1993 , apj , 410 , 56 marziani , p. , sulentic , j. w. , dultzin - hacyan , d. , calvani , m. , & moles , m. 1996 , apjs , 104 , 37 marziani , p. , calvani , m. , & sulentic j. w. 1997 , in accretions and related outflows , ed .
d. wickramasinghe , l. ferrario , & g. bicknell ( san francisco : astronomical society of the pacific ) , asp conference series , vol . 121 , in press netzer , h. , & peterson , b. m. 1997 , in astronomical time series , ed .
d. maoz , a. sternberg & e. m. leibowitz ( dordrecht : kluwer ) , in press obrien , p.t . , & gondhalekar , p.m. 1991 , mnras , 250 , 377 padovani , p. , & urry , c. m. 1992 , apj , 387 , 449 penston , m. v. 1991 , in variability of active galactic nuclei , ed . h. r. miller , & p. j. wiita ( cambridge : cup ) , 343 peterson , b. m. 1993 , pasp , 105 , 247 quinlan , g. d. 1996 , new astronomy , 1 , 35 romano , p. , zwitter , t. , calvani , m. , & sulentic j. w. 1996 , mnras , 279 , 165 sillanp , a. , et al .
1996 , apj , 305 , l17 sulentic , j. w. 1989 , apj , 343 , 54 sulentic , j. w. , marziani p. , zwitter , t , & calvani , m. 1995 , apj , 438 , l1 veilleux , s. , & zheng , w. 1991 , apj , 377 , 89 wanders , i. , & peterson , b. m. 1996 , apj , 466 , 174 llcccccccc la palma & 1988 may 5 & 2.5 & ids & 8000 & 1.6 & @xmath68 & 3 & 235 camera & a kpno & 1990 sep 20 & 2.1 & gold & 2700 & 1.5 & 0 & 6.57 & ti 800 @xmath69 800 & b kpno & 1992 jul 8 & 4.0 & @xmath68 & 1200 & 1.7 & 90 & 6 & tek 1024 @xmath69 1024 & c calar alto & 1994 apr 11 & 2.2 & b&ch & 4800 & 2.0 & 90 & 5.68 & tek 1024 @xmath69 1024 & d calar alto & 1995 may 1 & 2.2 & b&ch & 5400 & 1.6 & 90 & 6 & tek 1024 @xmath69 1024 & e asiago & 1996 apr 20 & 1.8 & b&ch & 7200 & 2.0 & 90 & 2 &
thomson 7882 & e san pedro martir & 1996 may 20 & 2.2 & b&ch & 8400 & 2.12.6 & 90 & 610 & ccdtek & e [ tab : observations ] lccccc 2447287 & 94 & 6.72 & 3.49 & 3.21 & 0.73 + 2448155 & 124 & 5.27 & 2.40 & 2.87 & 0.43 + 2448812 & 105 & 5.23 & 2.44 & 2.79 & 0.49 + 2449454 & 87 & 4.76 & 2.52 & 2.24 & 0.58 + 2449839 & 65 & 5.18 & 2.87 & 2.31 & 0.71 + 2450194@xmath70 & 75@xmath70 & 4.65@xmath70 & ... & ... & 0.49 + 2450224 & 120 & 5.03 & 2.37 & 2.66 & 0.53 + [ tab : lines ] | we present spectroscopic observations of the quasar pg 1512 + 370 , covering the line spectral range and collected at moderate resolution ( 2 - 7 fwhm ) from 1988 to 1996 .
the observations show that the blue wing of the broad profile component has changed significantly in flux and shape between 1988 and 1990 and between 1995 and 1996 .
a displaced blue peak on the profile , visible in 1988 , but not in the 1990 - 1995 spectra , is revealed again in one of the spectra obtained in 1996 .
the blue peak ( in both the 1988 and 1996 spectra ) is centered at @xmath0 from the rest frame defined by the narrow component of , and the @xmath14959,5007 lines .
we discuss several conflicting interpretations of the data .
we find that the variability of the blue wing is consistent with balmer line emission from regions whose motion is predominantly radial , if variations of the blue wing are a response to continuum changes . alternatively
, we note that observed line profile variations are consistent with a variable line component as in a `` binary black hole '' scenario .
more frequent observations of are needed to distinguish among these hypotheses .
@xmath2 49@xmath24959 |
models of structure formation and galaxy evolution have reached a state where the impact of large - scale baryonic feedback on galaxy evolution can no longer be neglected . however , the physics of such processes are complex , and as yet not very well understood , especially at high redshift , where the most massive galaxies formed most of their stars .
therefore , simulations of the ensemble of galaxies typically incorporate these processes , which originate on scales beyond the resolution of typical large scale simulations , as simple relationships and parameterizations , which are in turn tuned to approximately match observed galaxy properties . in spite of these limitations ,
baryonic feedback plays a crucial role for our understanding of galaxy evolution , perhaps most dramatically at high redshift ( e.g. , * ? ? ?
* ) the impact of feedback related to intense star - formation is now observationally well - established at low and high redshift ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , and simulations improved considerably when including it . but significant and likely fundamental discrepancies remain , e.g. , when comparing the _ observed _ mass function of galaxies with the _ predicted _ mass function of dark matter halos at the upper end of the mass function @xcite . from low - redshift studies , starburst - driven `` superwinds '' are known to be efficient in driving out gas only from the comparably shallow gravitational potentials of low mass galaxies ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
. however , these winds are mainly energy driven , therefore susceptible to radiative losses , and they probably lack the power , efficiency , and velocities necessary to remove significant gas fractions from the deep potential wells of the most massive galaxies and their dark - matter halos . as a consequence
, they can not explain the overabundance particularly of massive dark - matter halos compared to the co - moving galaxy density .
therefore another powerful feedback mechanism is recently gaining in popularity among researchers modeling and simulating the ensemble properties of galaxies , which are vigorous outflows driven by powerful active galactic nuclei at high redshift ( agn ; e.g. , * ? ? ?
* ; * ? ? ?
energy- and momentum - driven agn winds may reach very high energy output through radiation pressure and particle ejection .
the most powerful agn may eject enough energy and momentum to unbind significant fractions of the total interstellar medium ( ism ) of the host galaxy . as they slow and expand , such winds are able to efficiently heat gas in the inter - cluster medium ( icm ) and even the intergalactic medium ( igm
; e.g. , * ? ? ?
however , due to the clumpy nature of the ism in galaxies , rather small working surface of the jet , and agn producing `` light jets '' , they are not generally thought to couple strongly with the ambient ism in their hosts @xcite .
the observed characteristics of massive galaxies support the hypothesis that strong agn winds at high redshift may have a significant impact on galaxy evolution and their ensemble characteristics .
spheroidal and black hole masses in nearby massive systems follow the tight m@xmath6-@xmath7 relation @xcite , which perhaps signals co - eval growth of smbh and the bulge self - regulated through negative agn feedback @xcite .
the most massive galaxies , virtually all spheroidals , are `` old , red , and dead '' .
they are metal - rich and have luminosity weighted ages consistent with massive star - formation at high - redshift and subsequent passive evolution , perhaps indicating that the gas was effectively removed during the most massive burst ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
maybe the tightest constraint comes from the metal abundance ratios , especially @xmath8 $ ] , which indicate intense star - formation truncated by strong feedback within less than few @xmath9 yrs ( e.g. , * ? ? ? * ) .
in addition , the exponential cut - off in the bright end of the galaxy luminosity function may be a direct result of the efficiency of agn feedback @xcite .
furthermore , @xcite find evidence that more moderate feedback of radio - loud agn might balance gas cooling in the halos of the most massive galaxies at low redshift . a promising way to identify the fingerprints of a highly efficient mode of agn feedback at high redshift is by observing the gas kinematics in rapidly growing , massive galaxies at high redshift which host agn .
powerful high - z radio galaxies ( hzrgs ) fit these requirements : they reside in dense environments @xcite .
their magnitudes , colors , and continuum morphologies are consistent with large stellar masses and passive evolution at @xmath10 @xcite .
hzrgs at @xmath11 also contain large masses of molecular gas ( @xmath12 ; e.g. , * ? ? ?
* ) , dust ( @xmath13 ; e.g. , * ? ? ?
* ; * ? ? ?
* ) , and uv / optical line - emitting gas ( @xmath13 ; e.g. , * ? ? ?
at the highest redshifts , they are also likely to be forming stars at prodigious rates ( @xmath14 yr@xmath15 ; e.g. , * ? ? ?
curiously , hzrgs are known from longslit - spectroscopy to often have spatially extended broad emission lines with fwhm@xmath16 , indicative of extreme kinematics ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
however , with the slits typically aligned along the radio axis , the global role of these kinematics could not be examined more closely , highlighting the need for integral - field observations of hzrgs , preferably at rest - frame optical wavelengths ( i.e. , in the observed near - infrared ) to investigate the overall gas dynamics .
an excellent candidate for such a study is the @xmath17 mrc1138 - 262 , with @xmath18 and stellar mass @xmath19 .
vla interferometry reveals 2 radio jets with complex morphologies @xcite .
its nebulosity extends over @xmath20 kpc radius , @xmath21 , @xmath22 cosmololgy , size scales at @xmath17 are 8.3 kpc arcsec@xmath15 , the luminosity distance is 17070 mpc , the angular size distance is 1710 mpc .
cosmic age at that redshift is 3.0 gyrs . ] , and has a significant surrounding over - density of emitters which indicates it might be in the center of a forming galaxy cluster @xcite .
rest - frame uv spectroscopy implies star - formation rates of @xmath23 yr@xmath15 . by examining the spatially - resolved kinematics and spectral properties of the rest - frame optical emission line gas in mrc1138 - 262
, we wish to investigate the influence of the agn on the excitation and kinematics of the emission line gas using data obtained with the near - infrared integral - field spectrograph spiffi .
we observed mrc1138 - 262 in the near - infrared h and k bands using the integral - field spectrograph spiffi on ut1 of the vlt in april 2003 .
spiffi uses an image slicer to dissect the 8@xmath248 field of view on the sky into 32 slices or `` slitlets '' with individual pixel sizes in projection of 0.25 .
the spectrograph and image slicer have since become a facility instrument at ut4 as part of sinfoni . under good and stable weather conditions , we obtained a total of 140 min of data in k and 75 min in h with a pixel scale of 0.25and seeing of @xmath25 .
individual exposure times were 300s in h and 600s in k , respectively .
spectral resolution is r=@xmath26 at 1.6 and r=@xmath27 at 2.2 .
one `` off '' frame at a sky position well displaced from the target was taken for each `` on '' frame in an off - on - on - off pattern . to reduce the data , we used the iraf @xcite standard tools for the reduction of longslit - spectra , modified to meet the special requirements of integral - field spectroscopy .
data are dark frame subtracted and flat - fielded . for bad pixel correction we use the full three - dimensional information , identifying pixels that are 5@xmath7 deviant in the darks and flats .
rectification and wavelength calibration are done before night sky subtraction , to account for some spectral flexure between the frames .
curvature is measured and removed using an arc lamp , before shifting the spectra to an absolute ( vacuum ) wavelength scale with reference to the oh lines in the data .
the three - dimensional data cubes are then reconstructed , assuming that each slitlet covers exactly 32 pixels .
they are spatially aligned by cross - correlating the collapsed cubes , and then combined , clipping deviant pixels .
telluric correcton is applied to the combined cube .
flux scales are obtained from standard star observations . from the light profile of the standard star
, we measure the fwhm spatial resolution to be @xmath28 in right ascension and declination , respectively .
in fig . [ fig : specnuke ] we show h and k - band spectra of mrc1138 - 262 .
the upper panel shows the spectrum of the spatial pixel covering the agn , the lower panel shows the agn removed extended emission integrated over the central 30 kpc .
line profiles are not simple gaussians , but have irregular profiles indicating complex kinematics in the extended emission line gas .
the lines are generally broad , and [ nii]6583 and are blended , so that is not a reliable kinematic tracer . therefore we use [ oiii]5007 ( with fwhm@xmath29 in the integrated spectrum ) to trace the gas kinematics .
we discuss below that the emission line kinematics does not vary strongly with the specific emission line analyzed ( [ sec : iontempdens ] ) , within the accuracy we can reach given the line blending .
thus we do not expect and there is no evidence that this choice has a strong impact on our overall conclusions .
the luminous broad emission line is the most prominent spectral feature in the agn spectra , with fwhm@xmath30 , and a rest - frame equivalent width , eqw@xmath31 , indicating that the emission from the broad line region can not be neglected ( especially in the broad band near infrared magnitudes ) .
previous rest - frame uv spectroscopy of mrc1138 - 262 did not reveal nuclear broad lines , indicating that mrc1138 - 262 might harbor an obscured quasar , similar to mrc2025 - 218 @xcite .
broad emission in the h band is measured at much fainter flux : from the balmer decrement in the broad components , @xmath32 , using a galactic extinction law and @xmath33 .
the galactic extinction law might not be strictly adequate for broad agn emission lines , nonetheless , the large extinction underlines the interpretation that mrc1138 - 262 harbors an obscured quasar . to correct the k - band data for the agn contribution
, we fit and remove the nuclear spectrum from all other spatial pixels .
this may lead to a slight oversubtraction of extended line emission along the line of sight of the agn . in our later analysis , we aim at setting lower limits to the emission line kinematics and fluxes , therefore this procedure is nonetheless appropriate . in the k - band , we can use the luminous broad h@xmath5 emission line as an indicator for how well our algorithm removes the agn light .
agn removed spectra are shown in the lower panel of fig .
[ fig : specnuke ] to illustrate the success of the method .
to investigate the spatially resolved kinematics , we extracted spectra from 3@xmath243 pixel box apertures ( @xmath34 ; slightly larger than the estimated seeing in both reduced data cubes ) .
using the iraf task splot , we fitted [ oiii]5007 line emission with up to 3 kinematic components , requiring @xmath35 in each component . systematic errors of these measurements are possibly large , and the agn removal in the h band might be not very precise .
we thus use these spectra only to identify zones of rather uniform kinematics , and extract integrated spectra of these zones to further investigate their properties .
relative velocities of all [ oiii]5007 components are shown in fig .
[ fig : velvsdistplot ] as a function of projected radius .
they appear to be segregated by velocity into roughly 4 groups with sizes of @xmath36 ( @xmath37 kpc ) over the total radius of @xmath38 ( 20 - 25 kpc ) .
overlap between these groups in velocity is small . in fig .
[ fig : velocitymap ] we show the map of relative velocities of the brightest [ oiii]5007 line component in each spatial pixel .
at least three distinct kinematic zones , each separated in velocity space by @xmath39 1000 , but with rather uniform velocities internally , are evident .
velocities in zone 1 decline with increasing projected distance from the nucleus , gradually approaching the values in zone 2 at the eastern bound .
this is in agreement with fig .
[ fig : velvsdistplot ] , which also suggests that the velocities in zone 1 at its low - velocity end approach the velocities in zone 2 .
zone 3 has largely uniform intrinsic velocities , whereas velocities in zone 2 vary gradually with position angle , reaching a maximum at a position angle @xmath40 .
the relatively uniform velocities within each structure , ( fig .
[ fig : velocitymap ] ) , mirroring the regularity found in fig .
[ fig : velvsdistplot ] , are highly reminiscent of what would be expected for overlapping , edge - brightened `` bubbles '' ( such a scenario is also favored by the low filling factor of the emission line gas , [ sec : timescale ] and [ sec : heatgas ] ) .
however , we are using the term `` bubble '' somewhat loosely , referring mainly to the clearly distinct zones in velocity space .
their morphologies are nearly circular as if projected on the sky , and are consistent with expanding spheres or projected conical outflows
. they might be analogues of buoyantly rising , low - density regions in low - redshift radio galaxies , but our analysis does not rely on this interpretation . by using this term , we simply relate to the geometry and velocity structure of the gas , not necessarily the full astrophysical interpretation .
this interpretation is also supported by the spatial distribution of line widths . in fig .
[ fig : sigmamap ] we show a map of the fwhm of the brightest [ oiii]5007 component in each pixel .
line widths vary between fwhm@xmath41 .
most striking is a ring of broad line emission ( fwhm@xmath42 ) around zones 1 and 3 , shown as hatched area in the velocity map of fig .
[ fig : velocitymap ] .
this ring supports the above interpretation that the emission line regions with well separated velocities are geometrically and physical distinct entities , which originate from apparent broadening due to overlap between adjacent bubbles or from turbulence due to interactions along their boundaries .
integrated h and k band spectra of the 3 definable bubbles are shown in fig .
[ fig : z13spec ] .
emission lines are bright in all bubbles .
line profiles are complex , and and [ nii]6583 are blended in all zones . to obtain a consistent fit for all lines in a given zone
, we identify kinematic subcomponents from the [ oiii]5007 profile , and restrict all other line fits to have the same relative velocities and widths .
we leave the flux as a free parameter , except for the ratios of the [ nii]6548,6583 and [ oiii]4959,5007 emission line doublets , where we fix the line ratios to the ratios of their transition probabilities .
our data also include the [ sii]6717,6731 line doublet , which has line ratios sensitive to electron densities in the range of 100 to @xmath43 @xmath44 @xcite .
we identify two @xmath45 pixel apertures in bubble 2 where the two components are sufficiently narrow to be deblended robustly .
line ratios in these apertures are @xmath46 } \equiv f_{\lambda6716}/f_{\lambda6731}=
1.1\pm 0.1 $ ] .
this corresponds to a density range @xmath47 @xmath44 , and @xmath48 @xmath44 at best fit value , assuming a temperature @xmath49 k. ( the results of * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* indicate a most likely temperature range of @xmath50 k in agn narrow line regions , which is sufficiently accurate for our purpose . )
@xcite find densities of @xmath51 @xmath44 in the precursor regions of fast radiative shocks , constrained by radio observations , in good agreement with our result using an independent estimate .
the [ sii ] emission originates mainly in a `` partially ionized zone '' , where the thermal pressure is @xmath52 few @xmath2410@xmath53 n@xmath54 dynes @xmath55 and @xmath56 ( e.g * ? ? ? * ) .
we use these correction factors in calculations that make use of the total gas density , e.g. , the total mass estimate based on the recombination line flux of hydrogen .
moreover , we only use the line flux with the agn contribution removed , as described in [ sec : spectra ] .
the blending of [ nii ] and complicates the flux estimates , therefore we only aim at estimating robust lower limits to the kinetic energy in the extended emission line region , and hence minimize the contribution in our fits to the [ nii]@xmath57 line blend , typically obtaining [ nii]@xmath58 . with the adopted cosmology ,
the total observed flux in all zones corresponds to a total luminosity @xmath59 ergs s@xmath15 . to estimate the ionized gas mass in the diffuse nebula
, we assume case b recombination @xcite , and our previous gas density estimates : @xmath60 with luminosity @xmath61 in units of @xmath62 ergs s@xmath15 , planck constant @xmath63 , frequency @xmath64 , effective recombination coefficient , @xmath65 , proton mass @xmath66 , and electron number density @xmath67 in units of 100 s@xmath15 . with the electron density of @xmath68 @xmath44 ( [ sec : iontempdens ] ) this implies a total ionized gas mass @xmath69
. our analysis can be precise at the order - of - magnitude level only , and is dominated by astrophysical rather than the measurement uncertainties .
therefore we will adopt an hii mass of @xmath70 , corresponding to the measured luminosity and electron density at face value and not corrected for extinction . thus
this estimate is a robust low limit to the intrinsic luminosity and total emission line gas mass .
mrc1138 - 262 is not the first hzrg known to have broad , spatially resolved emission line gas ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , but we have not found another such galaxy in the literature with integral - field spectroscopy in the rest - frame optical . for galaxies with spatially resolved slit spectroscopy ,
the slit is typically aligned with the radio axis , so that the morphology of the high - velocity emission line gas and total energy and mass in the flow are difficult to estimate .
gas dynamics dominated by the large - scale gravitational potential of the host galaxy would have smooth velocity gradients and rather uniform line widths with fwhms of a few hundred .
@xcite discuss the relatively quiescent kinematics in the ly@xmath5 halo of the z=2.49 hzrg mrc2104 - 242 , with fwhm@xmath71 , and argue that the kinematics might be caused by rotation or infalling gas .
the kinematics we observe in mrc1138 - 262 , however , are different from mrc2104 - 242 , quantitatively and qualitatively .
our integral - field observations are uniquely suited for a more detailed investigation of what causes the extreme kinematics in mrc1138 - 262 , by relating the local bulk and turbulent velocities in the gas .
the right panel of fig .
[ fig : z13spec ] shows the fwhms as a function of velocity in each bubble .
velocities are relative to the mean velocity in each bubble .
fwhms and velocities are uncorrelated in zone 2 , in zone 1 and 3 they correlate tightly with correlation coefficients @xmath72 and @xmath73 , respectively .
positive correlations like in zone 3 are also found in starburst - driven winds , and indicate outflowing gas being accelerated and becoming more turbulent as the kinetic energy is being dissipated through shocks @xcite .
negative correlations like in zone 1 have been interpreted as deceleration shocks @xcite , when gas heated by the agn collides with the confining , higher density ism .
starburst powered outflows do not appear as good candidates to power an outflow with the velocity gradients and line widths observed .
typical outflow velocities in superwinds , as similarly estimated from optical emission lines , are a few@xmath74 @xcite .
these velocities are a factor of a few lower than in mrc1138 - 262 ( not including any correction for projection effects ) .
the velocities seen in starburst driven winds do not seem to be a function of either energy injection power or rate @xcite or perhaps redshift @xcite .
@xcite find that the outflow observed in the @xmath75 strongly star - forming ( sfr@xmath76 yr@xmath15 ) submillimeter galaxy smmj14011 + 0252 does not exceed velocities of @xmath77 and fwhms @xmath78 .
most importantly for this study , the line cores in smmj14011 + 0252 are not offset from systemic by large velocities .
in other words , the overall , large scale kinematics of smmj14011 + 0252 do not appear to be strongly influenced by the wind , similar to the integrated line profiles of local starburst galaxies .
moreover , pressures within the wind are similar to winds in local starburst galaxies , suggesting similar physical properties , which make it unlikely that the outflow properties would dramatically change with redshift ( see arguments in , e.g. , * ? ? ?
nonetheless , the presence of individual bubbles and large and irregular velocity offsets in mrc1138 - 262 do suggest that some sort of feedback mechanism is driving a massive gas outflow .
we will analyze the properties of this outflow in close analogy to @xcite and show that energy and timescale arguments strongly favour a wind driven by the agn as the only plausible explanation .
our data indicate that the gas reaches large distances , over 25 kpc for the high surface brightness emission , and might even completely escape the host potential . a way of estimating the time scale of the flow is to calculate the crossing time given the observed velocities .
with bulk motions of @xmath79 and an average observed radius of the emission line nebula @xmath80 kpc , the dynamical timescale is @xmath81 yrs , providing a rough timescale necessary to drive the flow over the observed distances .
if the outflow is powered by the radio source , then this timescale will be approximately similar to the age of the radio source . with estimated jet - head advance speeds of @xmath390.1c ( e.g. , * ? ? ? * with ranges of about 0.01c to about 0.2c ) , and the `` largest angular size '' of the radio source at @xmath82=1.41 ghz in mrc1138 - 262 ( 11.1 , or 92 kpc ; * ? ? ?
* ) , we estimate a dynamical timescale of @xmath393@xmath2410@xmath83 @xmath84 yrs . because of the unknown projection angle , this can only be a lower limit ( hence the factor @xmath84 ) , and the intrinsic timescale is likely a few times higher , up to a few @xmath2410@xmath85 yrs .
@xcite suggest that the ages of radio sources decrease with increasing redshift .
however , they use a flux limited sample , so that the average radio power will increase with redshift , whereas the size decreases , which might at least in part explain this trend .
using their relationship between radio power and redshift for the @xmath86 of mrc1138 - 262 implies an age of @xmath390.4 to 1@xmath2410@xmath85 yrs , similar to our previous estimate . in either approach
, the age of the radio source is not significantly different from the crossing time of the emission line nebula , and the radio jet and lobes reach a factor of @xmath392 larger radii than the ( detected ) high surface brightness emission line gas .
this alone can not prove that the radio jet is powering the outflow , but it is a necessary prerequisite . to estimate a mass outflow rate
, we simply assume that the outflow is being accelerated with an efficiency that is approximately constant during the jet lifetime .
this implies that the current estimate is representative for the average rate over the lifetime of the source .
given the large velocities and size of the nebula , this assumption is reasonable .
this very simple scenario implies @xmath87 yr@xmath15 .
of course , we do expect variations in the instantaneous outflow rate due to jet precession , variable jet energy or mass density , or perhaps changing coupling efficiency to the ism as might be expected for a jet interacting with a `` clumpy '' surrounding medium .
having measured the electron density and relative ionization ( assuming case b recombination ) , we can also constrain the volume filling factor of the emission line gas .
the total observed luminosity corresponds to a `` total emission volume '' @xmath88 , which we compare to the volume of the emission line nebula @xmath89 ( @xmath90 kpc@xmath91 , approximated by the total volume of 3 spherical bubbles with the observed radii of each zone ) .
this yields a filling factor @xmath92 .
we find a filling factor , @xmath93 , likely signaling that distributed in bright , line - emitting sheets , small clouds , and/or filaments .
low filling factors arise naturally when a sufficiently intense , expanding hot wind sweeps up , accelerates , and ionizes dense hydrogen clouds within the ism of the host galaxy .
relative to the lifetime of the host , an outflow with lifetime @xmath94 yrs is nearly explosive , implying impulsive energy injection ( but constant over the lifetime of the event ) . following , e.g.
, @xcite , expansion speed of the shell and energy content of the bubble are related as : @xmath95 where @xmath96 is the ( constant ) energy injection rate in units of 10@xmath97 ergs s@xmath15 , @xmath98=0.5 @xmath44 is the ambient ism density in @xmath44 and @xmath99 is the injection time in units of 10 myrs .
we approximate the velocity @xmath100 through the average of velocities in the line profile fits to the 3 bubbles discussed in
[ ssec : kinematics ] . from the analysis
, we find an average projected velocity of @xmath101 800 .
the required energy is then @xmath394.2@xmath2410@xmath102 ergs s@xmath15 . in a second approach , we model the ( constant ) energy injection rate into an energy conserving bubble expanding into a uniform medium with ambient density @xmath98 , which serves as a plausible upper bound to the energy injection rate , @xmath103 with the radius @xmath104 given in units of 10 kpc and velocity @xmath105 given in units of 1000 km s@xmath15 .
we estimated the half - light radius of the h@xmath5 emission , after removing the contribution from the broad line region , and found @xmath5210 kpc . for 10 kpc ,
a flux weighted mean velocity @xmath106 and ambient density @xmath107@xmath44 , the energy injection rate is @xmath108 ergs s@xmath15 .
this is likely a minimum energy required since the velocities are seen in projection .
if we use the total extent of the nebula for the radius , we find the energy injection rate is @xmath109 ergs s@xmath15 . alternatively
, if the bubble is momentum conserving , we estimate that , @xmath110 or @xmath111 dyn , of force must be injected to power the observed flow .
if the jet or radiation field of the agn is to be a viable driver of the observed gas dynamics in mrc1138 - 262 , it must inject at least 10@xmath112 ergs s@xmath15 of energy or alternatively , about 10@xmath113 dynes of force into the ambient interstellar medium to produce the outflow in mrc1138 - 262 .
each of these estimates is valid within a given model , although the underlying assumptions are very general , involving basic physical principles .
we can give a fully model - independent , strictly observational lower bound to the energy injection rate from the observed kinematics of the emission line regions . to this end
, we only account for the observed ionized gas masses , not correcting for projection effects lowering velocity measurements , extinction , and material in other phases , especially x - ray emitting or molecular gas , which will contribute significantly . with the measured gas masses and relative velocities ,
the kinetic energy is simply @xmath114 for the bulk velocities and @xmath115 for turbulent motion . for mrc1138 - 262 and summing over the velocities and dispersions in the individual bubbles , we estimate @xmath116 ergs , and @xmath117 ( with unknown correction for projection , @xmath118 ) . for our dynamical timescale of @xmath119 yrs
, this corresponds to an average injection rate @xmath120 ergs s@xmath15 over a lifetime @xmath94 yrs .
the results illustrate that energy injection rates at high redshift are not trivial to estimate , which limits the precision of our analysis to an order - of - magnitude level .
it is however encouraging that in spite of the large uncertainties , the values are broadly consistent .
we will in the following adopt e@xmath121 erg s@xmath15 .
this choice is motivated by the approximate range of values indicated by the simple estimates we have made and to the order - of - magnitude spirit in which they were made .
if the agn is to be a viable candidate for powering the observed outflow , then it must be able to at least provide the above energy injection rate .
it is difficult to know _ a priori _ what is powering the outflow , the overall bolometric luminosity , @xmath122 , which might couple to the ism through radiation pressure on dust grains , or directly through the kinetic luminosity of the jet @xmath123 .
we shall start by estimating @xmath122 .
following the scaling of x - ray to bolometric luminosity of @xcite and the x - ray luminosity of mrc1138 - 262 at @xmath124 kev , ( @xmath125=45.6 ergs s@xmath15 ; * ? ? ?
* ) , we estimate @xmath126 ergs s@xmath15
. we can also use the relationship between @xmath122 and rest - frame optical flux density of the agn in qsos @xcite , and the 5100 continuum flux in our h band spectrum of mrc1138 - 262 , @xmath127 w @xmath128 m@xmath129 , yielding @xmath130 ergs s@xmath15 . given the approximate nature of either estimate ,
these values are consistent .
we will in the following adopt @xmath131 ergs s@xmath15 since the estimate using the optical continuum is likely to be strongly affected by high extinction to the region of agn continuum emission in radio galaxies compared to qsos . estimating the jet kinetic luminosity as a measure of the power that the jet is capable of injecting in the ambient ism is more difficult and challenging and many authors have tried a variety of methods for estimating or constraining the mechanical jet luminosities .
we will therefore use and compare a variety of methods to estimate the mechanical luminosity of the jet in mrc1138 - 262 .
@xcite empirically calibrate the radio source characteristics in the 3cr as a function of radio power at 178 mhz . for mrc1138 - 262 the vla measurements at 4.86 ghz and the original selection of the molonglo reference catalog ( mrc ) at 408 mhz ( * ? ? ?
* and references therein ) , imply a radio spectral index @xmath5=1.34 , and a 178 mhz radio power , p@xmath132=10@xmath133 ergs s@xmath15 hz@xmath15 sr@xmath15 , in the rest - frame .
this corresponds to a jet kinetic luminosity of 10@xmath134 ergs s@xmath15 in the calibration of @xcite .
@xcite measured an integrated luminosity of the radio jet in mrc1138 - 262 in the rest - frame 0.1 - 1.0 ghz band , @xmath135 ergs s@xmath15 .
based almost solely on theoretical arguments , the ratio of mechanical to radio luminosity is typically argued to be about 10 to 1000 ( e.g. , * ? ? ?
* ; * ? ? ?
estimating that the radio luminosity is equal to @xmath136 , @xmath137 , of the total jet luminosity sets a lower limit to the jet kinetic luminosity of 10@xmath138@xmath137 ergs s@xmath15 .
obviously , it could be substatially higher .
this is , however , within the scatter of the @xcite estimate .
we can additionally estimate the jet power using analyses of x - ray cavities which are thought to be powered by the mechanical energy of the radio jet interacting with the surrounding hot gas .
following the analysis in @xcite , we estimate the mechanical luminosity of the jet in mrc1138 - 262 to be about 10@xmath139 ergs s@xmath15 .
we think that this estimate is valuable because it is similar to our approach and provides a robust lower limit to the mechanical luminosity .
however , @xcite study radio sources at lower redshift , which have different radio morphologies than mrc1138 - 262 and typically lower radio luminosity .
moreover , @xcite point out that their estimates might be systematically too low due to various physical effects . as a last approach , we can estimate the kinetic energy necessary to explain the observed expansion speeds of radio lobes which are of the order of 0.1 to 0.4c ( e.g. , * ? ? ?
these advance speeds and a plausible model for the medium into which the lobes are advancing , yields estimated jet kinetic powers of 10@xmath140 ergs s@xmath15 for powerful , high luminosity radio sources @xcite . for the subsequent analysis
, we will adopt a finducial jet kinetic luminosity of 10@xmath112 ergs s@xmath15 , but suggest that the true value could be a factor of a few higher .
thus , both jet power and bolometric luminosity estimates provide sufficient energy to power the observed outflow . both appear similarly powerful , thus their relative impact will depend on their coupling efficiency with the ambient ism and their ability to accelerate ambient interstellar material to large distance from the nucleus .
the importance of radiation pressure in driving an outflow depends on the intensity of the radiation field itself , i.e. , the bolometric luminosity , @xmath122 .
powerful agn have radiation fields that are likely intense enough to efficiently heat and accelerate the gas near their center through compton heating @xcite and dust opacity @xcite . through coulomb coupling of charged dust grains to the ionized gas
, radiation pressure can in principle trigger an outflow .
blue shifted x - ray absorption lines in quasar spectra indeed suggest parsec scale outflows , with mass loss rates of @xmath391 m@xmath141 yr@xmath15 and velocities @xmath390.l c ( e.g. , * ? ? ? * ) .
however , e.g. , @xcite find that this gas generally can not drive large scale outflows .
we describe a momentum driven agn wind , following , e.g. , @xcite and @xcite .
the equation of motion for an optically thick wind is , @xmath142 with acceleration @xmath143 , and agn luminosity @xmath144 .
@xmath145 denotes the gas mass as a function of radius .
@xcite argue that the velocity as a function of the distance for the black hole , @xmath146 , is set by , @xmath147 with a threshold luminosity @xmath148 above which the agn is able to drive a significant , optically thick outflow , gas fraction @xmath149 , velocity dispersion of the host galaxy @xmath7 , and gravitational constant @xmath150 .
@xmath151 is the radius where the acceleration sets on , namely the dust sublimation radius at a few pc from the black hole .
@xmath148 can be derived in close analogy to the eddington limit for the central agn , @xmath152 .
( see 5 of * ? ? ? * for the full derivation of this threshold . ) . for a @xmath153 galaxy , @xmath154 ergs s@xmath15 .
with @xmath155 ergs s@xmath15 , a reasonable gas fraction of 10 - 50% , and a velocity dispersion of 250 km s@xmath15 ( roughly corresponding to the stellar mass of mrc1138 - 262 ) , this suggests that @xmath156 ergs @xmath157 , well above the bolometric luminosity of mrc1138 - 262 . for a lower gas fraction and smaller velocity dispersion , radiation pressure might play a role , but it unlikely dominates in accelerating the gas to the high velocities we observe .
for instance , if @xmath158=@xmath159 ( in good agreement with the @xmath160 implied by * ? ? ?
* ) , the acceleration would have to have been maintained for a few @xmath24 10@xmath161 yrs in order to accelerate the emission line gas in mrc1138 - 262 to the observed velocities , likely about 1 - 2 orders of magnitude longer than the agn lifetime .
typical line velocities would be about @xmath162 those observed , perhaps less since we can not accurately quantify the effect of projection .
hence , radiation pressure does not seem sufficient to power the large - scale gas kinematics in mrc1138 - 262 , making the radio jet a more likely candidate to mediate the agn feedback .
we derive a simple estimate for the observed coupling efficiency @xmath163 between the energy input rate @xmath164 ergs s@xmath15 of the radio jet ( [ subsec : mrc.lbol ] ) and the energy injection rates derived in [ sec : injection ] , @xmath165 erg s@xmath15 . given the uncertainties in either estimate , and in the spirit of providing lower limits , we adopt a loosely constrained coupling efficiency @xmath166 , corresponding to the jet kinetic energy estimate common to all methods of [ subsec : mrc.lbol ] .
the `` dentist drill model '' @xcite is the simplest scenario explaining how the jet might interact with the ambient ism , which uses a termination shock to couple with the ism as the jet - head ploughs through the ambient medium .
small variations of the jet direction cause the jet to jitter across the working surface , and the resulting cavity will be somewhat larger than the jet diameter itself .
typical assumed values are @xmath167 @xcite .
thus in the `` dentist drill '' scenario , the interaction between radio jet and the surrounding medium is limited to the working surface of the jet , and the coupling efficiency is approximately the area of the working surface , divided by 4@xmath168 .
this can be parameterized as @xmath169 , or 0.045% for 15opening angle . to explain the @xmath170 erg s@xmath15 kinetic energy injected into the outflow of mrc1138 - 262 with the small @xmath171 implied by the dentist drill model would hence require an energy input of @xmath172 few @xmath173 erg s@xmath15 .
this is above even the most extreme estimates of jet kinetic energies discussed in [ subsec : mrc.lbol ] .
hence the observed kinematics in mrc1138 - 262 require a significantly more efficient coupling to the ambient ism than implied by the `` dentist drill '' model .
line widths and bulk motions in mrc1138 - 262 indicate shock speeds @xmath174 .
optical emission line properties of shocks with such speeds are very difficult to model , because emission arises predominantly through ionizing photons escaping upstream ahead of the shock into the `` precursor '' region . thus , accurately modelling the line emission from the shock depends crucially on the detailed properties of the ambient ism and the radiation field ( both background and that generated by the shock ) . following this argument and the modeling of @xcite , we use , @xmath175 to estimate the amount of shocked material from the measured h@xmath5 luminosity in mrc1138 - 262 ( [ sec : iontempdens ] ) . in this equation , @xmath176 is the number of recombinations per hydrogen atom .
at least @xmath177 yr@xmath15 of shock material are necessary to produce the entire h@xmath5 luminosity , not accounting for extinction and possibly a lower efficiency of the emission ( fewer recombinations per hydrogen atom for example ) . over the likely @xmath119 yrs lifetime of the radio jet ( [ sec : timescale ] ) , this implies an unphysically large @xmath178 of shocked gas .
on the other hand , the predicted surface brightness of a shock with 500 km s@xmath15 velocity is about 2@xmath2410@xmath179 ergs s@xmath15 kpc@xmath129 @xcite , and increases dramatically with velocity ( @xmath180 ) .
shock heating across the projected area of the emission line nebula in mrc1138 - 262 , @xmath181 kpc@xmath182 , would therefore produce a total h@xmath5 luminosity of @xmath392.4@xmath2410@xmath183 , @xmath184% of what we observe .
accounting for extinction will make this discrepancy yet more dramatic .
since the amount of shocked material necessary to power the nebula is unphysically large and the predicted surface brightness is too low , it seems that shock excitation is probably not viable as the sole excitation mechanism of the emission line gas , and ionization by the agn will play a large role .
@xcite reach a similar conclusion from longslit rest - frame uv spectra of 10 hzrgs . to estimate whether photoionization by the agn
can heat the gas , we assume a uniformly dense , radiation bounded nebula in ionization equilibrium . with the properties we observe in mrc1138 - 262 , this requires an ionizing photon rate of @xmath185 , or @xmath186 for @xmath187 .
the results of @xcite imply that on average each ionizing photon in powerful agn has an energy corresponding to @xmath188 times the ionizing potential of hydrogen ( i.e. , @xmath189 ) .
we therefore constrain the ionizing luminosity @xmath190 , corresponding to @xmath191 , or @xmath192 .
if @xmath193% of the bolometric luminosity of mrc1138 - 262 is ionizing @xcite , i.e. , @xmath194 ergs s@xmath15 ( [ subsec : mrc.lbol ] ) , the nebulosity of mrc1138 - 262 might well be mainly photoionized by the agn , if the escape fraction is @xmath195% .
this estimate also implies that the nebula could easily be matter bounded .
however , given the small filling factor ( @xmath196 , [ sec : timescale ] ) , a covering factor @xmath193% implies that the gas is distributed in elongated structures with large surface area and relatively small volume .
the gas must either be clumped into large numbers of small , approximately spherical clouds , long and thin `` string - like '' structures , or in thin , broad sheets , for either shock or excitation of the nebula .
whatever the exact geometry , our results are only consistent with a small volume filling factor , but a relatively large covering fraction .
thus , the mechanical and optical emission line properties of mrc1138 - 262 give direct evidence for an outflow which is driven by the expanding radio source , and heated mainly through photoionization by the agn .
our analysis of the kpc scale outflow in mrc1138 - 262 suggests that the agn is the best candidate for driving the outflow : the crossing time of the nebula ( @xmath197 yrs ) approximately equals the lifetime of the radio jet and typical estimates of agn lifetimes ( * ? ? ?
* and references therein ) , and the agn also provides the energy input rate necessary to power the observed outflow , given it is efficiently coupled to the ism , with @xmath198 . observed mass loss rates are @xmath199 yr@xmath15 ( [ sec : timescale ] ) .
population synthesis fits assuming exponentially declining star - formation histories with a variety of e - folding times ( including infinity for continuous star - formation ) over a wide range of maximum star - formation rates , suggest a stellar mass of @xmath200 .
the range of best fit masses for all of the models suggest an uncertainty of about a factor 2 - 3 .
the observed outflow rate therefore corresponds to a mass loss of @xmath193% of the stellar mass within @xmath201 years .
observations of carbon monoxide in high redshift radio galaxies suggest gas masses of @xmath202 @xcite , corresponding to @xmath203% of the stellar mass of mrc1138 - 262 . given the large uncertainties in these estimates , we assume a gas fraction of @xmath3% of the stellar mass , implying that agn feedback could remove the total ism of the host within @xmath204 yrs .
we emphasize that the mass loss rate is a strict lower bound estimated from the observed emission line gas mass .
the true outflow might be up to an order of magnitude higher , and might well be larger than the total needed to quench future star - formation .
the outflow in mrc1138 - 262 appears not aligned with the radio jets , but has a nearly spherical geometry , so that an adequate assessment of the dynamics in the extended emission line region clearly requires integral - field data , especially in the rest - frame optical , to be less sensitive to the source geometry , extinction , and the complex radiation transport in the rest - frame uv given the complex spatial structure of the emission line gas .
we do not believe that mrc1138 - 262 is a `` one - off '' special case .
broad forbidden emission lines in the nebulosities of hzrgs with fwhm@xmath16 are commonly found in longslit studies ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
simulating a longslit spectrum of mrc1138 - 262 from our spiffi data with the `` slit''aligned with the radio axis , yields similar fluxes as given in , e.g. @xcite .
we recently obtained sinfoni data of a small sample of hzrgs at z=2.5 - 3.5 , and find similarly extreme kinematics in the extended sources .
in the following we will therefore assume that mrc1138 - 262 is in an evolutionary stage , which is short , @xmath205 yrs , but common for massive galaxies , and in particular for those which go through a phase of powerful nuclear activity at high redshift .
a massive , agn driven outflow is a catastrophic event which can be expected to leave measurable imprints on the low - redshift descendents of the agn host galaxy .
semi - analytical models of massive gas - rich mergers with a multi - phase ism in the early universe are found to reproduce the local luminosity function of massive , red early - type galaxies if they include a phase of strong agn feedback @xcite .
@xcite find that these models can explain the tilt in the fundamental plane assuming dissipative , high redshift merging of galaxies with gas fractions @xmath206 , including star - formation and agn - related feedback , and @xcite are able to reproduce the observed m@xmath207@xmath7 relationship using this mechanism . in spite of these successes
, such models can not resolve the black hole accretion , therefore agn feedback is included in a rather schematic way , neglecting , e.g .
, radiatively inefficient black hole growth @xcite .
they assume isotropic thermal coupling between agn and ism with efficiencies of @xmath208% , and imply that star - formation is suppressed within a few @xmath9 yrs .
we note that this efficiency is lower than the coupling efficiency we find in mrc1138 - 262 ( @xmath209 ) , but the time scales agree surprisingly well , given the crudeness of the models and , by necessity , the `` order - of - magnitude '' nature of our analysis .
to set observational constraints on the impact of agn driven winds , we now investigate whether the outflow in mrc1138 - 262 can explain the observed properties of the low redshift massive galaxy population . a massive outflow related to major mergers at high
redshift is in overall agreement with a population of mostly spheroidal galaxies , stellar populations that have luminosity - weighted ages consistent with formation at redshifts @xmath210 @xcite , and with extremely low gas contents .
a more quantitative constraint comes from the relative enhancement of @xmath5 elements with respect to the iron content of these galaxies , @xmath8 $ ] ( e.g. , * ? ? ?
* ; * ? ? ?
the abundance of @xmath5-elements relative to iron traces the relative numbers of thermonuclear and core - collapse supernovae and effectively traces the time scale of star formation due to the relative delay in the thermonuclear supernova . @xcite
propose a top - heavy imf as alternative explanation for the observed @xmath8 $ ] overabundance .
however , as @xcite point out , the _ relative _
@xmath8 $ ] overabundance in fact is an _
absolute _ fe underabundance , which indicates a relative lack of contribution of the thermonuclear supernovae to the overall metal abundances rather than an enhancement of core - collapse supernovae . this argument does not favor explaining the relative @xmath8 $ ] enhancement compared to the solar value with a top - heavy imf .
moreover , @xcite find from semi - analytical modeling that the hypothesis of a top - heavy imf can not explain the observed tight correlation between @xmath8 $ ] enhancement and velocity dispersion in early - type galaxies , but agrees with a more pronounced early truncation of star - formation with increasing galaxy mass .
a mechanism which efficiently removes nearly all gas from the host in @xmath211 gyrs is therefore more promising , and was suggested by @xcite and @xcite for massive early - type galaxies .
outflows with the observed mass loss rate in mrc1138 - 262 , @xmath212 yr@xmath15 , fulfill this constraint , given that sufficient kinetic energy is injected into the ism during the agn luminous phase . in
[ subsec : simpleenergy ] we estimate that @xmath213 ergs or more are injected into the ism of mrc1138 - 262 during an assumed agn lifetime of @xmath119 yrs . the mechanical energy of the agn is able to entrain ambient material , propelling it outwards at the high velocities observed .
the maximum amount of ambient material ( if accelerated to escape velocity ) can be approximated by an energy conservation argument as : @xmath214 where @xmath215 is the energy of the outflow in units of @xmath216 ergs and @xmath217 is the escape velocity in units of 500 .
@xmath218 is difficult to constrain . for galaxies with nfw profiles
@xcite , there is a simple relationship between @xmath218 , virial radius and mean mass density of the halo over the background dark matter density ( see * ? ? ?
* ) . for a baryonic mass fraction of 10% ,
the escape velocity in mrc1138 - 262 is about 700 at the virial radius of about 120 kpc @xcite .
equation [ eqn : mejmax ] implies that mrc1138 - 262 may eject , @xmath219 9 @xmath24 10@xmath220 of material ( assuming @xmath215=9 .
, which is the maximum of our estimated values , and @xmath217=1.4 ) , or @xmath219 3 @xmath24 10@xmath220 assuming our fiducial value of @xmath215=3 ..
this is of course a very crude estimate , but it does suggest that , plausibly , much of the baryonic mass could escape the potential of the galaxy and thus will not be available for future star - formation .
however , this represents a strict upper limit to the total mass ejection .
we have not assumed a coupling factor between the energy and the ambient gas , and this factor is undoubtably less than one ( we estimate @xmath136 ) .
subsequently , we will adopt this maximum mass ejection estimate and a coupling efficiency of 10% as the upper boundary , and the simple observed mass loss rate from as a lower boundary to the most likely range of mass loss rates and to constrain the amount of mass and metals ejected by agn winds .
in addition to suppressing star formation within the short timescale set by the [ @xmath5/fe ] `` clock '' , the agn must also be able to inhibit the accretion of gas from the surrounding igm or icm to effectively halt future star formation .
@xcite find that it is necessary to incorporate such a mechanism into their simulation , to inhibit galaxy growth in a population of massive red galaxies , and to produce a mass function with the appropriate exponential cutoff at the high end . in a very interesting analysis of a large sample of radio loud agn based on data from the sloan digital sky survey ,
@xcite find that the dependence of the size of the radio - loud agn fraction on black hole mass mirrors that of the cooling rate within the hot atmospheres of elliptical galaxies .
@xcite further argue that the agn could be fueled by such cooling gas and that agn heating could plausibly balance the cooling of the gas over time .
could the energy output of mrc1138 - 262 balance gas accretion from the igm with a similar mechanism ? following the arguments in @xcite , we can estimate whether or not it is even plausible for the radio jets in mrc1138 - 262 to inhibit the accretion of gas onto the dark matter halo .
@xcite estimate the rate at which gas is accreted onto dark matter halos as a function of halo virial mass and redshift . for a halo with @xmath221 at @xmath222 , they give an accretion rate of @xmath223 yr@xmath15 . the gas is assumed to form into a hot , cooling halo of gas filling the dark matter halo , thus its energy can be estimated using simple kinetic arguments .
making this assumption , the rate at which energy is deposited in the halo with a virial velocity , @xmath224 , is approximately : @xmath225 over the lifetime of the agn , this will be , @xmath226 .
thus both the instantaneous ( @xmath22710@xmath112 ergs s@xmath15 ) and the total energy ejection rate ( @xmath2273 @xmath24 10@xmath228 ergs ) of the agn appear sufficient to stop the collapse of gas at the virial radius of the dark matter halo . if there is a connection between mass accretion and igniting agn activity in galaxies , then the output of a single radio loud agn cycle appears sufficiently energetic to suppress subsequent gas flow at the virial radius .
radio lobes at high redshift reach distances of @xmath229 kpc .
therefore the range and amount of energy injection seem sufficient to plausibly resist the gas accretion flow , and to influence the final baryonic mass of galaxies hosting powerful radio sources .
of course the energy output from the agn must couple to the infalling gas for it to be effective .
certainly the high efficiencies we have estimated for the gas within 10 - 20 kpc of the nucleus would be sufficient .
this is a fundamental difference to starburst - driven winds which will heat the gas on similar time scales , but lack the energy and power necessary to unbind significant amounts of gas from the most massive halos , and are very unlikely to stop the accretion flow . if the ejected gas rains back onto the galaxy , and this material supports further star - formation , the enhanced @xmath8 $ ] enhancement will not be preserved .
similar arguments disfavor models where star - formation has consumed most of the gas prior to the agn bright phase , e.g. , @xcite , and approaches where ultraluminous starbursts are triggered through _ positive _ agn feedback @xcite , provided that most of the stellar mass observed at @xmath230 are formed in such a burst .
the short lifetimes of powerful agn ( @xmath197 yrs ; * ? ? ?
* ) compared to a hubble time indicate that nuclear activity is likely episodic .
to provide an estimate of the cosmic energy density released through agn feedback , we will therefore assume that agn hosts have only one cycle of activity that lasts @xmath231 yrs with coupling efficiencies to the host s ism and outflow characteristics similar to mrc1138 - 262 .
the energy output during one such cycle follows from the previous analysis , @xmath232 ergs ( with @xmath233=10@xmath112 ergs s@xmath15 and @xmath234=10@xmath235 years ) . to estimate the agn duty cycle directly from the number of observed luminous agn
, we use the co - moving number density of luminous agn calculated by , e.g. , @xcite , @xmath236 mpc@xmath237 for optically selected high - redshift quasars brighter than @xmath238 .
if the `` unified model '' of agn is correct , then qsos are a subsample of all agn , and the observed number density of qsos will be a few times lower than the intrinsic number density of the full agn population .
we also assume that a `` radio loud '' episode is an intrisic part of the lifecycle of all powerful agn .
figure 9 of @xcite shows that 95% of all quasars are at @xmath239 , within an epoch of @xmath240 gyrs . for @xmath94 yrs , this yields
this is again a lower limit , and neglects radiatively inefficient black hole growth . the overall energy budget of powerful agn
can then be gauged by , @xmath242 as a result , powerful agn provide a cosmic energy density due to outflows of @xmath243 .
we note that this _ global _ estimate is relatively robust against uncertain agn lifetimes , because increased durations leads to lower duty - cycles and vice versa .
we can also use the local co - moving mass density of supermassive black holes ( smbh ) as a measure for the integral accreted black hole mass to constrain the cosmological significance of agn feedback . using the co - moving number density of spheroids and the m@xmath244 relation , @xcite estimate the local mass density of nuclear black holes , @xmath245 @xmath43 h@xmath246 mpc@xmath237 . to estimate the total kinetic luminosity corresponding to this black hole mass density
, we use the simple scaling , @xmath247 with the efficiency of the kinetic energy output relative to the radiative ( bolometric ) luminosity , @xmath248 , and the radiative efficiency relative to the accreted rest mass energy , @xmath249 .
the radiative efficiency of massive black holes has been estimated to be roughly 10% to 30% ( e.g. , * ? ? ? * and references therein ) , whereas there is a broad range of relative kinetic energy efficiency of 0.05 to 1 @xcite . adopting
this range of values suggests that the kinetic energy density is approximately 2 to 100 @xmath2410@xmath250 ergs mpc@xmath237 . for a fiducial @xmath248=1 ,
the estimated kinetic energy density ejected is about 5 @xmath24 10@xmath250 ergs mpc@xmath237 , @xmath251 higher than what we estimated from the qso co - moving space density and duty cycle .
this however implies that all black hole growth is related to gas accretion , and occurs during agn luminous phases .
both assumptions are unrealistic , as indicated , e.g. , by the large population of obscured quasars and the possibility of bh - bh mergers .
this result therefore sets an upper constraint to the energy density ejected by agn . in the spirit of setting lower constraints
we adopt the first value , and note that our calculation does not violate any low - redshift constraints from the low - redshift number densities of supermassive black holes .
simple energetic arguments alone show the potentially large significance of agn feedback on the evolution of @xmath252 local early type galaxies .
@xcite estimate a local co - moving space density of @xmath253 galaxies , 0.0020 h@xmath254 mpc@xmath237 , @xcite find a local co - moving space density of 0.0022 h@xmath255 mpc@xmath237 for m@xmath256 early type galaxies . using the co - moving density of m@xmath256 early type galaxies as a normalization factor , and the upper limit to the kinetic energy density in
[ subsubsec : simpleenergy ] , @xmath257 ergs mpc@xmath237 , we find that _ on average _ , each m@xmath256 early - type galaxy ejects @xmath258 ergs , roughly the binding energy of an elliptical galaxy with total mass @xmath259 ( or a galaxy about one magnitude brighter than @xmath260 ; * ? ? ?
this is of course a statistical estimate to relate the impact of such outflows to the population of massive galaxies . it does not require or imply that each m@xmath256 early - type galaxy indeed went through a mrc1138 - 262-like epoch .
given the large amount of ambient ism that the agn likely drives out from the galaxy and perhaps from the surrounding dark matter halo , it is pertinent to ask : if mrc1138 - 262 is typical , how much mass ( and mass of metals ) could such outflows eject ?
using either the observed outflow rate for @xmath119 yrs or the maximum total mass ejection , we have , @xmath261 ( [ subsec : alphafe ] ) , with the co - moving qso number density and the inverse of their duty cycle , we find that on average , @xmath262 of material will be swept out by agn feedback , where @xmath263 is the co - moving space density of qsos in units of 5@xmath2410@xmath264 mpc@xmath237 and @xmath265 is the inverse duty cycle in units of 300 .
this implies that approximately @xmath266 of material is ejected per m@xmath256 galaxy . within these loose constraints
, this is about the mass of ism in a low - redshift gas rich galaxy and also roughly the molecular gas mass in high - redshift submm and radio galaxies . lacking robust constraints on the gas - phase metallicity in hzrgs
, we use the metallicities of low - redshift massive galaxies to constrain the metal loss induced by agn feedback .
the local mass - metallicity relationship indicates that local @xmath267 galaxies have solar and greater metallicities @xcite . with our above analysis , and scaled to solar metallicity @xmath268 , this implies that powerful agn feedback expels up to @xmath269 of metals per m@xmath256 galaxy .
early - type nearby galaxies about a magnitude brighter than @xmath253 have total baryonic masses of @xmath270 , and super - solar metallicities by @xmath271 dex @xcite , which implies a total metal content of @xmath272 .
this indicates that mrc1138 - 262-like feedback could eject a substantial fraction of the metals contained in the stars of an @xmath273 elliptical in the local universe .
this metal - rich outflow may contribute significantly to the overall metallicity density at high redshift .
@xcite estimate the co - moving metallicity density at z@xmath274 from integrating the star - formation rate density , @xmath275 mpc@xmath237 .
if outflow rates in powerful agn are generally similar to mrc1138 - 262 , then our above estimate implies a total metal ejection of @xmath276 mpc@xmath237 .
@xcite were able to account for @xmath27750% of the total metallicity density at z@xmath278 in the known population of galaxies .
thus , agn feedback from massive galaxies may account for of - order a few percent to about 20% of all metals in the igm at z@xmath392 .
although these estimates are highly uncertain , it appears very likely that agn - driven outflows are a significant contributor to the overall metal distribution in the igm in the early universe .
we presented a study of the spatially - resolved dynamics in the optical emission line gas around the [email protected] powerful radio galaxy mrc1138 - 262 , using the near - infrared integral - field spectrograph spiffi on ut1 of the vlt .
the large - scale kinematics in the emission line nebula of this galaxy are not consistent with the signatures of large - scale gravitational motion or starburst - driven winds .
velocities and fwhms of @xmath279 indicate a vigorous outflow with an estimated total energy which is orders of magnitude higher than suggested for starburst - driven winds . based on timescale and energy arguments ,
we conclude that the observed outflow , which is a strict lower limit to the intrinsic amount of outflowing gas , is most plausibly driven by the agn .
this implies a relatively efficient interaction between agn and interstellar medium , with efficiency @xmath280 , mostly mediated by mechanical energy injection through the radio jet .
radiation - driven winds do not appear powerful enough to explain the observed velocities over the size of the optical line emitting gas . for mrc1138 - 262 we estimate from simple energetic arguments that the total energy needed to drive the observed outflow of optical emission line gas is of - order few @xmath24 @xmath216 ergs with a mass outflow rate , @xmath281 yr@xmath15 . if we assume that every luminous qso near the peak of their co - moving density at z@xmath392 has such an outflow phase with similar characteristics to mrc1138 - 262 we find that powerful agn eject @xmath282 ergs mpc@xmath237 of energy into the igm and @xmath283 mpc@xmath237 of gas , including @xmath284 mpc@xmath237 in metals .
these energy and masses are significant compared to the the total binding energy and mass budget of @xmath253 early - type galaxies in the local universe .
if mrc1138 - 262 is indeed archetypal , then agn winds have the potential to be of similar , perhaps even larger , cosmological signficance than starburst - driven winds , especially for the most massive galaxies .
they accelerate the outflowing gas to much higher velocities than starburst - driven winds , and much of this material and energy are likely to escape the potential of even the most massive individual galaxy halo .
outflows like in mrc1138 - 262 have the potential to explain the observed properties of low - redshift massive galaxies , namely their old , and narrow range of ages and low gas fractions .
moreover , the large mass loss rates and short timescales of their outflows can naturally explain the observed enhancement of [ @xmath5/fe ] relative metal abundances observed in the most massive galaxies at low redshift .
agn - driven outflows appear to be a plausible mechanism that efficiently suppresses star - formation within a few @xmath9 yrs .
they also have high enough energies to perhaps stop the accretion flow of matter from the igm into their dark matter halos .
thus , if accretion is followed by agn activity , then agn feedback may be an effective mechanism for regulating galaxy growth @xcite .
our study comprises only one source , but we have indirect evidence for mrc1138 - 262 being far from unique .
velocities of @xmath285 are not uncommon in the extended emission line gas of powerful radio galaxies , but are generally extracted from longslit spectra aligned with the radio jet axis only . simulating a longslit spectrum from the spiffi data cube , we find good agreement between mrc1138 - 262 and these samples .
the bolometric and radio jet luminosities of mrc1138 - 262 are not unusual , and the source follows the @xmath286 relationship of high - redshift radio galaxies , if correcting for the luminous broad line emission . we are currently acquiring rest - frame optical , integral - field data for a larger sample of hzrgs , to make statistically more robust predictions .
preliminary results strongly support the present analysis , giving further evidence that agn feedback plays a major role in the evolution of the most massive galaxies in the universe .
martini , p. 2004
, invited review to appear in `` carnegie observatories astrophysics series , vol . 1 : coevolution of black holes and galaxies , '' ed
. l. c. ho ( cambridge : cambridge univ . press ) , astro - ph/0304009 | to explain the properties of the most massive low - redshift galaxies and the shape of their mass function , recent models of galaxy evolution include strong agn feedback to complement starburst - driven feedback in massive galaxies . using the near - infrared integral - field spectrograph spiffi on the vlt , we searched for direct evidence for such a feedback in the optical emission line gas around the [email protected] powerful radio galaxy mrc1138 - 262 , likely a massive galaxy in formation .
the kpc - scale kinematics , with fwhms and relative velocities @xmath1 and nearly spherical spatial distribution , do not resemble large - scale gravitational motion or starburst - driven winds .
order - of - magnitude timescale and energy arguments favor the agn as the only plausible candidate to accelerate the gas , with a total energy injection of @xmath2 ergs or more , necessary to power the outflow , and relatively efficient coupling between radio jet and ism .
observed outflow properties are in gross agreement with the models , and suggest that agn winds might have a similar , or perhaps larger , cosmological significance than starburst - driven winds , if mrc1138 - 262 is indeed archetypal .
moreover , the outflow has the potential to remove significant gas fractions ( @xmath3% ) from a @xmath4 galaxy within a few 10 to 100 myrs , fast enough to preserve the observed [ @xmath5/fe ] overabundance in massive galaxies at low redshift . using simple arguments
, it appears that feedback like that observed in mrc1138 - 262 may have sufficient energy to inhibit material from infalling into the dark matter halo and thus regulate galaxy growth as required in some recent models of hierarchical structure formation . |
the star formation history of elliptical galaxies still remains , at many levels , a mystery .
the fundamental uncertainty afflicting attempts to determine the ages of such ( presumably ) old stellar populations lies in the unfortunate fact - of - life that age and metallicity have almost identical effects upon broad - band colours and on most line indices ( oconnell 1976 ) .
this age - metallicity _ degeneracy ( sometimes referred to as worthey s ( 1994 ) `` 3/2-rule '' , their broad - band colours and ( most ) line indices will be nearly identical . ] ) has led many ( oconnell 1976 ; worthey 1994 ; rose 1994 ) to search for individual absorption features which may be sensitive to either age _ or metallicity , but not both . _ _
unfortunately , line indices have not proven to be the panacea once hoped for . as worthey ( 1994 )
has shown ( see his table 6 ) , the vast majority of low - resolution lines defined by the lick ids survey ( worthey 1994 ) lie uncomfortably close to the 3/2-rule degeneracy `` zone '' . an encouraging result to come out of the worthey ( 1994 ) ids study , however , was the identification of one particularly useful , and strong , indicator , apparently free of any strong age influence .
fe4668 , whose bandpass spans @xmath3 , tracks not only fe , but contains a strong contribution from c@xmath4 , ti , cr , and mg , and as worthey , trager & faber ( 1996 ) note , is the best current candidate tracer for the mean metallicity `` z '' , almost exclusively sensitive to this alone .
on the other hand , identifying a matching `` pure '' age discriminator has proven more difficult . h@xmath5 and h@xmath6 have long been popular indices , in this regard ( oconnell 1976 ) , but as stressed by gonzlez ( 1993 ) , both lines can be easily distorted ( even filled in ) by emission from ionized gas , _ severely compromising their usage .
higher - order balmer emission lines ( h@xmath7 and h@xmath2 ) are admittedly weaker , yet free of such `` emission '' complications , and appear to provide a very sensitive age diagnostic . rose ( 1994 ) recognized this latter fact , and designed a set of higher - resolution line indices in order to take advantage of these age discriminating `` powers '' .
the rose system employs a 3.74 window over which the line indices are measured ; in contrast , the lick ids system ( typically ) uses an @xmath840 window .
one index in particular , h@xmath0 ( centred at 4340.5 ) , was strongly sensitive to age alone .
_ the near - orthogonal behaviour , in the age - metallicity plane , of the h@xmath0 and fe4668 line indices , was demonstrated by jones & worthey ( 1995 ; figure 2 ) , and suggests a viable mechanism for breaking the aforementioned age - metallicity degeneracy .
an extension of this lick+rose index comparison was undertaken by jones ( 1995,1996ab ) , but now concentrating solely upon pairs of high resolution rose indices jones examined the h@xmath0cai@xmath9 age - metallicity diagnostic diagram .
jones & worthey ( 1995 ) stress several caveats regarding their proposed optimal age - dating technique .
first , while the parameter - space coverage ( t@xmath10 , @xmath11 , and [ fe / h ] ) of their stellar library , upon which the model sequences are ultimately based , is good , it is not complete .
specifically , there is an admitted lack of metal - poor stars , especially hot horizontal branch stars , which could contribute to the balmer line absorption .
because of this , jones & worthey emphasize that their sequences are not directly applicable to systems with metallicities outside @xmath12}\le + 0.6 $ ] .
second , in order to apply this technique accurately to integrated spectra of spheroidal systems , very high signal - to - noise ratios ( @xmath13 ) , at moderately high spectral resolutions ( @xmath14 ) , are a necessity . to this end , jones & worthey ( 1995 ) present such a spectrum for m32 , and find an age of @xmath15 gyr , in agreement with oconnell ( 1980 ) and rose ( 1994 ) .
while this result is a necessary proof of the technique , it is not a sufficient one , as it effectively compares two population _ synthesis methods , both of which are _ indirectly determining an age estimate .
optimally , any testing of a new , and potentially exciting , technique should be completely `` independent '' , at some level .
unfortunately , there are simply no independently age - dated ellipticals known . on the other hand ,
there does exist another family of systems which are ideally suited to such a study the metal - rich globular clusters of the milky way .
_ metal - rich globular clusters are the only practical systems with which to independently assess the reliability of the h@xmath0fe4668 age - dating technique .
this comes about because accurate ages have already been determined for at least four clusters in the metal - rich regime to which the jones & worthey technique is applicable ( richer 1996 ) . with its high metallicity ( @xmath16 ) and colour - magnitude diagram ( cmd)-derived age of @xmath1 gyr ( richer 1996 ) ,
the southern globular cluster , 47 tuc , is an ideal laboratory within which to test this new age - dating technique .
_ _ _ existing high - resolution integrated spectra of 47 tuc in the literature include those of smith ( 1979 ) and rose ( 1994 ) . in both cases
though , the observations were restricted to @xmath17 , meaning the metallicity - discriminant fe4668 line index could not be measured . in section [ observations ] , we briefly describe our new observations of 47 tuc , which now provide simultaneous coverage of both the age - sensitive feature h@xmath0 , and the metallicity - sensitive features cai@xmath9 and
fe4668 . such data will allow us to test how accurately the spectroscopic age of this simple stellar population agrees with its accurately determined cmd age .
the results of our analysis are discussed in section [ discussion ] , and summarised in section [ summary ] .
two high signal - to - noise ( @xmath18 ) integrated spectra of 47 tuc were acquired during the nights of october 20 and 21 , 1996 , using the coud spectrograph , with the 32@xmath19 schmidt camera ( grating e ; 1@xmath20 order ) , on the 74@xmath19 telescope at mount stromlo observatory , was employed .
the @xmath21 long slit was scanned back and forth ( e - w ) across the slit , to limits of @xmath22 from the cluster center ( total coverage in excess of 2.5 core radii , ensuring stochastic fluctuations due to finite sampling were minimized ) , for @xmath23 minutes per exposure .
the 24 @xmath24 m pixels of the tek 2k@xmath252k ccd translated to 0.48 per pixel 983 of spectral coverage , which for a central wavelength setting of @xmath26 , allowed simultaneous determination of ca i@xmath9 , h@xmath0 , and fe4668 indices .
companion blank - sky exposures ( again , of @xmath23 minutes duration , taken @xmath27 to the north of the cluster core ) were acquired subsequent to each 47 tuc exposure .
the relative strength of the sky versus cluster signal was negligible . for calibration purposes , 13 stars ( spanning a range in metallicity and spectral type ) in common with jones coud feed spectral library ( see section 2.4 of leitherer 1996 and jones 1996b ) and
the lick / ids library ( worthey 1994 ) were observed , as were two spectrophotometric standards .
standard iraf tasks were employed for the flat - fielding , flux , and wavelength calibration .
observations taken throughout the run of bright , metal - poor halo giants imply that sky subtraction has not introduced any unforeseen spectral artifacts .
because the central one - dimensional velocity dispersion of 47 tuc is only @xmath28kms@xmath29 ( meylan & mayor 1986 ) , no correction for internal dispersion was applied to the spectra ; as trager ( 1998 ; figure 3 ) demonstrate , such a correction , for globular clusters , is negligible . before measuring the h@xmath0 and cai@xmath9 line indices , our observed standards and target 47 tuc spectra
were first smoothed with a @xmath30kms@xmath29 gaussian , and re - binned to 0.623 / pixel , in both cases to parallel jones ( 1996b ) .
the h@xmath0 and cai@xmath9 equivalent widths are measured as detailed in section 3 of jones & worthey ( 1995 ) . flanking blue and red pseudo - continuum peaks , usually associated with the immediately - adjacent blue and red shoulders of the relevant absorption line ,
were identified ; the equivalent widths were then measured relative to the continuum formed by connecting the two pseudo - continuum peaks . for h@xmath0 ,
the blue pseudo - continuum was associated with the local maximum between 4329 and 4336 , while the red pseudo - continuum was associated with the maximum in the 4346@xmath314350 window ; for cai@xmath9 , the correspondings windows were 4218@xmath314222 and , either 4230@xmath314232 ( early spectral types ) or 4242@xmath314246 ( later spectral types ) .
figure [ fig : fig1 ] shows the pair of integrated 47 tuc spectra in the region of both the h@xmath0 and cai@xmath9 indices ; the associated flanking pseudo - continua are similarly noted .
the equivalent width was then derived via a simple simpson s rule algorithm , summing the area associated with the seven 0.623 pixels centred on the minimum of the high resolution absorption feature gaussian - fitting to the line profiles was not employed in our analysis , again , in order to maintain consistency with jones ( 1996b ) .
the derived equivalent widths of figure [ fig : fig1 ] are consistent , from night - to - night , at the @xmath32% level .
for the fe4668 measurement , the 13 standards and 47 tuc spectra were smoothed with a @xmath33kms@xmath29 gaussian , to match the lick ids resolution .
adjacent blue and red continua were determined from the mean flux in the [email protected] and [email protected] windows , respectively . using the continuum defined by the line connecting the mean of these two windows ,
the equivalent width was then derived by integrating the absorption feature over the [email protected] window .
the adopted windows are marginally different from those defined by worthey ( 1994 ; table 1 ) , but reflect those adopted in the 1997 edition of the lick / ids stellar catalog ( trager 1998 ) .
comparison of the equivalent widths derived from our 13 standards with those tabulated in the coud feed spectral library ( h@xmath0 and cai@xmath9 ) and lick / ids library ( fe4668 ) revealed the two systems were on the scale , and no further corrections were needed to bring the systems into agreement
both groups were measuring identical equivalent widths for the same standards and therefore , our measured 47 tuc equivalent widths could be used _ as is in comparing against the model age - metallicity diagnostic diagrams of jones & worthey ( 1995 ) and jones ( 1996a ) .
evidence to this effect is presented in figure [ fig : fig2 ] , in which our standard star cai@xmath9 equivalent widths ( x - axis ) are plotted against those from jones ( 1996b ) coud feed spectral library ( y - axis ) - the mean offset ( in the sense of jones@xmath31present study ) is only @xmath34 , a factor of @xmath83@xmath25 smaller than the typical uncertainty associated with pseudo - continua placement and night - to - night variations ( recall figure [ fig : fig1 ] ) .
identical conclusions hold for the other line indices in our study .
our measured ( and adopted ) 47 tuc equivalent widths are therefore @xmath35 ( h@xmath0 ) , @xmath36 ( cai@xmath9 ) , and @xmath37 ( fe4668 ) , where the uncertainties reflect those due to pseudo - continuum placement , and , to a lesser degree , the scatter derived from repeat measurements of the set of standards . _
following the prescription outlined by jones & worthey ( 1995 ) and jones ( 1996b ) , the worthey ( 1994 ) population synthesis models were used to predict h@xmath0 , cai@xmath9 , and fe4668 line index strengths , as a function of age and metallicity [ fe / h ] , for a grid of single burst stellar populations .
twenty - four realizations , spanning ages @xmath38gyr and metallicities [ fe / h]=@xmath39 , were made ; the full set of 24 models were then incorporated into the h@xmath40cai@xmath9 age - metallicity diagnostic grid shown in figure [ fig : fig3 ] . while not perfectly orthogonal , this pair of high resolution rose - style indices do provide adequate separation of age and metallicity , with a minimum of degeneracy .
the four filled symbols represent the same set of low luminosity ellipticals shown in the earlier analysis of jones ( 1996a ) , but now smoothed to the same @xmath30kms@xmath29 resolution adopted by jones ( 1996b ) and the present study . as discussed previously ( jones 1996a ) , these four galaxies represent a ( luminosity - weighted ) solar metallicity age sequence spanning the range @xmath41gyr .
the open starred symbol represents our new 47 tuc data point ; unlike the case for the galaxies shown in figure [ fig : fig3 ] , 47 tuc has an independent age - element enhanced stellar models , have recently revised the age of 47 tuc from the canonical ( scaled - solar abundance ) richer ( 1996 ) value of @xmath1gyr , to the lower value of @xmath42gyr . on the surface
this only worsens the spectroscopic versus cmd age discrepancy of figure [ fig : fig3 ] , although the comparison is perhaps not a fair one , since worthey s ( 1994 ) synthesis models are ( necessarily ) tied to models and stellar data which are likewise of scaled - solar abundances .
similarly , the post - hipparcos globular cluster distance scale appears to favor a 2 - 3gyr reduction in the mean age of the oldest globulars ( chanoyer 1998 ) ; again , this only strengthens our conclusion regarding the spectroscopic age of 47 tuc , although to be fair , as before , we retain the richer ( 1996 ) result as the most representative one against which to compare our results . ]
( @xmath1gyr ) and metallicity ( [ fe / h]@xmath16 ) determination , derived from isochrone fitting to the cmd ( richer 1996 ; salaris & weiss 1998 )
. it should be immediately clear that the spectroscopic age of 47 tuc , as determined by the age - sensitive h@xmath0 line index is inconsistent with its cmd age , at the @xmath84@xmath43 level , being @xmath80.2 lower than expected for this 14gyr old simple stellar population .
this @xmath80.2 h@xmath2 discrepancy was previously noted by rose ( 1994 ; section 5.1 ) , although not in the context of age - dating its integrated spectrum . on the other hand , it is somewhat re - assuring that the metallicity inferred from figure [ fig : fig3 ] ( [ fe / h]@xmath44 - 0.5 ) is consistent with that derived from the cmd ( richer 1996 ) .
figure [ fig : fig4 ] shows the h@xmath0fe4668 age - metallicity diagnostic diagram as presented by jones & worthey ( 1995 ; figure 2 ) ; the same nineteen models used in their analysis are incorporated in our grid .
the only ( subtle ) difference between figure [ fig : fig4 ] , and that of jones & worthey , is that , in the latter , the spectra were smoothed to @xmath45kms@xmath29 ( to match the resolution of their m32 spectrum ) , while in the present study , recall that a @xmath30kms@xmath29 gaussian was employed ( to match the spectra and models of jones 1996b ) .
again , the model sequences of figure [ fig : fig4 ] clearly demonstrate the near perfect orthogonal nature of the h@xmath0 and fe4668 line indices in the age - metallicity plane . as in figure
[ fig : fig3 ] , the filled square represents the jones & worthey ( 1995 ) observation of m32 , while the open starred symbol shows our new 47 tuc data point . as was the case for the pair of rose - style high resolution line indices ( figure [ fig : fig3 ] ) , in the h@xmath0 versus fe4668 , the anomalously low h@xmath0 in the integrated spectrum of 47 tuc results in a spectroscopic age ( @xmath4620gyr ) which is clearly discrepant with cmd age ( @xmath1gyr ) .
is there some way in which to reconcile the spectroscopic versus cmd age discrepancy ? as discussed by rose ( 1994 ) , virtually every line index in the integrated spectrum of 47 tuc appears perfectly well - behaved , except the anomalously strong cn bands and anomalously weak h@xmath2 .
jones ( 1996b ) speculated that perhaps the h@xmath2 `` problem '' was being caused by ch contamination in the h@xmath2 line ; because the central regions of 47 tuc ( the region over which our integrated spectrum was obtained ) is replete with cn - strong giants ( norris & freeman 1979 ; da costa 1997 ) , and there exists the well - known cn - ch anticorrelation for cn - strong giants ( norris , freeman & da costa 1984 ) , a scenario was envisaged whereby h@xmath2 might be weaker in cn - strong ( ch - weak ) stars .
jones ( 1996b ) , though , was able to show , using a sample of 11 cn - strong and cn - weak giant and horizontal branch stars from m71 ( a cluster similar to 47 tuc ) , that this initial hypothesis was not supported by the data .
the most obvious speculative source of the discrepancy is that some unidentified source of h@xmath2 _ emission is acting to `` fill in '' the h@xmath2 _ absorption feature .
hesser & shawl ( 1977 ) provide an upper limit to the emission flux at h@xmath5 of f(h@xmath5)@xmath47erg@xmath48s@xmath29 . here
we have used their most restrictive dataset - @xmath49s@xmath29 ( from their 22/08/75 observations ) , coupled with the definition f(h@xmath5)@xmath50erg@xmath48s@xmath29 , from smith , hesser & shawl ( 1976 ) .
now , the core of 47 tuc , with m@xmath51 ( and ignoring reddening ) , corresponds to a flux of @xmath52erg@xmath48s@xmath29@xmath29 .
so , the hesser & shawl upper limit of @xmath53erg@xmath48s@xmath29 corresponds to @xmath54 of equivalent width at h@xmath5 .
several simplifying assumptions must be made to go from h@xmath5 to h@xmath2 - specifically , under case b conditions and a temperature of @xmath55k , radiative recombination leads to an intrinsic ratio l(h@xmath5)/l(h@xmath2)@xmath56 , which gets reduced by a factor 1.7 , reflecting the ratio of continuum flux at h@xmath5 to that at h@xmath2 ( employing the worthey 1994 models ) .
in other words , h@xmath2 is a factor of @xmath57 weaker than h@xmath5 ; recalling the constraint at h@xmath5 of @xmath58 , this corresponds to @xmath59 of equivalent width at h@xmath2 . _
_ from figures [ fig : fig3 ] and [ fig : fig4 ] , we can see that the h@xmath2 `` discrepancy '' amounts to @xmath60 , in that we _ a priori expected an equivalent width of @xmath61 , to match the known cmd age of 47 tuc , as opposed to the @xmath62 observed .
recalling that the inferred upper limit to emission fill - in at h@xmath2 ( from the previous paragraph ) was @xmath63 , we can see that the requisite @xmath64 of emission needed to reconcile the cmd and spectroscopic ages is not entirely excluded by the hesser & shawl ( 1977 ) observations , although little room is left to maneuver should new h@xmath5 emission observations push this limit even lower . _
the situation regarding the anomalously low h@xmath2 index in 47 tuc remains as it was at the time of the rose ( 1994 ) and jones ( 1996b ) studies - 47 tuc appears normal in virtually every line index , except h@xmath2 , and to date no satisfactory explanation has been put forth .
emission fill - in has not been entirely excluded , but the constraints are tight , as shown above . should such emission prove to be the ultimate resolution of the discrepancy though , it will beg the question as to exactly which mechanism is responsible , and why it differs from the local calibrating stars used to derive the model sequences of figures [ fig : fig3 ] and [ fig : fig4 ] .
from the outset , our goal was a simple one - to provide an independent test of the jones & worthey ( 1995 ) h@xmath0 fe4668 diagnostic ( and the jones 1996a h@xmath0cai@xmath0 diagnostic ) as a tool for breaking the long - standing age - metallicity degeneracy plaguing integrated spectra of stellar populations . such a test can only be done ( properly ) upon systems with independently - derived ages and metallicities ; local metal - rich globular clusters with accurate , deep , cmds , provide an ideal test - bed .
our new high @xmath65 , high resolution , integrated spectrum of 47 tuc shows anomalously low h@xmath2 absorption , which when tied to the diagnostic models of jones & worthey ( 1995 ) , shows , for the first time , that the inferred spectroscopic age is well in excess of 20gyr .
this is clearly at odds with the cmd age of @xmath1gyr .
no suitable explanation as to the source of this discrepancy has been found , despite the best efforts of rose ( 1994 ) and jones ( 1996b ) .
emission , amounting to @xmath60of equivalent width at h@xmath2 , is required to reconcile the discrepancy .
such emission is not excluded by the hesser & shawl ( 1977 ) observations of 47 tuc ( which set an upper limit of @xmath66 at h@xmath2 ) , but if it is the responsible mechanism , must be lurking just marginally below the limit set by these early fabry - perot observations . clearly , a deeper re - visit is called for . it is true that the jones & worthey ( 1995 ) h@xmath0fe4668 age - metallicity diagnostic will remain an invaluable tool for age - dating integrated spectra of early - type systems , but perhaps the most important conclusion to take from our simple test of the technique is that it is not infallible .
clearly , it will be imperative to extend our analysis to a larger sample of metal - rich globulars , galactic open clusters , and star clusters in the magellanic clouds such an analysis will demonstrate whether 47 tuc is a pathological case or the inconsistent spectroscopic age derived here is indicative of an endemic , and currently underappreciated , flaw in the technique .
we wish to thank guy worthey , jim rose , and ken freeman for their helpful advice .
a special word of thanks is aimed at the referee , who went above and beyond the call of duty in aiding us in the interpretation of the hesser & shawl ( 1977 ) observations .
dsm acknowledges the financial support provided through the anu / mssso summer research fellowship program . | high signal - to - noise integrated spectra of the metal - rich globular cluster 47 tuc , spanning the h@xmath0 and fe4668 line indices , have been obtained .
the combination of these indices has been suggested ( jones & worthey 1995 , apj , 446 , l31 ) as the best available mechanism for cleanly separating the age - metallicity degeneracy which hampers the dating of distant , unresolved , elliptical galaxies .
for the first time , we apply this technique to a nearby spheroidal system , 47 tuc , for which independent ages , based upon more established methods , exist .
such an independent test of the technique s suitability has not been attempted before , but is an essential one before its application to more distant , unresolved , stellar populations can be considered valid .
because of its weak series of balmer lines , relative to model spectra , our results imply a spectroscopic `` age '' for 47 tuc well in excess of 20 gyr , at odds with the colour - magnitude diagram age of @xmath1 gyr . the derived metal abundance , however , is consistent with the known value .
emission `` fill - in '' of the h@xmath2 line as the source of the discrepancy can not be entirely excluded by existing data , although the observational constraints are restrictive . # 1to 0pt#1 |
by now it has been well established that the x - ray emission from early - type galaxies has several components .
these galaxies can roughly be grouped into two categories based on the ratio of the x - ray to optical luminosity @xmath1 : the x - ray bright galaxies which have large values of @xmath1 , and the x - ray faint galaxies which have small values of @xmath1 . in
the x - ray bright elliptical and s0 galaxies , the x - ray emission is dominated by thermal emission from interstellar gas at a temperature of @xmath2 kev .
the spectra of their x - ray faint counterparts tend to require two components : a soft thermal component with @xmath3 kev , and a hard component which has been fit as either thermal bremsstrahlung with @xmath4 kev or a power - law ( fabbiano , kim , & trinchieri 1994 ; matsumoto et al .
1997 ; allen , di matteo , & fabian 2000 ; blanton , sarazin , & irwin 2001 ) .
the luminosity of the hard component varies roughly proportional to the optical luminosity , suggesting that the origin of this component is low - mass x - ray binaries ( lmxbs ) similar to those seen in our own galaxy ( trinchieri & fabbiano 1985 ) .
the high spatial resolution of the _ chandra _ x - ray observatory has allowed much of the hard component to be resolved into individual sources , thereby demonstrating that this hard component is indeed from individual sources ( e.g. , sarazin , irwin , & bregman 2000 , 2001 ; blanton et al . 2001 ) . in this paper , we present the results of a _
observation of the x - ray bright elliptical galaxy ngc 4649 ( m60 ) .
this is an e2 elliptical galaxy in the virgo cluster .
ngc 4649 has a close companion galaxy ngc 4647 , which is an sc galaxy .
this pair of galaxies is also referred to as arp 116 or vv 206 . with a third more distant galaxy ,
this pair forms a group of galaxies wbl 421 ( white et al .
ngc 4647 was the host of the type - i supernova sn1979a ( barbon et al .
1984 ) .
we adopt a distance for ngc 4649 of 16.8 mpc , based on the method of surface brightness fluctuations ( tonry et al . 2001 ) .
this is consistent with the corrected recession velocity distance in faber et al .
( 1989 ) if the hubble constant is 79 km s@xmath5 mpc@xmath5 .
unless otherwise noted , all the uncertainties quoted are at the 90% confidence level .
ngc 4649 was observed on 2000 april 20 on the acis - s3 ccd operated at a temperature of -120 c and with a frame time of 3.2 s. in addition to the s3 chip , the acis chips i2 , i3 , s1 , s2 and s4 were also turned on for the duration of the observation .
the pointing was determined so that the entire galaxy was located on the s3 chip and so that the center of the galaxy was not on a node boundary of the chip .
although a number of serendipitous sources are seen on the other chips , the analysis of ngc 4649 in this paper will be based on data from the s3 chip alone .
the data were telemetered in faint mode , and only events with asca grades of 0,2,3,4 , and 6 were included .
we excluded bad pixels , bad columns , and the columns next to bad columns and to the chip node boundaries .
we checked for periods of incorrect aspect solution , and none were found .
the total exposure for the s3 chip was 36,780 s. _ chandra _ is known to encounter periods of high background ( `` background flares '' ) , which especially affect the backside - illuminated s1 and s3 chips .
we determined the background count rate , using the s1 chip to avoid the enhanced flux due to the galaxy ngc 4649 on the s3 chip .
unfortunately , much of the exposure was affected by background flares ; the light curve of the total count rate ( 0.310 kev ) for the s1 chip is shown in figure [ fig : s1_lcurve ] .
we used the program lc_cleanwritten by maxim markevitch to remove periods of high background and data drop - outs .
the cleaned light curve for the s1 chip is shown as the filled squares in figure [ fig : s1_lcurve ] .
after cleaning , 19,303 of the total 36,780 s of exposure time remained . the expected quiescent rate for the s1 chip during this period in the 0.3 - 10.0 kev band is about1.41 cnt s@xmath5 .
after the removal of background flares , the mean quiescent s1 rate during our observation was 1.61 cnt s@xmath5 .
the other , front - side illuminated chips also show excess background . by examining different energy bands we determined that most of the excess background in the s1 chip was below 5 kev , with our quiescent rates being in excess of the expected rates by about 21% in the 0.3 - 1.0 kev band , 27% in the 1.0 - 5.0 kev band , and 9% in the 5.0 - 10.0 kev band .
we note that ngc 4649 is located both at the outer edge of the virgo cluster and at the edge of the galactic north polar spur , both of which may contribute to the soft x - ray background near ngc 4649 ( bhringer et al .
we examined _ rosat _ observations of the same region of the sky covered by the s1 chip and determined a background count rate of 0.031 @xmath6 0.01 cnt s@xmath5 in the _ rosat _ pspc r4-r7 band .
this roughly translates to 0.115 cnt s@xmath5 for the _ chandra _ s1 chip in the 0.310 kev energy band , thereby accounting for about 60% of the excess background emission we detected with the s1 chip . here
, pimms was used to convert the _
rosat _ count rate into a _
chandra _ 0.310 kev count rate by assuming a raymond - smith ( rs ) spectral model with temperature @xmath7 kev and solar abundance .
this model was chosen as a compromise between the spectrum of the north polar spur ( nps ) ( snowden et al .
1997 ) and that of diffuse emission from the virgo cluster ( bhringer et al .
the excess background at hard energies ( 9% ) is probably due to a slightly elevated particle background , even when the background flares are removed . if we assume that this is the case and renormalize the blank sky data used to determine the quiescent rate given above so that the count rate in the 510 kev band matches what we observed with the s1 chip ( see [ sec : spectra ] ) , then the excess emission from the virgo cluster and the nps determined from the rosat observations can completely account for the remaining excess background emission in the 0.35 kev energy band .
we therefore conclude that the excess background emission we detect is a combination of emission from the virgo cluster , the nps , and an elevated particle background .
three of the x - ray point sources detected on the s3 image have optical counterparts with positions listed in the the u.s .
naval observatory ( usno ) a2.0 optical catalog ( monet et al .
the optical and x - ray positions all agree to better than 2 .
thus , we believe that the absolute positions derived from the x - ray images are accurate to better than 2 .
the raw _ chandra _ s3 chip x - ray image is shown in figure [ fig : xray_whole ] for the cleaned 19 ksec exposure in the 0.310 kev energy band .
the sky background has not been subtracted from this image , nor has the exposure map been applied .
many discrete sources are evident . also visible
is bright diffuse emission at the center of the galaxy , which swamps emission from discrete sources in that region . in order to image the fainter , more diffuse emission , we adaptively smoothed the cleaned 19 ksec exposure _ chandra _ s3 x - ray image to a minimum signal - to - noise ratio of 3 per smoothing beam .
the resulting image is shown in figure [ fig : xray_smo ] .
the image was corrected for exposure and background .
this image shows rather extended diffuse emission from ngc 4649 as well as the point sources .
in addition to some extended features in the outer parts of the galaxy , the image appears to show faint radial `` fingers '' of emission reaching out from the center of ngc 4649 , which are discussed further in [ sec : radial_features ] .
the discrete x - ray source population on the acis s3 image ( figure [ fig : xray_whole ] ) was determined using a wavelet detection algorithm in the 0.310.0 kev band , and they were confirmed with a local cell detection method .
we used the ciao wavdetect and celldetect programs .
the high spatial resolution of _ chandra _ implies that the sensitivity to point sources is not affected very strongly by the background .
thus , the source detection was done using the entire exposure of 36,780 s , including periods with background flares .
we also did the source detection on the data after the removal of the background flares ; the stronger sources were all found in both cases , but the very faintest sources were not detected as significantly in the cleaned exposure at a level consistent with the much shorter exposure ( 19,303 vs. 36,780 s ) .
the wavelet source detection significance threshold was set at @xmath8 , which implies that @xmath91 false source ( due to a statistical fluctuation in the background ) would be detected in the entire s3 image .
this significance threshold approximately corresponds to requiring that the source flux be determined to better than 3-@xmath10 .
fluxes were corrected for exposure and the instrument point - spread - function ( psf ) using the standard ciao caldb psf tables .
all of the source detections were verified by visually examining the image to ensure that no obvious errors were made during the detection process .
table [ tab : src ] lists the 165 discrete sources detected by this technique , sorted in order of increasing distance @xmath11 from the center of ngc 4649 .
columns 1 - 7 give the source number , the iau name , the source position ( j2000 ) , the projected distance @xmath11 from the center of ngc 4649 , the count rate and the 1-@xmath10 error , and the signal - to - noise ratio snr for the count rate .
since we did not detect a distinct source at the center of the galaxy , we adopted the central position from 4.86 ghz radio observations of r.a .
= 12@xmath1243@xmath134002 , and dec .
= + 1133102 ( j2000 ; condon et al .
the accuracy of this position is @xmath14 . because of the uncertainty in the exact position of the center of the galaxy
, the values of @xmath11 might be off by @xmath152 .
the statistical errors in the positions of most of the sources are quite small ( @xmath16 ) , and the overall absolute errors are probably @xmath171 near the center of the field , with larger errors near the outside of the field .
these absolute errors are typical for observations during this period . over most of the image ,
the minimum detectable count rate was about @xmath18 cnt s@xmath5 ( @xmath19 ergs s@xmath5 at the ngc 4649 distance ) in the 0.310 kev band .
the detection limit is slightly higher ( by @xmath20 30% ) at large distances @xmath11 where the psf is larger , particularly at the northern edge of the field .
the minimum detectable flux is also much higher near the center of ngc 4649 , where the diffuse emission by interstellar gas is very bright .
this effect is obvious in table [ tab : src ] . with the exception of src . 1 , which is a detection of the extended peak in the diffuse emission ,
no sources are seen within 12 of the center .
if the x - ray source distribution followed the optical light in ngc 4649 , one would have expected at least 10 sources in this region . within the central 30 in radius , all of the sources are brighter than @xmath21@xmath22 cnt s@xmath5 , which is @xmath213 times the detection limit at larger distances . from the values of the diffuse plus background count rates found by the wavdetect algorithm
, we find that the source detection threshold is increased for @xmath23 . because the diffuse galactic x - ray emission is soft ( [ sec : diffuse_spec ] ) while the sources are mainly hard , we also did the source detection on a hard band ( 210 kev ) image both for the entire exposure and the cleaned exposure in the hope that this would allow sources to be detected more readily in the inner parts of the galaxy .
however , we did not detect any additional sources in the hard band . our detection limit for sources should result in @xmath91 false source ( due to a statistical fluctuation ) in the entire s3 field of view .
however , some of the detected sources are likely to be unrelated foreground or ( more likely ) background objects . based on the source counts in brandt et al .
( 2000 ) and mushotzky et al .
( 2000 ) , we would expect @xmath1510 such unrelated sources in our observation .
these should be spread out fairly uniformly over the s3 image ( figure [ fig : xray_whole ] ; see giacconi et al .
2001 ) , except for the reduced sensitivity at the center of ngc 4649 due to bright diffuse emission and at the outer edges of the field due to reduced exposure and increased psf .
thus , the unrelated sources should mainly be found at larger distances from the optical center of ngc 4649 ( the bottom part of table [ tab : src ] ) , while the sources associated with ngc 4649 should be concentrated to the center of the galaxy .
we searched for variability in the x - ray emission of the resolved sources over the duration of the _ chandra _ observation using the kolmogoroff - smirnov test ( see sarazin et al .
2001 ) . in most cases ,
the tests were inconclusive .
for only two of the sources ( srcs .
76 & 98 ) was the probability that they were constant @xmath24% ; these sources are marked with an `` h '' in the notes column of table [ tab : src ] .
76 appeared to brighten during the observation by a factor of @xmath17 1.6 , while src .
98 faded so as to be undetectable ( by a factor of @xmath21 1.5 ) .
as noted above , there was only one source , src . 1 , detected within 12@xmath25 of the center of ngc 4649 .
we believe that this detection does not represent an individual source , but rather a structural feature in the diffuse emission .
this source is 5.22 times wider ( fwhm ) than the psf of a point source at the same location , and is wider than any other detected source in the image .
we will therefore drop src . 1 from further discussion of the resolved sources .
we compared the positions of the x - ray sources with the digital sky survey ( dss ) image of this region ( fig .
[ fig : dss_field ] ) .
fourteen of the sources had possible faint optical counterparts on this image .
these are all marked with note `` d '' in table [ tab : src ] .
all of the possible x - ray / optical source matches occur in the outer part of the field away from the center of ngc 4649 since the optical detection of sources near the center of the galaxy is difficult .
three of the possible optical counterparts ( srcs .
136 , 162 , 165 ) have positions listed in the usno - a2.0 catalog ( monet et al .
these sources are indicated by a `` e '' in the notes column of table [ tab : src ] .
we find that two of the usno optical sources have positions which match those of x - ray sources to within 1 , and one matches to within 2 ( table [ tab : src ] ) .
given the density of usno sources in the field , one would expect about 0.3 misidentifications within 2 of the _ chandra _ sources .
we also compared the positions of the x - ray sources with those of globular clusters ( gcs ) in ngc 4649 . in the outer parts of the galaxy ( @xmath26 ) ,
a list is provided by hanes ( 1977 ) , based on ground - based optical images .
107 , 136 , 157 , 160 , & 162 correspond to hanes clusters 36 , 49 , 19 , 25 , & 17 , respectively .
based on the positional accuracy , the numbers of globular clusters , the number of x - ray sources , and the area of the sky covered , we estimate that @xmath170.7 x - ray sources would be expected to agree with hanes gcs by coincidence .
it is worth noting that , at the distance to ngc 4649 , globular clusters are not resolved in ground - based optical images . as a result , as many as half of these gcs which are located at large radii might be unrelated faint optical objects , rather than globular clusters .
kundu & whitmore ( 2001 ) have determined the globular cluster population of a region of ngc 4649 to the northwest based on _
hubble space telescope ( hst ) _ images .
confusion with background sources should not be significant in the _ hst _ data as these images slightly resolve the globular clusters .
a. kundu ( private communication ) very kindly provided the positions , magnitudes , and colors of these globulars .
we found that 24 of these gcs were within 09 of the positions of x - ray sources .
two of these may be associated with ngc 4647 .
there were 445 gcs associated with ngc 4649 and 47 x - ray sources in the region covered by the _ hst _ image and not associated with ngc 4647 ; one would expect @xmath93 associations to occur at random .
thus , it is likely that most of these identifications are real . within the region of ngc 4649 covered by the _ hst _
image , roughly 47% of the x - ray sources are identified with globular clusters .
a serendipitous _ rosat _ source rx j1244.1 + 1134 ( romer et al .
2000 ) was detected in the pspc pointed observation of ngc 4649 at a position which is consistent with srcs . 164 & 165 . from comparison of our _ chandra _ image with the _ rosat _ pspc image , we believe that this source is actually a blend of these two sources , plus another source ( comparable to src . 165 in brightness ) , but located just off the s3 chip on the s4 chip .
romer et al . also indicated that this source is probably a blend .
as shown in figure [ fig : dss_field ] , ngc 4649 has a nearby companion galaxy , the sc galaxy ngc 4647 .
some of the x - ray sources are projected on the optical image of this galaxy .
these sources are marked with note `` g '' in table [ tab : src ] .
in particular , src . 119 is located within 4 of the nucleus of ngc 4647 .
this galaxy also was the host of the type i supernova sn1979a ( barbon et al .
we were unable to find any very accurate position for the supernova in the literature ; however , none of the x - ray sources are within 20 of the approximate position given by barbon et al .
( 1984 ) .
the count rates for the sources were converted into unabsorbed luminosities ( 0.3 - 10 kev ) assuming that all of the sources were at the distance of ngc 4649 , which we take to be 16.8 mpc ( tonry et al .
we adopted the best - fit _
chandra _ x - ray spectrum of the resolved sources within the inner 1 effective radius ( @xmath27 , van der marel 1991 ; table [ tab : spectra ] below ) .
the factor for converting the count rate ( 0.310 kev ) into the unabsorbed luminosity @xmath28 ( 0.310 kev ) was @xmath29 ergs cnt@xmath5 .
the resulting x - ray luminosities are given in column 8 of table [ tab : src ] in units of @xmath30 ergs s@xmath5 , and range roughly from @xmath31 to @xmath32 ergs s@xmath5 .
we determined the luminosity function for the sources with @xmath33 , excluding a very small region near the chip edge with a low exposure .
this includes sources which may be associated with ngc 4647 ; however , when we compared the number of sources in the region of ngc 4647 to the number of sources in other regions of similar size at the same projected distance from ngc 4649 , we found no evidence for an over - abundance of sources near ngc 4647 . note that removing the region of low exposure removed src .
136 , which has the highest luminosity of any detected source .
we did not include the center of the galaxy ( @xmath34 ) because the brightness of the diffuse gaseous emission increased the minimum detectable source flux there .
the cumulative luminosity function of the sources from this region is shown as a histogram in figure [ fig : xlum ] , under the assumption that all of the sources are located at the distance of ngc 4649 .
we fit the luminosity function , using the same techniques as we used previously ( sarazin et al . 2000 , 2001 ;
blanton et al . 2001 ; irwin , sarazin , & bregman 2001 ) . in fitting the luminosity function
, we included a correction for unrelated sources using the deep source counts in brandt et al .
( 2000 ) and mushotzky et al .
we first tried a single power - law model for the differential luminosity function ( dashed curve in fig .
[ fig : xlum ] ) , but it did not provide a very good fit and could be rejected at the @xmath3595% level . a broken power - law model
, @xmath36 gave a good fit , where @xmath37 is the x - ray luminosity ( 0.310 kev ) in units of @xmath38 ergs s@xmath5 .
the best - fit values and 90% confidence uncertainties are @xmath39 , @xmath40 , @xmath41 , and a break luminosity of @xmath42 ergs s@xmath5 ( 90% errors ) .
note that the normalization @xmath43 only applies to the area of the galaxy used to derive the luminosity function ; the total population is probably about a factor of two larger , assuming that the x - ray sources are distributed similarly to the optical light in the galaxy .
the low - luminosity slope @xmath44 is very similar to that found in the elliptical ngc 4697 ( sarazin et al .
2000 , 2001 ) and the sa bulge ngc 1291 ( irwin et al .
in fact , irwin et al .
( 2001 ) argued that the slope and normalization ( relative to the optical luminosity ) of the low end of the luminosity function in early - types galaxies and spiral bulges might be universal .
the high luminosity slope @xmath45 is steeper than in ngc 4697 , but consistent within the errors . the break luminosity @xmath46 is larger than has been found in other early - type galaxies ( blanton et al .
2001 ; sarazin et al .
2001 ; finoguenov & jones 2002 ; kundu et al .
2002 ) , although it is consistent with these other measurements within the ( relatively large ) errors .
sarazin et al .
( 2000 ) suggested that @xmath47 corresponded to the eddington luminosity of a 1.4 @xmath48 neutron star ( @xmath49 ergs s@xmath5 ) , and that this value is universal .
the luminosity function in ngc 4649 may suggest that there are variations in @xmath46 from galaxy to galaxy , and/or that @xmath46 is larger than @xmath50 by a factor of @xmath171.52 .
alternatively , variations in @xmath46 might be due to distance errors . in ngc 4649
, the distance we adopted would need to be too large by @xmath1725% to give a best - fit break luminosity of @xmath51 ergs s@xmath5 .
this is larger than the stated statistical errors ( tonry et al .
2001 ) .
we determined x - ray hardness ratios for the sources , using the same techniques and definitions we used previously ( sarazin et al .
2000 , 2001 ; blanton et al . 2001 ; irwin et al .
hardness ratios or x - ray colors are useful for crudely characterizing the spectral properties of sources , and can be applied to sources which are too faint for detailed spectral analysis .
we define two hardness ratios as h21 @xmath52 and h31 @xmath53 , where @xmath54 , @xmath55 , and @xmath56 are the net counts in the soft ( 0.31 kev ) , medium ( 12 kev ) , and hard ( 210 kev ) bands , respectively .
the hardness ratios are listed in columns 9 & 10 of table [ tab : src ] for all of the resolved sources .
the errors in the hardnesses ratio are determined from the poisson errors in the original counts in the bands , and are carefully propagated so as to avoid mathematically impossible hardness ratios ; that is , the error ranges are limited to @xmath57 to 1 .
figure [ fig : colors ] plots h31 vs. h21 for all of the 107 sources with at least 20 net counts . for comparison ,
the hardness ratio for the sum of the sources is ( h21,h31 ) @xmath58 . as was also seen in ngc 4697 , ngc 1553 , and the bulge of ngc 1291 ( sarazin et al . , 2000 , 2001 ;
blanton et al . , 2001 ; irwin et al . ,
2001 ) , most of the sources lie along a broad diagonal swath extending roughly from ( h21,h31 ) @xmath59 to ( 0.3,0.2 ) . for example , these colors correspond to galactic absorption and power - law spectra with photon indices of @xmath60 to @xmath57 . in figure [
fig : colors ] , there are five sources with very hard spectra ( hardness ratios [ h21,h31 ] @xmath35 @xmath61 $ ] ) ; these are srcs . 23 , 52 , 77 , 84 , & 110
. these may be unrelated , strongly absorbed agns , similar to the sources which produce the hard component of the x - ray background , and which appear strongly at the faint fluxes in the deep _ chandra _ observations of blank fields ( brandt et al .
2000 ; mushotzky et al .
2000 ; giacconi et al .
however , of these five sources , two ( srcs .
23 & 52 ) have hardness ratio errors that include the full rage of possible values and two more ( srcs .
77 & 84 ) have error ranges that include @xmath62 .
therefore , src .
110 is the only source that can be said to be a strongly absorbed background agn with any certainty .
there also is a group of sources with hardness ratios of around @xmath63 ; these are srcs . 2 , 7 , 10 , 19 , 56 , 81 , 129 , 137 , 149 , 158 , & 162 .
these sources have very little hard emission , and many are at large radii . due to the lack of hard emission ,
some of these sources have very poorly constrained h31 values ( e.g. srcs .
56 , 81 , 158 , & 162 ) , although the h21 values are known to reasonable accuracy .
studies of other galaxies ( sarazin et al . 2001 ) and deep blank sky images ( e.g. , giacconi et al .
2001 ) suggest that many of these sources may also be unrelated background sources
. however , some of these sources ( particularly the ones at smaller radii ) may be sources in ngc 4649 .
the total number of sources with hardness ratios of @xmath17(1,1 ) or @xmath17@xmath63 , including faint sources not plotted in figure [ fig : colors ] with large errors in the hardnesses ratios , is @xmath1730 , which is considerably higher than the @xmath1710 background sources expected based on deep _ chandra _ images .
similarly , the total number of expected background sources based on deep _ chandra _ counts with these hardness ratios and @xmath64 net counts ( e.g. , those sources that would be plotted in figure [ fig : colors ] ) is @xmath65 , as compared to the 12 sources listed above with relatively well - known hardness ratios .
thus , it is likely that many of these sources are associated with ngc 4649 or ngc 4647 .
we used the ciaoscript psextract to extract spectra . for
extended or multiple sources , the response matrices were determined using the calcrmf / calcarf package written by alexey vikhlinin and jonathan mcdowell , which weighted them by the x - ray brightness in the 0.52 kev energy band over the corresponding image region .
we used the gain file acisd2000-01-29gainn0003.fits and the fef file acisd2000 - 01 - 29fef_pin0001.fits .
for the individual resolved sources , local backgrounds were used . for the diffuse spectra
, we used the blank sky background files provided by maxim markevitchto generate background spectra . since the count rate we detected on the s1 chip was in excess of the quiescent background rate expected ( see [ sec : obs ] ) , we renormalized the blank sky background spectra to match our observed s1 data in the 510 kev band .
the blank sky background spectra for the s3 chip , which was used as background for the spectral analysis of the diffuse emission throughout , was normalized by this same factor .
when fitting spectra we consider only the 0.710 kev energy band throughout since the response below 0.7 kev is uncertain .
each spectrum has been grouped to a minimum of 20 counts per pulse invariant ( pi ) channel so that @xmath66 statistics apply .
we used xspec to fit models to the spectra .
for all the fitted models discussed here , the absorption column was fixed at the galactic value ( @xmath67 @xmath68 ; dickey & lockman 1990 ) . allowing the absorption to vary did not improve the fits significantly .
one concern with normalizing the blank sky background spectra in this way is that the spectrum of the excess background in our observations may differ from that in the blank sky background observations .
for example , the excess background might be due to the virgo cluster or the north polar spur . to model this effect , we determined the spectrum of the excess background on the s1 chip , using the normalized blank sky data as a background . since we expect this excess emission to be mainly due to diffuse gas ( in the virgo cluster or north polar spur ) , we modeled this excess emission with a mekal model .
while we did find a statistically good fit , the temperature and abundance were only very poorly constrained .
we included this excess background component in our fits for the diffuse emission in ngc 4649 , freezing the spectral parameters to the best fit values obtained above .
the normalization was scaled with the area of the detector used . although including this component did somewhat affect the normalizations of the other model components , the fitted values for the temperature and abundance agreed with the results we obtained without including this component to well within the errors . including this excess emission component in the fit to the outermost annulus defined in table [ tab : spec_grad ] , where the effect is likely to be the largest due to the faintness of the diffuse emission there , only changed our results by an amount comparable to the errors ( for instance
, the best fit temperature decreased by 0.02 kev ) .
we conclude that the difference in the shape of the spectra between our true background and the blank sky background data is not large enough to significantly affect our spectral fitting results , and therefore do not include this component when doing spectral analyses of diffuse emission .
a summary of the best fitting spectral models is given in table [ tab : spectra ] for different components of the x - ray emission .
the first column gives the origin of the spectrum , the second column lists the spectral model used , the third and fourth columns give the temperature @xmath69 and abundances ( if relevant ) for the softer component of the spectrum , the fifth column gives the power - law photon spectral index @xmath70 or temperature @xmath71 of the harder component in the spectrum , the sixth column gives the value of @xmath66 and the number of degrees of freedom ( dof ) , and the last column gives the number of net counts ( after background subtraction ) in the spectrum .
figure [ fig : src_spec ] shows the spectrum of the sum of the sources on the s3 chip within one effective radius ( @xmath27 ) of the center of ngc 4649 .
if the lmxbs are distributed like the stars then roughly 50% of the lmxbs will be contained within this radius , whereas we expect to find @xmath201 background agn in this region . choosing this radius has the additional advantage that it facilitates comparisons of our results with results from optical observations of ngc 4649 . for this source spectrum
, we only considered photon energies of 0.78.0 kev since the bins with energies @xmath72 kev had very large errors .
we first considered models in which the spectrum of the sources was represented by a single , hard component .
initially , we tried a thermal bremsstrahlung model ( `` bremss '' in table [ tab : spectra ] ) with a temperature @xmath71 . while this model gave a statistically significant fit , with a @xmath66 per dof of @xmath73 ,
the temperature @xmath71 was poorly constrained . as an alternative
, we also fit the source spectrum with a power law with a photon index of @xmath70 .
this gave an acceptable fit to the spectrum which was slightly better than the fit given by the thermal bremsstrahlung model , with a @xmath66 per dof of @xmath74 .
we will adopt this as our best - fit model for the spectrum of the source population .
we also tried a model for the sources with a power - law hard component and a blackbody ( `` bbody '' ) soft component .
this gave a slightly better fit to the spectrum , but had very large uncertainties in the power - law index and blackbody temperature .
thus , we prefer the simpler power - law model . we extracted the diffuse emission within one @xmath75 of the center of ngc 4649 ( figure [ fig : diffuse_spec ] ) , excluding all sources except src .
the x - ray color of the diffuse emission in this region is soft , with hardness ratios of ( h21,h31 ) @xmath76 .
the spectrum shows strong emission lines .
the most prominent are the helium - like lines ( @xmath171.85 kev ) and lines ( @xmath172.45 kev ) . also present in the spectra
are lines and a blend of fe l lines .
the strong lines indicate that the diffuse radiation is mainly thermal emission from interstellar gas , as expected in this x - ray bright elliptical galaxy . based on the line emitting spectrum , we first tried a single temperature mekal model .
this did not provide an acceptable fit .
in particular , the residuals showed that there was a significant excess of hard x - ray emission .
since we would expect some hard x - ray emission from point sources below our detection limit , we tried modeling this hard component with a power - law spectrum .
this provided a much improved fit , and removed the excess hard emission .
the best - fit photon exponent was @xmath77 , which is very close to the value found for the resolved sources ; fixing the photon exponent to the best fit value found in
[ sec : src_spec ] , @xmath78 , does not significantly alter the fit so we assumed this value for consistency .
this best - fit mekal model , shown in figure [ fig : diffuse_spec ] , gave a gas temperature of @xmath79 and a heavy element abundance of about 0.58 solar .
we interpret this fit as representing soft emission from hot diffuse gas combined with a hard power - law emission from unresolved sources .
this model still has a @xmath80 which is larger than unity . examining the residuals
, one finds that a substantial portion of the @xmath66 originates in residuals at the positions of lines .
for example , the si and s lines in the model spectrum are too weak , and the fe lines are a bit too strong .
this suggests that the ratios of the heavy elements in this galaxy do not simply scale with the solar ratios . to include wider abundance variations
, we fit the soft component of the spectrum with a vmekal model , in which each of the heavy element abundances could vary independently .
we also fixed the normalization of the vmekal component to the normalization given for the mekal component from the soft mekal plus hard power - law model just discussed . allowing this normalization to vary
did somewhat improve the @xmath66 of the fit , but it gave abundance values which were unreasonably high . while the vmekal model improved the fit , most of the thirteen abundances considered in this model were very poorly constrained . examining
the abundances from this fit suggested that the elements could be divided into two groups : the light elements ( c through ar ) and the iron group elements ( ca , fe , and ni ) .
this model , with the normalization allowed to vary , gave a somewhat better fit to the spectrum , with a gas temperature of @xmath81 kev and an abundance of about 1.68 solar for the light elements and 1.12 solar for the iron group elements ( see also table [ tab : spectra ] ) .
it may be surprising that these abundance values are both higher than the abundance found above by the best - fit mekal model ( one might have expected the overall abundance to be an average of the abundances for the light element and iron group elements ) .
this occurs because separating the elements into two groups allows for more freedom in determining the normalizations of each model component . as a result the vmekal normalization
is decreased relative to the normalization of the mekal model found above and the abundances must be increased to match the observed line strengths . since this vmekal model still did not provide a good fit to the spectrum of the diffuse emission , we tried adding a cooling flow component ( mkcflow ) to the mekal plus power - law model described above , with the high temperature and abundance tied to the temperature and abundance values of the mekal component and the lower temperature frozen at @xmath82 kev . adding
the mkcflow component did not improve the @xmath66 of the fit .
we found an upper limit on the mass accretion rate of @xmath83 yr@xmath5 .
the spectral properties of the diffuse emission as a function of radius were examined to search for radial abundance and temperature gradients in the gas .
the area within 120 of the galactic center was broken up into 9 concentric annuli each containing roughly 2000 net counts .
the outer annuli have more gross counts ( but a higher fraction of background ) , so that these spectra have more grouped bins and a larger number of dofs . since the mekal and vmekal models provided roughly statistically equivalent fits to the total diffuse emission ( see [ sec : diffuse_spec ] ) , we choose to fit each region with the simpler mekal plus power - law spectral model with @xmath70 fixed at 1.78 .
the results of the fits are given in table [ tab : spec_grad ] .
although the temperature of the diffuse gas does appear to vary slightly as a function of radius , there is no clear trend with temperature or abundance as a function of radius .
if one excludes the central circle ( where the fit is poor ) , there is a stronger trend for the temperature to increase slightly with increasing radius .
while most of the fits in table [ tab : spec_grad ] are adequate , the spectral fit to the central circle is not .
this suggests that the central region may require an additional component or components to accurately model its spectrum .
note that the central circle contains enhanced diffuse emission due to a `` bar '' ( [ sec : central_bar ] ) .
an examination of the residuals of the mekal plus power - law fit to the central region revealed that the model was systematically fainter than the data at energies @xmath20 0.9 kev .
we therefore tried adding a cool component to the mekal plus power - law model to accommodate the data .
first we tried adding a second mekal component , with the abundances of the two thermal components being equal , which improved the fit . allowing the abundances to vary independently
did not significantly alter the fit .
the two temperatures in this model were @xmath84 and @xmath85 kev , with an abundance of @xmath86 solar .
the @xmath66 of the two - temperature fit is 67.9 for 44 degrees of freedom , and while this is a significant improvement over the single temperature model it still is not a particularly good fit .
we therefore also tried a mekal plus power - law plus cooling flow ( mkcflow ) model with the abundances and higher temperatures of the mekal and mkcflow components set equal to one another and the lower temperature set to @xmath82 kev . while the @xmath66 of this fit was 78.0 for 45 degrees of freedom , better than the mekal plus power - law model , the @xmath66 of this fit was not as good as that of the two mekal component model .
in addition , the abundance was very poorly constrained .
this fit gave a temperature of @xmath87 kev , and an upper limit on the mass accretion rate of @xmath88 yr@xmath5 .
we note that other authors have also found a preference for two - temperature models over cooling flow and single - phase models in the centers of elliptical galaxies ( e.g. , buote 2002 ) .
the smoothed image in figure [ fig : xray_smo ] shows some evidence for structure in the diffuse emission near the core of ngc 4649 .
specifically , there appeared to be faint `` fingers '' extending out from the core , between about 21 and 53 from the center of the diffuse emission .
these radial features are shown more clearly in figure [ fig : soft_nuc ] , which is an adaptively smoothed image of the center of ngc 4649 .
these features are far too strong to be due to the support structure of the telescope .
to check the statistical significance of these features , we defined a series of regions containing the most obvious fingers .
we considered only the raw data , unsmoothed and uncorrected for exposure , to ensure that the fingers were not an artifact of the smoothing process .
detected point sources were removed from all of the regions .
we then compared the average surface brightness of these regions to that of other regions in between the fingers at the same distance from the center of ngc 4649 .
for the regions containing the fingers , we found on average @xmath89 cnt s@xmath5 pix@xmath5 , and @xmath90 cnt s@xmath5 pix@xmath5 for the regions without fingers ( 1-@xmath10 errors ) .
the excess of counts in the regions containing fingers was therefore significant at the 4.1-@xmath10 level .
as a further test of the significance of these features , we examined the azimuthal brightness profile in an annulus ranging from 21 to 53 centered on the peak of the diffuse emission .
we then divided this annulus into 20 angular bins of 18 each and did a @xmath66 significance test comparing the counts in each bin to the mean number of counts per bin ( an azimuthal plot of the flux in each bin is given in figure [ fig : fingers ] ) .
we found a reduced @xmath66 of @xmath91 , which suggests that the azimuthal brightness profile is not well described by a constant .
it is therefore possible that the radial fingers we observed are real features in the diffuse emission .
the hardness ratios of the fingers , ( h21,h31 ) @xmath92 , and the hardness ratios of the regions between the fingers ( h21,h31 ) @xmath93 were both similar to the hardness ratios of the total diffuse emission within one effective radius ( see [ sec : diffuse_spec ] ) .
radial features have been produced in some numerical hydrodynamical simulations of cooling flows in elliptical galaxies ( kritsuk , bhringer , & mller 1998 , hereafter kbm ) .
kbm give results from two - dimensional axisymmetric hybrid inflow - outflow models which show such features .
the physical extent of their features ( @xmath94 kpc ) is about the same as those we observe ( @xmath95 kpc ) .
their model predicts an azimuthally - averaged temperature profile which is depressed in the central region of the galaxy and rises out to about 10 kpc ( see their fig .
we find some evidence for a temperature profile which rises with radius , excluding the central 4 of the galaxy where the x - ray spectrum may be complicated by the presence of the x - ray bar discussed in [ sec : central_bar ] ( see table [ tab : spec_grad ] ) .
the simulations of kbm predict outflowing jets of hot gas , surrounded by inflowing gas which is significantly cooler than the outflowing gas at the same distance from the galactic center ( see their fig . 7 ) .
we therefore fit spectra accumulated from the sum of the finger regions and from the regions in between the fingers , using the same mekal plus power - law model used to fit the spectrum of the diffuse emission .
the abundances in these fits were fixed at 0.6 solar . if the abundances were allowed to vary the resulting values were unreasonably high and had large errors . the temperatures and @xmath66 s given by the fits were not significantly affected by fixing the abundance values .
the results of these fits are given in table [ tab : spectra ] . while the bright fingers are slightly hotter than the regions between the fingers , the 90% confidence intervals overlap . in the kbm model
, one would expect larger temperature differences of about a factor of two , although projection effects might somewhat reduce these differences .
more seriously , in the kbm model the bright regions would be considerably cooler than the faint regions , whereas the bright regions are slightly hotter in the observations . thus , while some form of convective instability may explain the fingers in ngc 4649 , only the morphology is reproduced by the kbm model .
figure [ fig : bar ] shows an even smaller region at the nucleus of ngc 4649 from the same image shown in ( figure [ fig : soft_nuc ] ) .
there appears to be a small bar , approximately 5 long , located at the central peak in the diffuse emission . to assess the significance of the elongation of this feature
, we determined the net counts in an elliptical region corresponding to this bar and compared it to the net counts in an identical region perpendicular to and crossing the center of the bar .
the counts in the area where these two regions overlapped were excluded .
we found @xmath96 net counts in the bar region and @xmath97 net counts in the perpendicular off - bar region ( 1-@xmath10 errors ) . while this is not a large difference , it is marginally statistically significant ( a 2.5-@xmath10 detection )
the bar appears to be oriented roughly perpendicular to the x - ray isophotes at larger radii , which go from sw to ne .
it also appears to be somewhat asymmetric , with the sw end being somewhat brighter than the ne end ( if we identify the brightest point in the bar as its center ) .
the bar is included in the region given for src . 1 as detected by the source detection algorithm we used ( see [ sec : src_detect ] ) , although the region of src .
1 is not centered on nor rotationally aligned with the ( much smaller ) bar .
radio observations of the center of ngc 4649 with the very large array show an extended object , @xmath98 in size oriented 31 east of north ( condon et al .
1991 ) , making it coincident with the x - ray bar we observe .
the radio source has an integrated flux density of 18 mjy at 4.86 ghz .
a recent unrelated study of the _ chandra _ data has detected a soft ( 0.20.6 kev ) central source that the authors interpret as being a quiescent supermassive black hole ( soldatenkov , vikhlinin , & pavlinsky 2003 ) .
although no exact position for this source is given it is said to be at the peak of the optical emission , and therefore presumably near to or contained within the x - ray bar we observe .
unfortunately , there are too few counts from the bar to determine its spectral properties .
we can , however , determine the x - ray colors of the bar .
we find that the x - ray color is medium - soft , with ( h21,h31 ) @xmath99 , 1-@xmath10 errors ) .
this is somewhat harder than the color of the diffuse emission within one effective radius , ( h21,h31 ) @xmath100 ) .
it is also slightly harder than the color of the perpendicular off - bar region , although they are consistent within the errors .
however , the x - ray emission is probably not hard enough to indicate a strong shock in the region of the bar .
there are several possible explanations for this bar .
first , the bar may be an emission feature , or due to excess absorption perpendicular to the bar .
if the bar were due to excess absorption perpendicular to the bar , the off - bar region would have a particularly hard spectrum , whereas the hardness ratios indicate that the bar is slightly harder than the off - bar region .
in addition , when doing spectral fits to the central region of ngc 4649 , allowing the local absorption to vary did not improve the fit .
if the bar is an emission feature , it might be a rotating disk or torus viewed edge - on , perhaps produced by the cooling of rotating x - ray gas ( e.g. , brighenti & mathews 2000 ) .
however , if this were the case one would expect the bar to be cooler than the surrounding gas , which is not true here .
alternatively , the bar could be a shock associated with the expansion of a central radio source , as appears to have occurred in ngc 4636 ( jones et al .
although the x - ray colors of the bar are not hard enough to indicate a strong shock , they are somewhat harder than the surrounding diffuse emission .
furthermore , the bar appears to be coincident with the extended radio source at the center of the galaxy , although the radio emission is relatively weak .
although we did not detect a central x - ray source which could be associated with an agn , the recent detection by soldatenkov et al .
( 2003 ) , which they interpret as a quiescent supermassive black hole , could be responsible for the radio structure and the x - ray bar .
however , it should be noted that the interpretation of the central source as an lmxb can not be completely ruled out due to its low luminosity ( @xmath101 ergs s@xmath5 in the 0.20.6 kev band ) .
still , the interpretation of the bar as resulting from compression from the radio source is probably the most consistent explanation given the existing data .
we have presented results from a _
chandra _ observation of the x - ray bright elliptical galaxy ngc 4649 .
both bright diffuse emission and point sources are detected , with the bright diffuse emission dominating the overall emission .
a total of 165 discrete sources were detected .
fourteen of these sources had possible faint optical counterparts in the dss image of the same field .
of these fourteen , three had positions listed in the usno - a2.0 catalog . in a region to the nw of the center of ngc 4649
we find roughly 20 x - ray sources whose positions match those of globular clusters seen in hst images . in this region ,
roughly 45% of the x - ray sources are identified with globular clusters .
some of the sources may be associated with the companion sc galaxy ngc 4647 .
however , we do not detect x - rays from the type i supernova sn1979a in ngc 4649 .
we do detect the serendipitous _ rosat _ source rx j1244.1 + 1134 as a blend of three sources .
the luminosity function of the resolved sources is well described by a broken power - law , with a break luminosity which is somewhat larger than the eddington luminosity for a 1.4 @xmath102 neutron star .
this break luminosity is slightly larger than previously determined values for other early - type galaxies ( sarazin et al .
2000 ; blanton et al .
2001 ) , although it is consistent with previous measurements within the errors .
emission from diffuse gas swamps emission from discrete sources near the center of ngc 4649 .
no point sources are detected within 12 of the center of the galaxy .
we find no conclusive evidence for a central agn , although another study of the _ chandra _ data has found evidence for a central quiescent supermassive black hole .
we give an upper limit of @xmath28 ( 0.310 kev ) @xmath103 erg s@xmath5 for the x - ray luminosity of any agn .
the composite x - ray spectrum of the resolved sources within 1 @xmath104 is best described by a power - law with a photon spectral index of @xmath78 .
the spectrum of the diffuse emission is best fit by a mekal plus power - law model , with a gas temperature @xmath105 kev , an abundance of roughly 58% of solar , and @xmath70 frozen at 1.78 .
this argues that the diffuse emission is a combination of emission from diffuse gas and unresolved lmxbs .
there is some evidence that the abundances of iron group elements are lower than those of lighter elements when compared to the solar ratios .
there is also evidence for radial features in the diffuse emission extending out from the center of ngc 4649 , and for a central bar .
two - dimensional hydrodynamic simulations of cooling flows in elliptical galaxies by kritsuk et al .
( 1998 ) predict such radial features , which result from convective motions with relatively hot outflowing gas separated by regions of cooler inflowing gas .
however , only the morphology of the observed fingers match the specific predictions of the kbm model . the central bar is roughly perpendicular to the x - ray and optical isophotes at larger radii , is parallel to the extension in the weak central radio source , and may contain a recently detected quiescent supermassive black hole .
we suggest that the x - ray bar may be produced by an interaction between the central radio source and the surrounding x - ray gas .
we are very grateful to arunav kundu for providing us with his unpublished list of globular clusters in ngc 4649 , and for several very helpful conversations .
support for this work was provided by the national aeronautics and space administration primarily through @xmath106 award number go0 - 1141x , but also through go1 - 2078x , both issued by the @xmath106 x - ray observatory center , which is operated by the smithsonian astrophysical observatory for and on behalf of nasa under contract nas8 - 39073 .
lcccrrrrccl src .
& name & r.a . & dec .
& & & & & h21 & h31 & + no .
& & ( h : m : s ) & ( @xmath107:@xmath108:@xmath25 ) & & & & & & & notes + ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & & & & & ( 9 ) & ( 10 ) & ( 11 ) + 1&cxou j124340.0 + 113311&12:43:40.01&11:33:11.8 & 1.63&[email protected] & 46.21 & 1441.35 & & & a + 2&cxou j124339.2 + 113317&12:43:39.25&11:33:17.4 & 13.48 & [email protected] & 9.70 & 75.36&@xmath109&@xmath110&b + 3&cxou j124340.2 + 113324&12:43:40.22&11:33:24.7 & 14.79 & [email protected] & 7.04 & 45.94&@xmath111&@xmath112&b , f + 4&cxou j124340.0 + 113252&12:43:40.02&11:32:52.0 & 18.19 & [email protected] & 7.13 & 46.77&@xmath113&@xmath114&b + 5&cxou j124340.9 + 113325&12:43:40.99&11:33:25.2 & 20.65 & [email protected] & 5.23 & 29.47&@xmath115&@xmath116&b + 6&cxou j124338.6 + 113304&12:43:38.64&11:33:04.3 & 21.19 & [email protected] & 7.14 & 48.31&@xmath117&@xmath118&b , f + 7&cxou j124341.3 + 113320&12:43:41.34&11:33:20.5 & 21.91 & [email protected] & 4.97 & 23.66&@xmath119&@xmath120&b + 8&cxou j124339.6 + 113248&12:43:39.61&11:32:48.4 & 22.62 & [email protected] & 6.46 & 37.01&@xmath121&@xmath122&b + 9&cxou j124338.3 + 113305&12:43:38.38&11:33:05.4 & 24.64 & [email protected] & 6.20 & 35.85&@xmath123&@xmath124&b + 10&cxou j124341.1 + 113248&12:43:41.11&11:32:48.8 & 26.68 & [email protected] & 5.15 & 24.26&@xmath125&@xmath126&b + 11&cxou j124340.4 + 113243&12:43:40.45&11:32:43.3 & 27.61 & [email protected]&10.21 & 86.05&@xmath127&@xmath128&b + 12&cxou j124341.1 + 113333&12:43:41.14&11:33:33.3 & 28.33 & [email protected] & 5.02 & 22.36&@xmath129&@xmath130&b + 13&cxou j124342.0 + 113321&12:43:42.00&11:33:21.8 & 31.26 & [email protected] & 3.93 & 14.73&@xmath131&@xmath132&b + 14&cxou j124339.4 + 113238&12:43:39.40&11:32:38.0 & 33.47 & [email protected] & 5.22 & 26.66&@xmath133&@xmath134&b + 15&cxou j124338.8 + 113241&12:43:38.89&11:32:41.1 & 33.53 & [email protected] & 4.69 & 19.23&@xmath135&@xmath136&b + 16&cxou j124337.6 + 113309&12:43:37.68&11:33:09.9 & 34.46 & [email protected] & 3.16&9.13&@xmath137&@xmath138&b + 17&cxou j124339.4 + 113235&12:43:39.40&11:32:35.0 & 36.36 & [email protected] & 4.25 & 16.21&@xmath139&@xmath140&b + 18&cxou j124337.8 + 113328&12:43:37.83&11:33:28.4 & 37.04 & [email protected] & 7.41 & 46.71&@xmath141&@xmath115&b , f + 19&cxou j124337.9 + 113332&12:43:37.94&11:33:32.1 & 37.67 & [email protected] & 4.73 & 21.21&@xmath142&@xmath143&b + 20&cxou j124338.7 + 113342&12:43:38.71&11:33:42.6 & 37.73 & [email protected] & 6.45 & 34.17&@xmath144&@xmath145&b , f + 21&cxou j124339.6 + 113347&12:43:39.68&11:33:47.6 & 37.75 & [email protected] & 3.78 & 14.34&@xmath146&@xmath147&b + 22&cxou j124342.2 + 113329&12:43:42.25&11:33:29.8 & 38.13 & [email protected] & 3.61 & 11.59&@xmath148&@xmath149&b , f + 23&cxou j124337.4 + 113309&12:43:37.43&11:33:09.2 & 38.15 & [email protected] & 5.02 & 20.88&@xmath150&@xmath151&b , f + 24&cxou j124341.1 + 113234&12:43:41.11&11:32:34.6 & 39.00 & [email protected] & 3.63 & 12.40&@xmath152&@xmath153&b + 25&cxou j124338.0 + 113242&12:43:38.06&11:32:42.2 & 40.21 & [email protected] & 6.88 & 36.45&@xmath154&@xmath155&b + 26&cxou j124342.2 + 113335&12:43:42.21&11:33:35.0 & 40.58 & [email protected] & 4.03 & 15.05&@xmath156&@xmath157&b + 27&cxou j124338.6 + 113345&12:43:38.60&11:33:45.0 & 40.62 & [email protected] & 4.39 & 17.26&@xmath158&@xmath159&b + 28&cxou j124337.2 + 113304&12:43:37.25&11:33:04.3 & 41.20 & [email protected] & 4.29 & 16.61&@xmath113&@xmath160&b , f + 29&cxou j124338.2 + 113343&12:43:38.27&11:33:43.9 & 42.44 & [email protected] & 5.59 & 26.36&@xmath161&@xmath162&b , f + 30&cxou j124340.8 + 113226&12:43:40.85&11:32:26.3 & 45.53 & [email protected] & 3.54 & 12.27&@xmath163&@xmath164 & + 31&cxou j124341.6 + 113351&12:43:41.65&11:33:51.9 & 48.07 & [email protected] & 7.26 & 40.70&@xmath165&@xmath166&f + 32&cxou j124342.6 + 113340&12:43:42.62&11:33:40.7 & 48.84 & [email protected] & 4.26 & 16.94&@xmath167&@xmath168&f + 33&cxou j124336.9 + 113329&12:43:36.96&11:33:29.9 & 49.17 & [email protected] & 3.85 & 13.84&@xmath169&@xmath170&f + 34&cxou j124343.3 + 113319&12:43:43.34&11:33:19.2 & 49.55 & [email protected] & 3.48 & 10.89&@xmath171&@xmath172 & + 35&cxou j124343.0 + 113245&12:43:43.03&11:32:45.1 & 50.79 & [email protected] & 3.89 & 13.20&@xmath173&@xmath174 & + 36&cxou j124342.3 + 113231&12:43:42.31&11:32:31.7 & 51.08 & [email protected] & 3.05&8.02&@xmath175&@xmath176 & + 37&cxou j124340.0 + 113218&12:43:40.02&11:32:18.5 & 51.69 & [email protected] & 3.74 & 12.39&@xmath177&@xmath178 & + 38&cxou j124341.3 + 113221&12:43:41.39&11:32:21.9 & 52.29 & [email protected] & 3.58 & 11.67&@xmath179&@xmath180 & + 39&cxou j124343.5 + 113302&12:43:43.59&11:33:02.8 & 52.91 & [email protected] & 8.59 & 54.00&@xmath181&@xmath182 & + 40&cxou j124338.0 + 113355&12:43:38.08&11:33:55.0 & 53.15 & [email protected] & 3.62 & 11.76&@xmath183&@xmath184 & + 41&cxou j124336.6 + 113251&12:43:36.61&11:32:51.0 & 53.73 & [email protected] & 3.99 & 13.31&@xmath185&@xmath186 & + 42&cxou j124336.1 + 113313&12:43:36.19&11:33:13.1 & 56.43 & [email protected] & 5.67 & 25.58&@xmath187&@xmath188&f + 43&cxou j124341.6 + 113218&12:43:41.69&11:32:18.5 & 57.19 & [email protected] & 3.58 & 11.46&@xmath189&@xmath190 & + 44&cxou j124343.8 + 113255&12:43:43.82&11:32:55.7 & 57.62 & [email protected] & 3.44 & 11.05&@xmath191&@xmath192 & + 45&cxou j124343.3 + 113341&12:43:43.36&11:33:41.6 & 58.22 & [email protected] & 4.23 & 14.43&@xmath193&@xmath194&f + 46&cxou j124338.2 + 113405&12:43:38.20&11:34:05.4 & 61.38 & [email protected] & 4.28 & 15.66&@xmath195&@xmath196&f + 47&cxou j124336.7 + 113348&12:43:36.75&11:33:48.6 & 61.58 & [email protected] & 3.91 & 13.22&@xmath197&@xmath198&f + 48&cxou j124342.9 + 113356&12:43:42.96&11:33:56.1 & 63.00 & [email protected] & 5.09 & 19.66&@xmath199&@xmath200&f + 49&cxou j124340.3 + 113413&12:43:40.30&11:34:13.5 & 63.44 & [email protected] & 3.60 & 11.00&@xmath201&@xmath202 & + 50&cxou j124344.6 + 113315&12:43:44.61&11:33:15.8 & 67.62 & [email protected] & 3.39 & 10.72&@xmath203&@xmath204 & + 51&cxou j124344.8 + 113254&12:43:44.82&11:32:54.6 & 72.17 & [email protected] & 4.46 & 15.33&@xmath205&@xmath206 & + 52&cxou j124337.7 + 113205&12:43:37.78&11:32:05.3 & 72.79 & [email protected] & 3.88 & 12.00&@xmath207&@xmath208 & + 53&cxou j124335.8 + 113350&12:43:35.80&11:33:50.7 & 74.13 & [email protected] & 5.22 & 22.75&@xmath209&@xmath210&f + 54&cxou j124345.1 + 113316&12:43:45.10&11:33:16.2 & 74.83 & [email protected] & 3.40 & 10.08&@xmath211&@xmath212 & + 55&cxou j124341.2 + 113157&12:43:41.24&11:31:57.4 & 74.95 & [email protected] & 3.31&9.02&@xmath213&@xmath214 & + 56&cxou j124341.7 + 113154&12:43:41.73&11:31:54.0 & 80.20 & [email protected] & 6.30 & 31.00&@xmath215&@xmath216 & + 57&cxou j124341.5 + 113428&12:43:41.52&11:34:28.5 & 81.34 & [email protected] & 3.34&8.74&@xmath217&@xmath218&f + 58&cxou j124335.2 + 113350&12:43:35.20&11:33:50.5 & 81.56 & [email protected] & 5.30 & 20.97&@xmath219&@xmath220 & + 59&cxou j124344.9 + 113233&12:43:44.99&11:32:33.7 & 81.58 & [email protected]&12.01 & 97.19&@xmath221&@xmath222 & + 60&cxou j124345.5 + 113319&12:43:45.56&11:33:19.8 & 81.91 & [email protected] & 3.04&7.68&@xmath223&@xmath224 & + lcccrrrrccl + src . &
name & r.a . & dec .
& & & & & h21 & h31 & + no .
& & ( h : m : s ) & ( @xmath107:@xmath108:@xmath25 ) & & & & & & & notes + ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & & & & & ( 9 ) & ( 10 ) & ( 11 ) + 61&cxou j124340.9 + 113431&12:43:40.94&11:34:31.0 & 81.92 & [email protected] & 3.31&8.65&@xmath225&@xmath226 & + 62&cxou j124334.3 + 113310&12:43:34.39&11:33:10.6 & 82.81 & [email protected] & 3.65&9.92&@xmath227&@xmath228&f + 63&cxou j124334.6 + 113237&12:43:34.68&11:32:37.3 & 85.16 & [email protected] & 5.47 & 21.38&@xmath229&@xmath230&f + 64&cxou j124345.5 + 113340&12:43:45.50&11:33:40.9 & 86.12 & [email protected] & 4.04 & 12.81&@xmath231&@xmath232 & + 65&cxou j124343.5 + 113420&12:43:43.52&11:34:20.2 & 86.83 & [email protected] & 3.19&8.43&@xmath233&@xmath234&d + 66&cxou j124345.3 + 113232&12:43:45.37&11:32:32.6 & 87.09 & [email protected] & 5.92 & 26.05&@xmath235&@xmath236 & + 67&cxou j124343.2 + 113156&12:43:43.23&11:31:56.8 & 87.21 & [email protected] & 6.93 & 33.12&@xmath237&@xmath238 & + 68&cxou j124334.9 + 113356&12:43:34.98&11:33:56.4 & 87.36 & [email protected] & 3.81 & 10.94&@xmath239&@xmath240 & + 69&cxou j124341.7 + 113434&12:43:41.75&11:34:34.1 & 87.66 & [email protected]&11.30 & 85.84&@xmath241&@xmath242 & + 70&cxou j124334.7 + 113227&12:43:34.79&11:32:27.6 & 87.94 & [email protected] & 8.05 & 44.18&@xmath243&@xmath244&f + 71&cxou j124335.3 + 113405&12:43:35.32&11:34:05.0 & 88.23 & [email protected] & 3.36&9.47&@xmath245&@xmath246 & + 72&cxou j124339.0 + 113438&12:43:39.06&11:34:38.5 & 89.44 & [email protected] & 4.16 & 14.26&@xmath247&@xmath248 & + 73&cxou j124333.9 + 113319&12:43:33.91&11:33:19.1 & 90.30 & [email protected] & 3.02&7.52&@xmath249&@xmath250 & + 74&cxou j124337.9 + 113435&12:43:37.94&11:34:35.5 & 90.65 & [email protected] & 7.32 & 37.25&@xmath251&@xmath252 & + 75&cxou j124334.0 + 113342&12:43:34.03&11:33:42.2 & 93.73 & [email protected]&11.70 & 92.84&@xmath253&@xmath254 & + 76&cxou j124337.2 + 113143&12:43:37.27&11:31:43.9 & 95.31 & [email protected]&15.20 & 150.83&@xmath255&@xmath256&h + 77&cxou j124346.4 + 113337&12:43:46.40&11:33:37.4 & 97.56 & [email protected] & 5.30 & 22.47&@xmath257&@xmath258 & + 78&cxou j124342.8 + 113439&12:43:42.81&11:34:39.8 & 98.52 & [email protected] & 4.36 & 15.04&@xmath259&@xmath260 & + 79&cxou j124342.6 + 113441&12:43:42.65&11:34:41.3 & 98.94 & [email protected] & 5.94 & 25.79&@xmath261&@xmath262 & + 80&cxou j124335.8 + 113430&12:43:35.82&11:34:30.1&101.02 & [email protected] & 5.78 & 24.47&@xmath263&@xmath264&f + 81&cxou j124333.6 + 113348&12:43:33.66&11:33:48.9&101.23 & [email protected] & 5.20 & 20.55&@xmath265&@xmath266 & + 82&cxou j124335.1 + 113422&12:43:35.17&11:34:22.4&101.51 & [email protected] & 4.66 & 17.01&@xmath267&@xmath268&f , g + 83&cxou j124335.4 + 113427&12:43:35.48&11:34:27.5&102.17 & [email protected] & 8.98 & 54.64&@xmath269&@xmath270&f , g + 84&cxou j124333.3 + 113342&12:43:33.38&11:33:42.2&102.77 & [email protected] & 5.18 & 19.38&@xmath271&@xmath272 & + 85&cxou j124333.9 + 113401&12:43:33.94&11:34:01.4&103.05 & [email protected] & 7.50 & 39.88&@xmath273&@xmath274 & + 86&cxou j124333.3 + 113240&12:43:33.30&11:32:40.1&103.31 & [email protected]&10.26 & 70.10&@xmath275&@xmath276 & + 87&cxou j124344.5 + 113151&12:43:44.55&11:31:51.0&103.41 & [email protected] & 8.44 & 48.77&@xmath277&@xmath278 & + 88&cxou j124342.1 + 113130&12:43:42.19&11:31:30.9&104.26 & [email protected] & 4.26 & 13.52&@xmath279&@xmath280 & + 89&cxou j124333.0 + 113246&12:43:33.09&11:32:46.9&104.54 & [email protected] & 4.04 & 11.99&@xmath281&@xmath282 & + 90&cxou j124346.8 + 113234&12:43:46.89&11:32:34.2&107.12 & [email protected] & 8.29 & 47.09&@xmath283&@xmath284 & + 91&cxou j124333.2 + 113229&12:43:33.27&11:32:29.5&107.29 & [email protected] & 4.31 & 13.58&@xmath285&@xmath286 & + 92&cxou j124347.0 + 113237&12:43:47.02&11:32:37.0&108.03 & [email protected]&12.22 & 98.98&@xmath287&@xmath288 & + 93&cxou j124333.4 + 113212&12:43:33.41&11:32:12.8&112.89 & [email protected] & 3.19&8.02&@xmath289&@xmath290 & + 94&cxou j124341.5 + 113117&12:43:41.55&11:31:17.3&115.09 & [email protected] & 3.66 & 10.18&@xmath291&@xmath292 & + 95&cxou j124335.6 + 113447&12:43:35.63&11:34:47.3&116.63 & [email protected] & 3.52&9.45&@xmath293&@xmath294&d , g + 96&cxou j124333.9 + 113152&12:43:33.92&11:31:52.6&118.62 & [email protected] & 5.06 & 18.32&@xmath295&@xmath296 & + 97&cxou j124342.7 + 113118&12:43:42.70&11:31:18.2&118.69 & [email protected] & 3.12&7.54&@xmath297&@xmath298 & + 98&cxou j124334.9 + 113443&12:43:34.95&11:34:43.0&119.06 & [email protected] & 5.23 & 19.55&@xmath299&@xmath300&g , h + 99&cxou j124347.5 + 113222&12:43:47.54&11:32:22.0&120.50 & [email protected] & 4.04 & 12.05&@xmath301&@xmath302 & + 100&cxou j124344.3 + 113125&12:43:44.30&11:31:25.8&121.84 & [email protected] & 3.97 & 12.37&@xmath303&@xmath304 & + 101&cxou j124336.0 + 113119&12:43:36.04&11:31:19.8&124.96 & [email protected] & 5.80 & 23.84&@xmath305&@xmath306&d + 102&cxou j124343.4 + 113115&12:43:43.43&11:31:15.6&125.04 & [email protected] & 5.70 & 22.56&@xmath307&@xmath308 & + 103&cxou j124331.9 + 113350&12:43:31.95&11:33:50.0&125.17 & [email protected] & 3.30&8.29&@xmath309&@xmath310 & + 104&cxou j124348.6 + 113302&12:43:48.68&11:33:02.9&127.41 & [email protected] & 6.68 & 30.40&@xmath311&@xmath312 & + 105&cxou j124331.3 + 113303&12:43:31.34&11:33:03.2&127.83 & [email protected] & 6.61 & 30.08&@xmath313&@xmath314 & + 106&cxou j124340.9 + 113517&12:43:40.96&11:35:17.4&127.95 & [email protected] & 3.64 & 10.56&@xmath315&@xmath316 & + 107&cxou j124348.6 + 113241&12:43:48.61&11:32:41.3&129.44 & [email protected] & 7.38 & 37.03&@xmath317&@xmath318&d , f + 108&cxou j124334.0 + 113446&12:43:34.07&11:34:46.7&130.28 & [email protected] & 3.83 & 11.73&@xmath319&@xmath320&g + 109&cxou j124347.8 + 113206&12:43:47.83&11:32:06.5&131.20 & [email protected] & 3.93 & 11.50&@xmath321&@xmath322&d + 110&cxou j124332.0 + 113418&12:43:32.07&11:34:18.8&135.55 & [email protected] & 9.57 & 61.85&@xmath323&@xmath324&g + 111&cxou j124330.9 + 113338&12:43:30.97&11:33:38.0&135.94 & [email protected] & 3.94 & 11.01&@xmath325&@xmath326 & + 112&cxou j124349.3 + 113237&12:43:49.35&11:32:37.0&141.01 & [email protected] & 5.21 & 19.19&@xmath327&@xmath328 & + 113&cxou j124332.0 + 113146&12:43:32.02&11:31:46.9&144.14 & [email protected] & 7.62 & 71.19&@xmath329&@xmath330&c + 114&cxou j124340.6 + 113045&12:43:40.69&11:30:45.9&144.62 & [email protected] & 4.44 & 13.92&@xmath286&@xmath331 & + 115&cxou j124344.4 + 113056&12:43:44.45&11:30:56.0&149.12 & [email protected] & 3.06&6.93&@xmath332&@xmath333 & + 116&cxou j124329.8 + 113318&12:43:29.85&11:33:18.5&149.76 & [email protected] & 8.90 & 52.31&@xmath334&@xmath335 & + 117&cxou j124344.7 + 113056&12:43:44.79&11:30:56.9&150.57 & [email protected] & 4.46 & 15.38&@xmath336&@xmath191&d + 118&cxou j124334.8 + 113520&12:43:34.83&11:35:20.8&151.28 & [email protected] & 3.31&7.99&@xmath337&@xmath338&g + 119&cxou j124332.2 + 113450&12:43:32.27&11:34:50.5&151.82 & [email protected] & 4.93 & 17.95&@xmath339&@xmath340&g + 120&cxou j124349.8 + 113218&12:43:49.85&11:32:18.3&153.43 & [email protected] & 4.68 & 15.43&@xmath341&@xmath342 & + lcccrrrrccl src . &
name & r.a .
& dec . & & & & & h21 & h31 & + no . & & ( h : m : s ) & ( @xmath107:@xmath108:@xmath25 ) & & & & & & & notes + ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & & & & & ( 9 ) & ( 10 ) & ( 11 ) + 121&cxou j124341.3 + 113037&12:43:41.30&11:30:37.6&153.74 & [email protected] & 3.24&7.61&@xmath343&@xmath344 & + 122&cxou j124339.9 + 113036&12:43:39.95&11:30:36.2&153.99 & [email protected] & 3.57&9.24&@xmath345&@xmath346 & + 123&cxou j124343.2 + 113537&12:43:43.29&11:35:37.0&154.45 & [email protected] & 3.38&8.64&@xmath347&@xmath348 & + 124&cxou j124334.7 + 113527&12:43:34.79&11:35:27.6&157.48 & [email protected] & 3.16&7.66&@xmath349&@xmath350&g + 125&cxou j124340.3 + 113547&12:43:40.36&11:35:47.9&157.79 & [email protected] & 3.71 & 12.28&@xmath351&@xmath352 & + 126&cxou j124345.3 + 113531&12:43:45.34&11:35:31.7&161.64 & [email protected] & 3.38&9.90&@xmath353&@xmath354 & + 127&cxou j124348.4 + 113110&12:43:48.46&11:31:10.9&172.04 & [email protected] & 4.00 & 11.19&@xmath355&@xmath356 & + 128&cxou j124329.8 + 113440&12:43:29.89&11:34:40.6&174.23 & [email protected] & 3.52&8.87&@xmath357&@xmath358&g + 129&cxou j124335.6 + 113026&12:43:35.62&11:30:26.8&175.75 & [email protected] & 4.84 & 65.22&@xmath359&@xmath360&c + 130&cxou j124351.6 + 113357&12:43:51.68&11:33:57.0&177.56 & [email protected] & 6.90 & 33.66&@xmath361&@xmath362 & + 131&cxou j124347.2 + 113533&12:43:47.26&11:35:33.6&178.53 & [email protected] & 7.36 & 38.20&@xmath363&@xmath364 & + 132&cxou j124348.4 + 113519&12:43:48.45&11:35:19.8&179.24 & [email protected] & 6.05 & 27.39&@xmath365&@xmath366 & + 133&cxou j124341.2 + 113010&12:43:41.28&11:30:10.7&180.43 & [email protected] & 3.07&7.00&@xmath367&@xmath368 & + 134&cxou j124327.9 + 113402&12:43:27.96&11:34:02.3&184.80 & [email protected] & 4.84 & 16.04&@xmath369&@xmath370 & + 135&cxou j124342.9 + 113009&12:43:42.93&11:30:09.9&185.28 & [email protected] & 5.88 & 23.23&@xmath371&@xmath372 & + 136&cxou j124336.4 + 113009&12:43:36.49&11:30:09.3&188.20&359.49@xmath6
22.32&16.36 & 808.27&@xmath373&@xmath374&c , d , e , f + 137&cxou j124352.9 + 113312&12:43:52.97&11:33:12.2&190.25 & [email protected] & 5.05 & 19.40&@xmath375&@xmath376 & + 138&cxou j124335.5 + 113609&12:43:35.55&11:36:09.2&190.71 & [email protected] & 6.44 & 31.10&@xmath377&@xmath378 & + 139&cxou j124353.1 + 113301&12:43:53.17&11:33:01.2&193.39 & [email protected] & 3.32&9.27&@xmath379&@xmath380 & + 140&cxou j124344.2 + 113004&12:43:44.29&11:30:04.6&195.89 & [email protected] & 6.38 & 27.14&@xmath381&@xmath382 & + 141&cxou j124328.2 + 113443&12:43:28.25&11:34:43.9&196.78 & [email protected] & 4.87 & 16.57&@xmath383&@xmath384&g + 142&cxou j124347.0 + 113022&12:43:47.03&11:30:22.3&196.94 & [email protected] & 4.17 & 12.62&@xmath385&@xmath386 & + 143&cxou j124326.6 + 113321&12:43:26.63&11:33:21.5&197.18 & [email protected] & 3.19 & 21.26&@xmath387&@xmath388&c + 144&cxou j124352.0 + 113443&12:43:52.00&11:34:43.6&199.24 & [email protected] & 3.54 & 10.08&@xmath389&@xmath390 & + 145&cxou j124346.3 + 113011&12:43:46.32&11:30:11.6&201.13 & [email protected] & 4.67 & 15.35&@xmath391&@xmath392 & + 146&cxou j124330.5 + 113537&12:43:30.59&11:35:37.9&202.59 & [email protected] & 3.10&8.51&@xmath393&@xmath394&g + 147&cxou j124354.5 + 113242&12:43:54.51&11:32:42.0&214.74 & [email protected] & 6.27 & 28.72&@xmath395&@xmath396 & + 148&cxou j124354.9 + 113307&12:43:54.99&11:33:07.8&219.94 & [email protected]&12.08 & 98.82&@xmath397&@xmath398 & + 149&cxou j124337.0 + 113646&12:43:37.06&11:36:46.4&220.56 & [email protected] & 8.05 & 53.38&@xmath127&@xmath399 & + 150&cxou j124350.5 + 113558&12:43:50.53&11:35:58.7&228.53 & [email protected] & 3.71 & 12.99&@xmath400&@xmath401 & + 151&cxou j124326.7 + 113514&12:43:26.78&11:35:14.9&231.16 & [email protected] & 6.51 & 31.40&@xmath402&@xmath403&g + 152&cxou j124326.4 + 113521&12:43:26.40&11:35:21.7&239.55 & [email protected] & 3.52 & 10.04&@xmath404&@xmath405&g + 153&cxou j124357.3 + 113301&12:43:57.33&11:33:01.7&254.46 & [email protected] & 5.95 & 26.83&@xmath406&@xmath407 & + 154&cxou j124341.9 + 113728&12:43:41.93&11:37:28.9&260.22 & [email protected] & 4.08 & 13.51&@xmath408&@xmath409&d + 155&cxou j124342.1 + 113743&12:43:42.17&11:37:43.9&275.52 & [email protected] & 3.34&9.39&@xmath410&@xmath411 & + 156&cxou j124357.4 + 113450&12:43:57.48&11:34:50.9&275.57 & [email protected] & 4.14 & 15.20&@xmath412&@xmath413 & + 157&cxou j124347.7 + 113740&12:43:47.71&11:37:40.2&292.68 & [email protected] & 5.13 & 24.71&@xmath414&@xmath415&d , f + 158&cxou j124345.0 + 113803&12:43:45.07&11:38:03.1&302.15 & [email protected] & 4.14 & 18.70&@xmath261&@xmath416 & + 159&cxou j124400.5 + 113230&12:44:00.57&11:32:30.0&304.60 & [email protected] & 4.45 & 19.34&@xmath417&@xmath418 & + 160&cxou j124356.9 + 113629&12:43:56.99&11:36:29.3&319.05 & [email protected] & 3.42 & 13.72&@xmath419&@xmath420&d , f + 161&cxou j124349.6 + 113809&12:43:49.60&11:38:09.9&331.09 & [email protected] & 5.47 & 30.52&@xmath421&@xmath422&d + 162&cxou j124347.9 + 113826&12:43:47.91&11:38:26.2&336.58 & [email protected] & 3.80 & 16.67&@xmath423&@xmath424&d , e , f + 163&cxou j124355.6 + 113745&12:43:55.65&11:37:45.7&358.64 & [email protected] & 3.18 & 11.90&@xmath425&@xmath426 & + 164&cxou j124406.5 + 113420&12:44:06.53&11:34:20.2&395.75 & [email protected] & 6.29 & 44.06&@xmath427&@xmath428&d + 165&cxou j124408.9 + 113333&12:44:08.95&11:33:33.2&425.71&[email protected]&23.07 & 450.59&@xmath429&@xmath430&d , e + lcccccccc sources&bremss & & & 5.81@xmath431&49.5/51&804 + sources&powerlaw & & & 1.78@xmath432&42.4/51 & 804 + sources&powerlaw@xmath433bbody&1.51@xmath434 & & 2.53@xmath435&37.3/49 & 804 + diffuse&mekal&0.805@xmath436&0.33@xmath437&&275.5/154&17187 + diffuse&mekal@xmath433powerlaw&0.796@xmath436&0.58@xmath438 & 1.76@xmath439&203.1/152&17187 + diffuse&mekal@xmath433powerlaw&0.796@xmath436&0.58@xmath440 & ( 1.78 ) & 203.2/153&17187 + diffuse&vmekal@xmath433powerlaw&0.784@xmath441&1.68@xmath442/1.12@xmath443&(1.78 ) & 185.1/152&17187 + fingers&mekal@xmath433powerlaw&0.802@xmath444&(0.6 ) & ( 1.78 ) & 35.0/33&1019 + off fingers&mekal@xmath433powerlaw&0.756@xmath445&(0.6 ) & ( 1.78 ) & 27.9/25&663 + ccccc 0 - 4 & 0.917@xmath446&0.47@xmath447&89.4/46&1897 + 4 - 7 & 0.783@xmath448&0.74@xmath449&56.8/47&2258 + 7 - 11 & 0.757@xmath450&0.69@xmath451&42.1/45&2380 + 11 - 17&0.740@xmath452&0.76@xmath453&56.2/46&2318 + 17 - 26&0.771@xmath454&0.45@xmath455&41.7/46&2185 + 26 - 38&0.802@xmath456&0.57@xmath457&46.3/44&1930 + 38 - 54&0.824@xmath458&0.97@xmath459&59.7/55&1856 + 54 - 80&0.848@xmath460&0.69@xmath461&96.0/74&2291 + 80 - 120&0.908@xmath462&0.43@xmath463&90.7/109&2469 + | we present a _
chandra _ x - ray observation of the x - ray bright e2 elliptical galaxy ngc 4649 .
in addition to bright diffuse emission , we resolve 165 discrete sources , most of which are presumably low - mass x - ray binaries ( lmxbs ) . as found in previous studies ,
the luminosity function of the resolved sources is well - fit by a broken power - law . in ngc 4697 and ngc 1553 ,
the break luminosity was comparable to the eddington luminosity of a @xmath0 neutron star . one possible interpretation of this result is that those sources with luminosities above the break are accreting black holes and those below are mainly accreting neutron stars .
the total x - ray spectrum of the resolved sources is well - fit by a hard power - law , while the diffuse spectrum requires a hard and a soft component , presumably due to the relatively soft diffuse gas and the harder unresolved sources .
we also find evidence for structure in the diffuse emission near the center of ngc 4649 .
specifically , there appear to be bright `` fingers '' of emission extending from the center of the galaxy and a 5 long bar at the center of the galaxy .
the fingers are morphologically similar to radial features seen in two - dimensional hydrodynamic simulations of cooling flows in elliptical galaxies , and although their other properties do not match the predictions of the particular simulations used we conclude that the radial fingers might be due to convective motions of hot outflowing gas and cooler inflowing gas .
the bar is coincident with the central extended radio source ; we conclude that the bar may be caused by weak shocks in the diffuse gas from an undetected low - luminosity active galactic nucleus ( agn ) . |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.